Metformin potentiates the effect of arsenic trioxide suppressing intrahepatic cholangiocarcinoma: roles of p38 MAPK, ERK3, and mTORC1

The antiproliferative effects of combination treatment with metformin and ATO were investigated in CCLP-1, RBE, and HCCC-9810 cells using a system of real-time cell growth monitoring with an RTCA DP Analyzer. As shown in Fig. 1a, metformin (10 mM) in combination with ATO (3 μM) nearly abrogated the growth of ICC cells, according to the cell index. Moreover, to investigate whether metformin plus ATO acts synergistically, a median effect analysis [17] was performed, which showed that the combination index (CI) value trended toward values of less than one and then further decreased, as the combined antiproliferative effects increased, indicating the combination was synergistic (Fig. 1b). The concentrations of the two agents in each group and the corresponding combined antiproliferative effect values and Cis are listed in Additional file 1.

Next, we performed a cell cycle and apoptosis analysis. A significant increase in the number of cells arrested in G0/G1 phase was observed in ICC cells treated with a combination of metformin and ATO compared to the single-drug treatments, as shown in Fig. 1c, d. Similarly, the apoptosis assay, as shown in Fig. 2a, revealed that a significant increase in the number of apoptotic cells was observed in ICC cells with a combination of metformin and ATO treatment, compared to the single-drug treatment. Consistent with the greater apoptotic events, the combined treatment led to higher levels of cleaved PARP and cleaved Caspase-3 and Caspase-3/7 activity in the three ICC cell lines compared to the single-drug treatment alone (Fig. 2c, d). As ATO can induce mitochondrial-dependent ROS production in cancer cells [18], we further determined the influence of metformin on ATO-induced ROS production. As expected, the data in Fig. 2b shows that metformin (10 mM) readily promoted ATO (3 μM)-induced ROS production.

Collectively, these results revealed that combined treatment with metformin and ATO remarkably promotes antiproliferative effects by promoting cell apoptosis, inducing G0/G1 cell cycle arrest, and increasing intracellular ROS.

To analyze the potential molecular mechanism, we investigated the effects of metformin and ATO on the mTORC1 and AMPK/MAPK pathways in ICC cells. Metformin and ATO cooperated to abrogate the activation of mTORC1 (p-mTOR Ser2448, p-p70S6K Thr389, p-Raptor Ser792, p-4E BP1 Thr37/46), as shown in Fig. 3a, b. The inactivation effect of metformin (10 mM) on mTORC1 seemed to be more profound than that of ATO (3 μM). The consequences of combination treatment with metformin and ATO on AMPK/MAPK networks were further investigated (Fig. 3c, d), revealing that metformin and ATO could activate AMP-activated protein kinase (AMPK) while the p-ERK (T202/204) regulated by metformin, and ATO was inconsistent in the three cell lines used. However, activation of p38 MAPK was observed in ATO-treated cells, and this activation appeared to be almost completely abrogated by metformin.

Moreover, to explore the molecular mechanism in detail, we used a phospho-antibody array and screened more than 200 molecules clustered in twelve cancer-related pathways and found that total ERK3 exhibited the highest variation in ICC cells after treatment with metformin and ATO. The array data were quantified and are listed in Additional file 2. ERK3, encoded by MAPK6, is an atypical member of the MAPK family and is, thus far, primarily believed to be an antitumor molecule [19, 20]. Although, recent studies have reported that ERK3 can promote cancer cell migration and invasion [21, 22]. Thus, to further define the role of EKR3 in the metformin and ATO-induced antiproliferative effect on ICC cells, we demonstrated the upregulation of ERK3 in ICC cells treated with both metformin and ATO using a western blot assay (Fig. 3c, d). Phosphorylated ERK3 (detected by an anti-MAPK6 (phospho S189) antibody from Abcam (ab 74032)) was detected, and no significant variation was found when compared to total EKR3 (data not shown). Therefore, part of the following work focused on the total ERK3 variation in the mechanism studies.

To further validate our in vitro results showing the antiproliferative effects of combined treatment with metformin and ATO, we treated male athymic nude mice bearing palpable tumors (approximately 100 mm³) of CCLP-1 xenografts with control (vehicle-treated mice), metformin (200 mg/kg/day), ATO (3 mg/kg/day), and a combination of metformin and ATO (Figs. 4a–c) for 21 days. The tumor volumes of the combination treatment group were significantly reduced compared to the groups treated with either metformin or ATO alone. To evaluate the safety, the weights of the mice were obtained, and no significant variation was found (Fig. 4d), implying that treatment with a combination of metformin and ATO was very safe. Consistent with the in vitro data, immunohistochemistry and TUNEL analyses (Fig. 4e) of the xenograft tumors revealed that the metformin and ATO combination effectively inhibited the expression of Ki67, a marker representing tumor proliferation. Moreover, metformin and ATO promoted apoptosis in xenograft tumors, as evaluated by the cleaved caspase-3 staining and TUNEL assay.

Similarly, xenograft tumors were further examined for the status of mTORC1, AMPK/p38 MAPK, and ERK3 after treatment with the control, metformin, ATO, or a combination of both (Fig. 4f, g). Consistent with the in vitro results, metformin, and ATO cooperated to inactivate mTORC1, activate AMPK, and upregulate ERK3. Moreover, activation of p38 MAPK was also observed in mice in the ATO group, and this activation could be abrogated by metformin (Fig. 4f, g). These results supported the findings of the molecular network in the previous in vitro experiments.

To characterize the mechanism responsible for metformin-ATO synergism in ICC, we first thoroughly explored the roles of AMPK and p38 MAPK in the synergistic effect. Accordingly, pharmacological and genetic methods were used to regulate the expression level or active status of AMPK and p38 MAPK. As shown in Fig. 5a, b, and f, downregulation or inactivation of AMPK by short interfering RNA (siRNA) or by compound C [23] rescued the viability of ICC cells treated with the metformin and ATO combination. An AMPK activator, AICAR, significantly sensitized ICC cells to ATO (Fig. 5c), which further demonstrated the AMPK-dependence of the antiproliferative effect of metformin plus ATO on ICC cells. Additionally, similar methods were used for p38 MAPK, and the results (Fig. 5d–f) revealed that inactivation of p38 MAPK by SB203580 [24] or specific siRNAs promoted the effect of metformin and ATO on ICC cells, implying that metformin-induced abrogation of activated p38 MAPK by ATO may contribute to the metformin-ATO synergism in ICC cells.

As shown in Fig. 6a, b, downregulation of ERK3 by MAPK6-specific siRNAs rescued the viability of ICC cells treated with metformin and ATO, which implied that ERK3 is a suppressive molecule in ICC cells.

Inactivation of AMPK by compound C or specific siRNA partially rescued metformin-induced inhibition of p38 MAPK and partially abrogated metformin plus ATO-induced upregulation of ERK3 and inhibition of mTORC1 in ICC cells (Fig. 7a, c). This finding was also supported by the molecular regulation observed in cells treated with AICAR and ATO, shown in Fig. 7b. For p38 MAPK, inactivation of p38 MAPK by SB203580 or specific siRNA promoted the inactivation of mTORC1 in ICC cells treated with metformin and ATO (Fig. 7c); however, our results did not reveal the relationship between p38 MAPK and ERK3. (Fig. 7a) Moreover, downregulation of ERK3 by siRNA rescued the active status of mTORC1 inhibited by the metformin and ATO combination treatment (Fig. 7b), which implied that ERK3 is a suppressor of mTORC1.

Collectively, we propose a potential molecular mechanism in which metformin and ATO inhibit ICC development via modulation of a network involving the AMPK, p38 MAPK, ERK3, and mTORC1 pathways (Fig. 7e). Specifically, metformin and ATO cooperated to inhibit mTORC1 through activation of AMPK and upregulation of ERK3. Meanwhile, metformin abrogated the activation of p38 MAPK induced by ATO, which partially depended on AMPK activation. The western blot data of the three ICC cell lines are shown in Additional file 3.

The role of ERK3 in cancer is rarely explored. Therefore, we further explored the function of ERK3 in ICC. We found that overexpression of EKR3 in ICC cells inhibited cell proliferation (Fig. 8a) through inactivation of mTORC1 (Fig. 8b). Overexpression of EKR3 also inhibited the ICC xenograft growth (Fig. 8c). In addition, to explore whether ERK3 could be a promising molecular marker for predicting the prognosis of ICC patients, we detected the expression level of ERK3 in tumor samples from 73 ICC patients and compared the survival times of the patients with the expression level (high or low) of ERK3 (Additional file 4). As shown in Fig. 8d, high ERK3 expression is associated with a better prognosis in ICC patients. In addition, COX multivariate analysis showed that expression of ERK3 (low) and TMN stages (III and IV) are the independent risk factors for shorter overall survival times (Table 2 in Additional file 5). Collectively, these data implied that ERK3 is a suppressor of ICC development. Moreover, when correlated with clinical findings, we found that the expression level of ERK3, segregated as high or low, displayed a significant correlation with vessel invasion, implying that ERK3 has an antiangiogenic effect in ICC (Table 1 in Additional file 4), which is worth further exploration in future studies.

Three ICC cell lines, CCLP-1 [39], RBE, and HCCC-9810, were used. The CCLP-1 cell line was obtained from DSMZ (Braunschweig, Germany). RBE and HCCC-9810 cells were purchased from the Type Culture Collection of the Chinese Academy of Sciences (Shanghai, China). CCLP-1 cells were cultured in Dulbecco’s modified Eagle’s medium (DMEM)-high glucose (Gibco, USA). RBE and HCCC-9810 cells were cultured in RPMI-1640 (Gibco, USA) supplemented with 10% fetal bovine serum (FBS; Gibco) and 100 μg/ml each of penicillin and streptomycin (Invitrogen, USA) in 5% CO₂ at 37°°C.

Metformin (1,1-dimethylbiguanide hydrochloride, #D150959-5G) and AICAR(5-Aminoimidazole-4-carboxamide 1-β-D-ribofuranoside Acadesine N1-(β-D-Ribofuranosyl)-5-aminoimidazole-4-carboxamide, A9978) were purchased from Sigma-Aldrich (St. Louis, MO, USA). Compound C (Dorsomorphin, 6-[4-(2-Piperidin-1-ylethoxy)phenyl]-3-pyridin-4-ylpyrazolo[1,5-a]pyrimidine, P5499) and SB239063 (trans-1-(4-Hydroxycyclohexyl)-4-(4-fluorophenyl)-5-(2-methoxypyridimidin-4-yl)imidazole, S7741) were purchased from Selleck Chemicals (Houston, TX, USA). Arsenic trioxide (As2O3, ATO) was purchased from Shuanglu Pharmaceutical Co., Ltd. (Beijing, China). The cell counting kit-8 (CCK-8, KGA317), the Annexin V-FITC Apoptosis Detection Kit (KGA108), the cell cycle detection kit (KGA512), Caspase3/7 Activity Detection Kit (KGAS037), the terminal deoxynucleotidyltransferase-mediated deoxyuridine triphosphate (dUTP) nick-end labeling (TUNEL) Assay Kit (KGA707), and the ROS (reactive oxygen species) Detection Kit (KGT010) were purchased from KeyGen Biotech (Nanjing, China). The HistostainTM-Plus Kits (IgG/Bio, S-A/HRP, DAB) were purchased from Zhongshan Golden Bridge Co., Ltd. (Beijing, China).

The following antibodies were used for immunoblotting or immunohistochemical staining: β-actin (sc-47778) was from Santa Cruz Biotechnology, Inc. Sant (Santa Cruz, CA, USA). AMPKα (#2532) and phosphorylated AMPKα (Phospho-Thr172, #2535), mTOR (#2983), phosphorylated mTOR (Phospho-Ser2448, #5536), phosphorylated Raptor (Phospho-Ser792, #2083), phosphorylated p70 S6 kinase (Phospho-Thr389, #9234), phosphorylated 4E-BP1(Phospho-Thr37/46, #2855), cleaved PARP (#5626), cleaved caspase-3(#9661), ERK (#4696), phosphorylated ERK(Phospho-Thr202/Tyr204, #4370), p38 MAPK(#8690), phosphorylated p38 MAPK(Thr180/Tyr182, #4511), and Ki-67 (#9449) were from Cell Signaling Technology, Inc. (Danvers, MA, USA). ERK3(ab53277) was from Abcam (Cambridge, MA, USA). Goat anti-rabbit and goat anti-mouse IgG peroxidase-conjugated secondary antibodies (31460 and 31430) were from Thermo-Pierce (Rockford, IL, USA).

The real-time cell growth was monitored using a 16 × E-Plates RTCA DP Analyzer (ACEA Biosciences). After 5000 cells were seeded in a well of 16 × E-Plates for 16 to 24 h, the medium was replaced by fresh medium with certain agents (control, metformin (10 mmol/L), ATO (3 μmol/L) or a combination of both). Then, the cell index, representing the cell viability, was detected every 15 min for 72 h. Four replicates were prepared for each condition. The growth cures are shown.

Cell viability was determined using a CCK-8 assay according to the manufacturer’s instructions. The detailed procedure has been described previously [6].

The synergism of metformin and ATO was determined using the method of Chou [17] and the software package Calcusyn (Biosoft, Cambridge, United Kingdom). Combination indices (CI) were calculated, and a CI less than 1 was defined as synergism.

The cell cycle arrest, induction of apoptosis, and intracellular ROS levels were determined by flow cytometry. Portions of the detailed procedure have been described previously [6].

For the detection of intracellular ROS, an oxidation-sensitive fluorescent probe (DCFH-DA) was used. After treatment with agents for 24 h, a total of 1 × 10⁶ cells were trypsinized and pelleted by centrifugation and washed twice with PBS. Then, the cell pellets were resuspended in 1 mL of DMEM or RPMI-1640 (serum-free) with 10 μm/L DCFH-DA and incubated for 20 min at 37 °C. The stained cells were analyzed using a flow cytometer (Accuri Cytometers Inc.). The FL1-A received the fluorescence induced by DCF. For each sample, 10, 000 events were collected.

A Caspase3/7 activity detection kit was used according to the manufacturer’s instructions. Briefly, after treatment with agents for 48 h, 1 × 10⁶ cells were incubated in 1 μm/L working solution for 30 min at 37 °C. Finally, the optical density was measured using a microplate reader (Thermo, Varioskan Flash).

Phosphoprotein profiling by the cancer-signaling phospho-antibody microarray—the cancer-signaling phospho-antibody microarray PCS248, which was designed and manufactured by Full Moon Biosystems, Inc. (Sunnyvale, CA), contains 248 antibodies. Each of the antibodies has six replicates that are printed on coated glass microscope slides, along with multiple positive and negative controls. The antibody array experiment was performed by Wayen Biotechnology (Shanghai, China), according to their established protocol. Briefly, cell lysates obtained from CCLP-1 cells treated with single or double agents were biotinylated with an Antibody Array Assay Kit (Full Moon Biosystems, Inc.). The antibody microarray slides were first blocked in a blocking solution (Full Moon Biosystems, Inc.) for 30 min at room temperature, rinsed with Milli-Q grade water for 3–5 min, and dried with compressed nitrogen. The slides were then incubated with the biotin-labeled cell lysates (~100 μg of protein) in coupling solution (Full Moon Biosystems, Inc.) at room temperature for 2 h. The array slides were washed 4–5 times with 1X Wash Solution (Full Moon Biosystems, Inc.) and rinsed extensively with Milli-Q grade water before detection of bound biotinylated proteins using Cy3-conjugated streptavidin. The slides were scanned on a GenePix 4000 scanner, and the images were analyzed with GenePix Pro 6.0 (Molecular Devices, Sunnyvale, CA). The fluorescence signal (I) of each antibody was obtained from the fluorescence intensity of antibody-stained regions. A ratio computation was used to measure the extent of protein phosphorylation. The phosphorylation ratio was calculated as follows: phosphorylation ratio = phospho value/non-phospho value. The total proteome ratios were standardized by β-actin.

Cells after different treatments or tumor tissues from a xenograft were harvested for western blot analysis. The detailed procedure has been described previously [6]. Primary antibodies (described in the Antibodies section) were incubated at 4 °C overnight. The bands were visualized by chemiluminescence, imaged using a ChemiDoc XRS and analyzed using Image Lab (both from Bio-Rad).

To investigate the antiproliferative effect of metformin and ATO in combination treatment on ICC cells in vivo, a model of nude mice bearing ICC cell xenografts was established. Five-week-old male athymic nude mice were obtained from the Animal Facility of Zhejiang University. The mice were maintained under pathogen-free conditions and were provided with sterilized food and water. Briefly, 5 × 10⁶ CCLP-1 cells were injected subcutaneously into the right flank of each nude mouse. When mice exhibited palpable tumors (the tumor volume was approximately 100 mm³), they were randomly divided into control (100 μL of NS by intraperitoneal injection), metformin (150 mg/kg/day diluted in 100 μL of NS by intraperitoneal injection), ATO (2.5 mg/kg/day diluted in 100 μL of NS by intragastric administration), and a combination of both (metformin, 150 mg/kg/day plus ATO 3 mg/kg/day diluted in 100 μL of NS by intraperitoneal injection) groups (n = 6 animals per group). The treatments were performed five times per week for 3 weeks. The tumor volume was detected every week and was calculated using the following formula: volume = 1/2 (length × width²). After 4 weeks, all the mice were sacrificed, and the tumors were isolated. NS means normal saline.

To evaluate the antiproliferative effect of ERK3 in ICC in vivo, 5 × 10⁶ CCLP-1 cells with ERK3 overexpression or GFP transfection were injected subcutaneously into the right flank of each nude mouse. After 4 weeks, the tumor volume was determined and calculated, as previously described, all the mice were sacrificed, and the tumors were isolated.

The tumors isolated from the mice were paraffin embedded and cut into 10-μm-thick sections in a microtome cryostat (HM500 OM, Carl Zeiss, Germany). Immunohistochemical staining was conducted according to the manufacturer protocols described for HistostainTM-Plus Kits. Primary antibodies (described in the Antibodies section) were incubated at 4 °C overnight. Images were captured with a light microscope (Axiolab, Carl Zeiss, Germany), and five images/sample were prepared. Image-Pro Plus 4.5 Software was used to analyze the staining data.

ERK3 expression in the 73 cases was evaluated by two of the authors (Hai-Yang Xie and Fan Yang), who were blind to the clinicopathological data. Any discrepancy between the two individuals was resolved by a third individual. A semiquantitative scoring system was used. Brown granules in the cytoplasm representing ERK3 were considered positive staining. We scored the staining intensity as follows: 0, no staining; 1+, mild staining; 2+, moderate staining; 3+, intense staining. The area of staining was evaluated as follows: 0, no staining of cells in any microscopic field; 1+, <30% of tissue stained positive; 2+, between 30 and 60% of tissue stained positive; and 3+, >60% of tissue stained positive. ERK3 expression was evaluated by combining the staining intensity and area assessments. The minimum score when summed (intensity + extension) was 0, and the maximum score was 6. Scores ≤3 were defined as low expression, and scores >3 were defined as high expression.

In situ detection of apoptotic cells in the tumors isolated from the mice was performed with a TUNEL assay. The tumors were paraffin embedded and cut into 10-μm-thick sections in a microtome cryostat (HM500 OM, Carl Zeiss, Germany). The TUNEL assay was conducted according to the manufacturer’s protocols. 3,3-Di- aminobenzidine (DAB) was used as the substrate for the peroxidase. Images were captured with a light microscope (Axiolab, Carl Zeiss, Germany), and five images/sample were prepared. Image-Pro Plus 4.5 Software was used to analyze the staining data.

ICC cells were seeded in a 6-well plate at a concentration of 2 × 10⁵ cells per 2 ml in medium with FBS, penicillin, and streptomycin. After 24 h, the medium was replaced with fresh medium without antibiotics or FBS. Meanwhile, siRNA or a negative control oligonucleotide was transfected into the cells with Lipofectamine 3000, according to the manufacturer’s instructions. After 24 h, the medium was replaced with fresh complete medium. Then, the cells were cultured for subsequent experiments. All siRNAs were obtained from Shanghai GenePharma Co., Ltd. China; the sequences of siRNAs are listed in Additional file 6.

Lentiviral particles containing MAPK6 vector or empty vector were obtained from Shandong Vigenebio Co., Ltd. (Jinan, China). Cells were plated in a six-well plate at 2 × 10⁵ cells/well in supplemented medium 24 h before viral infection. Media were removed from well plates and replaced with media supplemented with Polybrene at a final concentration of 5 μg/ml. Next, cells were infected by adding lentiviral particles to the culture. Stable clones expressing MAPK6 or GFP were selected using puromycin dihydrochloride. Then, cells were collected for gene expression assays.

Tumor samples from a total of 73 ICC patients who underwent operations at our hospital (First Affiliated Hospital, Zhejiang University School of Medicine, Zhejiang, China) between 2010 and 2014 were used in the present study. These patients were diagnosed with ICC either before or after surgery. No treatment was done for the patients before surgery. The diagnosis was confirmed by histopathological examination, and complete clinical and laboratory data were collected before surgery and during follow-up. The distribution of clinicopathological data in the study cohort is given in Table 1 of Additional file 4. Specimens of cancer tissues and clinical information were available from these patients after obtaining informed consent.