Metformin (1,1-dimethylbiguanide monohydrochloride) was purchased from Dainippon Sumitomo Pharma (Tokyo, Japan). A Cell Counting Kit (CCK-8) was purchased from Dojindo Laboratories (Kumamoto, Japan), and all other chemicals were obtained from Sigma Chemical (Tokyo, Japan).

In this study, the following antibodies were used: anti-β-actin monoclonal antibody (Sigma-Aldrich, St. Louis, MO; A5441, used at 1:3,000), cyclin D1 (Thermo Fisher Scientific, Waltham, MA; RB-9041, used at 1:1,000), cyclin E (BD Biosciences, Franklin Lakes, NJ; used at 1:1,000), Cdk6 (Santa Cruz Biotechnology, Santa Cruz, CA; sc-177, used at 1:1,000), Cdk4 (Cell Signaling Technology, Danvers, MA; #2906, used at 1:1,000), Cdk2 (Santa Cruz Biotechnology, sc-163, used at 1:2,000), and secondary horseradish peroxidase-linked anti-mouse and anti-rabbit IgG antibodies (GE Healthcare UK, Buckinghamshire, UK; used at 1:2,000).

The human HCC cell lines Alex, HLE and Huh7 were obtained from the Japanese Cancer Research Resources Bank (Tokyo, Japan). Cells of a normal human liver cell line, Hc-Cells, were obtained from DS Pharma Biomedical (Osaka, Japan). The human HCC cells were grown in RPMI-1640 (Gibco Invitrogen, Carlsbad, CA, USA) and human normal liver cells were grown in a CS-C Complete Medium Kit R (DS Pharma Biomedical, Osaka, Japan). Both media contained 10% fetal bovine serum (533–69545; Wako, Osaka, Japan) and penicillin-streptomycin (100 mg/l; Invitrogen, Tokyo, Japan).

Cell proliferation assays were performed using the Alex, HLE, and Huh7 cells according to the manufacturer’s instructions. Cells from each line (1×10⁴) were seeded into a well of a 96-well plate and cultured in 100 μl of RPMI-1640 supplemented with 10% FBS. After 24 h seeding cells were treated by addition of 1 or 5 mM metformin into the culture medium or were left untreated. At the indicated time points, the medium was exchanged for 110 μl of RPMI-1640 with the cell lines reagent (10 μl CCK-8 and 100 μl RPMI-1640), and the cells were incubated for 2 h. The absorbance was measured for each well at a wavelength of 450 nm using an auto-microplate reader.

The cell lysate assay was performed according to the methods described in our previous reports (16–18). All steps were carried out at 4°C. Protein concentrations were measured using a dye-binding protein assay based on the Bradford method (19).

Samples were electrophoresed using 10% sodium dodecyl sulfatepolyacrylamide gels (SDS-PAGE) (18), and the proteins were transferred to nitrocellulose membranes. The membranes were incubated with primary antibodies after blocking. Then, the membranes were incubated with horseradish peroxidaseconjugated secondary antibodies (20). The immunoreactive proteins were visualized with an enhanced chemiluminescence detection system (Perkin-Elmer Co., Waltham, MA) on X-ray film.

To evaluate the mechanism of growth inhibition by metformin, the cell cycle profile was analyzed after treatment with metformin. Huh7 cells (1.0×10⁶ cells in a 6-well plate dish) were treated with 10 mM metformin or without metformin for 24–72 h. After treatment, the cells were harvested and fixed in 80% ethanol. The fixed cells were washed with PBS and then stored at −20°C until flow cytometric analysis was performed. On the day of analysis, cells were washed and centrifuged using cold PBS, suspended in 100 μl PBS and 10 μl RNase A solution (250 μg/ml) and incubated for 30 min at 37°C. Then, 110 μl propidium iodide (PI) stain (100 μg/ml) was added to each tube, and the tubes were incubated at 4°C for at least 30 min prior to analysis.

Flow cytometric analysis was performed using a Cytomics FC 500 flow cytometer (Beckman Coulter, Brea, CA) appointed with an argon laser (488 nm). The percentages of cells in different phases of the cell cycle were analyzed by using FlowJo software (Tree Star, Ashland, OR). All experiments were performed in triplicate to assess for consistency of response.

Animal experiments were performed according to the guidelines of the Committee on Experimental Animals of Kagawa University and use of animals by National Cancer Research Institute (NCRI) (21). We purchased 30 male athymic mice (BALB/c-nu/nu; 8 weeks old; 20–25 g) from Japan SLC (Hamamatsu, Japan). The animals were maintained under specific pathogen-free conditions using a laminar airflow rack and had continuous free access to sterilized food (gamma-ray-irradiated food, CL-2; Clea Japan, Tokyo, Japan) and autoclaved water. Each mouse was subcutaneously inoculated with Huh7 cells (5×10⁶ cells/animal) in the flank region. After the xenografts were identifiable as a mass of more than 6 mm in maximal diameter in all recipients, the animals were randomly assigned to three groups. The metformin-treated group was injected 5 times a week intraperitoneally (i.p.) at 1 mg/body or 2 mg/body per day for 2 weeks. The control group was administered only PBS (n=10) for 2 weeks. After the initiation of the metformin administration, the tumor growth was monitored by the same investigators (H.M. and T.M.), and the tumorigenesis of HCC was monitored everyday. Tumor size was measured weekly by measuring the two greatest perpendicular tumor dimensions. To examine the significance of the differences between growth curves in this study, all the measurements of tumor volume for each growth curve from the start of the treatment to the end, typically about 30 observations, were analyzed by one-way analysis of variance (ANOVA). The tumor volume was calculated as follows: tumor volume (mm³) = [tumor length (mm) × tumor width (mm)²]/2 (22). All animals were sacrificed on day 14 after treatment. All animals were alive during the observation.

The RayBio™ Human Phospho Array Kit (cat. no. ARY 001) was purchased from RayBiotech Inc. (Norcross, GA). The assay for the p-RTK array was performed according to the manufacturer’s instructions. Briefly, p-RTK array membranes were blocked with 5% BSA/TBS (0.01 M Tris-HCl, pH 7.6) for 1 h. The membranes were then incubated with 2 ml of lysate prepared from cell lines or tumorous tissues after normalization with equal amounts of protein. After extensive washing with TBS including 0.1% v/v Tween-20 (3 washings of 10 min each), and TBS alone (2 washings of 10 min each) to remove unbound materials, the membranes were incubated with anti-phospho-tyrosine-HRP antibody for 2 h at room temperature. The unbound HRP antibody was washed out with TBS including 0.1% Tween-20. Finally, each array membrane was exposed to X-ray film using a chemiluminescence detection system (Perkin-Elmer Co.).

To assess whether molecules important in angiogenesis are regulated by the antitumor effect of metformin, the RayBio Human Angiogenesis Antibody Array (RayBiotech Inc.) was used according to the protocol. This method is a dot-blot-based assay which enables detection and comparison of 19 different angiogenesis-specific cytokines. Briefly, angiogenesis protein array membranes were blocked with 5% BSA/TBS (0.01 M Tris-HCl, pH 7.6) for 1 h. The membranes were then incubated with 2 ml of lysate prepared from tumor tissues after normalization with equal amounts of protein. After extensive washing with TBS including 0.1% v/v Tween-20 (3 washings of 5 min each) to remove unbound materials, the membranes were incubated with HRP angiogenesis antibody for 2 h at room temperature. Unbound HRP antibody was washed out with TBS including 0.1% Tween-20. Finally, each array membrane was exposed to X-ray film using a chemiluminescence detection system (Perkin-Elmer Co.).

The densitometric densities of p-EGFR and p-IGF-1R in the cell lines and p-EGFR, Mer and angiogenin in tumorous tissue were visualized in black and white, respectively. The densities of p-EGFR, p-IGF-1R and Mer obtained from the membrane array were analyzed by means of a Kodak Image Station (Kodak, Tokyo, Japan).

The samples of cancer cell lines were processed for total RNA extraction with an miRNeasy mini kit (Qiagen, Hilden, Germany) according to the manufacturer’s instructions. RNA samples typically showed A260/280 ratios of between 1.9 and 2.1, as determined using an Agilent 2100 Bioanalyzer (Agilent Technologies, Santa Clara, CA). After RNA measurement with an RNA 6000 Nano kit (Agilent Technologies), the samples were labeled using a miRCURY Hy3/Hy5 Power labeling kit and were hybridized on a human miRNA Oligo chip (v.14.0; Toray Industries, Tokyo, Japan). Scanning was performed with the 3D-Gene Scanner 3000 (Toray Industries). 3D-Gene extraction version 1.2 software (Toray Industries) was used to read the raw intensity of the image. To determine the change in miRNA expression between metformin-treated and control samples, the raw data were analyzed via GeneSpringGX v 10.0 (Agilent Technologies). Samples were first normalized relative to 28sRNA and baseline corrected to the median of all samples. Replicate data were consolidated into two groups: those from metformin-treated animals and those from control animals, and were organized by using the hierarchical clustering and analysis of variance (ANOVA) functions in the GeneSpring software. Hierarchical clustering was done by using the clustering function (condition tree) and Euclidean correlation as a distance metric. Two-way ANOVA analysis and asymptotic p-value computation without any error correction on the samples were performed to search for the miRNAs that varied most prominently across the different groups. The p-value cutoff was set to 0.05. Only changes >50% on at least one of the time points for each sample were considered significant. All the analyzed data were scaled by global normalization. The statistical significance of differentially expressed miRNAs was analyzed by Student’s t-test. All our micoroarray data in this study were submited as a complete data set to the NCBI Gene Expression Omnibus (GEO), no. 44204 (http://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?token=ljglzioeieoqijc&acc=GSE44204).

To evaluate the effect of the cell growth of metformin on human HCC cells in vitro, we examined metformin’s effect on proliferation in three HCC cell lines, namely, Alex, HLE and Huh7. Cells were grown in 10% fetal bovine serum (FBS) and treated with 1 or 5 mM metformin, or without metformin as a control. To discern the direct relationship between the decrease in cell viability and the inhibition of cell proliferation, we followed the course of proliferation over 3 days after the addition of metformin. Metformin (0, 1 and 5 mM) led to a decrease in cell proliferation in a dose- and time-dependent manner in all three cell lines (Fig. 1), with the reduction being especially prominent in Huh7. On the other hand, metformin did not suppress the proliferation of normal hepatocytes. Thus the results showed that metformin inhibited the proliferation of HCC cells, but not the proliferation of normal hepatocytes (Hc cells).

To study whether or not metformin affects the cell cycle in Huh7 cells, we examined the expression of various cell cycle-related molecules in Huh7 cells with and without metformin treatment using western blot analysis. The samples were cells treated with 10 mM metformin or left untreated for 24–72 h. The most remarkable changes were the loss of cyclin D1, Cdk4 and cyclin E, a key protein implicated in the transition of the G0/G1 phase. In short, the cyclin D1, Cdk4 and cyclin E levels had already declined rapidly at 24 h after the addition of metformin, and were no longer detectable at 24 h (Fig. 2A).

As shown in Fig. 2A, we also studied the expression of two other cell cycle-related proteins, Cdk6 and Cdk2, implicated in the G0/G1 transition. The results showed that the catalytic subunit of cyclin E, Cdk2 and the catalytic subunit of cyclin D1, Cdk6 were almost unchanged at all time points after the addition of metformin (Fig. 2A). Similar things occurred in other cancer cell lines, such as Alex and HLE (data not shown). As an internal control, the amount of β-actin was also measured and was almost the same in each sample (Fig. 2A).

Next, to further investigate the inhibition of Huh7 cell proliferation in the presence of metformin, the cell cycle progression was examined using flow cytometry. We treated proliferating Huh7 cells with 10 mM metformin for different durations. After the addition of 10 mM metformin, an increasing number of cells started to accumulate in G0/G1, 60.7% after 24 h, 59.7% after 48 h and 67.7% by 72 h (Fig. 2B). In parallel, after the addition of 10 mM metformin, we observed reductions in the percentage of cells in the S phase and G2/M phase. These data suggest that metformin inhibits the cell cycle progression from G0/G1 into S phase, resulting in G1 cell cycle arrest.

In order to determine whether metformin could affect tumor growth in vivo, metformin was injected daily intraperitoneally (i.p.) at 1 or 2 mg per day in xenografted mice. Animals in the control group received i.p. administration of the vehicle (PBS). As shown in Fig. 3A and B, animals in the control group developed rapidly-growing subcutaneous HCC. On the basis of the integrated values of the tumor growth curves, i.p. administration of metformin led to a substantial inhibition of tumor growth, by 25% (1 mg/day) and 52% (2 mg/day) (Fig. 3A and B). These growth rates were significantly above those of the control (both p<0.01, respectively). In addition, 2 mg metformin treatment significantly inhibited tumor growth compared with that in 1 mg metformin-treated mice (p<0.05). All animals remained alive throughout the experiment, and metformin exhibited no apparent changes in mice and did not affect their weight (data not shown).

To determine whether metformin also affected cell cycle-regulatory protein levels in vivo, we performed a western blot analysis in tumorous tissues obtained from the xenograft experiments. Metformin (1–2 mg dose) reduced the levels of cell cycle-related molecules (cyclin D1, Cdk4, cyclin E) in treated tumors compared with controls (Fig. 3C). However, the expressions of Cdk2 and Cdk6 in tumorous tissue remained unchanged. These data were similar to the results of the in vitro observation (Fig. 2A). In summary, in the in vivo study, as in the in vitro study, metformin decreased tumor growth by reducing cell cycle-regulatory protein levels, especially in cyclin D1, Cdk4 and cyclin E, resulting in G1 cell cycle arrest.

We used a phosphorylated-RTK array system to identify the key RTKs associated with the antitumor effect of metformin. By using an antibody array (Fig. 4A), we simultaneously screened the expressions of 42 different activated RTKs in Huh7 cells and tumorous tissues with or without metformin. In the in vitro study, the expressions of phosphorylated-epidermal growth factor (p-EGFR) and phosphorylated-insulin-like growth factor-1 receptor (p-IGF-1R) in Huh7 cells were reduced by treatment with metformin (Fig. 4B), while in the in vivo study, the levels of p-EGFR and Mer in tumorous tissues (Fig. 4C) were reduced by the treatment with metformin. The densitometric ratios of the p-EGFR and p-IGF-1R spots of the metformin-treated cell lines to the non-treated cell lines were 51 and 12%, respectively (Fig. 4D). In addition, the ratios of p-EGFR and Mer of the metformin-treated tumorous tissue to the non-treated tissue were 67 and 55%, respectively (Fig. 4E).

We used an angiogenesis array system to identify the key angiogenesis-related molecules associated with the antitumor effect of metformin. The antibody array (Fig. 5A) allowed us to simultaneously screen the expressions of 20 different angiogenesis molecules in Huh7 cells (Fig. 5B) and tumor tissues (Fig. 5C) with or without metformin. Among the 20 angiogenesis molecules screened, angiogenin was reduced in in vitro and in vivo study with metformin treatment. The densitometric densities on angiogenin in cell line and tumorous tissue were visualized in black and white, respectively. The densitometric ratios of the angiogenin spots from metformin-treated cell line and tumorous tissues to those of cell line and tumorous tissues not treated with metformin were 51% (Fig. 5D) and 54% (Fig. 5E).

Using a custom microarray platform, we analyzed the expression levels of 985 human miRNA probes in the cell lines treated with metformin and the control cell line. As shown in Table I, when the expression of miRNAs was studied in Huh7 cells treated with 10 mM metformin and without metformin in vitro, 33 of 985 miRNAs were significantly upregulated in Huh7 cells at 72 h after metformin treatment (Table I), while 18 of 985 miRNAs downregulated (Table II).

Unsupervised hierarchical clustering analysis, using Pearson’s correlation, showed that cell lines treated with metformin clustered together and separately from the untreated cell lines (Fig. 6). These subsets of 119 microRNAs in cell lines exhibited >1.5-fold alterations in expression levels between the metformin-treated and control groups.