Article Text

PDF

Original article
Metformin decreases hepatocellular carcinoma risk in a dose-dependent manner: population-based and in vitro studies
  1. Hsiao-Ping Chen1,2,
  2. Jeng-Jer Shieh3,4,
  3. Chia-Che Chang3,5,
  4. Tzu-Ting Chen4,
  5. Jaw-Town Lin2,6,7,
  6. Ming-Shiang Wu2,
  7. Jeng-Horng Lin1,
  8. Chun-Ying Wu4,5,8
  1. 1Department of Life Sciences, National Chung Hsing University, Taichung, Taiwan
  2. 2Division of Gastroenterology, National Taiwan University Hospital, Taipei, Taiwan
  3. 3Institute of Biomedical Sciences, National Chung Hsing University, Taichung, Taiwan
  4. 4Department of Education and Research and Division of Gastroenterology, Taichung Veterans General Hospital, Taichung, Taiwan
  5. 5Graduate Institute of Basic Medical Science and Department of Public Health, China Medical University, Taichung, Taiwan
  6. 6Department of Internal Medicine, E-Da Hospital and I-Shou University, Kaohsiung, Taiwan
  7. 7Center for Health Policy Research and Development, National Health Research Institutes; Miaoli, Taiwan
  8. 8Faculty of Medicine, School of Medicine, National Yang-Ming University, Taipei, Taiwan
  1. Correspondence to Professor Chun-Ying Wu, Faculty of Medicine, School of Medicine, National Yang-Ming University, Taipei, Taiwan, 155, Sec. 2, Linong Street, Taipei 112, Taiwan; chun{at}vghtc.gov.tw

Abstract

Objective Type 2 diabetes mellitus is associated with a higher risk of hepatocellular carcinoma (HCC), which is attenuated by the use of metformin. However, there are no studies addressing the effect of metformin on hepatocarcinoma cells from the antitumoural perspective.

Design In the nationwide case-control study, the authors recruited 97 430 HCC patients and 194 860 age-, gender- and physician visit date-matched controls. The chemopreventive effects of metformin were examined by multivariate analysis and stratified analysis. The in vitro effects of metformin on cell proliferation and cell cycle were studied in HepG2 and Hep3B hepatoma cell lines.

Results The OR of diabetes in HCC patients was 2.29 (95% CI 2.25 to 2.35, p<0.001). Each incremental year increase in metformin use resulted in 7% reduction in the risk of HCC in diabetic patients (adjusted OR=0.93, 95% CI 0.91 to 0.94, p<0.0001). In the multivariate stratified analysis, metformin use was associated with a reduced risk of HCC in diabetic patients in nearly all subgroups. Cell line studies showed that metformin inhibits hepatocyte proliferation and induces cell cycle arrest at G0/G1 phase via AMP-activated protein kinase and its upstream kinase LKB1 to upregulate p21/Cip1 and p27/Kip1 and downregulate cyclin D1 in a dose-dependent manner, but independent of p53. Combined treatment of oral metformin with doxorubicin functioned more efficiently than either agent alone, in vivo.

Conclusions Use of metformin is associated with a decreased risk of HCC in diabetic patients in a dose-dependent manner, via inhibition of hepatoma cells proliferation and induction of cell cycle arrest at G0/G1 phase.

  • Hepatocellular carcinoma
  • diabetes
  • metformin
  • cell cycle
  • dose-dependent
  • Helicobacter pylori
  • molecular biology
  • cell biology
  • gastric lymphoma
  • gastric cancer
  • Helicobacter pylori–pathogenesis
  • molecular oncology
  • gastrointestinal neoplasia
  • gene expression
  • oxidative stress
  • cancer epidemiology
  • bleeding
  • aspirin

Statistics from Altmetric.com

Significance of this study

What is already known on this subject?

  • Type 2 diabetes mellitus is associated with increased risks for hepatocellular carcinoma (HCC).

  • Metformin has been found to lower HCC risk in clinical observational studies.

  • In vitro studies show that metformin is an activator of AMP-activated protein kinase signalling and reduces mTOR pathway.

What are the new findings?

  • In the nationwide population study, we found that use of metformin is associated with a decreased risk of hepatocellular carcinoma (HCC) in diabetic patients in a dose-dependent manner.

  • Each incremental year increase in metformin use results in 7% reduction in the risk of HCC.

  • In vitro study shows that metformin inhibits hepatocyte proliferation and induces cell cycle arrest at G0/G1 phase via AMP-activated protein kinase and its upstream kinase LKB1 to upregulate p21/Cip1 and p27/Kip1 and downregulate cyclin D1 in a dose-dependent manner, but independent of p53.

  • In vivo study shows that combined treatment of oral metformin with doxorubicin functioned more efficiently than either agent alone.

How might it impact on clinical practice in the foreseeable future?

  • Using metformin in diabetic patients to decrease the risk of hepatocellular carcinoma should be recommended.

Introduction

Type 2 diabetes mellitus (DM) is associated with increased risks for cancers of the liver, pancreas and endothelium, among others.1 For hepatocellular carcinoma (HCC), a twofold to threefold higher risk in diabetic patients has been reported, regardless of other HCC risk factors, such as alcoholic liver disease or viral hepatitis.2 ,3 Several factors may explain the association between diabetes and HCC. First, diabetes enhanced the risk for non-alcoholic fatty liver disease and non-alcoholic steatohepatitis, which can lead to cirrhosis and HCC.4 ,5 Increasing evidence has suggested that insulin resistance plays a role in hepatic carcinogenesis via activation of IGF-1 signalling axis and increased fat accumulation in hepatocytes.6 On the other hand, liver cirrhosis may further promote glucose intolerance and overt diabetes in certain patients through impaired insulin secretion or insulin resistance.7

Metformin, a drug that improves insulin sensitivity, inhibits hepatic gluconeogenesis and decreases glycogenolysis8 has been found to lower HCC risk more than insulin or insulin secretogogue in recent studies. In a case-control study, Donadon et al reported an inverse relationship between HCC and metformin use.9 In a prospective cohort of cirrhotic patients with hepatitis C, Nkontchou et al found that metformin use is associated with a significantly reduced HCC risk.10 Metformin is an activator of AMP-activated protein kinase (AMPK) signalling,11 ,12 and reduces mTOR pathway,13 which may explain its chemopreventive effects. Metformin also inhibits cancer cell growth by inducing cell cycle arrest and enhancing apoptosis.14 ,15

Although previous studies have provided important evidence for the chemopreventive roles of metformin, the dose-dependent effect of metformin in HCC remains unclear. Furthermore, there are no relevant studies in the literature addressing the effect of metformin on hepatocarcinoma cells from the antitumoural perspective. Thus, the aims of this study are to examine the dose-dependent chemopreventive effect of metformin in a nationwide population and the antitumoural cell biological effects of metformin on hepatoma cell lines.

Patients and methods

Study design and population

To carry out a population-based case-control study, we recruited HCC patients and matched controls based on data from Taiwan's National Health Insurance Research Database (NHIRD). The NHIRD consists of detailed healthcare data from nearly the entire Taiwan population of 23 million, which has been described in our previous studies.16–18 International classification of diseases-9 (ICD-9) codes were used to define diseases. HCC diagnosis was defined according to the Registry for Catastrophic Illness Patient Database (RCIPD). RCIPD is a subpart of the NHIRD and requires specific registration procedures to confirm the diagnosis of HCC. Most HCC patients in the RCIPD were histologically approved by surgery or biopsy. For those not suitable for surgery or biopsy, typical computerised tomography images were required to registry as HCC patients. We have even validated the cases in RCIPD by reviewing a selected sample of 354 medical records with 100% accuracy in our previous study.18 ,19

HCC subjects and matched controls

All newly diagnosed HCC patients in the RCIPD between 1 January 1997 and 31 December 2008 (ICD-9 codes: 155, 155.0, 155.2) were recruited. Each HCC patient was matched with two matched controls from the Longitudinal Heath Insurance Database 2000 and 2005 (LHID2000 and LHID2005), which contained 1 000 000 randomly sampled subjects from 2000 to 2005, respectively. The subjects who were sampled repeatedly in LHID2000 and LHID2005 were excluded. The controls were matched to cases on age (±1 year), gender and date of physician visit (within 1 year of HCC diagnosis date for the matched case). Patients with previous malignancies, younger than 20 years of age, or who had undergone liver surgery were excluded. We designated the index date as the date of HCC diagnosis for cases or the matched date of physician visit for controls. The observation period was defined as the 5 years before the index date.

Diabetes diagnosis, duration and controls

In Taiwan, the diagnosis of DM was based on one of the following criteria with one repeat: fasting plasma glucose ≥126 mg/dl, 2 h plasma glucose ≥200 mg/dl during an oral glucose tolerance test and a random plasma glucose ≥200 mg/dl in a patient with classic symptoms of hyperglycaemia or hyperglycaemic crisis. Once DM was diagnosed, non-pharmacological therapy, including weight reduction, diet control and increased physical activity, will be suggested in the first 3 months. If lifestyle modifications did not succeed, pharmacological therapy will be initiated. According to these diabetic care policies, we defined patients as having diabetes if they fulfilled one of the following two criteria: diagnosed three times as diabetes in outpatient clinics or admitted once due to diabetes before the index date and had taken or were taking oral hypoglycaemic agents (OHAs) for more than 28 days. Diabetic duration was defined as the period between the date of diagnosis as diabetes and the index date. Diabetic control was assessed by the frequency of physician visits per year.

Metformin and other diabetic medications

For diabetic patients, exposure to metformin and other diabetic medications was analysed according to prescription records in NHIRD. Patients who ever used metformin, insulin, thiazolidinediones or other OHAs during the observation periods were defined as metformin users, insulin users, thiazolidinediones users and other OHA users, respectively. Patients who ever used more than one kind of medications were classified into more than one group independently. The total durations of diabetic medications use were also analysed to examine the dose-dependent antitumoural effect.

Hepatitis, cirrhosis and comorbidities

Patients were defined as having hepatitis B (ICD-9: 070.2 and 070.3), hepatitis C (ICD-9: 070.41, 070.44, 070.51, 070.54) or liver cirrhosis (ICD-9: 571.5), if they fulfilled one of the following two criteria: three diagnoses coded in outpatient clinics or one diagnosis on admissions prior to the index date. Comorbidities included previous ischaemic heart disease, cerebral infarction, hypertension, chronic obstructive lung disease and hyperlipidaemia diagnosed on admissions before the index date.

Cell culture

Human hepatoma cell lines, HepG2 and Hep3B, were maintained in DMEM medium (Invitrogen, Carlsbad, California, USA) at 37°C in a 5% CO2 atmosphere at constant humidity. The medium was supplemented with 10% fetal bovine serum (Invitrogen) and 1% antibiotic solution (Invitrogen) and was changed every other day. For subculturing, cells were rinsed once with phosphate buffered saline (PBS) and incubated in 0.25% trypsin containing 0.02% EDTA (Invitrogen) in PBS for 5 min.

Cell viability

The effect of metformin on cell viability was determined on an MTS (3-(4,5-dimethylthiazol-2-yl)-5-(3-carboxymethoxyphenyl)-2-(4-sulfophenyl)-2H-tetrazolium) assay. Briefly, cells were seeded onto 96-well plates at 1×103 cells per well. The cells were treated with 100 μl of medium containing various concentrations (0, 0.5, 2, 5, 10 or 20 mM) of metformin for 48 h, or treated with metformin 20 mM for 0, 1, 2, 3 days. In the co-treatment assay, we first added metformin 20 mM after 24 h and then added 1 μg/ml doxorubicin for another 24 h. Twenty microlitres of MTS (5 mg/ml in PBS) was added to each well and incubated at 37°C for 1 h. Then, the absorbance was measured at 490 nm.

Colony formation

For assay of colony formation, 200 cells were seeded into 6 cm culture dishes and allowed to attach for 16 h. The medium was removed and the cells were treated with 0, 0.5, 2, 5, 10 or 20 mM of metformin. The culture dish was incubated for 15 days under the same conditions used for culturing and the cells were fixed with methanol and stained with haematoxylin. Colonies containing more than 50 cells originating from single cells were counted under inverted microscope.

For soft agar colony-forming assay, 1×103 exponentially growing cells were suspended in complete growth medium containing 0.7% agar (Lonza, Rockland, Maine, USA) and with 0, 0.5, 2, 5, 10 or 20 mM of metformin and overlaid on 1% agar in 6 cm culture dishes. The dishes were maintained at 37°C in a humidifier incubator with 5% CO2 for 2–3 weeks. The colonies with the diameter larger than 50 μm were scored.

Cell cycle analysis

The cells were harvested, washed two times with 1×PBS, fixed with 75% ethanol for 30 min and stained with propidium iodide (Invitrogen). For each concentration of metformin, at least 1×105 cells were analysed by FAC Sort flow cytometry (BD Biosciences, San Jose, California, USA). The proportions of cells in G0/G1, S and G2/M phases were estimated using ModFit LT analysis software Verity software House, Topsham, Maine, USA.

siRNA transfection and p53 shRNA insertion

A total of 1×105 cells were seeded into 6 cm culture dishes overnight, the cells were transfected with 2.2 pmoles AMPK, LKB1 or scrambled control small interfering RNA (siRNA) (Santa Cruz, California, USA) by INTERFERin siRNA Transfection reagent (Polyplus, Illkirch, France). After 24 h incubation, the transfected cells were treated with metformin 20 mM for 48 h. To establish the p53 RNAi vector, DNA fragment containing a human U6 promoter followed by a p53 short hairpin RNA (shRNA) sequence was recovered from CMV-/Hu6-RNAi by HindIII and EcoRI double digestion.16 The p53 shRNA sequence (5′-GACTCCAGTGGTAATCTACTTCAAGAGAGTAGATTACCACTGGAGTCTTTTT-3′) was inserted into a mammalian expression vector pcDNA3, resulting in a plasmid carrying a U6 promoter that drives the expression of p53 shRNA and allowing for selection of stable transfected clones.

Western blotting

The HepG2 and Hep3B cells were treated with various concentrations of metformin for the desired time and then collected, washed twice with PBS and detached in PBS containing 0.02% EDTA. For western blotting, 30–50 μg of protein was used.

In vivo xenograft experiments

The tumourigenicity potential of the cell lines were tested by subcutaneous injection of 1×108 cells suspended in 0.2 ml of DMEM culture medium into the right flank of 5- or 6-week-old female BALB/c nude mice. All mice developed tumours in 10 days with a size of approximately 55 mm3. For each experiment, mice were randomly distributed into each groups (five mice per group) that were untreated or treated by intratumoural injections with 2.5 mg/kg doxorubicin every 5 days (four cycles), 200 μg/ml metformin (diluted in drinking water and present throughout the experiment starting at day 10), or combinations that included doxorubicin and metformin. These mice were observed for 30 days. Xenograft tumours were resected, fixed in 10% formalin and embedded in paraffin for sectioning (5 μm) on a rotary microtome, followed by slide mounting and H&E staining.

Statistical analysis

For the NHIRD studies, we first conducted multivariate analysis to examine whether metformin use was an independent protective factor against HCC development in diabetic patients by adjusting age, gender, hepatitis B, hepatitis C, liver cirrhosis, ESRD (end stage renal disease), DM duration, DM control and other diabetic agents. Then we investigated the dose-dependent antitumoural effect of metformin by analysing the attenuated risk of HCC by each incremental year of metformin use. The chemopreventive effect of metformin was further examined in different strata according to age, gender, DM duration, DM follow-up frequency, use of insulin, thiazolidinediones, other OHAs, hepatitis B, hepatitis C, liver cirrhosis and ESRD after adjusting other confounders.

For cell line studies, the differences between the vehicle control and the different doses of metformin were compared by Student t test. All data management and analyses were performed using SAS V.9.1 software (SAS Institute, Cary, North Carolina, USA).

Results

Demographic data

During the period 1997 to 2008, all eligible 97 430 HCC patients and 194 860 age-, gender- and physician visit date-matched controls were recruited. The demographic data, including age, gender, liver diseases, diabetes, metformin use and comorbidities, are shown in table 1. Diabetes was diagnosed in 22 047 (22.6%) HCC patients and 25 773 (13.2%) controls. The OR of diabetes in HCC patients was 2.29 (95% CI 2.25 to 2.35, p<0.001). There were significantly higher incidences of liver diseases including hepatitis B, hepatitis C, liver cirrhosis and alcoholic liver diseases in the HCC case group.

Table 1

Baseline characteristics of HCC cases and matched controls

Chemopreventive effects of metformin

In table 2, we found metformin use was associated with decreased risk of HCC in diabetic patients (adjusted OR=0.79, 95% CI 0.75 to 0.83, p<0.0001) after adjusting age, gender, hepatitis B, hepatitis C, liver cirrhosis, E SRD, DM duration, DM control and other diabetic agents. In table 3, we examined the dose-dependent chemopreventive effect of metformin and found each incremental year increase in metformin use was associated with 7% decrease of HCC risk in diabetic patients (adjusted OR=0.93, 95% CI 0.91 to 0.94, p<0.0001).

Table 2

Multivariate analysis for estimation of the chemopreventive effect of metformin in diabetic patients

Table 3

Multivariate analysis for estimation of the chemopreventive effect of metformin in diabetic patients by incremental years

Metformin as an independent protective factor for HCC development

In the multivariate stratified analysis, metformin use was associated with reduced risk of HCC in diabetic patients in nearly all subgroups according to age, gender, DM duration, DM follow-up frequency, insulin use, liver cirrhosis and ESRD. However, the chemopreventive effect of metformin for HCC development was not found in patients using thiazolidinediones, and those with hepatitis B or hepatitis C.

Metformin inhibited proliferation of hepatoma cell lines in a dose-dependent manner

To determine whether metformin influences cell growth of hepatoma in vitro, we examined the effect of metformin on cell viability in HepG2 and Hep3B hepatoma cell lines using the MTS test. Metformin significantly decreased cell viability in a dose- and time-dependent manner (figure 1A,B). Metformin also suppressed colony formation and anchorage-independent cell growth in both hepatoma cell lines (figure 1C,D). This suggests that metformin inhibits both short-term and long-term cell growth on hepatoma cell lines.

Figure 1

Effects of metformin (MET) on cell growth and cell cycle in hepatoma cell lines. (A and B) HepG2 and Hep3B cells were treated with various concentrations of MET (0, 0.5, 2, 5, 10 or 20 mM) for 48 h or exposed to 20 mM of MET for 0, 24, 48, or 72 h. Cell viability was determined by the MTS assay and was calculated as a percentage of the viable vehicle-treated cells. (C) HepG2 and Hep3B cells were treated with different concentrations of MET for 15 days, and then the cells were fixed with methanol and stained with haematoxylin. Colonies containing more than 50 cells that originated from single cells were counted under inverted microscope. (D) HepG2 and Hep3B cells were treated with different concentrations of MET in soft agar for 2–3 weeks. The colonies with the diameter larger than 50 μm were scored. (E) HepG2 cells were treated with MET (0, 0.5, 2, 5, 10 or 20 mM) for 48 h, and the treated cells were collected for DNA content analysis by flow cytometry. (F) HepG2 cells were treated with 20 mM of MET or co-treated with compound C (10 μM) for 48 h or (G) transfected with AMPK siRNA for 24 h and then treated with 20 mM of MET for another 48 h. At the end of each treatment, cells were collected and stained with propidium iodide. Cell cycle phases were analysed by flow cytometry. Data are expressed as the mean ± SEM of three independent experiments. *p<0.05 and **p<0.001 when compared with vehicle-treated cells (Student t test). AMPK, AMP-activated protein kinase; siRNA, small interfering RNA.

To determine the mechanism responsible for metformin-mediated antiproliferation, cell cycle distribution was evaluated using flow cytometric analysis (figure 1E and Supplement figure 1). Metformin treatment increased the cell population in the G0/G1 phase in a dose-dependent manner from 27.08% to 48.82% and from 46.15% to 74.79% in HepG2 cells and Hep3B cells, respectively. However, we observed that high dose of metformin (80 mM) caused apoptosis of both cell lines (Supplement figure 3A,B). To examine whether the AMPK mediated the antiproliferation effect of metformin, we pharmaceutically and genetically inhibited AMPK activity. Co-treatment with compound C, an AMPK inhibitor, or knockdown AMPK by siRNA rescued the hepatoma cells from metformin-induced G0/G1 arrest (figure 1F,G, Supplement figure 4A,B). These results suggested that metformin inhibited cell growth by attenuating cell cycle progression caused by G0/G1 arrest and AMPK is involved in this process.

In figure 1A,B, we found that HepG2, a p53 wild-type hepatoma cell line, is more sensitive to metformin-induced growth inhibition than Hep3B, a p53 mutant hepatoma cell. To evaluate whether the metformin-induced growth inhibition is p53-dependent, we examined the effect of metformin on HepG2 cells deficient in p53 function using p53 shRNA lentiviral constructs. We found no significant difference in reduction of cell numbers between control shRNA HepG2 cells and p53-knockdown HepG2-derived cell lines. Our results indicated that metformin-induced growth inhibition is p53 independent (Supplement figure 5).

Metformin-induced AMPK-dependent cell cycle arrest was via LKB1

We further characterised the molecular mechanisms of metformin-induced G0/G1 cell cycle arrest. Hepatoma cells were treated with 20 mM of metformin, cell extracts were prepared at various time intervals and cell cycle regulatory proteins, including cyclin D1, p21/Cip1 and p27/Kip1, were assessed by immunoblotting (figure 2A,B, quantitated and statistically analysed in Supplement figure 2A and B). Metformin enhanced AMPK phosphorylation, upregulated p21/Cip1 and p27/Kip1, and reduced expression of cyclin D1 in HepG2 and Hep3B cell lines in a time-dependent manner. Furthermore, the AMPK knockdown by siRNA inhibited the metformin-induced upregulation of p21/Cip1 and p27/Kip1 and the decline of cyclin D1 in HepG2 and Hep3B cell lines (figure 2C,D, quantitated and statistically analysed in Supplement figure 2C and D). This suggested that AMPK determines the metformin-induced G0/G1 cell cycle arrest via upregulation of p21/Cip1 and p27/Kip1, and downregulation of cyclin D1 in hepatoma cells.

Figure 2

Metformin (MET)-induced G0/G1 cell cycle arrest was via AMPK and its upstream kinase LKB1 to upregulate p21/Cip1 and p27/Kip1 and downregulate cyclin D1 in hepatoma cells. (A and B) HepG2 and Hep3B cells were treated with 20 mM of MET for 0, 6, 12, 24, 36 and 48 h. Equal amounts of protein were analysed by western blotting using antibodies that recognise the indicated cell cycle regulatory proteins. β-Actin was detected as a loading control. (C and D) HepG2 and Hep3B cells transfected with AMPK siRNA were treated with 20 mM of MET for 48 h and the proteins were analysed by western blotting using antibodies that recognise cyclin D, p27/Kip1 or p21/Cip1. (E and F) HepG2 cells treated with CaMKKβ or TAK1 inhibitor (Sto-609 5 μg/ml, 5-Z-7-oxozeaenol (5Z-7) 1 μM) and treated with 20 mM MET for 48 h. The phosphorylated AMPK and total AMPK were analysed by western blotting (E). The cell cycle was analysed by flow cytometry (F). (G and H) HepG2 cells transfected with LKB1 siRNA were treated with 20 mM of MET for 48 h and the proteins were analysed by western blotting (G). The cell cycle was analysed by flow cytometry (H). Data are expressed as the mean ± SEM of three independent experiments. *p<0.05 when compared with vehicle-treated cells (Student t test). AMPK, AMP-activated protein kinase; siRNA, small interfering RNA.

Next, we used Sto-609 and 5-Z-7-oxozeaenol, CaMKKβ and TAK1 inhibitor, respectively, to determine whether the metformin activated AMPK activity via these two upstream kinases. Both CaMKKβ and TAK1 inhibitors (Sto-609 5 μg/ml, 5-Z-7-oxozeaenol 1 μM) could not prevent metformin-induced AMPK phosphorylation (figure 2E and Supplement figure 6A, quantitated and statistically analysed in Supplement figures 2E and 6Aa) and metformin-induced cell cycle G0/G1 arrest (figure 2F and Supplement figure 6B). However, knockdown LKB1 expression by siRNA could prevent AMPK phosphorylation by metformin (figure 2G and Supplement figure 6C, quantitated and statistically analysed in Supplement figures 2G and 6Cc) and rescue cells from G0/G1 arrest (figure 2H and Supplement figure 6D). This suggested that metformin-induced AMPK activation was via LKB1.

Combination of metformin and doxorubicin accelerates hepatoma regression in mouse xenograft model

Metformin has been reported to potentate the effect of chemotherapy in lymphoma and improve patient response following neo-adjuvant chemotherapy for breast cancer.20 ,21 To examine the possibility of the chemo-sensitising effect of metformin in hepatoma, HepG2 and Hep3B cells were simultaneously treated with metformin and doxorubicin. We found that combination of metformin and doxorubicin significantly reduced cell viability (figure 3A) and increased the cell population in the subG1 phase when compared with either agent alone (figure 3B). In xenograft modal, nude mice bearing tumours 10 days after injection of the HepG2 cells were treated with 2.5 mg/kg doxorubicin (four cycles of intratumoural injection on days 10,15, 20 and 25), 200 μg/ml metformin (in drinking water) or the combination. Mice treated with metformin alone or doxorubicin alone had reduced tumour growth in comparison with the untreated mice, and after day 30, there was a relatively mild regression of the tumour. However, the combination of doxorubicin and metformin remarkably suppressed tumour growth (figure 3C). Thus, metformin may have chemo-sensitising effect in combination with doxorubicin to accelerate hepatoma regression in xenograft model.

Figure 3

Metformin (MET) combines with doxorubicin to suppress hepatoma growth in a xenograft animal model. (A and B) In addition to MET or doxorubicin (DOXO) treatment alone, HepG2 and Hep3B cells were treated with 20 mM of MET for 24 h, and following co-treated DOXO 1 μg/ml for another 24 h. Cell viability was determined by the MTS assay and was calculated as a percentage of the viable vehicle-treated cells (A). The subG1 analysis was also determined by flow cytometry (B). (C) In tumour bearing nude mice, experiment group were not treated (NT) or treated by intratumoural injections (days 10, 15, 20 and 25) with 2.5 mg/kg doxorubicin, or treated continuously with MET in drinking water (200 μg/ml) or both starting at day 10. Tumour volume (left panel) and weight (right panel) were monitored at 30 days after tumour injection. Data are expressed as the mean ± SEM of three independent experiments. *p<0.05 and **p<0.001 when compared with vehicle-treated cells (Student t test).

Discussion

Compared with non-diabetic subjects, diabetic patients not using metformin had significantly higher HCC risk (OR=1.95), which was consistent with the findings of previous studies. In two large-scale studies, El-Serag et al and Davila et al found that diabetes is associated with a twofold to threefold increased HCC risk, regardless of the presence of other HCC risk factors.2 ,3 Hassan et al and Donadon et al reported that diabetes is associated with a 2.5–4.2-fold higher risk of HCC.9 ,23

The chemopreventive effect of metformin observed in the present study appears to be less strong than that observed in previous reports.9 ,10 Different durations of diabetes may explain the discrepancy. Donadon et al found that the use of metformin is associated with an OR of 0.15 for HCC risk with a median follow-up duration of 10–12 years.9 Nkontchou et al reported that metformin use is associated with an HR of 0.19 for HCC occurrence. Their study subjects were followed up for 16 years.10 In the present study, we found a reduction of 7% in the risk of HCC for each incremental year increase in metformin use. Once our patients were followed up for 12–16 years, a 84%–90% reduction of HCC risk should be observed, which was consistent with previous studies. Since our observations were based on a nationwide population, the results were generalisable to other populations, at least in the Asian countries.

The chemopreventive effect of metformin on risk of HCC was observed in nearly all subgroups of diabetic patients, but not found in thiazolidinediones users and patients infected with hepatitis B or C. Several reasons may explain these interesting observations. First, thiazolidinediones use alone was reported to be associated with reduced risk of HCC.22 In the present study, thiazolidinediones use was found to be an independent protective factor for HCC development (adjusted OR=0.76). Once the diabetic patients had been protected with thiazolidinediones, the chemopreventive effect of metformin could not be observed. Second, hepatitis B and C were both strong risk factors for HCC development in diabetic patients (adjusted OR=14.58 for HBV and adjusted OR=18.02 for HCV found in the present study). The chemopreventive effect of metformin might be attenuated by these two factors and became statistically non-significant.

The dose-dependent chemopreventive effect of metformin was confirmed in hepatoma cell lines. To the best of our knowledge, this is the first time that metformin has been demonstrated to inhibit cell growth through cell cycle G0/G1 arrest in hepatoma cell lines. Downregulation of cyclin D1 in response to metformin has been reported in prostate cancer cells.23 Cell cycle arrest in response to metformin requires CDK inhibitors, p21/Cip1 and p27/Kip1, in addition to cyclin D1 downregulation and AMPK activation in breast cancer cell lines.24 Consistent with previous studies, we found that metformin treatment is associated with a reduction in cyclin D1 expression and increases in AMPK phosphorylation, p21/Cip1 and p27/Kip1 expressions, which mechanistically contribute to G1 checkpoint arrest. Our results also showed that compound C treatment, a specific inhibitor of AMPK, and AMPK knockdown by siRNA could block metformin-induced cell cycle G0/G1 arrest. This suggested that antigrowth effect of metformin is at least partially mediated by AMPK pathway.

Activation of AMPK by metformin requires its upstream kinase LKB1. The LKB1 also mediates glucose homeostasis in liver and therapeutic effects of metformin.25 Thus, it is reasonable that the antiproliferation effect of metformin was mediated by LKB1 in hepatoma cells. In our study, we demonstrated that LKB1 was necessary for metformin-induced AMPK activation and for induction of cell cycle G0/G1 arrest in hepatoma cell lines, but independent with CaMKKβ and TAK1. Metformin activates AMPK by two LKB1-dependent mechanisms. First, metformin inhibits complex I of the mitochondrial respiratory chain, which results in the generation of reactive nitrogen species (ONOO), activates PKCζ and phosphorylates LKB1.26 Second, metformin increases intracellular AMP, which activates AMPK by an LKB1-dependent mechanism. Further study is needed to demonstrate which pathway mediates the LKB1-dependent AMPK activation by metformin. It is also interesting to evaluate the risk of hepatoma in type 2 diabetes patients with regular metformin treatment and who carry the LKB1 mutants.

It has been reported that activation of AMPK-induced phosphorylation of p53 is required for AMPK-dependent cell cycle arrest in mouse embryo fibroblasts.27 However, some evidence has indicated that metformin can selectively inhibit p53 negative tumour cell growth.15 Moreover, metformin inhibits cell growth and induces cell cycle arrest independently of p53 expression.28 Our data revealed that metformin inhibits cell growth through cell cycle G1 arrest regardless of the p53 status of the hepatoma cell lines. This suggests that metformin has broad therapeutic efficiency in hepatoma treatment.

Metformin enhanced the effect of chemotherapy in lymphoma, ovarian cancer and breast cancer.19 ,20 Our in vitro study demonstrated that doxorubicin worked efficiently with the metformin to induce apoptosis of hepatoma cells. In vivo tumour inhibitory effect of doxorubicin was significantly enhanced by metformin treatment in drinking water. The concentration of metformin used in cell line studies and in xenograft experiments is relatively comparable with the dose used to treat human patients with type 2 diabetes.29 In addition, the oral metformin treatment has been reported to relate to its ability to selectively kill cancer cells.30 Thus, these findings further support the effect of metformin in combination with other chemotherapeutic agents to target different cancer cells.

Our study has several possible limitations. First, we did not have access to some important clinical risk factors in these patients, such as body mass index, smoking, lipid levels and liver function test. However, our estimations seemed to be consistent with those of previous studies. Second, our estimations were based on a case-control study. Although the controls were matched by age, gender and physician visit date to balance some demographic characteristics and to conduct multivariate and stratified analyses to confirm the robustness of our observations, unmeasured confounders may exist. Third, we did not analyse the dose of metformin taken every day. However, we found most patients taking metformin 2000 mg per day in divided doses in the present study. The numbers of metformin using days could be analysed as the total dose of metformin taken.

In conclusion, diabetic patients are associated with a higher HCC risk, which is attenuated by the use of metformin in a dose-dependent and duration-dependent manner. We estimated that each incremental year increase in metformin use reduces HCC risk approximately 7% in diabetic patients. We also demonstrated that metformin inhibits cell proliferation and induces cell cycle arrest at G0/G1 phase in a dose-dependent and time-dependent manner via LKB1 activated AMPK pathway, but independent of p53 status in hepatoma cells. Metformin also contains chemo-sensitising effect in combination with doxorubicin to accelerate hepatoma regression in a xenograft model. Thus, AMPK activation by metformin, a generally considered safe, well tolerated and relatively inexpensive drug for type 2 diabetes, may represent a new strategy to improve treatment of hepatoma.

References

View Abstract

Footnotes

  • Strobe guideline regarding bias and missing data has been followed.

  • Funding This work was supported in part by the Taiwan National Health Research Institutes (Grant number: PH-100-PP54, PH-101-PP-23) and Taichung Veterans General Hospital (Grand number: TCVGH-1013305C). The Taiwan National Health Research Institute provided the NHIRD and financial support for hiring research assistants.

  • Competing interests None.

  • Ethics approval Study approval: The present study used secondary database and cell lines, not involving human subjects. However, it was approved by the Taiwan National Health Research Institute.

  • Provenance and peer review Not commissioned; externally peer reviewed.

Request permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

Linked Articles

  • Digest
    Emad El-Omar William Grady Alexander Gerbes