Metformin Reduces Endogenous Reactive Oxygen Species and Associated DNA Damage

See Perspective on p. 503

Metformin is a biguanide widely used in the treatment of type II diabetes (1). While details of its mechanism of action remain an active area of research, there is prior evidence that its primary effect is in mitochondria, where it interferes with respiratory complex I and reduces ATP production (2), leading to the activation of AMP kinase (AMPK; ref. 3). In type II diabetes, metformin action in the liver results in inhibition of gluconeogenesis, reducing blood glucose concentration (4, 5), and secondarily reducing the elevated insulin levels characteristic of this condition. Retrospective epidemiologic evidence (e.g., refs. 6–10, reviewed in refs. 11–13) suggests that type II diabetics receiving metformin have substantially lower cancer incidence and mortality than those not receiving this agent, with some reports showing approximately 50% reduction.

Although these epidemiologic data are retrospective and must be considered hypothesis generating rather than conclusive, they have motivated laboratory research to evaluate antineoplastic activities of metformin. Several in vitro and in vivo experimental systems have shown that metformin reduces growth rates of experimental tumors (e.g., refs. 14–20, reviewed in refs. 12, 21). One class of proposed mechanisms is indirect, and involves the well documented action of metformin on the liver, which results in reduction of the hyperglycemia and hyperinsulinemia characteristic of type II diabetes (14, 15, 18), leading to reduced insulin receptor activation and reduced proliferation of the subset of neoplasms for which hyperinsulinemia provides a growth advantage. There is also evidence for direct actions of metformin on neoplastic cells (16, 17, 19) secondary to various AMPK-dependent antiproliferative effects such as inhibition of mTOR (22), or, for those neoplasms that have impaired ability to survive under conditions of energy stress, cell death related to ATP deficiency (18, 20). However, these mechanisms may not be sufficient to account for the cancer risk reduction reported in pharmacoepidemiologic investigations or observed experimentally. For example, in a tobacco carcinogen-induced lung cancer model, metformin treatment was associated with more than 70% decrease in tumor incidence despite observations that it led to only modest decreases in signaling downstream of the insulin receptor and minimal activation of AMPK (23).

AMPKα^+/+ and AMPKα^−/− mouse embryonic fibroblasts (MEF) created by Dr. B. Viollet were provided from the laboratory of Dr. Russell Jones (McGill University, Montreal, QC, Canada). Cells were cultured in 10% FBS Dulbecco's modified Eagle's medium (DMEM; Wisent) with glutamine. Human mammary epithelial cells (HMEC) were obtained from Lonza and grown in 90% mammary epithelial growth medium (MEGM) completed with bovine pituitary extract (BPE), human recombinant epidermal growth factor (hEGF), insulin, hydrocortisone, gentamicin/amphotericin as provided in the MEGM Bullet kit (CC-3150; Lonza), and 10% DMEM (Wisent) supplemented with 10% FBS (Wisent). Identity of cell lines was not reauthenticated as part of the research described here.

AMPKα^+/+ and AMPKα^−/− MEFs were plated in 2% FBS DMEM with or without metformin for 48 hours. For analysis, cells were trypsinized, washed in serum-free media, incubated in 10 μmol/L dichlorofluorescein diacetate (DCF-DA; Molecular Probes) with 1 mmol/L H₂O₂ or 500 μmol/L paraquat (1,1′-dimethyl-4,4′-bipyridium dichloride) for 30 minutes, and analyzed by flow cytometry with a FACSCalibur equipped with CellQuest software.

AMPKα^+/+, AMPKα^−/− MEFs, or primary mammary epithelial cells were plated on glass cover slips in their appropriate medium. Twenty-four hours later, the medium was changed to 2% FBS DMEM and selected groups were treated with 5 mmol/L metformin (Sigma) for 12 hours at which time groups were treated with either 500 μmol/L H₂O₂ or 250 μmol/L paraquat for 36 hours. Cells were washed twice in cold PBS and fixed in 4% paraformaldehyde. Cells were permeabilized with 0.5% Triton X-100 in PBS with 3% BSA and were incubated overnight with a mouse monoclonal anti-phospho-Histone H2AXS134 (1:200; clone JBW301, #05-636, Millipore). Then cells were washed 3 times in PBS with 3% BSA and incubated with a goat anti-mouse secondary antibody (1:1,000; AlexaFluor 488, A-11001, Molecular Probes; Invitrogen) for 1 hour at room temperature. Finally, cells were rinsed 3 times with PBS and were mounted on slides with Vectashield (Vector Laboratories Inc.). Images were captured with an Olympus FV300 confocal laser microscope and were processed with Metamorph and ImageJ.

AMPKα^+/+ and AMPKα^−/− fibroblasts were treated with 5 mmol/L metformin (Sigma) for 48 hours followed by treatment of 500 μmol/L H₂O₂, 250 μmol/L paraquat, or 15 μmol/L antimycin for 30 minutes. Primary antibodies against p-ERK, extracellular signal-regulated kinase (ERK) total, AMPK, and β-actin were purchased from Cell Signaling.

The SupF plasmid pSP189 and Escherichia coli strain MBM7070 were kindly provided by Dr. M. Seidman (National Institute of Aging, NIH, Bethesda, MD). MEFs were trypsinized and washed twice with PBS. Then, 2.5 × 10E6 cells in 500 μL of PBS were placed in a 4-mm electroporation cuvette and pulsed during 20 ms at 240 V in the presence of 10 μg of plasmid DNA. The plasmid was introduced into MEFs by electroporation. Cells were then plated and treated with 5 mmol/L metformin or vehicle for 15 hours. Then, 25 μmol/L paraquat was added for another 36 hours; plasmids were purified and incubated with the restriction endonuclease DpnI. This endonuclease cleaves plasmids methylated in bacteria focusing the analysis only to plasmid DNA that replicated in mammalian cells. Plasmids were purified again by phenol–chloroform and ethanol precipitation and introduced into MBM7070 E. coli strain by electroporation. Bacteria cells were plated into 5-bromo-4-chloro-3-indolyl-β-d-galactopyranoside (X-Gal) indicator plates and blue and white colonies were counted.

Normal human diploid fibroblasts IMR90 were obtained from American Type Culture Collection and cultured in DMEM (Wisent) supplemented with 10% FBS (Wisent) and 1% penicillin G/streptomycin sulfate (Wisent). Retroviral vectors pBabe and pBabeRasV12 and retroviral-mediated gene transfer were carried out as described in the legend to Fig. 5. Growth curves were obtained from estimations of cell numbers according to a crystal violet retention assay. To measure reactive oxygen species (ROS), cells were incubated with dichlorodihydrofluorescein diacetate (H2DCFDA), dihydroethidium (DHE), or MitoSox from Molecular Probes. Fluorescence was measured by FACS. Immunofluorescence was measured as described earlier.

Male C57BL/6 mice were purchased from Jackson Laboratories at 10 weeks of age. Mice were randomized to receive either metformin (Sigma), 50 mg/kg/d intraperitoneally, for 10 days or PBS of equal volume. On day 11, mice were injected with 50 mg/kg paraquat dichloride (Sigma). To obtain serum, a subset of animals was sacrificed 24 hours after paraquat injection. For survival analysis, viability was recorded every 12 hours after paraquat injection and the survival curve was generated with GraphPad5 software.

Blood was collected from mice (n = 10) 24 hours following paraquat injection by cardiac puncture. Serum was used to measure free 8-isoprostane (Cayman Chemical) following the manufacturer's instructions.

MEFs were treated with 5 mmol/L metformin for 48 hours, harvested by centrifugation, and resuspended in 10 mmol/L HEPES, pH 7.4, 1.5 mmol/L MgCl₂, 10 mmol/L KCl, and 0.05 mmol/L dithiothreitol (DTT). The cells were sonicated and debris pelleted by centrifugation at 4°C. Protein concentration was determined by Bradford assay and NAD(P)H measured by absorbance at 340 nm.

The primary site of metformin action has been identified as complex I of the respiratory chain, where it inhibits oxidative phosphorylation in a manner distinct from classic complex I inhibitors such as rotenone (2). As complex I is an important source of ROS (24), we studied the influence of metformin on levels of ROS in cells treated with H₂O₂, which acts as an exogenous source of ROS, or paraquat, which stimulates the endogenous production of ROS by complex I (25). Given that several effects of metformin in glucose metabolism and diabetes can be explained by activation of the AMPK (16, 17, 19), we used AMPKα^+/+ and AMPKα^−/− MEFs to evaluate the role of metformin-induced AMPK activation on the observed effects (Fig. 1A). We observed a modest increase in intracellular ROS levels following exposure to metformin in both cell lines (Fig. 1B). As expected, 30-minute exposure to H₂O₂ increased ROS levels, as detected by DCF-DA flow cytometry, and this was not altered when cells were pretreated with metformin for 48 hours, regardless of AMPK expression (Fig. 1B). Exposure to paraquat for 30 minutes increased ROS levels in both cell lines, in keeping with increased endogenous ROS production (Fig. 1B–D). In contrast to the observations with H₂O₂-treated cells, metformin reduced the elevated ROS levels associated with paraquat exposure in both AMPKα^+/+ and AMPKα^−/− MEFs, providing evidence that metformin reduces endogenous ROS levels, in an AMPKα-independent manner.

ROS increases ERK signaling (26). We measured p-ERK in AMPKα^+/+ or AMPKα^−/− MEFs following exposure to H₂O₂, paraquat, or antimycin. Antimycin binds to cytochrome c reductase and leads to the formation of large quantities of ROS (27). As shown in Fig. 1E, treatment with paraquat or antimycin for 30 minutes led to a marked increase in p-ERK, consistent with increases in ROS. We observed this effect of H₂O₂, paraquat, and antimycin on ERK phosphorylation in both AMPKα^+/+ and AMPKα^−/− cell lines. In contrast, when cells were pretreated with metformin for 48 hours, we observed an increase in p-ERK following treatment with H₂O₂ exclusively, as 5 mmol/L metformin prevented the increase in ERK phosphorylation following treatment with either paraquat or antimycin in both AMPKα^+/+ and AMPKα^−/− cell lines. The observation that both paraquat and antimycin-induced p-ERK was attenuated by metformin suggests that the drug decreases ROS levels when they are produced by either complex I, the site of action of paraquat (25) or III (the site of action of antimycin; ref. 27). Mechanistically, this is consistent with induction of ROS defenses by metformin, inhibition of ROS production by the mitochondrial electron transport chain, or a direct scavenging action. The fact that metformin itself induces a modest increase in ROS and fails to reduce ROS related to H₂O₂ exposure argues against a scavenger or increased ROS detoxification mechanisms. In aqueous solutions, metformin is a poor scavenger of hydroxyl radicals and does not react with either H₂O₂ or superoxide (28). Hence, the inhibition of ROS production by complex I remains the most plausible mechanism to explain the observed metformin actions. Blocking NADH oxidation by complex I leads to accumulation of NADH (2), a phenomenon that we confirmed in MEFs treated with metformin (Supplementary Fig. S1). Inhibition of complex I reduces entry of electrons to the electron transport chain, and therefore, would reduce ROS production by both complex I and III, consistent with our observations. The modest increase in ROS following metformin exposure suggests NADH-dependent generation of superoxide by the pyruvate dehydrogenase and α-ketoglutarate dehydrogenase complexes (24). These findings provide evidence that metformin acts not as a classic antioxidant but rather as a mitochondrial regulator that decreases ROS production associated with oxidative phosphorylation but not ROS produced by nonmitochondrial sources that are required for normal cell signaling and cellular defenses (29).

Although ROS can damage a variety of cellular components, DNA is a critical target because it can lead to base modifications, abasic sites and double strand breaks, all of which can alter the information content of cells (30). While mitochondrial ROS production would be expected to preferentially damage mitochondrial as compared with nuclear DNA, nuclear DNA mutations are related to ROS levels (31). This does not necessarily require direct action of ROS on nuclear DNA, as it has been shown that nuclear mutation may arise due to direct oxidation of the nucleotide pool by ROS (31). Using γH2AX staining, we quantified DNA double-strand breaks in AMPKα^+/+ and AMPKα^−/− fibroblasts following exposure to H₂O₂ or paraquat, with or without metformin (Fig. 2). Consistent with the observed increases in ROS levels, we observed a significant increase in γH2AX-positive foci in both cell lines following exposure to H₂O₂ and paraquat. Pretreatment with metformin had no effect on H₂O₂-induced DNA damage; however, we observed significant reduction in the number of γH2AX-positive foci following paraquat treatment in the metformin group and, consistent with our observations in Fig. 1, these results were independent of AMPKα expression. To confirm the generality of our findings obtained in mouse fibroblasts, we repeated this experiment using primary HMEC. Metformin also reduced DNA damage induced by paraquat in these cells (Supplementary Fig. S2).

The relationship between DNA damage and carcinogenesis is complex. The DNA damage response constitutes a barrier for tumor progression (32), and tumors arise when these barriers are inactivated by genetic or epigenetic mechanisms. We considered the possibility that cancer risk reduction by metformin could be attributed at least in part to inhibition of mutagenesis. To evaluate this hypothesis, we measured the rate of DNA mutation by the SupF forward mutation assay (33). This method uses a suppressor tRNA (SupF) expression plasmid (pSP189) capable of replicating in mammalian cells. Mutations in the tRNA can be detected in the plasmids purified from mammalian cells using an indicator E. coli strain bearing a lacZ gene with a stop codon in its open reading frame. We introduced the SupF shuttle vector into MEFs by electroporation and treated with 5 mmol/L metformin for 15 hours. Then, we added paraquat or vehicle for 36 hours and recovered the plasmids. We found that paraquat induced mutations in the SupF tRNA, impairing its ability to suppress the stop codon in the lacZ indictor strain. Metformin prevented the mutagenic effect of paraquat in this assay (Fig. 2D), indicating that its ability to prevent ROS accumulation is associated with a reduction in the mutation rate in mammalian cells.

Having established that metformin reduces ROS accumulation, DNA damage and mutations in experimental systems involving mitochondrial toxins, we next addressed the question of whether metformin could also prevent these processes when induced by naturally occurring human oncogenes. To investigate this, we introduced oncogenic Ras into primary human fibroblasts where it is known to induce ROS production, DNA damage, and cell senescence (34). Consistent with our prior data, 5 mmol/L metformin modestly increased ROS accumulation in the control cells as measured both with superoxide and H₂O₂-sensitive probes (Fig. 3A and B). Metformin attenuated the substantial increase in ROS production observed in Ras-expressing cells. Importantly, this was associated with a significant decrease in the number of DNA damage foci (normalized for cell number; Fig. 3C), but had no effect on Ras signaling (Fig. 3D) and was associated with reduced proliferation (Fig. 3E). Of note, Ras-expressing cells have high levels of superoxide dismutase and catalase despite the fact that they accumulate ROS and associated oxidative DNA damage (34). The ability of metformin to reduce ROS accumulation in these cells is consistent with an effect of reducing ROS formation rather than increasing ROS detoxifying mechanisms, which were previously found to be insufficient to cope with the amounts of ROS produced in Ras-expressing cells (34). The finding that metformin attenuates Ras-induced ROS production and associated DNA damage is important from the perspective that genome instability is an “enabling characteristic” for neoplasia (35) that can be a consequence of oncogene activation (36). For example, in cases where carcinogens activate Ras (37, 38), the induction of further genetic instability and DNA damage is necessary to bypass Ras-induced senescence, and this is facilitated by Ras-induced ROS production.

Next, we determined whether metformin influences in vivo paraquat toxicity, which is a consequence of increased oxidative stress (39). Twenty mice were randomized to receive paraquat, with or without prior treatment with metformin. Figure 4A shows survival to 120 hours after paraquat injection, at which point all of the mice in the paraquat-only treatment group had died. The mice treated with metformin prior to paraquat administration displayed significantly improved survival. Figure 4B shows the expected increase in levels of free 8-isoprostanes (a marker of oxidative stress in vivo; ref. 40) in the serum of mice 24 hours following paraquat injection. This increase was significantly attenuated by metformin, showing that metformin reduces oxidative stress induced by paraquat in vivo.

Early pharmacoepidemiologic data suggest a reduction in cancer risk associated with administration of metformin to patients with diabetes. Although these studies are retrospective, and may or may not have implications for subjects without diabetes, the magnitude of the reported protective effect clearly justifies further research. While experimental investigation of the antineoplastic activity of metformin has documented growth inhibitory activity for established cancers, review of the epidemiologic data suggests that the dominant effect of metformin on neoplastic disease may involve reduction in risk rather than improvement in prognosis, as the magnitude of the reported decline in mortality is similar to the magnitude of the decline in incidence. Our finding that sequellae of the previously described (2) mitochondrial actions of metformin include reduced endogenous ROS production, reduced oxidative stress, reduced DNA damage, and reduced mutagenesis in normal somatic cells or their variants expressing activated oncogenes provide a novel mechanism to explain reduced cancer incidence associated with metformin therapy and raise the possibility of novel applications in prevention of cancer and other diseases associated with cellular damage caused by mitochondrial ROS production.

No potential conflict of interest were disclosed.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.