Metformin-induced ROS upregulation as amplified by apigenin causes profound anticancer activity while sparing normal cells

Metformin increased cellular ROS levels in AsPC-1 pancreatic cancer cells, with minimal effect in HDF, human primary dermal fibroblasts. Metformin reduced cellular ATP levels in HDF, but not in AsPC-1 cells. Metformin increased AMPK, p-AMPK (Thr172), FOXO3a, p-FOXO3a (Ser413), and MnSOD levels in HDF, but not in AsPC-1 cells. p-AMPK and p-FOXO3a also translocated from the cytosol to the nucleus by metformin in HDF, but not in AsPC-1 cells. Transfection of si-FOXO3a in HDF increased ROS levels, while wt-FOXO3a-transfected AsPC-1 cells decreased ROS levels. Metformin combined with apigenin increased ROS levels dramatically and decreased cell viability in various cancer cells including AsPC-1 cells, with each drug used singly having a minimal effect. Metformin/apigenin combination synergistically decreased mitochondrial membrane potential in AsPC-1 cells but to a lesser extent in HDF cells. Metformin/apigenin combination in AsPC-1 cells increased DNA damage-, apoptosis-, autophagy- and necroptosis-related factors, but not in HDF cells. Oral administration with metformin/apigenin caused dramatic blocks tumor size in AsPC-1-xenografted nude mice. Our results suggest that metformin in cancer cells differentially regulates cellular ROS levels via AMPK-FOXO3a-MnSOD pathway and combination of metformin/apigenin exerts anticancer activity through DNA damage-induced apoptosis, autophagy and necroptosis by cancer cell-specific ROS amplification.

Cellular ATP production was measured in ASPC-1 and HDF cells after incubation with metformin (0, 0.05, 0.5, and 5 mM) for 48 h. Three different measurements were performed for each sample. Statistical significance is indicated as *p < 0.05, **p < 0.01, ***p < 0.001, and ****p < 0.0001 compared with HDF group. Transfection with wild type FOXO3a (wt-FOXO3a) and si-FOXO3a RNA suggests that FOXO3a activation plays a key role in determining cellular ROS levels. To examine whether FOXO3a activation contributes to cellular ROS levels, AsPC-1 cells were either mock transfected or transfected with wt-FOXO3a, both subsequently treated with 10 mM metformin. For HDF cells, they were mock transfected or transfected with si-FOXO3a RNA and then treated with 10 mM metformin as indicated (Fig. 3A). The intracellular ROS levels for this experiment were detected with CellROX staining of the cells and the cell lysates were also analyzed by western blotting (Fig. 3A,B). Cellular ROS levels were dramatically decreased in wt-FOXO3atransfected AsPC-1 cells when compared with mock transfected cells (p < 0.0001) (Fig. 3A). Expression of FOXO3a and MnSOD proteins were also highly increased in wt-FOXO3a-transfected AsPC-1 cells compared with mock transfected cells (p < 0.0001) (Fig. 3B). On the other hand, cellular ROS levels were dramatically increased in si-FOXO3a RNA-transfected HDF cells (p < 0.001) when compared with mock transfected HDF cells and expression of FOXO3a and MnSOD were largely decreased in wt-FOXO3a-transfected HDF cells (p < 0.001) (Fig. 3A,B).
Co-treatment with metformin and apigenin affects cell survival, apoptosis and cellular ROS levels, and inhibits mitochondrial potential. To examine the anticancer activity of co-treatment with metformin and apigenin, AsPC-1 cells and HDF cells were treated with either metformin (5 mM), apigenin (1 AsPC-1 and HDF cells were measured by CellROX Green staining. AsPC-1 cells were transfected with si-FOXO3a RNA while HDF cells were transfected with wt-FOXO3a. The transfected cells were treated with metformin (10 mM) for 24 h; then cellular ROS levels were measured by CellROX Green staining. (B) Protein levels of FOXO3a and MnSOD were measured in the transfected and metformin-treated cells by western blot analysis using anti-FOXO3a and MnSOD antibodies. Three different measurements were performed for each sample. Statistical significance is indicated as *p < 0.05, **p < 0.01, ***p < 0.001, and ****p < 0.0001. www.nature.com/scientificreports/ or 20 µM), or both metformin and apigenin up to 120 h. Cell growth/viability changes was examined by the MTT assay. The results showed that metformin or apigenin alone causes little change in cell growth/viability for AsPC-1 cells; however, co-treatment of the cells with metformin and apigenin cells led to a significant inhibition of cell growth/viability (Fig. 4A). In contrast, for HDF cells co-treatment with metformin and apigenin did not significantly affect cell growth/viability (Fig. 4A). Cell cycle analysis via flowcytometry and cellular ROS level changes via CellROX were also performed in response to varying levels of metformin and in the presence or absence of apigenin (20 µM) in cancer cells treated for 24 h. Cell cycle analysis showed that co-treatment with metformin and apigenin induces cell death in majority of AsPC-1 cells (Fig. 4B). As shown in Fig. 4C, co-treatment with metformin (0.05, 0.5 or 5 mM) and apigenin (20 µM) dramatically increased cellular ROS levels in AsPC-1 cells (p < 0.0001). However, the same co-treatment did not affect cellular ROS levels or extent of cell death in HDF cells (Fig. 4B,C). In addition, the effect of N-acetyl cysteine (NAC), an ROS scavenger, was gauged when co-treating the cancer cells in the presence of metformin and apigenin. From cell cycle analysis, NAC blocked the metformin/apigenin co-treatment-induced cell death in AsPC-1 cells (Fig. 4B) and NAC also blocked ROS increases seen with co-treatment with metformin and apigenin in the same cells (p < 0.0001) (Fig. 4C). Combination of metformin (0.05, 0.5 or 5 mM) and apigenin (20 µM) also synergistically inhibited mitochondrial membrane potential in AsPC-1 cells (p < 0.0001) (Fig. 4D), whereas the same treatment had a lesser effect in HDF cells (p < 0.001) (Fig. 4D). In addition to AsPC-1 cells, co-treatment with metformin and apigenin decreased cell viability and increased ROS levels (p < 0.01) in a synergistic manner in other cancer cells such as MIAPaCa-2, DU145, LNCaP and HCC1195 cells (Fig. 4E,F).

Combination of metformin and apigenin leads to DNA damage-induced apoptosis, autophagy and necroptosis in AsPC-1 cells but not in HDF cells.
To examine the mechanism involved in combination of metformin and apigenin-induced cell death, expression of DNA damage-related proteins was measured by western blot analysis in AsPC-1 and HDF cells. Levels of p-ATM, γ-H2AX, and DNA damage markers were increased by combination of metformin and apigenin in AsPC-1 cells (p < 0.05), indicating that amplified ROS induced severe DNA damage (Fig. 5A). DNA damage appeared to induce apoptosis as the levels of p-p53, Bim, Bid, Bax, cleaved PARP, caspase 3, caspase 8, and caspase 9 were also significantly increased by combination of metformin and apigenin in AsPC-1 cells (p < 0.05), and not HDF cells (Fig. 5B). Cytochrome C was also released from mitochondria in AsPC-1 cells, along with Bcl-2, an anti-apoptotic marker, becoming decreased in AsPC-1 cells (p < 0.01) (Fig. 5B). Interestingly, autophagy-related proteins (AIF, P62 and LC3B) and necroptosisrelated proteins (MLKL, p-MLKL, RIP3 and p-RIP3) were also increased by combination of metformin and apigenin (p < 0.05), suggesting that autophagy and necroptosis were also involved ( Fig. 6A,B). In comparison, DNA damage markers, apoptosis-, autophagy-, and necroptosis-related proteins were not altered by combination of metformin and apigenin in HDF cells (Fig. 6A,B).

Combination of metformin and apigenin effectively reduces tumor growth in an in vivo model.
To test the effect of combination of metformin and apigenin on tumor growth, AsPC-1 (1 × 10 7 cells) cells were injected into athymic nude mice to generate a xenograft cancer model. When the xenografts had reached about 80 mm 3 in size, the mice were randomized into treatment groups of control (vehicle treated), metformin (75 or 125 mg/kg), apigenin (5 or 40 mg/kg), or metformin/apigenin combination. The control/ drugs were given orally and twice daily as described in the "Materials and methods" section. As depicted in Fig. 7A,D, the treatments were continued for a period of 4 weeks, with the tumor sizes monitored as the control group reached an average of 1000 mm 3 in size (starting from 80 mm 3 in size). As seen in Fig. 7A, administration of metformin (75 mg/kg) or apigenin (5 mg/kg) alone caused a little change of tumor size, but a combination of two drugs decreased tumor size and weight in a synergistoical manner (Fig. 7B,C). Administration with higher dose of metformin (125 mg/kg) or apigenin (40 mg/kg) caused a reduction of tumor size compared to the control group (Fig. 7D). However, oral administration of combination of metformin and apigenin decreased tumor weight profoundly (p < 0.01) (Fig. 7E,F).

Discussion
In the present study, metformin increases ROS production in AsPC-1 cancer cells, but not in the HDF normal cells. In addition, for AsPC-1 cells, ATP levels were not changed, whereas ATP levels were decreased in HDF cells.
Metformin-induced ROS production in normal cells appears to be primarily associated with signal molecules such as AMPK, FOXO3a, and MnSOD. Our study clearly demonstrated that in normal cells metformin increases AMPK and p-AMPK levels and in turn, elevations in FOXO3a and p-FOXO3a, finally leading to increases in MnSOD levels. MnSOD causes a decrease in existing levels of ROS. For normal cells, metformin leads to AMPK being activated from a decrease in ATP production in mitochondria. In contrast to the normal cells, metformin did not affect AMPK, FOXO3a, and MnSOD levels in AsPC-1 cancer cells. Thus, increased ROS levels in cancer cells appear to be due to the lack of MnSOD action. AMPK-induced activation of FOXO3a is a key step in allowing a differential response between normal and cancer cells via metformin. AMPK-mediated phosphorylation of FOXO3a S413 activates FOXO3a. When FOXO3a levels were reduced via si-FOXO3a transfection, ROS increases were also seen in normal cells. We also found that when FOXO3a levels are raised by transfection via wt-FOXO3a, ROS levels becomes undetectable in cancer cells. This result clearly suggests that FOXO3a is a key molecule in bringing about this anticancer activity seen with metformin. In normal cells, many drugs are known to bind to mitochondria, generate ROS and decrease production of ATP upon producing electron leakage from the mitochondrial ETC. Some of these agents like metformin and apigenin result in a mild leakage in the ETC 30 www.nature.com/scientificreports/ cells and only mildly affect mitochondrial membrane potential are combined, they may still be excellent for amplifying ROS levels and inducing apoptosis in target cancer cells. One such agent is apigenin that is known to bind mitochondria and only mildly decrease the membrane potential in the treated cells 30,31 . Treatment with metformin or apigenin alone in normal fibroblasts did not affect cell viability. For the same cells, the combination of metformin and apigenin decreased mitochondrial membrane potential greatly but it did not affect cellular integrity and cell viability. In the present study, 5 mM of metformin itself inhibited MMP just partially in HDF cells. This observation is in part correlated with finding that IC50 value of metformin for MMP inhibition reported previously is 19 mM 34 . However, what we observed in the present is the finding that 5 mM of metformin could strongly inhibit MMP when metformin (0.05, 0.5 or 5 mM) was combined with apigenin (20 µM). This finding suggests that metformin interacts with apigenin for MMP inhibition in a synergistic manner. Based on our results, the overall hypothetical diagram depicting differential cell death by combination of metformin and apigenin between normal and cancer cells is described (Fig. 8). The exact reason for the synergistic increase on ROS production with the combination of metformin and apigenin in many cancer cells is not currently clear. However, it is speculated that metformin and apigenin may act via different mechanisms on the electron transport system in mitochondria and these additively produce ROS in mitochondria in cancer cells. Our findings suggest that the decrease of membrane potential by metformin and apigenin appeared to be synergistic when compared to treatment by each drug alone. Our finding is in line with a previous study in that the decrease of membrane potential by treatment metformin is well related with the increase of ROS production 35 .
We also found the combination of metformin and apigenin causing a reduction in cell growth in AsPC-1, MIAPaCa-2, LNCaP, DU145 and HCC1195 cells in a synergistic manner as revealed by the MTT assay and an induction of apoptosis in AsPC-1 cells as revealed by cell cycle analysis. This synergistic interaction was not observed in HDF cells, suggesting that cell growth inhibition and apoptosis induction by combination of metformin and apigenin are cancer cell specific. The nullifying effects of NAC against metformin/apigenin-induced ROS increase and apoptosis suggest that overproduction of ROS level appears to be primarily responsible for the cell death. Excessive amounts of ROS can cause oxidative damage to lipids, proteins, and DNA 5 . It is documented that if increases in ROS reaches a certain threshold that is detrimental to the cell with ROS exerting a cytotoxic effect. For cancer cells, this may lead to cell death and thus limit cancer progression [36][37][38] . In our observations, a combination of metformin and apigenin generated much higher levels of ROS over that perceived threshold, bringing about irreversible DNA damage.
The current study indicate that cell death induced by the combination of metformin and apigenin is mediated through apoptosis, autophagy, and necroptosis as there were increases in the levels of the proteins for each of these processes. For normal cells, these changes were not seen by the combination of metformin and apigenin, suggesting that the anticancer activity induced by these two agents might be achieved by selective activation of apoptosis, autophagy, and necroptosis pathways only for cancer cells.
From our in vivo experiment, we found that the individual oral administration with a lower dose of metformin (75 mg/kg) or apigenin (5 mg/kg) alone for 4 weeks did not affect much on tumor volume and weight. However, a combination of metformin and apigenin for 4 weeks caused a synergistic reduction in tumor volume. In addition to this finding, we found that individual oral administration of metformin (125 mg/kg) or apigenin (40 mg/kg) alone at higher doses, to some extent, reduced tumor growth in the ASPC-1 xenograft. However, oral administration of a combination of both metformin and apigenin almost completely inhibited tumor growth, suggesting that a combination of the two drugs exerts a more profound anticancer effect in vivo.
Several studies have reported that apigenin exerts an inhibition on tumor growth in several cancer xenograft models [39][40][41] . In addition, administration of metformin has also been documented to have a decrease in tumor growth in several animal cancer models 42,43 . However, as metformin and apigenin individually do not exert sufficiently robust antitumor activity, our findings suggest that combining metformin with apigenin may be useful to further test in preclinical models of pancreatic and other cancer types and potentially for a new class of normal-cell sparing anticancer drugs.

(C) Cellular ROS levels of metformin-treated (for 24 h) in AsPC-1 and HDF cells were measured by CellROX
Green staining when cells (AsPC-1 and HDF) were treated with metformin (0.05, 0.5 or 5 mM), apigenin (20 µM) or combination of metformin and apigenin for 24 h. (D) Mitochondrial membrane potentials were measured when cells (AsPC-1 and HDF) were treated with metformin (0.5 or 5 mM), apigenin (20 µM) or combination of metformin/apigenin for 24 h. Effect of NAC (N-acetyl cysteine, 10 µM) pretreatment on ROS production (A) and cell death (B) induced by combination of metformin/apigenin was examined. The effects of combination of metformin (5 mM)/apigenin (1 and 20 μM) on cell proliferation/viability (E) and cellular ROS level (F) in MIAPaCa-2, DU145, LNCaP and HCC1195 cells. Three different measurements were performed for each sample. Statistical significance is indicated as *p < 0.05, **p < 0.01, and ****p < 0.0001 compared with control group. #### p < 0.0001 compared with metformin (0.05, 0.5 or 5 mM)/apigenin (20 µM) treated group. + p < 0.05, +++ p < 0.01, and ++++ p < 0.0001 compared with ASPC-1 control group. ^^p < 0.01, ^^^p < 0.001 and ^^^^p < 0.0001 compared with HDF control group.   Immunofluorescence. The cellular localization of p-FOXO3a (Thr172) or p-AMPK (Ser413) proteins in AsPC-1 and HDF cells was determined by immunofluorescence antibody staining. First, the cells were treated with 0.5, 1, 10 and 20 mM of metformin, grown on glass slides for 24 h and were then washed with 1 × PBS. The cells were fixed in 4% formaldehyde at room temperature for 15 min; they were then washed with 1xPBS and incubated in blocking buffer (1% BSA in PBS) for 1 h. The slides were again washed with 1 × PBS and then diluted anti-FOXO3a, p-FOXO3a (Ser413), AMPK, p-AMPK (Thr172) antibodies were added to the cells on the slides. The slides were incubated in dark at 4 °C overnight. After washing the slides twice with 1 × PBS, the diluted fluorophore-conjugated secondary antibody was added and incubated at room temperature in dark for 1 h. After washing the slides twice with 1 × PBS, the slides were incubated with the DAPI solution (0.1 µg/ml double distilled water) for 2 min and washed with 1 × PBS for three times. One drop of mounting reagent was added to the slides, and were covered with a cover glass and incubated at room temperature overnight. The photographs were taken using a confocal microscope (Carl Zeiss LSM710).

ROS measurement.
To determine the cellular ROS levels in drug-treated AsPC-1 and HDF cells, the cells were grown for 24 h and then washed with 1 × PBS. The cells were incubated with the fluorogenic probe, Cell-ROX Green, for 2 h and were then fixed with 4% paraformaldehyde. The photographs were taken using a fluorescence microscope (absorption 485 nm, emission 520 nm) (Zeiss Axiovert 200, Zeiss, Oberkochen, Germany). ROS intensity of each picture was quantified using Photoshop CS4 (Adobe Systems, San Jose, CA) software after subtraction of background fluorescence measured in the nucleus. Cellular ROS level was calculated through dividing ROS intensity by cell numbers in a picture and plotted in bar.

Figure 8.
Hypothetical diagram depicting mechanism of differential cell death induced by combination of metformin and apigenin between HDF and AsPC-1 cells. www.nature.com/scientificreports/ Western blots analysis. The cells were first harvested by trypsinization and then washed twice with cold 1 × PBS. The cells were resuspended in the extraction buffer (150 mM NaCl/50 mM EDTA, pH 8.0/1% Nondiet p-40) containing the mixture of protease inhibitors (1 mM phenylmethyl methyl sulfonyl fluoride, 5 µg/ml aprotinin, 5 µg/ml leupeptin). After centrifugation at 14,000 rpm, 4 °C for 30 min, the supernatant was collected and used for western blot analysis, with the remaining lysate stored at − 70 °C. The protein concentration of each lysates was determined using the Lowry protein assay reagent (Thermo Fisher Scientific, Waltham, MA). The protein bands of the cell lysates were resolved in 12% SDS-PAGE gels and they were then electrically transferred onto a PVDF membrane for western probing. The membrane was blocked in 5% nonfat powder milk in TBST (50 mM Tris, pH 7.5, 150 mM NaCl, 0.1% Tween 20) for 2 h. The membrane was incubated overnight with the probing antibody diluted in TBST, washed and incubated with the diluted-anti-mouse IgG-horseradish peroxidase for 2 h. The resulting protein bands were visualized with the Immobilon reagent according to the manufacture's protocol (Merck Millipore, Burlington, MA). The protein band intensity for each membrane probing was captured and measured using an image analyzer (Fusion FX7, Vilber Korea, Seoul, Korea). The final figures were prepared from original protein bands using Adobe Photoshop CS4 program.
In vivo test of anticancer activity. All procedures were conducted in accordance with the 'Guide for Care and Use of Laboratory Animals' published by the National Institute of Health. This work was carried out in compliance with the ARRIVE guidelines 45 . Four-week-old female athymic nude mice (Koatech, Seoul, Korea) were received and allowed to acclimatize for a week. AsPC-1 cells (1 × 10 7 cells/100 µl) were then injected subcutaneously at the flank region of mice. Drug treatment was initiated at 7-days post injection of the cancer cells with the tumors being palpable to a mean volume of 80 mm 3 . The animals were randomly allocated to four groups (8 mice per group); these included the control group (vehicle treated); metformin group; apigenin group; and the combination group. Control (vehicle treated), metformin (75 or 125 mg/kg), apigenin (5 or 40 mg/kg) or the combination of two drugs were dissolved in 0.5% carboxymethyl cellulose (CMC); Drugs were twice daily, orally administrated. Tumor volume and body weight were measured once a week. The tumor volume assessment was with a Vernier caliper measurement along two perpendicular axes of the tumor lump, using the formula of total volume being equal to (length × width 2 )/2. After 28 days of treatment, the mice were sacrificed by cervical vertebral dislocation and the tumors were extracted immediately. Tumor weights at the sacrifice time were then measured. All animal procedures and experimental protocols were approved by Laboratory Animal Committee of Hallym University (Hallym 2020-23).

Statistical analysis.
Statistical analysis was carried out by student t test using the GraphPad Prism Version 4.0 software for Windows (GraphPad Software, San Diego, CA). P-values of less than 0.05 were considered to indicate statistical significance. All values were expressed as mean ± S.E.M.