Metformin ameliorates the Phenotype Transition of Peritoneal Mesothelial Cells and Peritoneal Fibrosis via a modulation of Oxidative Stress

Phenotype transition of peritoneum is an early mechanism of peritoneal fibrosis. Metformin, 5′-adenosine monophosphate-activated protein kinase (AMPK) activator, has recently received a new attention due to its preventive effect on organ fibrosis and cancer metastasis by inhibiting epithelial-to-mesenchymal transition (EMT). We investigated the effect of metformin on EMT of human peritoneal mesothelial cells (HPMC) and animal model of peritoneal dialysis (PD). TGF-β1-induced EMT in HPMC was ameliorated by metformin. Metformin alleviated NAPDH oxidase- and mitochondria-mediated ROS production with an increase in superoxide dismutase (SOD) activity and SOD2 expression. Metformin inhibited the activation of Smad2/3 and MAPK, GSK-3β phosphorylation, nuclear translocalization of β-catenin and Snail in HPMCs. Effect of metformin on TGF-β1-induced EMT was ameliorated by either AMPK inhibitor or AMPK gene silencing. Another AMPK agonist, 5-amino-1-β-D-ribofuranosyl-imidazole-4-carboxamide partially blocked TGF-β1-induced EMT. In animal model of PD, intraperitoneal metformin decreased the peritoneal thickness and EMT with an increase in ratio of reduced to oxidized glutathione and the expression of SOD whereas it decreased the expression of nitrotyrosine and 8-hydroxy-2′-deoxyguanosine. Therefore, a modulation of AMPK in peritoneum can be a novel tool to prevent peritoneal fibrosis by providing a favorable oxidant/anti-oxidant milieu in peritoneal cavity and ameliorating phenotype transition of peritoneal mesothelial cells.

tubular cells and breast cancer cells 15,16 , and also alleviated hepatic and cardiac fibrosis by blocking TGF-β signaling pathway via AMPK-dependent manner 17,18 .
In this study, we investigated the effect of metformin and another AMPK agonist, 5-amino-1 -β-D-ribofuranosyl-imidazole-4-carboxamide (AICAR) on EMT of peritoneal mesothelial cells with an exploration of an alteration in oxidative stress and AMPK activation. Effect of metformin on peritoneal fibrosis was also examined in animal model of peritoneal dialysis. Metformin decreased both NADPH oxidase (NOX)-mediated and mitochondrial ROS generation, and also alleviated EMT and peritoneal thickening via AMPK-independent and -dependent mechanism. This data suggest the potential therapeutic use of metformin or other AMPK agonists for the prevention and/or treatment peritoneal fibrosis.

Results
Metformin ameliorated TGF-β1-induced EMT of HMPCs. TGF-β1 (1 ng/mL) induced EMT shown as a morphologic change from a cuboidal shape to elongated spindle-shaped cells (Fig. 1A) with an altered expression of epithelial and mesenchymal cell markers (Fig. 1). Immunofluorescence staining, real time PCR and western blot analysis demonstrated a decreased expression of ZO-1 and E-cadherin, the markers of epithelial cells, with an increase in the expressions of α-SMA, collagen type I and fibronectin, the markers of mesenchymal cell, upon TGF-β1 stimulation. Metformin (10 μM) alleviated TGF-β1-induced alteration in cell morphology and the expression of epithelial and mesenchymal cell markers (Fig. 1).

Metformin inhibited TGF-β1-induced Snail Expression by blocking the phosphorylation of GSK-3β and nuclear translocation β-catenin.
To further understand the mechanisms responsible for TGF-β1-induced EMT in HPMCs, we investigated the activation of GSK-3β, nuclear translocation β-catenin, and Snail signaling pathway, which were known as a key mechanism of E-cadherin down-regulation and EMT [19][20][21] . TGF-β1 increased phosphorylation GSK-3β, nuclear translocation of β-catenin, followed by an increased expression of Snail in HPMCs ( Fig. 3A and B). Metformin blocked the phosphorylation of GSK-3β, nuclear translocation of β-catenin, and Snail expression in HPMCs induced by TGF-β1 ( Fig. 3B and C).

Metformin blocked TGF-β1-induced ROS Production by NOX activation and Mitochondrial
Dysfunction. One of the earliest changes in TGF-β1-exposed HPMCs was an increased ROS production with an enhanced NOX activity, which was found in 15 minutes of TGF-β1 stimulation ( Fig. 5A and B). Interestingly, metformin ameliorated ROS production in TGF-β1-exposed HPMCs ( Fig. 5C and D). TGF-β1 also increased mitochondrial ROS production shown as an increased Mito-Sox staining in 6 hours of TGF-β1 stimulation in HPMCs, which was blocked by metformin (Fig. 5E).
Metformin increased Anti-oxidant Activities in TGF-β1-exposed HPMCs. In addition to an amelioration of TGF-β1-induced oxidative stress in HPMC, metformin reversed a decrease in anti-oxidant activity by TGF-β1. Metformin increased an activity of superoxide dismutase (SOD) and the expression of manganese-dependent superoxide dismutase, which is also known as SOD2, with a release of reduced glutathione (GSH/GSSG) in HPMCs exposed to TGF-β1 ( Fig. 5F-H).
Interestingly, there was a timely differential activation of AMPK by metformin and AICAR. AICAR activated AMPK of HPMCs from 15 minutes whereas the effect of metformin on AMPK phosphorylation was observed from 24 hours ( Supplementary Fig. 2), which was delayed compared to AICAR.
AICAR also alleviated TGF-β1-induced EMT (Fig. 6A,B), which was reversed by co-treatment with an AMPK inhibitor, compound C (Fig. 7). Interestingly, AICAR treatment did not induce a significant down-regulation of Smad2/3 and MAPK phosphorylation in contrast to metformin (Fig. 6C). Compound C (20 μg/mL) did not affect the effect of metformin on EMT at 12 hours, however partially reversed the effect of metformin on TGF-β1-induced EMT at 48 hours ( Fig. 7) attributable to a delayed activation of AMPK by metformin ( Supplementary Fig. 2).
Gene silencing of AMPK by siRNA also blocked or alleviated the effect of metformin on TGF-β1-mediated EMT (Fig. 8).

Metformin ameliorated EMT and Peritoneal Thickening in Animal Model of Peritoneal
Dialysis. In 8 weeks of peritoneal dialysis, mean thickness of parietal peritoneum in PD group (26.2 ± 4.3 μm) was significantly higher compared with control (12.1 ± 1.0 μm). Intraperitoneal metformin injection (50 mg/kg/ day) resulted in a decrease in peritoneal thickness (16.2 ± 0.3 μm) compared to dialysis group (group D) (Fig. 9A). Increased peritoneal thickness in group D was associated with an evidence of EMT shown as a decreased cytokeratin with an increased α-SMA staining in mesothelial lining and an appearance of cytokeratin (+)/α-SMA (+) cells in submesothelial zone (Fig. 9B). Rats on dialysis and metformin (group D + M) demonstrated a decrease in cytokeratin (+)/α-SMA (+) cells in mesothelial and submesothelial area, suggesting that metformin protected peritoneum from EMT. Western blot analysis also showed EMT in group D, which was alleviated by metformin treatment (Fig. 9C).

Metformin restored Anti-oxidant/Oxidant Balance in Peritoneum.
Eight-week-peritoneal dialysis in rats resulted in a decrease in SOD2 with an increased oxidative stress shown as an intense staining of nitrotyrosine (NT) and 4-hydroxynonenal (4-HNE) in peritoneal membrane (Fig. 10A). Metformin treatment alleviated an altered expression of SOD2, NT and 4-HNE in rats on PD.
In peritoneal effluents, the levels of anti-oxidants such as SOD and GSH/GSSG were significantly decreased in group D with an increase in a marker of oxidative stress, 8-OH-dG (Fig. 10B). Particularly, SOD activity in peritoneal effluent was not detected in all 4 rats of group D. Metformin significantly restored the change in the markers of anti-oxidant/oxidative stress in peritoneal effluent (Fig. 10).

Discussion
In this study, we demonstrate that (i) metformin ameliorates TGF-β1-induced EMT of peritoneal mesothelial cells and animal model of PD; (ii) the beneficial effect of metformin on EMT and peritoneal fibrosis is attributed to an inhibition of ROS generation, Smad2/3, ERK and p38 MAPKinase activation; (iii) the effect of metformin is independent of AMPK activation at early time points, however can be mediated by AMPK at later time points; (iv) metformin also protects the peritoneum from EMT and fibrosis by reinforcing local anti-oxidant activity. This is the first study to demonstrate the effect of metformin and AMPK agonist on peritoneal EMT both in in-vitro and in-vivo experiment.
The most important finding of this study is a validation of antifibrotic effect of metformin in peritoneal mesothelium. EMT is one of the earliest mechanisms of peritoneal fibrosis 4,5,22 . Mesothelial cells are known to undergo EMT to become matrix-producing myofibroblasts under pathologic conditions and participate in the pathogenesis of peritoneal fibrosis 4, 5, 7, 23-25 . Therefore, pharmacological prevention and/or reversal of EMT may serve as one of the possible therapeutic approaches to peritoneal fibrosis. In this study, TGF-β1 induced EMT with an activation of Smad2/3 and MAPKinase, followed by phosphorylation of GSK-3β and nuclear translocalization of β-catenin. It was already known that inactivation of GSK-3β by phosphorylation at threonine residue resulted in a mobilization of β-catenin into nucleus and an enhanced expression of transcription factors such as snail to down-regulate E-cadherin expression 26 . We confirmed a role of TGF-β1-induced activation of β-catenin/TCF signaling pathway assessed by TOPflash assay, nuclear staining and quantitation of β-catenin and real-time PCR of MMP-7 in EMT of HPMCs, which was blocked by metformin. Earlier studies have revealed that metformin inhibits EMT of breast cancer cells, renal tubular cells, and lung epithelial cells 11,15,27,28 . A modulation of EMT by metformin in cancer cells is proposed as a mechanism explaining a reduced risk of developing cancer [29][30][31] . However, there have been no reports on the effect in peritoneal mesothelial cells, and this study is the first demonstrating the beneficial effect of metformin on peritoneal EMT and fibrosis.
The earliest phenomenon observed in TGF-β1-exposed peritoneal mesothelial cells was an activation of NOX which induced ROS generation. In this study, we confirmed not only membranous NOX but mitochondria also contributed to an accumulation and release of ROS in HPMCs. Mitochondrial ROS generation was observed at later time points, in 6 hours of TGF-β1 stimulation compared to an early activation of membranous NOX at 15 minutes. Oxidative stress is known as a major mechanism of high glucose or TGF-β1-induced EMT of peritoneal mesothelial cells 32 . Consistent to the result of previous report 24 , we observed anti-oxidants, including N-acetyl cysteine (NAC, ROS scavenger), apocynin (NOX inhibitor), mitoQ (an inhibitor of mitochondrial electron transfer chain subunit I) alleviated TGF-β1-induced EMT of HPMCs ( Supplementary Fig. 3). Interestingly, metformin decreased H 2 O 2 generation by inhibiting both NOX activation and mitochondrial ROS production. The primary target of metformin in intact cells is the mitochondrion, where it inhibits respiratory chain complex I. Metformin was also reported to reduce mitochondrial damage expressed as an increased mitochondrial DNA copy number, cytochrome c release, caspase 3/9 activation, and mitochondrial ROS production [33][34][35][36] . In addition to an amelioration of oxidative stress, metformin also enhanced anti-oxidant activity and the expression of SOD2 in mesothelial cells, which was demonstrated for the first time in this study (Fig. 5). SOD is an important antioxidant enzyme present in nearly all living cells. SOD2, one of the most abundant SODs in HPMCs in our preliminary experiment, was known to be located in mitochondria, and is reported to play a pivotal role in cardioprotection in ischemia-reperfusion injury to myocardium 37 . Our study also confirmed a favorable effect of metformin on oxidative stress in in-vivo study. Peritoneal dialysis for 8 weeks resulted in an increased accumulation of nitrotyrosine and 4-HNE in peritoneal membrane, which was alleviated by metformin (Fig. 10). Peritoneal dialysis also reduced GSH/GSSG, SOD activity with an increased 8-hydroxy deoxyguanosine in peritoneal dialysate. Particularly, SOD activity in peritoneal effluent was almost not detected in rats on peritoneal dialysis in this study. Recent paper demonstrated metformin blocked TGF-β, angiotensin II and high glucose-induced EMT of renal tubular cells via an upregulation of heme oxygenase and thrioredoxin 15 . Taken together, metformin seems to provide a favorable oxidant/anti-oxidant microenvironment in peritoneal cavity by restoration of oxidative stress/antioxidant balance.
Metformin is an agonist of AMPK. A decrease in ATP production by metformin activates the energy sensor AMPK 38 . In addition to its traditional role as a modulator of cellular and whole body energy homeostasis, AMPK is recently shown to be involved in an amelioration of inflammation, angiogenesis, and organ fibrosis [39][40][41] . In this study, we confirmed a constitutive expression of AMPK in HPCMS. Metformin activated AMPK in HPMCs from 24 hours, which was delayed compared to another AMPD agonist, AICAR which activated AMPK in 15 minutes. Therefore, the effects of metformin on ROS generation (from 15 minutes), the activation of intracellular signal pathways (from 30 minutes), and a down-regulation of E-cadherin (from 12 hours) seem to be independent of AMPK activation. Considering both the protective effect of AICAR on TGF-β1-induced EMT and the partial reversal of beneficial effect of metformin by AMPK inhibitor, compound C or AMPK gene silencing, AMPK is thought to play a role in modulation of peritoneal EMT and metformin alleviates peritoneal EMT via both AMPK-dependent (late) and independent (early) pathways.
Despite the beneficial effects of metformin not related to glycemic control, the prescription of metformin has been limited by a concern for the development of metabolic acidosis 42 . Metformin is eliminated unchanged by the kidney, and expected to accumulate in patients with reduced renal function. However, the most recent Cochrane review found no evidence of an increased incidence of lactic acidosis with metformin based on the data including 347 comparative trials and cohort studies including the patients with impaired renal function 43 . Recent follow-up study in peritoneal dialysis patients also showed no case of lactic acidosis with metformin (0.5-1.0 g/day) for 4 weeks in 37 automated PD patients 44 . Based on this observation, the authors suggested that metformin could be   Metformin concentration in our study was determined by both preliminary experiment to exclude the toxic level to increase the release of lactic dehydrogenase in HPMCs ( Supplementary Fig. 4). We also considered the peak plasma concentration reported in diabetic patients, 4-15 μM 45 . The concentration used in this study (10 μM) was within physiologic range in contrast to previous studies done in other cells, suggesting the possibility of safe use of metformin in PD patients with careful monitoring.
In this study, an improvement of peritoneal morphology with an induction of favorable oxidant/anti-oxidant balance in animal model of PD provided by metformin was not associated with a functional improvement of peritoneal transport (Supplementary Fig. 5). It is partly due to complicated interaction of multiple factors to determine peritoneal function in PD patients.
Our findings provide the first evidence that metformin ameliorate EMT and peritoneal fibrosis. Metformin decreased ROS generation with an enhanced anti-oxidant activity in peritoneum. Effect of metformin on peritoneal EMT seems to be mediated by both AMPK-dependent and -independent pathways. Our data suggests a potential therapeutic use of metformin or other AMPK agonists in prevention/treatment of peritoneal fibrosis.

Preparation of protein extraction and western blotting.
For the preparation of whole cell protein fraction, cells were collected with a cell scraper after washing cold PBS, and centrifuged at 15000 rpm for 5 min. After discarding supernatant, the cold RIPA buffer with protease inhibitor was added to cell pellet. Cells and RIPA buffer mixture were kept on ice for 15 min, and spun at 15000 rpm for 15 min at 4 °C.
The nuclear protein extraction was prepared using an NE-PER Nuclear Cytoplasmic Extraction Reagent kit (Pierce, Rockford, IL, USA) according to the manufacturer's instruction. Briefly, the treated cells were washed twice with cold PBS and centrifuged at 500 g for 3 min. The cell pellet was suspended in 200 μl of cytoplasmic extraction reagent I by vortexing. The suspension was incubated on ice for 10 min followed by the addition of 11 μl of cytoplasmic extraction reagent II, vortexed for 5 s, incubated on ice for 1 min and centrifuged for 5 min at 16 000 g. The supernatant fraction (cytoplasmic extract) was transferred to a pre-chilled tube. The insoluble pellet fraction, which contains crude nuclei, was resuspended in 100 μl of nuclear extraction reagent by vortexing for 15 s and incubated on ice for 10 min, then centrifuged for 10 min at 16 000 g. The resulting supernatant, constituting the nuclear extract, was used for the subsequent experiments.
Protein lysates were run in SDS-PAGE for Western blot analysis. The blot was incubated overnight at 4 °C with primary antibodies directed to the following antigens: E-cadherin (BD Bioscience, Bedford, MA, USA), α-SMA, Collagen type I and SOD2 (Abcam), ZO-1 (Invitrogen), Snail, Phospho-GSK3β, GSK3β (Cell signaling, Danvers, MA, USA), fibronectin, β-catenin, Lamin B1, Phospho-Smad2, Phospho-Smad3, Smad2, Smad3, Phospho-ERK1/2, Phospho-p38, ERK1/2, p38, and β-actin (Santa Cruz Biotechnology). After washing the blot with PBS with Tween 20, blots were incubated with horse raddish peroxidase-conjugated secondary antibodies corresponding to each primary antibody followed by enhanced chemiluminescence detection (Santa Cruz Biotechnology). Positive immunoreactive bands were quantified by densitometry and compared with the expression of human β-actin.  Measurement of NADPH Oxidase (NOX) Activity. NOX activity was measured by a luminescence assay in a 50 mM phosphate buffer containing 1 mM EGTA, 150 mM sucrose, 5 μM lucigenin as the electron acceptor, and 100 μM NADPH as the donor with an addition of 100 μL of cell homogenate. Superoxide production was expressed as the rate of relative chemiluminescence units per milligram of protein. Protein content was measured using the Bio-Rad protein assay reagent (Bio-Rad Laboratories, Hercules, CA, USA).

Gene Silencing of AMPK.
To determine the effect of AMPK gene silencing on TGF-β1-induced EMT of HPMCs, we treated HPMCs with human AMPK siRNA obtained from Santa Cruz Biotechnology. The scrambled siRNA control was also selected from nontargeting siRNA pool (Santa Cruz Biotechnology). For transfection of siRNA, HPMCs were seeded into 6 wells for 24 hours at about 80% confluence, and then transfection of siRNA was performed using Lipofectamine RNAiMAX TM (Invitrogen) according to the manufacturer's protocol. After 24 hour of transfection media was changed, TGFβ was added, and the cells were incubated for 48 hours.
Animal Model of Peritoneal Dialysis. Sprague-Dawley rats (250-300 g) were anesthetized with isoflurane. A 9-French silicone catheter (Access Technologies, Skokie, IL, USA) was inserted into the peritoneum through a small incision in the abdominal wall. The catheter was tunneled subcutaneously to an implanted port (Access Technologies). In one week after catheter implantation, daily infusion of dialysate was initiated. The dialysis solution of 25 ml (4.25% Dianeal ® , pH 5.2; Baxter Healthcare, Deerfield, IL, USA) was infused twice daily (at 09:00 and 18:00 hr), 7 days per week, for 8 weeks. Control group (group C, n = 4) had a catheter inserted, but no fluid instilled. Dialysis group (group D, n = 4) was infused with a conventional dialysis solution (4.25% Dianeal ® , pH 5.2; Baxter Healthcare). Rats in group D + M (n = 4) was infused with a dialysate and metformin (50 mg/kg/ day).
Morphologic and Immunofluorescence Analyses of Peritoneum. The parietal peritoneum of abdominal wall was fixed with 4% paraformaldehyde (pH 7.4) and embedded in paraffin. Parietal peritoneum sections of 2 μm thickness were stained with Masson's trichrome. The thickness of the parietal peritoneum, including the mesothelium and submesothelial interstitium, was measured with an Aperio ScanScope CS Slide Scanner system (Aperio Technologies, Vista, CA, USA); whole-slide digital images were captured with X20 objective. For immunofluorescence microscopy, 3 μm tissue sections were incubated with primary antibodies against cytokeratin (Thermo Fisher Scientific, Waltham, MA, USA) or α-SMA (Abcam) at 4 °C overnight after antigen retrieval in boiling citrate buffer (pH 6.0) for 10 minutes. Sections were then incubated for 1 hours with fluorescein-conjugated secondary antibodies (Alexa Fluor 488 and Alexa Fluor 594; Molecular Probes, Eugene, OR, USA). The nuclei were counterstained with 4′,6-diamidino-2-phenylindole (DAPI; Molecular Probes), and slides were mounted with antifade mounting reagent (Molecular Probes). These sections were viewed under a confocal scanning laser microscope using LSM 5 EXCITER (Carl Zeiss).

Immunohistochemistry of Peritoneal Oxidative Stress and Anti-oxidant Activity.
Paraformaldehyde (4%)-fixed and paraffin-embedded tissue section (4 μm in thickness) were processed for immunohistochemical staining. Sections were de-paraffinized in xylene, rehydrated with graded ethanol, and washed with 0.1 M phosphate-buffered saline (PBS, pH 7.4). After microwave antigen retrival, sections were placed in a 0.3% H 2 O 2 blocking solution for 15 minutes and then in 5% normal serum (Jackson Immuno Research Laboratories, West Grove, PA, USA) for 1 hour. Sections were incubated overnight, 4 °C with each of the following primary antibodies: SOD 2 (Abcam), NT (Santa Cruz Biotechnology), and 4-HNE (Abcam). Sections were then washed in 0.1 M PBS and incubated in corresponding biotinylated-secondary antibodies (Santa Cruz Biotechnology) for 1 hour. After washing, sections were incubated in ABC solution (Vector Vectastain Elite ABC kit, Vector, Burlingame, CA, USA) for 30 minutes, followed by a visualization with a diaminobenzidine (DAB) solution (DAB Peroxidase Substrate Kit, Vector). After counterstaining with Mayer's hematoxylin, sections were dehydrated in graded ethanol, cleared in xylene, and coverslipped using Permount solution (Thermo Fisher Scientific). Sections were viewed and photographed with X20 objective under a light microscope (Zeiss Axioskop 2; Carl Zeiss)

Measurement of oxidant/antioxidant status in HPMCs and peritoneal dialysate.
To assess antioxidant status, the activities of glutathione reductase and SOD were measured in cell lysates or dialysate effluent using Glutathione Assay Kit (Cell Biolabs, San Diego, CA, USA) and Superoxide Dismutase Activity Assay (Cell Biolabs), respectively. Glutathione content was expresses as reduced/oxidized glutathione (GSH/GSSG). 8-hydroxy-2′-deoxyguanosine (8-OH-dG) content in dialysate were measured using Oxidative DNA Damage ELISA Kit (Cell Biolabs).