Tubular Injury in a Rat Model of Type 2 Diabetes Is Prevented by Metformin: A Possible Role of HIF-1α Expression and Oxygen Metabolism

Metformin was provided by Dainippon Sumitomo Pharma (Osaka, Japan). AICAR, MG-132, and rapamycin were purchased from Calbiochem (San Diego, CA). Anti–AMPK-α, anti–AMPK-α1, anti–AMPK-α2, anti–phosphorylated (p)-AMPK-α (Thr172), anti–p-acetyl-CoA carboxylase (ACC) (Ser79), and anti–p-mTOR (Ser2448) antibodies were obtained from Cell Signaling Technology (Beverly, MA). The primary antibodies are monoclonal mouse antibodies for human HIF-1α (BD Biosciences, Benford, MA) and for rat HIF-1 α (Novus Biological, Littleton, CO). d-glucose was purchased from Wako Pure Chemical Industries (Osaka, Japan). ON-TARGETplus SMARTpool siControl (D-001810), AMPK-α1 (NM-006251), AMPK-α2 (NM-006252), and small-interference RNAs (siRNAs) were purchased from Dharmacon (Lafayette, CO). Alexa Fluor 594 donkey anti-mouse secondary antibody was purchased from Invitrogen (Gaithersburg, MD). Other chemicals and antibodies were obtained from Sigma-Aldrich (St. Louis, MO), unless otherwise indicated.

Human proximal tubular epithelial cells (HPRTECs) were purchased as once- or twice-passaged tubular cells from Lonza Walkersville (Walkersville, MD). The tubular cells were cultured in renal epithelial cell growth medium, according to the manufacturer’s instructions, as described before (24). After 24 h of starvation with serum-free DMEM, the cells were exposed to reagents under normoxic (21% O₂) or hypoxic (1% O₂) conditions for ~4–24 h and were then harvested for experiments. In all, experiments were deprived of serum.

Total RNA extraction and cDNA synthesis were preformed as described previously (24). Each cDNA sample was analyzed for gene expression by quantitative real-time PCR with an ABI 7300 Sequence Detector (Applied Biosystems, Foster City, CA) using the TaqMan Universal PCR Master Mix (Applied Biosystems), as described previously (24). Unlabeled specific primers were purchased from Applied Biosystems for detecting the human plasminogen activator inhibitor (PAI)-1(Serpine) gene (assay ID no. HS00167155-ml), the human SLC2A1 gene (assay ID no. HS00197884-ml), the human HIF-1α gene (assay ID no. HS00153153-ml), the human vascular endothelial growth factor (VEGF) gene (assay ID no. HS00900054-ml), and the 18S gene (assay ID no. HS 9999991-sl).

Silencing of AMPK-α1/-α2 gene expression in HRPTECs was achieved by the siRNA technique, as described previously (24). Forty-eight hours after transfection, HRPTECs were serum starved for an additional 24 h and subsequently treated as indicated.

Total cellular extracts from HRPTECs were prepared, and Western blot was carried out using a denaturing 8% Novex Tris-glycine gels (Invitrogen, Carlsbad, CA) or 10% NuPage Bis-Tris SDS-PAGE gels under reducing conditions, as described previously (24). Membranes were washed and reprobed with an antibody against α-actinin (Sigma) to control for small variations in protein loading and transfer. Images were acquired using the Adobe Photoshop program (Adobe Systems, San Jose, CA) and processed using Multi Guage (Fuji Film, Tokyo, Japan) for densitometric analysis. Signal intensities in control lanes were arbitrarily assigned a value of 1.00.

Cells were incubated under normoxic (21% O₂) or hypoxic (1% O₂) conditions in medium containing 5.5 mmol/L (low) or 25 mmol/L (high) glucose and then resuspended in normoxic medium. Average oxygen consumption rates in HRPTECs treated with reagents in normoxia or hypoxia for 4 h were measured in a sealed chamber using a Clark-type electrode.

HRPTECs were incubated with reagents under normoxic or hypoxic conditions for 4 h. ATP production was monitored by glucose-6-phosphate formation. Briefly, cells were extracted with perchloric acid (6%) and centrifuged (8,000g for 10 min). Subsequently, the extract was neutralized with K₂CO₃ (5 mol/L) neutralized to pH 7. NADP⁺ (0.5 mmol/L) and glucose 6-phosphate dehydrogenase (0.25 units) were then added and ATP production was monitored from the NADPH content by spectrophotometry at 340 nm. Cell proteins were determined in parallel dishes for the normalization of the ATP values.

The oxidative fluorescent dihydroethidium (DHE) (Sigma) was used to evaluate the intracellular production of superoxide (O₂⁻) (25). In brief, after incubation overnight, cells with or without 1 mmol/L metformin or 1 mmol/L AICAR under normoxic and hypoxic conditions were washed with serum-free and phenol-red–free DMEM and loaded with 5 μmol/L DHE. After incubation for 10 min in the dark, the cells were washed with PBS and were subjected to fluorescence microscopy.

NADPH content was determined using a NADP/NADPH Quantification kit (BioVision, Mountain View, CA) and the protocol supplied by the manufacturer.

HRPTECs were cultured on four-chamber glass slides (BD Biosciences) to reach 80% confluence. After exposure to 1 mmol/L metformin or 1 mmol/L AICAR for 4 h under normoxic or hypoxic conditions, the cells were fixed with 100% ethanol for 10 min and were incubated with an anti–HIF-1α antibody (1:100; BD Biosciences) at 4°C overnight. Then, cells were rinsed in PBS and subsequently incubated with Alexa Fluor 594 donkey anti-mouse secondary antibody (Invitrogen) at 1:200 dilution overnight at 4°C. Finally, slides were analyzed by cofocal laser–scanning microscopy.

Cellular hypoxia was detected by adding pimonidazole hydrochloride (200 mmol/L hypoxyprobe-1; Hydroxyprobe, Burlington, MA), which binds to cells or tissues with pO₂ levels <10 mmHg, to HRPTECs that were treated with 1 mmol/L metformin or 1 mmol/L AICAR and exposed to hypoxia (1% O₂) for 4 h. To detect hypoxic conditions in each group of rats, pimonidazole (60 mg/kg) was injected intraperitoneally 1 h before they were killed. Staining was performed according to the manufacturer’s instructions.

Male ZDF/Gmi-fa/fa rats and their heterozygous (ZDF/Gmi-+/fa) lean littermates were purchased from Charles River Japan. Animals purchased at 7 weeks of age were given an ad libitum commercial pellet diet (CE-2; CLEA, Tokyo, Japan) and tap water. ZDF rats were randomly assigned to four groups at the age of 9 weeks. A first group did not receive an active pharmacological treatment and was used as a control (ZDF control). A second group received metformin in standard diet at a concentration of 3.3 g/kg (ZDF+M group). Taking into account an average daily food intake of 32.4 g per rat, a dose of ~253.2 mg ⋅ kg⁻¹ ⋅ day⁻¹ was achieved. The metformin dose is compatible with those used in previous studies in rats (26,27). A third group was treated with a sustained-release insulin implant (~2 units per day per implant) (ZDF+I group) (Lin Shin Canada, Ontario, Canada). Heterozygous animals without active treatment were used as nondiabetic, lean controls (ZL group). Tail postprandial glucose in whole blood was quantified using a glucose analyzer (One Touch; Life Scan, Johnson & Johnson, Milpitas, CA). Hemoglobin A_1c was measured using the DCA 2000 analyzer (Siemens Medical Solutions Diagnostics, Tokyo, Japan). Renal function was assessed by measuring urinary albumin excretion. Urinary albumin was quantified using the Nephrat kit (Exocell, Philadelphia, PA), according to manufacturer’s instructions. Animals were killed at the age of 17 weeks for pimonidazole staining and at the age of 39 weeks for assessment of morphology and immunohistochemistry for HIF-1α by overanesthetization with isoflurane and cardiac puncture. The kidneys were fixed with perfusion, as described in a previous study (28). The Asahikawa Medical College Research Center for Animal Life Science approved all experiments.

To evaluate HIF-1α or pimonidazole expression in the kidney, immunohistochemistry was performed with mouse monoclonal anti–HIF-1α antibody (ESEE 122, 1:100) or mouse antipimonidazole (1:1000), as described previously (24). The positive immunoreactivity for nuclear HIF-1α protein or cytosolic pimonidazole was estimated as absent (0) for 0%, low (1) for 1–20%, intermediate (2) for 21–50%, and high (3) for more than 50% of the tubular cells in the juxtamedullary cortical and outer medullar regions. Assessment of tubulointerstitial injury was evaluated in the juxtamedullary cortical and outer medullar regions using periodic acid–Schiff staining and a semiquantitative scoring system evaluating interstitial fibrosis, inflammation, tubular atrophy, tubular dilation, debris accumulation, and cast formation. A score of 0 means normal tubulointerstitium, 1 means injury in <25%, 2 means injury in up to 50%, and 3 means injury in >50% of the biopsy specimen.

Three separate experiments at least were performed per protocol. Each treatment group was assayed in duplicate for real-time RT-PCR. The values shown represent the means ± SD. Statistical analysis was performed by ANOVA and Bonferroni post hoc tests. Values of P < 0.05 were considered statistically significant.

We investigated the impact of metformin on hypoxia-induced HIF-1α expression. HRPTECs faintly expressed HIF-1α protein under normoxic condition (Fig. 1A). The treatment of hypoxia (1% O₂) markedly induced HIF-1α protein accumulation in HRPTECs, and metformin, at the concentration of 1 mmol/L, inhibited hypoxia-induced HIF-1α protein expression (Fig. 1A). Densitometric analysis showed that hypoxia significantly induced HIF-1α protein expression compared with the control in normoxia (P < 0.01) and a significant ~85% attenuation in hypoxia-induced HIF-1α protein accumulation in HRPTECs treated with 1 mmol/L metformin, compared with hypoxia-treated controls (P < 0.01) (Fig. 1B). The inhibitory effect of metformin was detected at the concentration of 10 μmol/L, which was a therapeutic concentration as described in previous studies (11,29) (Fig. 1C and D), and for the incubation time of ~4–24 h (data not shown). In contrast, HIF-1α mRNA levels were not reduced by metformin in hypoxia, suggesting that metformin decreased HIF-1α protein via posttranslational mechanisms (Fig. 1E).

We examined the effects of metformin on the expression of HIF-1 target genes in HRPTECs (Fig. 1F–H). Real-time PCR results showed that hypoxia significantly promoted PAI-1, VEGF, and Glut-1 gene expression in HRPTECs. Treatment with 10 mmol/L metformin significantly reduced the enhancing effects of hypoxia on these mRNA expressions.

As shown in Fig. 1A and C, metformin inhibited hypoxia-induced HIF-1α protein accumulation in HRPTECs, accompanied by increased AMPK-α phosphorylation. We also determined the effects of an AMPK activator, AICAR, on hypoxia-induced HIF-1α expression (Fig. 1A). Although AICAR stimulated the phosphorylation of AMPK and ACC and inhibited the phosphorylation of mTOR-like metformin under normoxic and hypoxic conditions, AICAR failed to suppress hypoxia-induced HIF-1α expression (Fig. 1A). To study the role of AMPK on the inhibitory effects of metformin on HIF-1α expression, we have taken a genetic approach to inhibit the AMPK activity under hypoxic conditions. Immunoblots using antiphosphorylated (Thr172)-AMPK-α1 and -α2 antibody increased significantly in response to metformin dose dependently in siAMPK-α1–treated cells, suggesting that the presence of AMPK-α2 compensated for the absence of AMPK-α1 (Fig. 2A). In contrast, total p–AMPK-α content was markedly suppressed in siAMPK-α2–transfected cells treated with metformin (Fig. 2C), indicating that metformin could predominantly augment the phosphorylation of AMPK-α2 in HRPTECs, which was consistent with the previous study (30) investigating the effects of AICAR, hypoxia, and rotenone on isoform-specific AMPK activity. Thus, lowering the AMPK-α protein using specific siRNA for AMPK-α1 or -α2 did not affect hypoxia-induced HIF-1α expression and failed to abolish the inhibitory effects of metformin on hypoxia-induced HIF-1α expression (Fig. 2A–D).

To examine the posttranslational regulation of HIF-1α protein by metformin, we used the proteasomal inhibitor MG-132. As shown in Fig. 2E, MG-132 restored the inhibitory effect of metformin on hypoxia-induced HIF-1α accumulation, indicating that metformin exhibited its inhibitory effect on HIF-1α accumulation by promoting proteasomal HIF-1α degradation. Although an AMPK inhibitor, compound C (20 μmol/L) failed to antagonize the inhibitory effect of metformin on HIF-1α protein expression (Fig. 2E).

As shown in Fig. 3A, high glucose attenuated the stimulatory effects of metformin on the phosphorylation of AMPK-α and ACC and restored the inhibitory effect of metformin on p-mTOR expression. These results were similar to those of a recent study (31) in diabetic rats. However, the treatment of high glucose did not blunt the inhibitory effect of metformin on HIF-1α expression (Fig. 3A and B), concomitantly suggesting that the activation of AMPK-α is insufficient for the inhibitory effect of metformin on hypoxia-induced HIF-1α expression.

To determine the mechanism implicated in the regulation of HIF-1α expression in HRPTECs, subsequent experiments were achieved by the use of inhibitors of mitochondrial pyruvate transport (0.1 mmol/L α-cyano-4-hydroxycinnamate [CHC]), mTOR (100 nmol/L rapamycin), mitochondrial respiratory complex I (0.25 μmol/L, 25 μmol/L rotenone), and mitochondrial respiratory complex III (10 ng/mL, 1 μg/mL antimycin A). The expression of HIF-1α induced by hypoxia was unaffected by CHC and rapamycin (Fig. 4A), whereas it was completely blocked by rotenone and antimycin A (Fig. 4B). Moreover, rotenone and antimycin A increased the phosphorylation of AMPK-α and ACC and decreased the expression of p-mTOR (Fig. 4B).

Having identified that metformin had similar effects on the expressions of hypoxia-induced HIF-1α protein and AMPK as mitochondrial respiratory inhibitors, we then directly measured the oxygen consumption and intracellular ATP levels in HRPTECs treated with 1 mmol/L metformin and 1 mmol/L AICAR under normoxic and hypoxic conditions in medium containing 5.5 mmol/L (low) or 25 mmol/L (high) glucose for 4 h. As shown in Fig. 4C, hypoxia decreased oxygen consumption down to 76.5% of the control in normoxia, which was consistent with the previous study (32), whereas this response to hypoxia was blunted by the treatment of 25 mmol/L glucose (Fig. 4D). Notably, metformin significantly inhibited oxygen consumption to <50% of controls, independent of the oxygen conditions or the glucose milieu (Fig. 4C and D). In addition, the results show that the trend was for cell ATP to decrease during metformin treatment under either normoxic or hypoxic conditions in contrast to the enhancement of ATP levels by the treatment of AICAR (Fig. 4E). Consistent with a previous study (33), hypoxia did not induce ATP depletion in HRPTECs because the metabolic shift to anaerobic glycolysis seems to have been sufficient to maintain ATP in hypoxic cells (Fig. 4E). In contrast, 25 mmol/L glucose attenuated ATP levels and the inhibitory effects of metformin on ATP levels in either normoxia or hypoxia (Fig. 4F).

Furthermore, we examined the effects of the NADPH oxidase inhibitor diphenylene iodonium (DPI) (10 μmol/L) and the antioxidant N-acetylcysteine (NAC) (5 mmol/L) on the expression of HIF-1α and the phosphorylation of AMPK-α and mTOR under normoxic and hypoxic conditions for 4 h. DPI, but not NAC, inhibited hypoxia-induced HIF-1α expression; increased the phosphorylation of AMPK-α and ACC; and decreased the phosphorylation of mTOR (Fig. 5A). Although DPI also has been known to inhibit the mitochondrial complex I (34,35), we measured NADPH content in order to evaluate whether the inhibition of NADPH oxidase by metformin was involved in the inhibitory effects of metformin on HIF-1α expression (Fig. 5B). Both metformin and AICAR recovered hypoxia-decreased NADPH content, suggesting that the inhibition of NADPH oxidase could not be attributed to the inhibitory effect of metformin.

To investigate whether reactive oxygen species (ROS) production may be implicated in the inhibitory effects of metformin on HIF-1α expression, DHE staining was examined in HRPTECs. As shown in Fig. 5C, upper panel, hypoxia decreased DHE-associated fluorescence in the low-glucose medium compared with that under normoxic conditions. In the medium containing low glucose, metformin increased DHE staining independent of oxygen tension, as reported in the previous study (36), in contrast to AICAR (Fig. 5C). High glucose enhanced ROS production compared with low glucose, and metformin decreased high-glucose–induced ROS production, especially in normoxia (Fig. 5C, lower panel). Thus, these observations suggested that intracellular ROS production was not involved in the inhibitory effect of metformin on HIF-1α expression.

Because metformin decreased oxygen consumption in HRPTECs, we investigated whether metformin increases intracellular oxygen tension using a hypoxia-sensitive dye, pimonidazole (Fig. 5D). Notably, metformin, but not AICAR, rescued the hypoxic state in HRPTECs under hypoxic conditions. In addition, immunocytochemical analysis confirmed that hypoxia apparently induced the nuclear expression of HIF-1α in HRPTECs compared with those in normoxia, and metformin inhibited hypoxia-induced HIF-1α nuclear stainings (Fig. 5D). AICAR did not evidently alter the expression of HIF-1α in HRPTECs under hypoxic conditions (Fig. 5D). We also examined the effects of metformin on the state of hypoxia in the kidney (Fig. 6A and B). The intensity of pimonidazole staining was increased in tubular cells of the kidney cortex of ZDF rats, which was significantly attenuated by metformin but not by insulin.

To confirm the effects of metformin on tubular HIF-1α expression in vivo, we treated ZDF rats with metformin for 30 weeks. As expected, ZDF rats were hyperglycemic, and administration of metformin to ZDF rats significantly decreased HbA_1c by 24% (P < 0.01) (Table 1). ZDF rats had increased urinary volume and albumin excretion compared with lean controls, as described previously (23). Treatment of ZDF rats with metformin significantly prevented an increase in urinary albumin excretion to 58.3% of ZDF rats (P < 0.05) (Table 1). Consistent with the lack of albuminuria, ZL rats as lean controls showed neither glomerulosclerosis nor tubular injury (Fig. 6C, d and e). ZL rats exhibited the immunostainings of HIF-1α only in the inner medullar tubules (Fig. 6C, b). Intriguingly, ZDF rats showed strong nuclear HIF-1α expression, especially in outer medullar and cortical proximal tubules in addition to the inner medulla (Fig. 6C, f–h), which is consistent with previous data in human diabetic kidney (8) and in the kidney of streptozotocin-induced diabetic rats (7). In addition, kidneys of ZDF rats showed glomerulosclerosis and tubulointerstitial injuries, as described previously (23) (Fig. 6C, i and j). Notably, positive immunostaining for HIF-1α was evidently decreased in kidneys of metformin-treated ZDF rats (Fig. 6C, k and m), accompanied by amelioration of the tubular injury (Fig. 6C, n and o), which was in contrast to those in kidneys of insulin-treated rats, as described in the previous study (37) (Fig. 6C, p, r–t). Semiquantitative assessment for HIF-1α immunostaining and injury in cortical and outer medullary tubules revealed that chronic metformin treatment significantly decreased HIF-1α expression and ameliorated tubular injury in ZDF rats (Table 1).

In this study, we found that metformin inhibits hypoxia-induced HIF-1α protein expression and thus the expression of its targeted genes, probably via the inhibition of mitochondrial respiration in HRPTECs.

Metformin has been found to activate AMPK, a major cellular regulator of lipid and glucose metabolism (16–18). Previous studies also have shown that metformin is an inhibitor of complex I of the mitochondrial respiratory chain (38,39), independent of the AMPK pathway (17,40).

To investigate the underlying mechanism(s) of the action of metformin, we first assessed the role of AMPK activation in hypoxia-induced HIF-1α expression, utilizing AICAR, compound C, and siRNA for AMPK-α1 or -α2. We found that metformin-induced AMPK phosphorylation was not related to HIF-1α inhibition. In addition, we show that metformin suppresses mitochondrial respiratory function, suggesting that the redistribution of intracellular oxygen is involved in the inhibitory effects of metformin on HIF-1α expression, as described in a previous study (41). Collectively, our results suggest that metformin inhibits HIF-1α expression through the suppression of mitochondrial respiration described as inhibition of oxygen consumption and intracellular ATP levels, which subsequently activates AMPK pathways, proposing that AMPK-α is a downstream regulator of the mitochondrial respiratory chain. That is probably the reason that an AMPK activator, AICAR, fails to inhibit hypoxia-induced HIF-1α expression.

AMPK activated by metformin has been reported to increase glucose transporter (GLUT-1 and GLUT-4) and glycolytic enzymes, thereby enhancing glycolytic glucose utilization (42), and to directly stimulate glycolysis in hypoxic or anoxic cells (43). Thus, metformin could promote Pasteur effects, which include decreased oxidative phosphorylation and an increase in anaerobic fermentation (44) by inhibiting mitochondrial respiration and increasing glucose uptake mediated by AMPK activation. Although metformin inhibits hypoxia-induced Glut-1 expression in HRPTECs, an HIF-1α–mediated increase in glucose uptake and metabolism during severe hypoxia is known not to be required for cell survival using cultured mouse primary renal proximal tubular cells (33). Intriguingly, high glucose increases oxygen consumption even under hypoxic conditions but fails to augment ATP production, indicating that high glucose abolishes Pasteur effects, which are adaptive responses to hypoxic stress, and that high glucose decreases mitochondrial efficiency by uncoupling oxygen consumption from ATP production. An HIF-1–dependent block to oxygen utilization results in increased oxygen availability and decreased cell death when total oxygen is limiting (32). Therefore, our findings suggest that in chronic hypoxia, as seen in diabetic nephropathy, metformin may rescue the renal proximal tubular cells from hypoxia, oxidative stress, and energy depletion by suppressing oxygen consumption via inhibition of mitochondrial respiratory chain complex I, instead of HIF-1, subsequently restoring Pasteur effects, which are impaired by hyperglycemia (Fig. 7).

In chronic kidney injury, including diabetic nephropathy, hypoxia promotes the development of renal fibrosis by inducing connective tissue growth factor (CTGF) (45) and EMT (8,9,46) through a HIF-1–dependent pathway in renal tubular cells by augmenting apoptosis of glomerular endothelial cells (47) and by upregulating matrix production and decreasing turnover in renal fibroblasts (48). Very recently, Kimura et al. (10) induced interstitial fibrosis in the kidney of aged VHL^−/− mice, in which pronounced HIF-1α expression was observed in cortical tubular epithelial cells. They also demonstrated that the anti–HIF-1α agent [3-(5′-hydroxymethyl-2’-furyl)-1-benzyl indazole] (YC-1) ameliorated interstitial fibrosis in unilateral obstruction kidneys (10). We have shown that a preclinical study using ZDF rats confirmed the expression of HIF-1α in diabetic nephropathy, which has been described as the presence of hypoxia in previous studies (4–8), and explored the renoprotective effects of metformin in vivo.

In conclusion, these results in vitro and in vivo indicate, for the first time, that the antidiabetes drug metformin inhibits renal tubular HIF-1α expression. These data provide a novel mechanism of effects of metformin to improve microalbuminuria in diabetic nephropathy (14), protecting from hypoxia-induced renal fibrosis by attenuating the expression of HIF-1α (Fig. 7).

No potential conflicts of interest relevant to this article were reported.

Y.T. researched data and wrote the manuscript. T.H. contributed to discussion. J.W., Y.F., and J.H. researched data. N.S. and Y.M. contributed to discussion. M.H. contributed to discussion and reviewed and edited the manuscript.

The authors are grateful to Ken Inoki (University of Michigan, Ann Arbor, MI) and Shin-ichi Araki (Shiga University of Medical Science) for helpful discussions. The authors also express their appreciation to Shinji Kume (Shiga University of Medical Science) for his technical assistance.