|Home | About | Journals | Submit | Contact Us | Français|
Metformin is a widely used drug for treatment of type 2 diabetes with no defined cellular mechanism of action. Its glucose-lowering effect results from decreased hepatic glucose production and increased glucose utilization. Metformin’s beneficial effects on circulating lipids have been linked to reduced fatty liver. AMP-activated protein kinase (AMPK) is a major cellular regulator of lipid and glucose metabolism. Here we report that metformin activates AMPK in hepatocytes; as a result, acetyl-CoA carboxylase (ACC) activity is reduced, fatty acid oxidation is induced, and expression of lipogenic enzymes is suppressed. Activation of AMPK by metformin or an adenosine analogue suppresses expression of SREBP-1, a key lipogenic transcription factor. In metformin-treated rats, hepatic expression of SREBP-1 (and other lipogenic) mRNAs and protein is reduced; activity of the AMPK target, ACC, is also reduced. Using a novel AMPK inhibitor, we find that AMPK activation is required for metformin’s inhibitory effect on glucose production by hepatocytes. In isolated rat skeletal muscles, metformin stimulates glucose uptake coincident with AMPK activation. Activation of AMPK provides a unified explanation for the pleiotropic beneficial effects of this drug; these results also suggest that alternative means of modulating AMPK should be useful for the treatment of metabolic disorders.
Metformin is widely used for the therapy of type 2 diabetes mellitus (DM2) (1). Metformin ameliorates hyperglycemia without stimulating insulin secretion, promoting weight gain, or causing hypoglycemia (2, 3). In addition, metformin has beneficial effects on circulating lipids linked to increased cardiovascular risk (2–4).
Although used as a drug since 1957, the mechanism(s) by which metformin lowers glucose and lipids remains an enigma. Two effects, decreased hepatic glucose production (2, 5, 6) and increased skeletal myocyte glucose uptake (7, 8), have been implicated as major contributors to glucose-lowering efficacy. Metformin also decreases hepatic lipids in obese mice (9). Metformin is, however, a low-potency compound that is used at high doses, resulting in only modest net efficacy; in addition, significant side effects can occur (2). Thus, an understanding of the molecular basis for metformin’s effects on glucose and lipid homeostasis is a critical focus of research directed toward improved therapeutic approaches to DM2 and related disorders.
AMP-activated protein kinase (AMPK) provides a candidate target capable of mediating the beneficial metabolic effects of metformin. AMPK is a multisubunit enzyme that is recognized as a major regulator of lipid biosynthetic pathways due to its role in the phosphorylation and inactivation of key enzymes such as acetyl-CoA carboxylase (ACC) (10). More recent data strongly suggest that AMPK has a wider role in metabolic regulation (10, 11): this includes fatty acid oxidation, muscle glucose uptake (12–14), expression of cAMP-stimulated gluconeogenic genes such as PEPCK and G6Pase (15), and glucose-stimulated genes associated with hepatic lipogenesis, including fatty acid synthase (FAS), Spot-14 (S14), and L-type pyruvate kinase (16). Chronic activation of AMPK may also induce the expression of muscle hexokinase and glucose transporters (Glut4), mimicking the effects of extensive exercise training (17). Thus, it has been predicted that AMPK activation would be a good approach to treat DM2 (11). In this report we tested the hypothesis that activation of AMPK mediates the beneficial metabolic effects of metformin.
Hepatocytes were isolated from male Sprague Dawley (SD) rats by collagenase digestion (18). For the AMPK assay, cells were seeded in six-well plates at 1.5 × 106 cells/well in DMEM containing 100 U/ml penicillin, 100 μg/ml streptomycin, 10% FBS, 100 nM insulin, 100 nM dexamethasone, and 5 μg/ml transferrin for 4 hours. Cells were then cultured in serum-free DMEM for 16 hours followed by treatment for 1 hour or 7 hours with control medium, 5-amino-imidazole carboxamide riboside (AICAR), or metformin at concentrations indicated. For a 39-hour treatment, cells for both control and metformin (10 or 20 μM) groups were cultured in DMEM plus 5% FBS and 100 nM insulin, and the fresh control and metformin-containing medium were replaced every 12 hours (last medium change was 3 hours before harvest). After treatment, the cells were directly lysed in digitonin-containing and phosphatase inhibitor–containing buffer A (19), followed by precipitation with ammonium sulfate at 35% saturation. AMPK activity was determined by measurement of phosphorylation of a synthetic peptide substrate, SAMS (HMRSAMSGLHLVKRR) (20). For ACC assay, the 35% ammonium sulfate precipitate from digitonin-lysed hepatocytes (4 μg each) was used for determination of ACC activity via 14CO2 fixation in the presence of 20 mM citrate as done previously (19). For fatty acid oxidation, the oxidation of 14C-oleate to acid-soluble products was performed as done previously (21), but in medium M199 in the absence of albumin.
Liver AMPK was partially purified from male SD rats (22) to the blue-Sepharose step. The 100-μl reaction mixture contained 100 μM AMP, 100 μM ATP (0.5 μCi 33P-ATP per reaction), and 50 μM SAMS in a buffer (40 mM HEPES, pH 7.0, 80 mM NaCl, 0.8 mM EDTA, 5 mM MgCl2, 0.025% BSA, and 0.8 mM DTT). The reaction was initiated with addition of the enzyme. After 30-minute incubation at 30°C, the reaction was stopped by addition of 80 μl 1% H3PO4. Aliquots (100 μl) were transferred to 96-well MultiScreen plates (MAPHNOB50; Millipore Corp., Bedford, Massachusetts, USA). The plate was washed three times with 1% H3PO4 followed by detection in a Top-count. The in vitro AMPK inhibition data obtained with compound C — (6-[4-(2-Piperidin-1-yl-ethoxy)-phenyl)]-3-pyridin-4-yl-pyyrazolo[1,5-a] pyrimidine — were fit to the following equation for competitive inhibition by nonlinear regression using a least-squares Marquardt algorithm in a computer program written by N. Thornberry of Merck Research Laboratories: Vi/Vo = (Km + S)/[S + Km × (1 + I/Ki)], where Vi is the inhibited velocity, Vo is the initial velocity, S is the substrate (ATP) concentration, Km is the Michaelis constant for ATP, I is the inhibitor (compound C) concentration, and Ki is the dissociation constant for compound C.
Rat hepatocytes were seeded and starved in M199 medium as described (16). Cells were then treated for 6 hours with the same serum-free and hormone-containing medium containing 25 mM glucose in the presence or absence of AICAR or increasing concentrations of metformin as indicated. Total RNAs were extracted from cultured hepatocytes using the guanidine thiocyanate method (TRIZOL; Life Technologies Inc., Gaithersburg, Maryland, USA). The mRNA was quantitated using a TaqMan One Step Gold RT-PCR kit (Applied Biosystems, Branchburg, New Jersey, USA). Primers were as follows: for S14, S14p (6FAM-TGGTGATGATCCCCAGCCTTCTGAG), S14F (TGTGGTGCGGAACATGGA), and S14R (CTCCGGACCCACTCAGCTC); for FAS, FASp (6FAM-TCCGCC AGAGCCCTTTGTTAA-TTGG), FASF (AACTGAACGGCATTACT-CGGTC), and FASR (GTGTCCCATGTTGGATTTGGT); for sterol regulatory element-binding protein 1 (SREBP1), SREBP1p (6FAM-TCCACCATCGGCACCCACTGCT), SREBP1F (AGGACCCAAGGTGACACCTG), SREBP1R (GCCGGACGGG-TACATCTTT).
Hepatocytes from 24-hour starved rats were incubated at 2 × 106 cell/ml in bicarbonate-buffered saline medium containing 10 mM L-lactate, 1 mM pyruvate, 0.3 μM glucagon, with a gas phase of O2/CO2 (19:1). Glucose was measured using glucose oxidase kit from Sigma Chemical Co. (St. Louis, Missouri, USA).
Isolated rat epitrochlearis muscles were incubated for 3 hours with metformin (2 mM) or control medium followed by measurement of AMPKα1 or AMPKα2 activities as described (23). For glucose uptake, insulin (300 nM) was present where indicated for the last 30 minutes of the 3-hour incubation. Then, 3-0-methylglucose uptake was measured using a 10-minute incubation in the absence or presence of metformin and/or insulin as described previously (23).
Oral gavage was used to administer 1 ml of metformin (100 mg/ml) or water alone to male SD rats (300–350 g, n = 7–8). Rats were treated once (see Table Table1and1and Figure Figure5b)5b) or twice (see Figure Figure6)6) a day for 5 days. Rats were starved for 20 hours and then re-fed for 2 hours before the final dose; 4 hours after final dose, the animals were anesthetized and livers rapidly removed by freeze clamping followed by blood withdrawal. RNA was prepared from the freeze-clamped liver by Ultraspec RNA isolation reagent (Biotecx Laboratories Inc., Houston, Texas, USA). Nuclear extracts were prepared from a pool of seven rat livers (24). Glucose levels were determined using the standard glucose oxidase assay kit; β-hydroxybutyrate concentrations were assayed by measuring the reduction of NAD to NADH with a standard assay kit (Sigma Chemical Co.). FFA levels were measured with the assay kit from Boehringer Mannheim GmbH (Mannheim, Germany). Triglyceride levels were assayed with a kit from Roche Diagnostics (Indianapolis, Indiana, USA). Insulin concentrations were measured with the enzyme-immunoassay kit from ALPCO Ltd. (Windham, New Hampshire, USA). All in vivo experiments were approved by the Institutional Animal Care and Use Committee.
For phospho-AMPK (Thr172) detection, the 35% ammonium sulfate precipitate from treated rat hepatocytes was used for Western blot analysis using polyclonal Ab’s to phospho AMPKα (Thr172; Cell Signaling Technology, Beverly, Massachusetts, USA). For SREBP-1, 100 μg nuclear extracts from metformin-treated rat livers were similarly analyzed, but using monoclonal anti–SREBP-1 Ab (prepared from hybridoma CRL-2121; American Type Culture Collection [ATCC], Rockville, Maryland, USA).
Ten micrograms of 35% ammonium sulfate precipitate (containing AMPK) from AICAR- or metformin-treated rat hepatocytes was immunoprecipitated using polyclonal Ab’s raised against AMPKα1 (NH2-DFYLATSPPDSFLDDHHLTR-OH) or AMPKα2 (NH2-MDDSAMHIPPGLKPH-OH), followed by AMPK assay.
Although steady-state plasma levels of metformin in humans are reported to be around 10 μM (25, 26) to as high as 40 μM (27), studies in rats showed that levels in liver are severalfold higher than in plasma (28). Thus, liver levels of greater than 180 μM can be achieved in rats after a 50 mg/kg dose (29); this dose is lower than doses typically required for efficacy in diabetic rats (30, 31). In addition to the fact that high concentrations may be achieved in target tissues, membrane permeability of metformin is a time-dependent and slow process (32). Therefore, we anticipated that high concentrations and/or longer-term experiments might be required in order to see effects of metformin on AMPK in cells or tissues. In primary cultured hepatocytes from both rats and humans, metformin activated AMPK in a concentration- and time-dependent manner (Figure (Figure1a1a and data not shown). AICAR was used as positive control. AICAR is a cell-permeable adenosine analogue that can be phosphorylated to ZMP, an AMP analogue and known AMPK activator (10). After a 1-hour treatment, 500 μM metformin was required to significantly activate AMPK, whereas after a 7-hour treatment, 50 μM was sufficient to significantly activate the enzyme. Maximal AMPK stimulation, comparable to the effect of 500 μM AICAR, was achieved with metformin concentrations of 2,000 μM (1 hour) or 500 μM (7 hours). To ensure that lower concentrations could mediate a similar effect, rat hepatocytes were incubated with 10 μM or 20 μM metformin for 39 hours. Both 10 μM and 20 μM metformin produced significant AMPK activation (1.3-fold, P = 0.0062, and 1.6 fold, P = 0.0007, respectively; Figure Figure1a).1a). Given the time-dependent effects of metformin, we relied on relatively high concentrations in several shorter-term experiments described below in order to augment the ability to see relevant effects. In Figure Figure1b,1b, we used an immunoprecipitation-AMPK assay to show that metformin as well as AICAR activated AMPK in both AMPKα1 and AMPKα2 complexes efficiently.
AMP activates AMPK by promoting its phosphorylation at Thr 172 and by direct activation via an allosteric AMP site. Since enzyme activity shown in Figure Figure1,1, a and b, was measured after precipitation from hepatocyte lysates and in the presence of AMP, it is likely that metformin promotes AMPK phosphorylation. In keeping with this notion, we observed that treatment of hepatocytes with either metformin or AICAR resulted in slower electrophoretic mobility of AMPKα1 and AMPKα2, which is consistent with increased Ser/Thr phosphorylation (data not shown). More specifically, both metformin and AICAR treatment induced AMPK Thr172 phosphorylation demonstrated by using anti–phospho-AMPK Ab (Figure (Figure11c).
Recent observations suggest that metformin can impair oxidative phosphorylation by inhibiting mitochondrial phosphorylation complex 1 (32, 33). Although some have reported that very high metformin concentrations (10 mM) can suppress total cellular ATP levels (25, 33), Owen et al. (32) pointed out that more subtle changes in the free ATP/ADP ratio might occur with concentrations of metformin that do not suppress total ATP, but do inhibit gluconeogenesis. We also found that measured total ATP concentrations from metformin-treated (500–2,000 μM) rat hepatocytes were not affected (data not shown). Thus, it is still plausible that metformin could produce a subtle decline in the free ATP/ADP and ATP/AMP ratios or a change in local concentrations of nucleotides, serving as a stimulus for AMPK phosphorylation and activation. Alternatively, metformin could directly activate AMPK kinase (AMPKK) or promote AMPK phosphorylation by binding to AMPK and making it a better substrate for AMPKK or worse substrate for phosphatase PP2C (10). The possibility that metformin could be a direct allosteric AMPK activator was excluded by showing that, when incubated with partially purified rat liver AMPK, metformin (from 4 μM to 12 mM), unlike AMP, did not activate the enzyme (Figure (Figure11d).
As a result of AMPK activation, hepatocyte ACC activity was markedly suppressed by metformin treatment, paralleling concentrations and time points associated with AMPK activation (Figure (Figure1e).1e). ACC catalyzes the biosynthesis of malonyl-CoA from acetyl-CoA. Malonyl-CoA is an initial substrate for de novo fatty acid biosynthesis, and it serves as a potent inhibitor of carnitine palmitoyl transferase I (CPT-I), a rate-limiting step for fatty acid oxidation (34). Thus, decreased malonyl-CoA concentrations, as a result of ACC inactivation, could lead to decreased lipid synthesis and an increased rate of fatty acid oxidation. Indeed, like AICAR, metformin stimulated fatty acid oxidation in isolated rat hepatocytes (Figure (Figure11f).
AMPK is also involved in gene regulation (10). As noted above, activation of AMPK can suppress the expression of several glucose-activated lipogenesis-associated genes. Similar to the effect of AICAR, progressive inhibition of glucose-induced FAS and S14 expression was measured in hepatocytes in response to increasing concentrations of metformin (Figure (Figure2).2). As a control, expression of HMG-CoA reductase (HMGR), a gene not affected by AICAR, was not affected by metformin.
To firmly establish the role of AMPK in mediating metformin’s effects, a high-throughput in vitro assay was used to identify small-molecule AMPK inhibitors. Screening of a compound library containing more than 10,000 molecules lead to the discovery and characterization of compound C (inset in Figure Figure3a).3a). Experiments conducted using variable ATP concentrations revealed that compound C is a potent reversible inhibitor that is competitive with ATP, with Ki = 109 ± 16 nM in the absence of AMP (Figure (Figure3a).3a). In in vitro assays, compound C did not exhibit significant inhibition of several structurally related kinases including ZAPK, SYK, PKCθ, PKA, and JAK3. Thus, to our knowledge, this is the first description of a potent and selective small-molecule AMPK inhibitor that may be useful as a means to further assess the physiologic effects of this pathway. Incubation of cultured hepatocytes with compound C inhibited ACC inactivation by either AICAR or metformin (Figure (Figure3b).3b). Compound C also attenuated AICAR and metformin’s effect to increase fatty acid oxidation or suppress lipogenic genes in hepatocytes (not shown).
It is believed that metformin-mediated inhibition of hepatic glucose production (HGP) plays a major role in its glucose-lowering efficacy (2, 3, 6). Here, we determined that both metformin and AICAR were able to inhibit cumulative glucose production in primary cultured rat hepatocytes stimulated with glucagon (Figure (Figure3c3c and data not shown). Importantly, coincubation of hepatocytes with compound C was able to attenuate the effects of metformin to decrease glucose production in these cells (Figure (Figure3c),3c), demonstrating that AMPK activation was required for metformin’s inhibitory effect on hepatocyte glucose production.
Since AMPK activation is implicated as a mechanism for stimulation of glucose uptake in skeletal muscle (12–14), we assessed the effect of metformin on glucose uptake and AMPK activity in intact rat epitrochlearis muscles. Incubation of isolated muscles with metformin resulted in an increase in the activity of both catalytic subunits of AMPK (Figure (Figure4a).4a). This was coincident with a significant increase in glucose uptake that was also observed to be additive with the effect of insulin stimulation (Figure (Figure4b).4b). In previous studies, AMPKα2 was stimulated to a greater extent than AMPKα1 by AICAR and other stress (23). Further studies will be required to assess this difference.
SREBP-1c is an important insulin-stimulated transcription factor that is implicated in the pathogenesis of insulin resistance, dyslipidemia, and DM2 (35, 36). Target genes that are induced by SREBP-1 include those that encode lipogenic enzymes, such as FAS and S14. Under conditions where SREBP-1 was induced by insulin, incubation of rat hepatocytes with metformin or AICAR strongly suppressed SREBP-1 mRNA expression (Figure (Figure5a).5a). Thus, a new pathway by which AMPK can mediate effects on lipogenic gene expression was also identified via these experiments.
To assess whether selected effects of metformin described above also occurred in vivo, SD rats were studied (Table (Table1).1). Rats were orally dosed with metformin or vehicle (H2O) for 5 days. Rats were starved for 20 hours and then re-fed for 2 hours before the final dose. Four hours after the final dose, tissue and blood samples were obtained for analysis (see Methods). During starvation, there should be very little lipid synthesis. Upon refeeding, hepatic lipid synthesis should be dramatically induced. Metformin’s effects were examined under re-fed conditions. Along with modest decreases in plasma insulin and triglycerides, a small, but significant increase in β-hydroxybutyrate was present, suggesting that hepatic fatty acid oxidation was induced in metformin-treated rats. Furthermore, metformin treatment produced significant decreases in hepatic expression of mRNAs for SREBP-1, FAS, and S14 that were consistent with effects documented in cells (Table (Table11).
The mature SREBP-1 protein in rat liver nuclear extracts was examined using an anti-SREBP1 Ab (Figure (Figure5b).5b). As anticipated (35), SREBP-1 mature-form protein was not detected in hepatic nuclear extracts from starved animals. In re-fed animals, mature SREBP-1 protein had accumulated consistent with an increase in lipid synthesis under this condition. Treatment with metformin prevented this accumulation. Additional results obtained using hepatic nuclear extracts from re-fed rats after treatment with AICAR (500 mg/kg/day) also showed that the presence of SREBP-1 mature-form protein was ablated (data not shown).
Measurement of AMPK activation in liver ex vivo is difficult because brief hypoxia is known to produce marked activation of the enzyme (ref. 37 and data not shown). Thus we used liver tissue derived from metformin-treated rats to determine that ACC activity was decreased significantly at several tested citrate concentrations (Figure (Figure6).6). The greatest ACC activity reduction was at a citrate concentration of 1 mM (from 54.6 ± 11.8 to 35.6 ± 7.7 nmol/mg/min; P < 0.01). These results are consistent with metformin having produced in vivo AMPK activation and ACC inactivation.
To summarize, results presented here are consistent with a model (Figure (Figure7)7) in which increased phosphorylation and activation of AMPK by metformin leads to the effects on glucose and lipid metabolism as follows. Phosphorylation and inactivation of ACC, as a result of AMPK activation, serves to inhibit the proximal and rate-limiting step of lipogenesis. Reduced synthesis of the ACC product, malonyl-CoA, is also predicted to relieve inhibition of CPT-1, resulting in increased fatty acid oxidation. These effects are likely to contribute to metformin’s in vivo ability to lower triglycerides and VLDL.
AMPK mediates a decrease in SREBP-1 mRNA and protein expression. Known target genes for SREBP-1, which include FAS and S14, are also downregulated in liver, further contributing to metformin’s effects to modulate circulating lipids and to reduce hepatic lipid synthesis and fatty liver. It should be noted that increased SREBP-1 is postulated as a central mediator of insulin resistance in DM2 and related metabolic disorders (35, 36) and that increased liver lipid content is implicated in hepatic insulin resistance (38).
Metformin-mediated effects on hepatic glucose production contribute to its glucose-lowering efficacy. We demonstrated (using an AMPK inhibitor, Figure Figure3)3) that AMPK activation is required for inhibition of hepatocyte glucose production by metformin under the test conditions we employed. Additional studies will be required to further elucidate precise mechanism(s) by which metformin-stimulated AMPK activation could result in inhibition of hepatic glucose production.
AMPK activation is implicated as a mechanism for the induction of skeletal muscle glucose uptake; this effect is also additive with insulin (12). Thus, the observed association of increased glucose uptake and AMPK activation in isolated skeletal muscles suggests that metformin’s effect to augment muscle insulin action in vivo may be attributed to AMPK as well.
A major mechanism (AMPK activation) by which metformin produces beneficial metabolic effects has been characterized, along with discovery of the mechanism (SREBP-1 suppression), by which AMPK inhibits the expression of lipogenic genes. Attempts to generate novel therapies for metabolic disease via AMPK activation may be worthy of pursuit.
We thank Shiying Chen for characterizing anti-AMPK Ab’s, Marcie Donnelly for technical support related to animal studies, Mark Fraley for compound synthesis, and Georgianna Harris and Denis McGarry for helpful discussions and intellectual support.
See the related Commentary beginning on page 1105.