AMP-activated Protein Kinase Is Required for the Lipid-loweringEffect of Metformin in Insulin-resistant Human HepG2 Cells*

Received for publication, July 19, 2004, and in revised form, September 7, 2004Published, JBC Papers in Press, September 14, 2004, DOI 10.1074/jbc.M408149200

Mengwei Zang‡§, Adriana Zuccollo‡§, Xiuyun Hou‡§, Daisuke Nagata§, Kenneth Walsh§,Haya Herscovitz¶, Peter Brecher‡§, Neil B. Ruderman�, and Richard A. Cohen‡§**

From the ‡Vascular Biology Unit, §Whitaker Cardiovascular Institute, �Diabetes and Metabolism Unit,Department of Medicine, and the ¶Department of Physiology and Biophysics, Boston University School of Medicine,Boston, Massachusetts 02118

The antidiabetic drug metformin stimulates AMP-ac-tivated protein kinase (AMPK) activity in the liver andin skeletal muscle. To better understand the role ofAMPK in the regulation of hepatic lipids, we studied theeffect of metformin on AMPK and its downstream effec-tor, acetyl-CoA carboxylase (ACC), as well as on lipidcontent in cultured human hepatoma HepG2 cells. Met-formin increased Thr-172 phosphorylation of the � sub-unit of AMPK in a dose- and time-dependent manner. Inparallel, phosphorylation of ACC at Ser-79 was in-creased, which was consistent with decreasing ACC ac-tivity. Intracellular triacylglycerol and cholesterol con-tents were also decreased. These effects of metforminwere mimicked or completely abrogated by adenoviral-mediated expression of a constitutively active AMPK�or a kinase-inactive AMPK�, respectively. An insulin-resistant state was induced by exposing cells to 30 mM

glucose as indicated by decreased phosphorylation ofAkt and its downstream effector, glycogen synthase ki-nase 3�/�. Under these conditions, the phosphorylationof AMPK and ACC was also decreased, and the level ofhepatocellular triacylglycerols increased. The inhibi-tion of AMPK and the accumulation of lipids caused byhigh glucose concentrations were prevented either bymetformin or by expressing the constitutively activeAMPK�. The kinase-inactive AMPK� increased lipidcontent and blocked the ability of metformin to decreaselipid accumulation caused by high glucose concentra-tions. Taken together, these results indicate thatAMPK� negatively regulates ACC activity and hepaticlipid content. Inhibition of AMPK may contribute tolipid accumulation induced by high concentrations ofglucose associated with insulin resistance. Metforminlowers hepatic lipid content by activating AMPK,thereby mediating beneficial effects in hyperglycemiaand insulin resistance.

AMP-activated protein kinase (AMPK)1 is a phylogeneticallyconserved intracellular energy sensor that has been implicated

in the regulation of glucose and lipid homeostasis (1–4). AMPKis activated by physiological stimuli, such as exercise, musclecontraction, and hormones including adiponectin and leptin, aswell as by pathological stresses, glucose deprivation, hypoxia,oxidative stress, and osmotic shock (2, 5). AMPK serine/threo-nine protein kinase is a heterotrimeric complex consisting of acatalytic subunit (�) and two regulatory subunits (� and �) (5).Regulation of AMPK activity is complex; it involves allostericactivation by AMP, which increases during states of stresswhere ATP is depleted, and phosphorylation via the presump-tive upstream activator AMPK kinase (6–9), which may also beallosterically activated by AMP (5). Moreover, phosphorylationof Thr-172 within the activation loop of the catalytic domain ofthe � subunit is necessary for AMPK activity because site-directed mutagenesis of Thr-172 to Ala completely abolishesAMPK activity (10, 11). Once activated, AMPK phosphorylatesits downstream substrates to reduce ATP-consuming anabolicpathways, including cholesterol, fatty acid, and triacylglycerolsynthesis, and increases ATP-generating catabolic pathways,including fatty acid oxidation and lipolysis. Phosphorylation byAMPK of two key substrates, 3-hydroxy-3-methylglutaryl-co-enzyme A reductase and acetyl-CoA carboxylase (ACC) (12),which are the rate-limiting enzymes in cholesterol and fattyacid biosynthesis, respectively, results in their inactivation andthus reduces cellular ATP consumption during metabolicstress.

Type II diabetes is associated with hyperinsulinemia andinsulin resistance leading to elevated hepatic glucose produc-tion, hyperglycemia, and hyperlipidemia (13, 14). Administra-tion of the AMPK activator, 5-amino-4-imidazolecarboxamideriboside (AICAR), improves glucose tolerance and lipid profilesin the insulin-resistant Zucker rat (15), suggesting that AMPKactivity regulates insulin sensitivity and the associated dyslipi-demia. Metformin is an oral biguanide antidiabetic drug thatimproves insulin sensitivity and reduces plasma glucose andlipids in patients with type II diabetes (16–18). In fructose-fedrats with hyperinsulinemia and hyperglycemia, metformin de-creases liver triacylglycerol and free fatty acid levels and in-creases lipoprotein lipase activity (19). Metformin may increaseinsulin sensitivity through up-regulation of insulin signalingas reflected by increased tyrosine phosphorylation of the insu-lin receptor and insulin receptor substrate 1 (20). However, theprecise mechanism by which metformin lowers lipids is un-known. Recently, activation of AMPK by metformin has been

* This work was supported by National Institutes of Health GrantsP01 HL68758 (to M. Z., N. B. R., and R. A. C.) and AR40197 (to D. N.and K. W.) and by a Strategic Alliance with Institut de RecherchesServier. The costs of publication of this article were defrayed in part bythe payment of page charges. This article must therefore be herebymarked “advertisement” in accordance with 18 U.S.C. Section 1734solely to indicate this fact.

** To whom correspondence should be addressed: Vascular BiologyUnit, X708, Boston University School of Medicine, 650 Albany St.,Boston, MA 02118-2393. Tel.: 617-638-7115; Fax: 617-638-7113; E-mail:racohen@bu.edu.

1 The abbreviations used are: AMPK, AMP-activated protein kinase;

Ad, adenoviral vector; CA-AMPK, constitutively active AMPK; DN-AMPK, dominant-negative AMPK; ACC, acetyl-CoA carboxylase;AICAR, 5-amino-4-imidazolecarboxamide riboside; GSK, glycogen syn-thase kinase; DMEM, Dulbecco’s modified Eagle’s medium; GFP, greenfluorescent protein.

THE JOURNAL OF BIOLOGICAL CHEMISTRY Vol. 279, No. 46, Issue of November 12, pp. 47898–47905, 2004© 2004 by The American Society for Biochemistry and Molecular Biology, Inc. Printed in U.S.A.

This paper is available on line at http://www.jbc.org47898

at UT

Southw

estern Medical C

enter Library, on May 12, 2011

ww

w.jbc.org

Dow

nloaded from

shown to decrease glucose production and increase fatty acidoxidation in the liver (21, 22).

The present studies were performed to determine 1) theextent to which the effect of metformin on hepatocellular lipidsis mediated by AMPK and 2) whether AMPK regulates lipidaccumulation in insulin-resistant states. We utilized the cul-tured human hepatoma HepG2 cell line as a model system andpresented biochemical evidence that AMPK serves as a criticalregulator of hepatocellular lipid content. Metformin increasedACC phosphorylation in an AMPK-dependent manner, whichin turn decreased intracellular triacylglycerol and cholesterolcontents. Importantly, the ability of metformin to reduce lipidswas mimicked by adenoviral-mediated expression of a consti-tutively active AMPK� and was blocked by a kinase-inactiveAMPK�, suggesting that AMPK is required for the hypolipi-demic actions of metformin. In addition, exposing cells to highglucose concentrations induced a model of insulin resistance inwhich AMPK was inhibited and lipids accumulated. Expres-sion of the kinase-inactive AMPK� also led to triacylglycerolaccumulation, suggesting a pivotal role for AMPK in lipid ac-cumulation associated with insulin resistance. This study alsoprovides a model for developing potential therapeutic agents totarget AMPK in insulin resistance and dyslipidemia in type IIdiabetes.

MATERIALS AND METHODS

Materials—Metformin (1,1-dimethylbiguanide), Nonidet P-40, apro-tinin, leupeptin, phenylmethylsulfonyl fluoride, and 3-(4,5-dimethyl-thiazol-2-yl)-2,5-diphenyltetrazolium bromide were purchased fromSigma. InfinityTM triacylglycerol and cholesterol reagents were fromThermo DMA (Louisville, CO). The cytotoxicity detection kit used tomeasure lactate dehydrogenase release was from Roche Applied Sci-ence. Phospho-AMPK� (Thr-172), phospho-Akt (Ser-473), and phospho-GSK3�/� (Ser-21/9) antibodies were purchased from Cell SignalingTechnology (Beverly, MA). AMPK antibodies against the �1 or �2isoform were from Bethyl Laboratories, Inc. (Montgomery, TX). Phos-pho-ACC (Ser-79) antibody was purchased from Upstate Biotechnology(Lake Placid, NY). Anti-Myc (9E10) antibody and GSK3� antibody werefrom BD Biosciences. �-Actin antibody was from Abcam Inc. (Cam-bridge, MA). Anti-mouse and anti-rabbit antibodies conjugated tohorseradish peroxidase were obtained from Santa Cruz Biotechnology,Inc. (Santa Cruz, CA). Dulbecco’s modified Eagle’s medium (DMEM)and fetal bovine serum were purchased from Invitrogen. All otherreagents were of analytical grade.

Cell Culture—Human hepatoma HepG2 cells were obtained from theAmerican Type Culture Collection (Manassas, VA). HepG2 cells werecultured in DMEM containing 10% fetal bovine serum, 100 units/mlpenicillin, 100 �g/ml streptomycin, and 5.5 mM D-glucose. The cells wereincubated in a humidified atmosphere of 5% CO2 at 37 °C and passagedevery 3 days by trypsinization. For experiments, HepG2 cells wereincubated in complete medium with 10% fetal bovine serum in 100-mm-diameter dishes, grown to 70% confluence, and maintained in serum-free DMEM overnight as described elsewhere (23–25). Cells weretreated with metformin as indicated in the legends of Figs. 2, 3, 4, and6. To ensure cell viability during prolonged incubation of metformin inserum-free medium, cell viability was assessed with the 3-(4,5-dimeth-ylthiazol-2-yl)-2,5-diphenyltetrazolium bromide assay performed ac-cording to the manufacturer’s suggestions (Sigma). In addition, cyto-toxicity was assessed by measuring lactate dehydrogenase release intothe culture medium using the cytotoxicity detection kit (Roche AppliedScience) as described previously (26). Under the conditions of our stud-ies, treatment of HepG2 cells with metformin had no detectable celltoxicity.

Induction of Insulin Resistance Caused by High Glucose Concentra-tions—To develop a model of insulin resistance, HepG2 cells wereincubated in serum-free DMEM containing either normal concentra-tions of glucose (5.5 mM D-glucose) or high concentrations of glucose (30mM D-glucose) for 24 h, and the response to insulin (100 nM for 10 min)was tested as described (25). The phosphorylation of Akt and its down-stream target, GSK3�/�, was analyzed with anti-phospho-Ser-473 Aktand anti-phospho-Ser-21/9 GSK3�/� antibodies. Insulin stimulated 6-and 3-fold increases in the phosphorylation of Akt and GSK3�/�, re-spectively, over the basal level in cells maintained in normal glucoseconcentrations (Fig. 1, lanes 1 and 2). In contrast, high glucose concen-

trations diminished the insulin-induced phosphorylation of Akt as wellas both basal and insulin-stimulated GSK3�/� phosphorylation. Therewas no change in the expression of total Akt and GSK3� protein or�-actin (Fig. 1, lanes 2 and 4).

Adenoviral-mediated Gene Transfer—A replication-defective adeno-viral vector expressing green fluorescence protein (Ad-GFP) was usedas a control (27). The adenoviral vector expressing a dominant-negativemutant of AMPK�2 (Ad-DN-AMPK�2) was constructed from AMPK�2bearing a mutation of lysine 45 to arginine (K45R) as described previ-ously (27–29). To generate the adenoviral vector expressing a constitu-tively active mutant of AMPK�1 (Ad-CA-AMPK�1), a rat cDNA encod-ing residues 1–312 of AMPK�1 and bearing a mutation of threonine 172into aspartic acid (T172D) was subcloned into a shuttle vector (p-shuttlecytomegalovirus). The c-Myc epitope tag was fused in frame to the 5�terminus of the coding sequence. The resulting plasmid was linearizedby digestion with PmeI and co-transformed into Escherichia coliBJ5183 with the adenoviral backbone plasmid, pAdEasy-1. Homologousrecombinants were selected with kanamycin. The linearized recombi-nant plasmid was transfected into transformed human embryonic kid-ney 293 cells. Recombinant adenoviruses were amplified on humanembryonic kidney 293 cells and purified by two ultracentrifugationsteps on cesium chloride gradients. The number of viral particles wasassessed by measuring the optical density at 260 nm. HepG2 cells weretransfected overnight with 30–100 plaque-forming units/cell in serum-free medium. After fresh serum-free medium was added, cells weretreated with metformin (0.5–2 mM) for the indicated times. Preliminarystudies revealed that within 48 h of transfection with control Ad-GFP,80–90% of HepG2 cells expressed green fluorescent protein.

Western Blot Analysis—Western blot analysis was carried out asdescribed previously (30, 31). In brief, HepG2 cells were lysed in buffer(20 mM Tris-HCl, pH 8.0, 1% Nonidet P-40, 1 mM EDTA, 1 mM EGTA,1 mM sodium orthovanadate, 1 mM dithiothreitol, 1 mM phenylmethyl-sulfonyl fluoride, 2 �g/ml aprotinin, 2 �g/ml leupeptin, and 1 �g/mlpepstatin). Cell debris was removed by centrifugation at 14,000 � g for15 min at 4 °C, and the resulting supernatant (cell lysate) was used forWestern blotting and analysis of lipid content. Protein concentrationsin cell lysates were measured using a Bio-Rad protein assay kit. ForWestern blotting, 20–50 �g of protein were separated by 8% SDS-polyacrylamide gel electrophoreses (SDS-PAGE) and electrophoreti-cally transferred to polyvinylidene difluoride membranes in a transferbuffer consisting of 20 mM Tris-HCl, 154 mM glycine, and 20% metha-nol. The membranes were blocked with 5% nonfat dry milk in Tris-buffered saline with 0.1% Tween 20 and incubated with specific anti-bodies, followed by incubation with horseradish peroxidase-conjugated

FIG. 1. Induction of insulin resistance caused by high glucoseconcentrations as indicated by the decrease in insulin-inducedAkt and its downstream GSK3�/� phosphorylation. HepG2 cellswere quiesced in serum-free medium overnight and incubated in serum-free medium containing either normal (5.5 mM) or high (30 mM) con-centrations of D-glucose (high glucose) for an additional 24 h. Beforeharvesting, cells were stimulated with 100 nM insulin for 10 min. Totalcell extract was analyzed for the phosphorylation of Akt and GSK3�/�by immunoblots with phospho-Ser-479 Akt (�-pAkt), total Akt (�-Akt),phospho-Ser-21/9GSK3�/� (�-pGSK3�/�), total GSK3� (�-GSK3�), orloading control �-actin (�-�-Actin) antibodies, respectively. A represent-ative blot from three independent experiments is shown.

Metformin Regulates Hepatic Lipids via AMP Kinase 47899

at UT

Southw

estern Medical C

enter Library, on May 12, 2011

ww

w.jbc.org

Dow

nloaded from

secondary antibodies. Immunoreactive bands were visualized by theenhanced chemiluminescence detection system. The intensity of bandswas quantified using a model GS-700 Imaging Densitometer (Bio-Rad).

Determination of Triacylglycerol and Cholesterol Contents—Triacyl-glycerol and total cholesterol contents were determined in cell lysatesusing a colorimetric assay and expressed as �g of lipid/mg of cellularprotein as described (32). In brief, HepG2 cells were maintained inserum-free medium overnight and incubated with either normal or highglucose concentrations in the absence or presence of metformin. Celllysates were prepared as described above. Triacylglycerols and totalcholesterol levels in cell lysates were measured according to the man-ufacturer’s instructions for InfinityTM reagents.

Statistics—Results are expressed as the means � the standard errorof the mean (S.E.). Significance was analyzed using a two-tailed un-paired Student’s t test. p � 0.05 was considered significant.

RESULTS

Metformin Stimulates AMPK and ACC Phosphorylation, De-creases Lipid Content, and Attenuates the Hyperlipidemic Ef-fects of High Glucose Concentrations in Insulin-resistantHepG2 Cells—We first determined the phosphorylation state ofthe � subunit of AMPK (AMPK�) by using a specific anti-phospho-Thr-172 AMPK antibody. In HepG2 cells exposed toincreasing concentrations of metformin for various times (Fig.2A), phosphorylation of AMPK� at Thr-172 was stimulated byas much as 3.6-fold by metformin (2 mM) at 24 h (Fig. 2B). Nochange in the expression of endogenous AMPK� protein wasnoted by immunoblotting with AMPK�2 antibody (Fig. 2A). At

FIG. 2. Metformin stimulates AMPK� and ACC phosphorylation and decreases intracellular lipid content in a dose- and time-de-pendent manner in cultured human hepatoma HepG2 cells. HepG2 cells were quiesced in serum-free medium overnight and treated withincreasing doses of metformin (0.5–2 mM) for the indicated times. A, representative immunoblots of AMPK and ACC phosphorylation by metformin.Cell lysates were prepared, and 20–50 �g of protein were resolved by 8% SDS-PAGE as described under “Materials and Methods.” Thephosphorylation of AMPK and ACC was immunoblotted with anti-phospho-Thr-172 AMPK� (�-pAMPK) and anti-phospho-Ser-79 ACC (�-pACC)antibodies, respectively, and the expression of total AMPK protein was reprobed with anti-AMPK�2 antibody (�-AMPK�2). B, dose-response effectof metformin on AMPK and ACC phosphorylation. The phosphorylation levels of AMPK and ACC in cells treated with metformin for 24 h werequantified using model GS-700 Imaging Densitometer (Bio-Rad) and normalized to total AMPK protein. Data were expressed as mean � S.E.relative to the basal phosphorylation level from at least three independent experiments. *, p � 0.05, compared with control. C, time course of theeffect of metformin on AMPK� phosphorylation at Thr-172. D, time course of the effect of metformin on ACC phosphorylation at Ser-79. E,dose-response effect of metformin on hepatocellular lipid content. Levels of intracellular triacylglycerol and cholesterol in cells treated withmetformin for 24 h were measured using spectrophotometric assays and expressed as �g of lipid/mg of protein as described under “Materials andMethods.” The data are represented as the mean � S.E. (n � 4). *, p � 0.05, compared with the untreated control.

Metformin Regulates Hepatic Lipids via AMP Kinase47900

at UT

Southw

estern Medical C

enter Library, on May 12, 2011

ww

w.jbc.org

Dow

nloaded from

the lower concentrations used (0.5 mM and 1 mM), no change inphosphorylation was evident at 6 h, but there was a significantincrease at 24 h (Fig. 2C). Metformin (2 mM) caused a morerapid and potent increase in AMPK� phosphorylation, with asignificant increase noted at 6 h, reaching levels comparablewith those caused by 1 mM at 24 h (Fig. 2C).

To assess AMPK activity, we next determined the phospho-rylation of its downstream target, ACC. Western blot analysisshowed that metformin markedly induced phosphorylation ofACC at Ser-79 in a dose- and time-dependent fashion, whichparalleled the stimulation of Thr-172 phosphorylation ofAMPK� at 6 and 24 h (Fig. 2, A and D). When cells were treatedwith metformin for 24 h, phosphorylation of ACC was signifi-cantly increased by 3-fold at 1 mM and up to 5.5-fold at 2 mM

(Fig. 2B). Metformin at 0.5 mM had little effect, whereas theeffect of metformin at 1 mM caused a 3.2-fold rise at 6 h and wassustained for 24 h. Metformin at 2 mM caused a robust andsustained increase in ACC phosphorylation over 24 h (Fig. 2D).

Intracellular levels of triacylglycerol and cholesterol inHepG2 cells exposed to metformin for 1, 6, and 24 h were alsomeasured. Increasing concentrations of metformin (0.5–2 mM)decreased intracellular triacylglycerol and cholesterol contentat 24 h in a dose-dependent manner (Fig. 2E). Treatment for1 h had little effect, but significant reductions in both triacyl-glycerol and cholesterol content were observed at 6 h, althoughthey were less pronounced than those shown at 24 h (data notshown). At the highest metformin concentration utilized (2mM), there was approximately a 30 and 40% decrease in tri-acylglycerol content and a 21 and 33% decrease in cholesterolcontent at 6 and 24 h, respectively (Fig. 2E and data notshown).

Using a model of insulin resistance induced by high concen-trations of glucose (Fig. 1), we determined the effects of thesehigh glucose concentrations on the phosphorylation of AMPK�

at Thr-172, on the phosphorylation of ACC at Ser-79, and onlipid levels. Exposure of HepG2 cells to glucose (30 mM, 24 h)decreased phosphorylation of AMPK and ACC without achange in total AMPK protein (Fig. 3A, lanes 1 and 3). Inconcert with these changes in phosphorylation of AMPK andACC, triacylglycerol content dramatically increased by almost3-fold in insulin-resistant hepatocytes without a significantchange in cholesterol content (Fig. 3, B and C).

To determine whether metformin reverses lipid accumula-tion induced by high glucose concentrations, HepG2 cells wereincubated with either normal or high glucose concentrations inthe absence or presence of metformin (2 mM) for 24 h. Consist-ent with our earlier observations, the phosphorylation ofAMPK� and ACC was markedly up-regulated by metformin innormal concentrations of glucose (Fig. 3A, lanes 1 and 2), andthe intracellular contents of triacylglycerol and cholesterolwere lowered �40% by metformin (Fig. 3, B and C). Moreover,the inhibition of AMPK and ACC phosphorylation in cells ex-posed to high glucose concentrations was restored by met-formin (Fig. 3A, lanes 3 and 4). Consistent with this, the intra-cellular contents of triacylglycerol that were increased by highglucose concentrations were reduced 40% by metformin (Fig. 3,B and C).

The Effect of Metformin on Hepatic Lipids Is Mimicked by aConstitutively Active Form of AMPK—To confirm the role of thecatalytic � subunit of AMPK in regulating lipid metabolism, wedetermined the effect of adenoviral-mediated expression ofa Myc-tagged constitutively active mutant of AMPK�1 (Ad-CA-AMPK�1) on the changes in ACC phosphorylation and lipidlevels. As expected, after adenoviral transfection, the recombi-nant AMPK�1 protein that contained a T172D mutation andwas truncated at residue 312 (�31 kDa) was expressed as

detected by immunoblotting for the Myc tag at its N terminus(Figs. 4A and 5A, lanes 3 and 4). In contrast, endogenous �1 or�2 AMPK protein level was not obviously affected by the trans-fection (Figs. 4A and 5A). As predicted, ACC phosphorylationwas higher in the basal state following transfection with theCA-AMPK (Figs. 4A and 5A, lane 3). It was evident that thecells remained sensitive to metformin because expression of theCA-AMPK potentiated the effect of metformin on ACC phos-phorylation over the effect observed in Ad-GFP transfectedcells (Fig. 4A, lanes 2 and 4). Similar to the effect of metforminin GFP-transfected cells, the CA-AMPK transfection also re-duced triacylglycerol and cholesterol levels by �35%, and the

FIG. 3. Metformin reverses the suppression of AMPK� andACC phosphorylation and the accumulation of triacylglycerolby high concentrations of glucose. HepG2 cells were quiesced inserum-free medium overnight and incubated in DMEM containing ei-ther normal (5.5 mM) or high (30 mM) glucose concentrations in theabsence or presence of 2 mM metformin for an additional 24 h. A,metformin restores the suppression of AMPK signaling by high glucoseconcentrations. Total cell extracts were subjected to immunoblot anal-ysis with phospho-Thr-172 AMPK (�-pAMPK), total AMPK�2 (�-AMPK�2), and phospho-Ser-79 ACC (�-pACC) antibodies, respectively.All Western blots shown represent three independent experiments. Band C, metformin lowers the lipid accumulation caused by high glucoseconcentrations. Levels of intracellular triacylglycerols and cholesterolwere measured as in Fig. 1. Each bar represents the mean � S.E. (n �4). *, p � 0.05, compared between two groups as indicated.

Metformin Regulates Hepatic Lipids via AMP Kinase 47901

at UT

Southw

estern Medical C

enter Library, on May 12, 2011

ww

w.jbc.org

Dow

nloaded from

levels were further significantly reduced by metformin (Fig. 4,B and C), suggesting that there is a synergistic effect of met-formin and the CA-AMPK on the phosphorylation of ACC andreduction of lipids.

The effect of the CA-AMPK transfection was also tested incells exposed to high concentrations of glucose. Although theCA-AMPK had no appreciable effect on total or phosphorylatedforms of endogenous AMPK, the phosphorylation of ACC wasincreased, and the level in cells exposed to high glucose con-centrations was comparable with that in cells exposed to nor-mal glucose concentrations and transfected with GFP (Fig. 5A,lanes 1 and 4). Exposing CA-AMPK-transfected cells to highglucose concentrations elevated triacylglycerols level signifi-cantly, but the level was significantly less than that in cellstransfected with GFP and exposed to high glucose concentra-tions (Fig. 5B). The CA-AMPK also reduced the cholesterol

FIG. 4. Expression of a constitutively active form of AMPK�mimics the effect of metformin on ACC phosphorylation andlipids. HepG2 cells were transfected with adenoviral vectors encodingGFP (Ad-GFP) or a Myc-tagged constitutively active form of AMPK�1(Ad-CA-AMPK) in serum-free medium for 24 h, followed by incubationwith or without 2 mM metformin for an additional 24 h. A, expression ofthe CA-AMPK is sufficient to increase baseline ACC phosphorylation aswell as enhance metformin-induced ACC phosphorylation. Expressionof the CA-AMPK recombinant protein (�31 kDa) was detected by im-munoblots using anti-Myc antibody (�-Myc). The phosphorylation ofAMPK and ACC and expression of total AMPK protein were analyzedby immunoblots with the phospho-Thr-172 AMPK� antibody (�-pAMPK), phospho-Ser-79 ACC antibody (�-pACC), and anti-AMPK�2antibody (�-AMPK�2), respectively. Immunoblots shown are represent-ative of at least three independent experiments. B and C, inhibition ofintracellular triacylglycerol and cholesterol contents by metformin wasmirrored by expression of the CA-AMPK. Each bar represents themean � S.E. (n � 4). *, p � 0.05, compared between two groups asindicated.

FIG. 5. Modulation of AMPK activity is sufficient to alter hep-atocellular lipid contents. Ad-GFP or Ad-CA-AMPK were trans-fected into HepG2 cells for 24 h. Cells were subsequently incubated inDMEM containing either normal (5.5 mM) or high (30 mM) glucoseconcentrations for another 24 h. A, suppression of AMPK signaling byhigh glucose concentrations is prevented by expression of the CA-AMPK. Total cell extracts were subjected to immunoblot with phospho-Thr-172 AMPK� (�-pAMPK), total AMPK�1 (�-AMPK�1), totalAMPK�2 (�-AMPK�2), phospho-Ser-79 ACC (�-pACC), and anti-Mycantibodies, respectively. A representative blot from three independentexperiments is shown. B and C, expression of the CA-AMPK is suffi-cient to decrease intracellular triacylglycerol and cholesterol content.Each bar represents the mean � S.E. (n � 4). *, p � 0.05, comparedbetween two groups as indicated.

Metformin Regulates Hepatic Lipids via AMP Kinase47902

at UT

Southw

estern Medical C

enter Library, on May 12, 2011

ww

w.jbc.org

Dow

nloaded from

content in cells exposed to high glucose concentrations, al-though cholesterol levels were not statistically affected by highglucose concentrations (Fig. 5C). These results indicate thatthe expression of the CA-AMPK promotes the phosphorylationof ACC and lowers lipid accumulation caused by high glucoseconcentrations in insulin-resistant HepG2 cells, thus mimick-ing the effect of metformin.

Metformin Reduces Lipid Accumulation Caused by High Glu-cose Concentrations in an AMPK-dependent Manner—Thefunctional relationship between AMPK activation, ACC phos-phorylation, and lipid content was further examined by over-expression of a dominant-negative form of AMPK�2. AMPKactivity was inhibited by adenoviral-mediated expression of acatalytically inactive AMPK�2 bearing a K45R mutation,which has a dominant-negative effect on both �1 and �2 AMPK(29). As shown in Fig. 6A, transfection with the DN-AMPKincreased expression of the �64-kDa �2 subunit mutant, asestimated by immunoreactivity with anti-AMPK�2 or anti-Mycantibodies. Expression of the DN-AMPK suppressed levels ofphosphorylated AMPK and ACC in cells exposed to normalglucose concentrations and further suppressed these levels incells exposed to high glucose concentrations (Fig. 6A). Further-more, similar to the effect of high concentrations of glucose,cells expressing the DN-AMPK displayed an �50% increase inthe triacylglycerol levels and a smaller but significant increasein cholesterol content (Fig. 6, B and C). These results suggestnot only that exposure to high concentrations of glucose inhib-its AMPK and ACC phosphorylation, but also that modulationof AMPK by high concentrations of glucose contributes to theincrease in lipid content that accompanies insulin resistancecaused by high concentrations of glucose.

To ascertain whether AMPK activity is required for met-formin to phosphorylate ACC and lower lipids, HepG2 cellswere transfected with Ad-DN-AMPK or Ad-GFP, followed byincubation with either normal or high glucose concentrations inthe absence or presence of metformin (2 mM) for 24 h. InGFP-transfected cells, the effects of metformin on AMPK andACC phosphorylation, as well as on triacylglycerol and choles-terol content, were similar to the effects observed in non-trans-fected cells as shown in Fig. 3. However, the ability of met-formin to phosphorylate AMPK and ACC was diminished bythe DN-AMPK (Fig. 6A), indicating that metformin up-regu-lates ACC phosphorylation in an AMPK-dependent manner.Moreover, although metformin significantly decreased the tri-acylglycerol content by �40% in cells exposed to normal or highglucose concentrations, the expression of the DN-AMPK abol-ished the inhibitory effect of metformin on triacylglycerols (Fig.6B). The DN-AMPK also blocked the smaller inhibitory effect ofmetformin on cholesterol content in cells exposed to normal orhigh glucose concentrations (Fig. 6C). Together, these datasupport our hypothesis that AMPK mediates the effect of met-formin on hepatic lipid content.

DISCUSSION

The primary goal of this study was to assess the role ofAMPK activation by metformin in the regulation of hepatocel-lular lipids. We showed not only that metformin increasedphosphorylation of AMPK and its downstream lipid regulator,ACC, but also that its actions were mimicked by overexpressionof a constitutively active AMPK (CA-AMPK). In addition, anAMPK dominant-negative AMPK (DN-AMPK) blocked boththe increase in ACC phosphorylation and the decrease in lipidcontent of HepG2 cells caused by the antidiabetic agent. Fur-thermore, in cells in which an insulin-resistant, hyperlipidemicstate was induced by high glucose concentrations, the lipid-lowering effects of metformin were also mimicked by overex-pressing the CA-AMPK or prevented by the DN-AMPK. These

studies strongly suggest that the salutary effects of metforminon hepatocellular lipids are mediated predominantly by AMPK.

Stimulation of AMPK and Lipid-lowering Effects of Met-formin—Considerable evidence has been accumulated to showthat the phosphorylation of Thr-172, the major stimulatoryphosphorylation site of the � subunit, is essential for AMPKactivity (10, 11). Several groups have generated antibodies that

FIG. 6. Metformin prevents lipid accumulation induced byhigh glucose concentrations in an AMPK-dependent manner.HepG2 cells were transfected with the adenovirus vector encodingeither Ad-GFP or a Myc-tagged dominant-negative AMPK�2 (Ad-DN-AMPK) in serum-free medium for 24 h, followed by incubation with orwithout 2 mM metformin in medium containing either normal (5.5 mM)or high (30 mM) glucose concentrations for an additional 24 h. A, theability of metformin to phosphorylate AMPK and ACC is diminished bythe DN-AMPK in cells exposed to normal or high glucose concentra-tions. Expression of the DN-AMPK recombinant protein (�64 kDa) wasdetected by immunoblots using anti-Myc antibody (�-Myc) or anti-AMPK�2 antibody (�-AMPK�2), respectively. The phosphorylation ofAMPK and ACC was immunoblotted with the phospho-Thr-172AMPK� (�-pAMPK) or phospho-Ser-79 ACC (�-pACC) antibodies. AllWestern blots shown represent three independent experiments. B andC, inhibition of lipid levels by metformin was abrogated by the DN-AMPK. Each bar represents the mean � S.E. (n � 4). *, p � 0.05,compared between two groups as indicated.

Metformin Regulates Hepatic Lipids via AMP Kinase 47903

at UT

Southw

estern Medical C

enter Library, on May 12, 2011

ww

w.jbc.org

Dow

nloaded from

specifically recognize the phosphorylated form of Thr-172 andhave demonstrated that the phosphorylation status of Thr-172mirrors AMPK activity under all conditions tested (6, 7, 33, 34).In the present study, it was observed that the time- and con-centration-dependent effects of metformin increase the phos-phorylation of AMPK at Thr-172, which confirms previous ob-servations that metformin increases the catalytic activity ofAMPK in rat primary hepatocytes, the H4IIE hepatocyte cellline, skeletal muscle, and rat pancreatic � cells (21, 22, 35, 36).

In the present study, the phosphorylation state of ACC atSer-79 was also used to assess AMPK activity. ACC, whichplays a pivotal role in hepatic lipid metabolism, is controlled byallosteric regulation by citrate and glutamate and by covalentmodification by phosphorylation (37). AMPK inhibits ACC byphosphorylation of Ser-79 (38–40). Although other protein ki-nases can phosphorylate ACC, the increase in ACC phospho-rylation at Ser-79 caused by metformin in this study was in-hibited by the overexpression of the DN-AMPK, which isconsistent with AMPK being the mediator. This finding is inagreement with the observation that metformin negatively con-trols ACC activity via AMPK in skeletal muscle (36). ACCcatalyzes the biosynthesis of malonyl-CoA, which serves as theinitial substrate for fatty acid biosynthesis as well as a potentinhibitor of carnitine palmitoyltransferase I, the rate-limitingstep for mitochondrial fatty acid oxidation. The important roleof ACC in lipid metabolism has been demonstrated by in-creased fatty acid oxidation and leanness in mice deficient inACC2 (41). Previous studies have shown that AMPK activationby either AICAR or metformin stimulates fatty acid oxidation,and AICAR reduces [14C]oleate and [3H]glycerol incorporationinto triacylglycerol in rat hepatocytes (21, 42). This suggeststhat the reduction in triglyceride levels caused by metforminobserved in HepG2 cells, which was coincident with stimula-tion of ACC phosphorylation by metformin, can be explained byincreased fatty acid oxidation and/or decreased fatty acid syn-thesis. In addition to regulating ACC, it is also possible thatmetformin regulates other factors that affect fatty acid synthe-sis or oxidation, such as sterol regulatory element-binding pro-teins, which could affect the lipid levels measured in the pres-ent study. In metformin-treated rats and hepatocytes, gene andprotein expression of sterol regulatory element-binding protein1 and other lipogenic enzymes is suppressed (21). Furtherstudies will be needed to delineate critical mediators of thelipid-lowering effect of metformin via AMPK.

Because the increase in ACC phosphorylation, as well as thereduction in triacylglycerols caused by metformin, were inhib-ited in the present study by overexpression of the DN-AMPK, itis likely that ACC is the main determinant of the decrease inlipids. This conclusion is also strengthened by the fact thatoverexpression of the CA-AMPK mimicked both the increase inACC phosphorylation and the decrease in triacylglycerolscaused by metformin. In addition, the AMPK activator, AICAR,decreases synthesis of triacylglycerols and their precursor, di-acylglycerol, in hepatocytes (42). Metformin also decreasedHepG2 cell content of cholesterol. The effect of metformin oncholesterol content may be explained by the fact that 3-hy-droxy-3-methylglutaryl-coenzyme A reductase, the rate-limit-ing enzyme in cholesterol biosynthesis, is also phosphorylatedand inhibited by AMPK (43). Like the decrease in triacylglyc-erols, the effect of metformin on cholesterol was mimicked byoverexpression of the CA-AMPK and blocked by the DN-AMPK, suggesting a direct action of metformin on hepatocel-lular cholesterol metabolism via AMPK. Thus, the effects ofmetformin on both ACC and 3-hydroxy-3-methylglutaryl-coen-zyme A reductase via AMPK can explain its inhibitory effectson lipid content of HepG2 cells.

Actions of Metformin in a Hepatocellular Model of InsulinResistance—To investigate the importance of AMPK in theactions of metformin in insulin-resistant states, HepG2 cellswere exposed to high glucose concentrations for 24 h, andresistance to insulin signaling was demonstrated by decreasedinsulin-induced Akt phosphorylation. This observation is inagreement with the finding that high concentrations of glucoseattenuate insulin-induced phosphorylation of tyrosine residuesof the insulin receptor substrate 1 as well as association ofinsulin receptor substrate 1 and phosphatidylinositol 3-kinase(25). Exposure of HepG2 cells to high glucose concentrationsalso decreased AMPK and ACC phosphorylation, which wassimilar to findings in pancreatic � cell lines MIN-6 and INS-1as well as in multiple hypothalamic regions (44, 45). In addi-tion, high glucose concentrations dramatically increased thetriacylglycerol content of HepG2 cells, supporting the role ofACC in hepatocellular triacylglycerol accumulation. It is nota-ble that overexpression of the DN-AMPK also increased tri-acylglycerol content of HepG2 cells. This supports the role ofdecreased ACC phosphorylation, which indicates increased ac-tivity, in those cells as well as those exposed to high glucoseconcentrations as an important cause for the increased tri-acylglycerol content. No change in AMPK expression oc-curred in this short term model of insulin resistance, sug-gesting that changes in phosphorylation of AMPK were theprimary mediators of the effects of high glucose concentra-tions on triacylglycerol accumulation. In contrast to triacyl-glycerols, cholesterol content did not increase in HepG2 cellsexposed to high concentrations of glucose, suggesting that thechanges in AMPK and ACC phosphorylation were not suffi-cient to increase the levels; this was perhaps also reflected inthe smaller changes in cholesterol content that occurred inresponse to overexpression of the DN-AMPK.

Metformin overcame the decrease in phosphorylation ofAMPK and ACC caused by high glucose concentrations, andlowered the elevated triacylglycerol levels in cells exposed tohigh glucose concentrations to levels that were observed inuntreated cells exposed to normal glucose concentrations. Theeffects of metformin could be attributed to changes in AMPK,as they were completely abrogated by overexpression of theDN-AMPK and mimicked by the CA-AMPK. It is possiblethat the effects of metformin are also mediated by the alteredexpression of other enzymes that regulate lipid biosynthesis,such as fatty acid synthase and sterol regulatory element-binding proteins (21, 46, 47). Nevertheless, the fact thatDN-AMPK prevented the effects of metformin is strong evi-dence that its actions are primarily mediated by AMPK. Thisis also consistent with accumulating evidence that AMPKregulates the key enzymes that control lipid biosynthesis,such as glycerol-3-phosphate acyltransferase and malonyl-CoA decarboxylase (42, 48, 49). The fact that the actions ofmetformin in HepG2 cells were completely abrogated by theDN-AMPK makes this hepatocellular model of insulin resist-ance ideal for identifying other agents that affect lipid me-tabolism via AMPK.

In summary, this study provides strong biochemical evidencethat the effects of metformin on the lipid content of HepG2 cellsdepend on activation of AMPK. Despite the fact that the mech-anism by which metformin stimulates AMPK remains contro-versial, this study demonstrates that AMPK is the principalmediator of the effects of metformin on lipid biosynthesis. Ourresults also suggest that the salutary effects of metformin onthe elevated lipids associated with insulin-resistant states alsodepend on AMPK. Therefore, this study demonstrates an ex-cellent model both to explore the mechanisms by which AMPKregulates lipids as well as to identify more potent agents than

Metformin Regulates Hepatic Lipids via AMP Kinase47904

at UT

Southw

estern Medical C

enter Library, on May 12, 2011

ww

w.jbc.org

Dow

nloaded from

metformin that are capable of stimulating AMPK and control-ling lipid biosynthesis.

Acknowledgments—We thank Dr. Bingbing Jiang and Tyler Heibeckfor helpful suggestions.

REFERENCES

1. Hardie, D. G. (2003) Endocrinology 144, 5179–51832. Kemp, B. E., Stapleton, D., Campbell, D. J., Chen, Z. P., Murthy, S., Walter,

M., Gupta, A., Adams, J. J., Katsis, F., van Denderen, B., Jennings, I. G.,Iseli, T., Michell, B. J., and Witters, L. A. (2003) Biochem. Soc. Trans. 31,162–168

3. Carling, D. (2004) Trends Biochem. Sci. 29, 18–244. Ruderman, N., and Prentki, M. (2004) Nat. Rev. Drug Discov. 3, 340–3515. Hardie, D. G., Scott, J. W., Pan, D. A., and Hudson, E. R. (2003) FEBS Lett.

546, 113–1206. Shaw, R. J., Kosmatka, M., Bardeesy, N., Hurley, R. L., Witters, L. A.,

DePinho, R. A., and Cantley, L. C. (2004) Proc. Natl. Acad. Sci. U. S. A. 101,3329–3335

7. Woods, A., Johnstone, S. R., Dickerson, K., Leiper, F. C., Fryer, L. G.,Neumann, D., Schlattner, U., Wallimann, T., Carlson, M., and Carling, D.(2003) Curr. Biol. 13, 2004–2008

8. Lizcano, J. M., Goransson, O., Toth, R., Deak, M., Morrice, N. A., Boudeau, J.,Hawley, S. A., Udd, L., Makela, T. P., Hardie, D. G., and Alessi, D. R. (2004)EMBO J. 23, 833–843

9. Hong, S. P., Leiper, F. C., Woods, A., Carling, D., and Carlson, M. (2003) Proc.Natl. Acad. Sci. U. S. A. 100, 8839–8843

10. Crute, B. E., Seefeld, K., Gamble, J., Kemp, B. E., and Witters, L. A. (1998)J. Biol. Chem. 273, 35347–35354

11. Stein, S. C., Woods, A., Jones, N. A., Davison, M. D., and Carling, D. (2000)Biochem. J. 345, 437–443

12. Hardie, D. G., and Carling, D. (1997) Eur. J. Biochem. 246, 259–27313. Gerich, J. E. (1998) Endocr. Rev. 19, 491–50314. Winder, W. W., and Hardie, D. G. (1999) Am. J. Physiol. 277, E1–E1015. Buhl, E. S., Jessen, N., Pold, R., Ledet, T., Flyvbjerg, A., Pedersen, S. B.,

Pedersen, O., Schmitz, O., and Lund, S. (2002) Diabetes 51, 2199–220616. Fedele, D., Tiengo, A., Nosadini, R., Marchiori, E., Briani, G., Garotti, M. C.,

and Muggeo, M. (1976) Diabetes Metab. 2, 127–13317. Hermann, L. S. (1979) Diabetes Metab. 5, 233–24518. Davidson, M. B., and Peters, A. L. (1997) Am. J. Med. 102, 99–11019. Anurag, P., and Anuradha, C. V. (2002) Diabetes Obes. Metab. 4, 36–4220. Yuan, L., Ziegler, R., and Hamann, A. (2003) Acta Pharmacol. Sin. 24, 55–6021. Zhou, G., Myers, R., Li, Y., Chen, Y., Shen, X., Fenyk-Melody, J., Wu, M.,

Ventre, J., Doebber, T., Fujii, N., Musi, N., Hirshman, M. F., Goodyear,L. J., and Moller, D. E. (2001) J. Clin. Investig. 108, 1167–1174

22. Hawley, S. A., Gadalla, A. E., Olsen, G. S., and Hardie, D. G. (2002) Diabetes51, 2420–2425

23. Yu-Poth, S., Yin, D., Zhao, G., Kris-Etherton, P. M., and Etherton, T. D. (2004)J. Nutr. 134, 68–71

24. Payne, K. L., Loidl, N. M., Lim, C. F., Topliss, D. J., Stockigt, J. R., and Barlow,J. W. (1997) Eur. J. Endocrinol. 137, 415–420

25. Nakajima, K., Yamauchi, K., Shigematsu, S., Ikeo, S., Komatsu, M., Aizawa,T., and Hashizume, K. (2000) J. Biol. Chem. 275, 20880–20886

26. Riebeling, C., Forsea, A. M., Raisova, M., Orfanos, C. E., and Geilen, C. C.(2002) Br. J. Cancer 87, 366–371

27. Nagata, D., Mogi, M., and Walsh, K. (2003) J. Biol. Chem. 278, 31000–3100628. Zou, M. H., Hou, X. Y., Shi, C. M., Nagata, D., Walsh, K., and Cohen, R. A.

(2002) J. Biol. Chem. 277, 32552–3255729. Mu, J., Brozinick, J. T., Jr., Valladares, O., Bucan, M., and Birnbaum, M. J.

(2001) Mol. Cell 7, 1085–109430. Zang, M., Waelde, C. A., Xiang, X., Rana, A., Wen, R., and Luo, Z. (2001)

J. Biol. Chem. 276, 25157–2516531. Zang, M., Hayne, C., and Luo, Z. (2002) J. Biol. Chem. 277, 4395–440532. Wang, Y. X., Lee, C. H., Tiep, S., Yu, R. T., Ham, J., Kang, H., and Evans, R. M.

(2003) Cell 113, 159–17033. Yamauchi, T., Kamon, J., Minokoshi, Y., Ito, Y., Waki, H., Uchida, S.,

Yamashita, S., Noda, M., Kita, S., Ueki, K., Eto, K., Akanuma, Y., Froguel,P., Foufelle, F., Ferre, P., Carling, D., Kimura, S., Nagai, R., Kahn, B. B.,and Kadowaki, T. (2002) Nat. Med. 8, 1288–1295

34. Inoki, K., Zhu, T., and Guan, K. L. (2003) Cell 115, 577–59035. Kefas, B. A., Cai, Y., Kerckhofs, K., Ling, Z., Martens, G., Heimberg, H.,

Pipeleers, D., and Van de, C. M. (2004) Biochem. Pharmacol. 68, 409–41636. Musi, N., Hirshman, M. F., Nygren, J., Svanfeldt, M., Bavenholm, P.,

Rooyackers, O., Zhou, G., Williamson, J. M., Ljunqvist, O., Efendic, S.,Moller, D. E., Thorell, A., and Goodyear, L. J. (2002) Diabetes 51,2074–2081

37. Witters, L. A., Watts, T. D., Daniels, D. L., and Evans, J. L. (1988) Proc. Natl.Acad. Sci. U. S. A. 85, 5473–5477

38. Winder, W. W., Wilson, H. A., Hardie, D. G., Rasmussen, B. B., Hutber, C. A.,Call, G. B., Clayton, R. D., Conley, L. M., Yoon, S., and Zhou, B. (1997)J. Appl. Physiol 82, 219–225

39. Dyck, J. R., Kudo, N., Barr, A. J., Davies, S. P., Hardie, D. G., and Lopaschuk,G. D. (1999) Eur. J. Biochem. 262, 184–190

40. Ruderman, N., and Flier, J. S. (2001) Science 291, 2558–255941. Abu-Elheiga, L., Matzuk, M. M., Abo-Hashema, K. A., and Wakil, S. J. (2001)

Science 291, 2613–261642. Muoio, D. M., Seefeld, K., Witters, L. A., and Coleman, R. A. (1999) Biochem.

J. 338, 783–79143. Carling, D., Zammit, V. A., and Hardie, D. G. (1987) FEBS Lett. 223, 217–22244. da Silva Xavier, G., Leclerc, I., Varadi, A., Tsuboi, T., Moule, S. K., and Rutter,

G. A. (2003) Biochem. J. 371, 761–77445. Minokoshi, Y., Alquier, T., Furukawa, N., Kim, Y. B., Lee, A., Xue, B., Mu, J.,

Foufelle, F., Ferre, P., Birnbaum, M. J., Stuck, B. J., and Kahn, B. B. (2004)Nature 428, 569–574

46. Foretz, M., Carling, D., Guichard, C., Ferre, P., and Foufelle, F. (1998) J. Biol.Chem. 273, 14767–14771

47. Woods, A., Azzout-Marniche, D., Foretz, M., Stein, S. C., Lemarchand, P.,Ferre, P., Foufelle, F., and Carling, D. (2000) Mol. Cell. Biol. 20, 6704–6711

48. Saha, A. K., Schwarsin, A. J., Roduit, R., Masse, F., Kaushik, V., Tornheim, K.,Prentki, M., and Ruderman, N. B. (2000) J. Biol. Chem. 275, 24279–24283

49. Sambandam, N., Steinmetz, M., Chu, A., Altarejos, J. Y., Dyck, J. R., andLopaschuk, G. D. (2004) Eur. J. Biochem. 271, 2831–2840

Metformin Regulates Hepatic Lipids via AMP Kinase 47905

at UT

Southw

estern Medical C

enter Library, on May 12, 2011

ww

w.jbc.org

Dow

nloaded from