Metformin Lowers Plasma Triglycerides by Promoting VLDL-Triglyceride Clearance by Brown Adipose Tissue in Mice

  1. Bruno Guigas2,3
  1. 1Department of General Internal Medicine, Endocrinology, and Metabolic Diseases, Leiden University Medical Center, Leiden, the Netherlands
  2. 2Einthoven Laboratory for Experimental Vascular Medicine, Leiden University Medical Center, Leiden, the Netherlands
  3. 3Department of Molecular Cell Biology, Leiden University Medical Center, Leiden, the Netherlands
  4. 4Department of Human Genetics, Leiden University Medical Center, Leiden, the Netherlands
  5. 5Gaubius Laboratory, Netherlands Organization for Applied Scientific Research-Metabolic Health Research, Leiden, the Netherlands
  6. 6INSERM, U693, Faculté de Médecine Paris-Sud, Université Paris-Sud, Le Kremlin-Bicêtre, France
  7. 7Department of Parasitology, Leiden University Medical Center, Leiden, the Netherlands
  1. Corresponding author: Bruno Guigas, b.g.a.guigas{at}
  1. J.J.G., M.R.B., P.C.N.R., and B.G. contributed equally to this study.


Metformin is the first-line drug for the treatment of type 2 diabetes. Besides its well-characterized antihyperglycemic properties, metformin also lowers plasma VLDL triglyceride (TG). In this study, we investigated the underlying mechanisms in APOE*3-Leiden.CETP mice, a well-established model for human-like lipoprotein metabolism. We found that metformin markedly lowered plasma total cholesterol and TG levels, an effect mostly due to a decrease in VLDL-TG, whereas HDL was slightly increased. Strikingly, metformin did not affect hepatic VLDL-TG production, VLDL particle composition, and hepatic lipid composition but selectively enhanced clearance of glycerol tri[3H]oleate-labeled VLDL-like emulsion particles into brown adipose tissue (BAT). BAT mass and lipid droplet content were reduced in metformin-treated mice, pointing to increased BAT activation. In addition, both AMP-activated protein kinase α1 (AMPKα1) expression and activity and HSL and mitochondrial content were increased in BAT. Furthermore, therapeutic concentrations of metformin increased AMPK and HSL activities and promoted lipolysis in T37i differentiated brown adipocytes. Collectively, our results identify BAT as an important player in the TG-lowering effect of metformin by enhancing VLDL-TG uptake, intracellular TG lipolysis, and subsequent mitochondrial fatty acid oxidation. Targeting BAT might therefore be considered as a future therapeutic strategy for the treatment of dyslipidemia.


Metformin is one of the most widely used glucose-lowering agents for the treatment of type 2 diabetes (1) and is now considered the first-line drug therapy for patients (2). This antidiabetic drug from the biguanides family is prescribed for its effective antihyperglycemic action, mostly achieved through a potent reduction of hepatic glucose production secondary to inhibition of gluconeogenesis (3). Interestingly, another important but often overlooked property of metformin relies on its beneficial effect on the blood lipid profile, which is characterized by a significant reduction in circulating triglycerides (TGs) and VLDL cholesterol and increased HDL cholesterol levels (4). This metabolic feature might partly be involved in its cardioprotective effect observed in obese patients treated with the drug (5). Despite extensive efforts during the last years (6), the exact molecular mechanism(s) of action of metformin still remains incompletely understood, especially the one by which the drug exerts its lipid-lowering action. In 2001, Zhou et al. (7) were the first to report that metformin activates hepatic AMP-activated protein kinase (AMPK), emphasizing the putative role of this energy-sensing kinase in the mechanism of action of the drug.

AMPK is a well-conserved serine/threonine protein kinase that plays a crucial role in the regulation of catabolic/anabolic pathways by acting as a cellular energy and nutrient sensor (8,9). AMPK consists of a heterotrimeric complex containing a catalytic α subunit and two regulatory β and γ subunits. Each subunit has several isoforms (α1, α2; β1, β2; γ1, γ2, γ3) that are encoded by distinct genes, giving multiple heterotrimeric combinations with tissue-specific distribution (8,9). The α subunit contains a threonine residue (Thr 172) whose phosphorylation by upstream kinases, such as the liver kinase B (LKB1), is required for AMPK activation. The β subunit acts as a scaffold to which the two other subunits are bound and also allows AMPK to sense energy reserves in the form of glycogen (8,9). Binding of AMP and/or ADP to selective domains on the γ subunit leads to AMPK activation via a complex mechanism involving direct allosteric activation, phosphorylation on Thr172 by AMPK upstream kinases, and inhibition of dephosphorylation of this residue by specific protein phosphatases that remain to be identified (8,9). Interestingly, the mechanism by which metformin activates AMPK, involving specific inhibition of the mitochondrial respiratory chain complex 1 (10,11), was recently clarified (12,13), although the contribution of the LKB1/AMPK axis in its hepatic effects still remains controversial (1418).

The objective of this study was to investigate the molecular mechanisms underlying the effects of metformin on lipoprotein metabolism by using APOE*3-Leiden.CETP (E3L.CETP) transgenic mice, a well-established model of human-like lipoprotein metabolism (19) that also responds to lipid-lowering pharmacological interventions (2023). Collectively, our data show that treatment of E3L.CETP mice with metformin is able to recapitulate the lipid-lowering effect of the drug evidenced in humans, i.e., causing a reduction in plasma VLDL-TG associated with a parallel mild increase in HDL cholesterol. Remarkably, this effect is mediated not by apparent changes in hepatic VLDL-TG production but rather by a selective increase in VLDL-TG clearance by the brown adipose tissue (BAT). At the molecular level, we found an increase in AMPKα1 activity and protein expression of both hormone-sensitive lipase (HSL) and mitochondrial respiratory chain complexes, suggesting that metformin, on top of increasing VLDL-TG uptake, also promotes intracellular TG lipolysis and subsequent mitochondrial fatty acid (FA) oxidation in BAT.

Research Design and Methods


All chemicals were purchased from Sigma-Aldrich (St. Louis, MO).


All mouse experiments were performed in accordance with the Institute for Laboratory Animal Research Guide for the Care and Use of Laboratory Animals and have received approval from the university ethical review boards (Leiden University Medical Center).

Animals, Diet, and Metformin Treatment

Homozygous human CETP transgenic mice were crossbred with hemizygous APOE*3-Leiden (E3L) mice at our Institutional Animal Facility to obtain E3L.CETP mice, as previously described (19). In this study, 12-week-old E3L.CETP female mice, housed under standard conditions in conventional cages with ad libitum access to food and water, were fed a Western-type diet containing 0.1% (weight for weight [w/w]) cholesterol (Hope Farms, Woerden, the Netherlands) for 4 weeks. Upon randomization according to body weight, plasma total cholesterol (TC), and TG levels, mice next received a Western-type diet with or without 200 mg/kg body weight/day (0.2%, w/w) metformin for 4 weeks. Unless otherwise mentioned, experiments were performed after 4 h of fasting at 1:00 p.m. with food withdrawn at 9:00 a.m.

Plasma Lipid and Lipoprotein Analysis

Plasma was obtained via tail vein bleeding and assayed for TC, TG, and phospholipid (PL) using the commercially available enzymatic kits 236691, 11488872, and 1001140 (Roche Molecular Biochemicals, Indianapolis, IN), respectively. Free FAs were measured using the NEFA-C kit from Wako Diagnostics (Instruchemie, Delfzijl, the Netherlands). The distribution of lipids over plasma lipoprotein fractions was determined using fast protein liquid chromatography. Plasma was pooled per group, and 50 μL of each pool was injected onto a Superose 6 PC 3.2/30 column (Akta System, Amersham Pharmacia Biotech, Piscataway, NJ) and eluted at a constant flow rate of 50 μL/min in 1 mmol/L EDTA in PBS, pH 7.4. Fractions of 50 μL were collected and assayed for TC and TG as described above.

Hepatic VLDL-TG and VLDL-Apolipoprotein B Production

Mice were fasted for 4 h prior to the start of the experiment. During the experiment, mice were sedated with 6.25 mg/kg acepromazine (Alfasan, Woerden, the Netherlands), 6.25 mg/kg midazolam (Roche, Mijdrecht, the Netherlands), and 0.31 mg/kg fentanyl (Janssen-Cilag, Tilburg, the Netherlands). At t = 0 min, blood was taken via tail bleeding and mice were intravenously injected with 100 μL PBS containing 100 μCi Trans35S label (ICM Biomedicals, Irvine, CA) to measure de novo total apolipoprotein B (apoB) synthesis. After 30 min, the animals received 500 mg tyloxapol (Triton WR-1339; Sigma-Aldrich) per kilogram body weight as a 10% (w/w) solution in sterile saline, to prevent systemic lipolysis of newly secreted hepatic VLDL-TG. Additional blood samples were taken at t = 15, 30, 60, and 90 min after tyloxapol injection and used for determination of plasma TG concentration. After 90 min, the animals were killed and blood was collected by orbital bleeding for isolation of VLDL by density-gradient ultracentrifugation, as previously described (1923). 35S-apoB was measured in the VLDL fraction, and VLDL-apoB production rate was calculated as dpm/h, as previously reported (1923).

In Vivo Clearance of VLDL-Like Emulsion Particles

Mice were fasted overnight with food withdrawn at 6:00 p.m. During the experiment, mice were sedated as described above. At t = 0 min, blood was taken via tail bleeding and mice received a continuous intravenous infusion of glycerol tri[3H]oleate-labeled emulsion particles mixed with albumin-bound [14C]oleic acid (4.4 µCi [3H]TG and 1.2 µCi [14C]FA; both from GE Healthcare Life Sciences, Little Chalfont, U.K.) at a rate of 100 µL/h for 2.5 h, as previously described (24,25). Blood samples were taken using chilled paraoxon-coated capillaries by tail bleeding at 90 and 120 min of infusion to ensure that steady-state conditions had been reached. Subsequently, mice were killed and organs were quickly harvested and snap frozen in liquid nitrogen. Retention of radioactivity in the saponified tissues was measured per milligram of tissue and corrected for the corresponding plasma-specific activities of [3H]FA and [14C]FA, as previously described (24).

Hepatic Lipid Composition

Liver lipids were extracted as previously described (20). In brief, small liver pieces were homogenized in ice-cold methanol. After centrifugation, lipids were extracted by addition of 1,800 µL CH3OH:CHCl3 (1:3 volume for volume [v/v]) to 45 µL homogenate, followed by vigorous vortexing and phase separation by centrifugation (14,000 rpm; 15 min at room temperature). The organic phase was dried and dissolved in 2% Triton X-100 in water. TG, TC, and PL concentrations were measured using commercial kits as described above. Liver lipids were expressed as nanomoles per milligram protein and were determined using the BCA protein assay kit (Pierce, Rockford, IL).


Interscapular BAT was removed and fixed directly in 4% paraformaldehyde, dehydrated, and embedded in paraffin. Hematoxylin and eosin staining was performed using standard protocols. The area of intracellular lipid vacuoles in BAT was quantified using ImageJ (NIH, Bethesda, MD).

Cell Culture and Brown Adipocyte Differentiation

T37i cells were cultured and differentiated as described previously (26). Cells were next treated with metformin or vehicle (PBS) for 8 h. Then, supernatant was collected for determination of glycerol (Instruchemie, Delfzijl, the Netherlands) and cells were harvested in ice-cold lysis buffer, as described below.

Western Blot Analysis

Snap-frozen liver and BAT samples (∼50 mg) or T37i cells were lysed in ice-cold buffer containing the following: 50 mmol/L HEPES (pH 7.6), 50 mmol/L NaF, 50 mmol/L KCl, 5 mmol/L NaPPi, 1 mmol/L EDTA, 1 mmol/L EGTA, 1 mmol/L dithiothreitol, 5 mmol/L β-glycerophosphate, 1 mmol/L sodium vanadate, 1% NP40, and protease inhibitor cocktail (Complete; Roche). Western blots were performed as previously described (13). All the primary antibodies used are listed in Supplementary Table 1. Bands were visualized by enhanced chemiluminescence and quantified using ImageJ.

AMPK Assay

AMPK activity was assayed after immunoprecipitation with specific antibodies directed against α1- or α2-AMPK catalytic subunits (Kinasource, Dundee, Scotland), as previously described (14).

RNA/DNA Purification and qRT-PCR

RNA was extracted from snap-frozen liver or BAT samples (∼25 mg) using Tripure RNA Isolation Reagent (Roche). Total RNA (1–2 µg) was reverse transcribed, and quantitative real-time PCR was then performed with SYBR Green Core Kit on a MyIQ thermal cycler (Bio-Rad). mRNA expression was normalized to CypD mRNA content and expressed as fold change compared with control mice using the ∆∆CT method. Genomic DNA was extracted using the Qiagen Tissue and Blood Kit (Qiagen, Hilden, Germany). For mitochondrial DNA copy number, ND1 (mitochondrial) and LPL (nuclear) copy numbers were quantified by qRT-PCR. All the primer sequences are listed in Supplementary Table 1.

Statistical Analysis

All data are expressed as mean ± SEM. Statistical analysis was performed using SPSS 17.0 software package for Windows (SPSS, Chicago, IL) with two-tailed unpaired Student t test. Differences between groups were considered statistically significant at P < 0.05.


Metformin Reduces Plasma Cholesterol and TG Levels

To investigate the effect of metformin on lipoprotein metabolism, E3L.CETP mice were first fed a cholesterol-rich (0.1%) Western-type diet for 4 weeks and next treated with or without metformin (200 mg/kg body weight/day) added to the diet for another 4 weeks. As compared with the control group, metformin did not affect body weight and composition, food intake and plasma glucose, and insulin and FA levels throughout the intervention period (Supplementary Fig. 1). However, metformin rapidly reduced both plasma TC (−27 and −36% at weeks 2 and 4, respectively; P < 0.05) and TG (−26 and −38% at weeks 2 and 4, respectively; P < 0.05) in a time-dependent manner (Fig. 1A and C). Plasma lipoprotein profile analysis showed that this lipid-lowering effect mostly resulted from a reduction of VLDL particles. In addition, a slight shift in plasma cholesterol profile, from VLDL cholesterol to HDL cholesterol (−37 and 37%, respectively), was evidenced (Fig. 1B and D).

Figure 1

Effect of metformin on plasma cholesterol and TG levels and lipoprotein distribution. Blood samples from 4 h–fasted control (black bars) and metformin-treated (open bars) mice were collected by tail bleeding using chilled paraoxon-coated capillaries at different time points. Plasma TC (A) and TG (C) levels were determined. The plasma samples collected after 4 weeks of treatment were pooled group wise and size fractionated by fast protein liquid chromatography. The individual fractions were analyzed for cholesterol (B) and TG (D). Data are means ± SEM (n = 9 per group). *P < 0.05 vs. control.

Metformin Does Not Affect Hepatic VLDL-TG Production

Plasma VLDL-TG levels are determined by the balance between VLDL-TG production by the liver and VLDL-TG clearance by peripheral organs. Therefore, we first assessed the effect of metformin on hepatic VLDL-TG and -apoB production by injecting Trans35S and tyloxapol in 4 h–fasted control and metformin-treated E3L.CETP mice. Despite the significantly lower basal plasma TG levels (1.72 ± 0.26 vs. 2.65 ± 0.36 mmol/L, P < 0.05; data not shown), metformin did not affect the time-dependent accumulation of plasma TG after tyloxapol injection when compared with control E3L.CETP mice (Fig. 2A). Therefore, the VLDL-TG production rate, calculated from the slope of the curve, was not significantly different (Fig. 2A, inserted panel), although a trend for a slight decrease can eventually be suggested. The rate of VLDL-apoB production (Fig. 2B), the ratio of TG-apoB (Fig. 2C), as well as the composition of the VLDL particles secreted (Fig. 2D) were not significantly altered, indicating that metformin did not affect the hepatic lipidation of VLDL particles. In line with these results, the TG, TC, and PL content in the liver from E3L.CETP mice did not significantly differ between the control and metformin groups, although hepatic TC content tended to be decreased in the metformin-treated group (−21%, P = 0.07) (Supplementary Fig. 2). Furthermore, in our experimental conditions, metformin treatment did not affect hepatic AMPK activity, as assessed by phosphorylation of Thr172-AMPK and Ser79-acetyl-CoA carboxylase (ACC), the main downstream target of AMPK (Supplementary Fig. 2). Finally, we found that hepatic expression of key genes involved in FA/TG uptake, synthesis, and oxidation were not affected, whereas Lrp1 and Scarp1, both involved in cholesterol uptake, were significantly downregulated by metformin (Supplementary Table 1). In addition, the expression of Abca1, Lcat, and Pltp was also found to be significantly downregulated by metformin, suggesting that part of the HDL-enhancing effect of the drug could result from subtle changes in hepatic lipoprotein metabolism.

Figure 2

Effect of metformin on hepatic VLDL-TG production. After 4 weeks of treatment, 4 h–fasted control (black squares/bars) and metformin-treated mice (open squares/bars) were injected with Trans35S label (t = −30 min) and tyloxapol (t = 0 min), and blood samples were drawn up to 90 min after tyloxapol injection. Plasma TG concentrations were determined and plotted as the increase in plasma TG as compared with baseline (A). The rate of TG production was calculated from the slopes of the curves from the individual mice (A, inset). After 120 min, mice were exsanguinated and the total VLDL fraction was isolated by ultracentrifugation. The rate of newly synthesized VLDL-35S-apoB (B), the TG-to–35S-apoB ratio (C), as well as the amount of TG, TC, and PL per mg VLDL protein (D) were measured. Data are means ± SEM (n = 5–8 per group).

Metformin Promotes VLDL-TG Clearance by BAT and Influences BAT Mass and Composition

As clearance of TG from plasma is the other major determinant of TG metabolism, the effect of metformin on whole-body lipid partitioning was investigated next. For this purpose, the tissue-specific retention of FA derived from both [3H]TG-labeled VLDL-like emulsion particles and albumin-bound [14C]FA was determined after continuous tracer infusion for 2.5 h. Strikingly, metformin did not affect the uptake of [3H]TG-derived FA by liver, heart, skeletal muscle, and various WAT depots but markedly increased 3H retention in BAT (+58%, P < 0.05) (Fig. 3A). The uptake of albumin-bound [14C]FA was not different for any of the organs studied (Fig. 3B), suggesting that metformin does not affect FA uptake per se but rather promotes lipoprotein lipase (LPL)-mediated VLDL-TG hydrolysis in BAT. Interestingly, BAT mass (−29%) and intracellular lipid droplet content (−91%) were found to be reduced in metformin-treated mice (Fig. 3CE), both pointing toward more active BAT (27). However, neither UCP1 mRNA expression nor protein content was significantly affected (Fig. 3F and G).

Figure 3

Effect of metformin on peripheral VLDL-TG clearance and BAT. Four hour–fasted control (black bars) and metformin-treated (open bars) mice were continuously infused with [3H]TG-labeled VLDL-like emulsion particles mixed with albumin-bound [14C]FA for 2.5 h. Blood samples were taken using chilled paraoxon-coated capillaries by tail bleeding at 90 and 120 min of infusion to ensure that steady-state conditions had been reached. Subsequently, mice were killed and organs were quickly harvested and snap frozen in liquid nitrogen. Plasma levels of TGs and FAs were determined in plasma, and uptake of the radioactively [3H]TG-labeled emulsion particles (A) and albumin-bound [14C]FA (B) was determined in the organs. In separate experiments, BAT from control and metformin-treated mice was collected and weighed (C). Hematoxylin and eosin staining of BAT sections was performed, and representative pictures are shown (D). Relative content of lipid vacuoles in BAT tissue sections (n = 3–4 per group) was quantified (E). mRNA (F) and protein (G) expression of UCP1 were determined by qRT-PCR and Western blot, respectively. Data are means ± SEM (n = 7–8 per group). *P < 0.05 vs. control. BW, body weight; gWAT, gonadal white adipose tissue; sWAT, subcutaneous white adipose tissue; vWAT, visceral white adipose tissue. (A high-quality color representation of this figure is available in the online issue.)

Metformin Increases AMPK Activity, Lipolytic Machinery, and Mitochondrial Content in BAT

To further investigate the molecular mechanism by which metformin increased VLDL-TG clearance by BAT, we first showed that the organic cation transporter 1 (OCT1), which is crucial for intracellular transport of metformin (28), was expressed in BAT at both transcript (data not shown) and protein levels (Fig. 4A). We next determined the mRNA expression of genes involved in FA/lipoprotein uptake, FA metabolism, mitochondrial functions, and BAT differentiation but did not find any significant effect of metformin treatment in our experimental condition (Supplementary Table 1). We confirmed that BAT mostly expressed AMPKα1 (Fig. 4A), as previously reported (29,30). Interestingly, we found that metformin selectively increased the activity of α1- (19%, P < 0.05) but not of α2-containing AMPK heterotrimers in BAT (Fig. 4B) without affecting the whole-tissue energy state, as assessed by the AMP-ATP ratio (Fig. 4C). This was associated with a significant increase in both Thr172 phosphorylation (21%, P < 0.05) and expression of the AMPKα catalytic subunits (21%, P < 0.05), whereas the phospho-to-total ratio was not affected (Fig. 4DF). However, a trend for increased phosphorylation of the AMPK downstream target ACC was evidenced (22%, P = 0.07) (Fig. 4DF). The only AMPK subunit significantly increased by metformin was the α1 isoform (38%, P < 0.05), whereas the other subunits were not affected (Fig. 4G), suggesting that the higher AMPK activity in BAT from metformin-treated mice was mostly due to an increase in AMPKα1 content. We next examined whether some of the key players involved in the regulation of TG lipolysis and FA oxidation in BAT were affected by metformin. Interestingly, the protein expression of the lipolytic enzyme HSL, but not that of adipose TG lipase (ATGL) and adipose differentiation–related protein (ADRP), was significantly increased by metformin in BAT (30%, P < 0.05) (Fig. 5A and B). In addition, although the mitochondrial DNA content was not affected (data not shown), we found a marked increase in protein expression of two of the main regulators of mitochondrial biogenesis, endothelial nitric oxide synthase (eNOS) and PGC-1α (23 and 127%, respectively; P < 0.05), together with a significantly higher content of citrate synthase (CS) and of most of the mitochondrial electron transport chain complexes in BAT from metformin-treated E3L.CETP mice (Fig. 5A and CE). Of note, the ratio of mitochondrial respiratory chain complex 2 to complex 1 expression was also increased by metformin in BAT (+17%, P < 0.05) (Fig. 5F).

Figure 4

Effect of metformin on AMPK expression, activity, and downstream signaling in BAT. BAT from 4 h–fasted mice was collected after 4 weeks of treatment with (open bars) or without metformin (control, black bars) and immediately snap frozen in liquid nitrogen. The protein expression of OCT1, AMPKα1, and AMPKα2 was determined by Western blot in BAT from control mice (A). AMPK activity was measured by kinase assay after immunoprecipitation of either AMPKα1 or -α2 catalytic subunits with specific antibodies (B). Adenine nucleotide concentrations in BAT were measured by high-performance liquid chromatography, and the AMP-to-ATP ratio was calculated (C). The protein expression of BAT AMPK catalytic (α) and regulatory (β and γ) subunits, ACC, and the phosphorylation states of Thr172-AMPK and Ser79-ACC were assessed by Western blot (D), followed by densitometric quantification (EG). Tubulin expression was used as internal housekeeping protein. Data are means ± SEM (n = 8 per group). *P < 0.05 vs. control. IP, intraperitoneal; ND, not detectable; p/T, phospho-to-total ratio.

Figure 5

Effect of metformin on expression of key lipolytic and mitochondrial proteins in BAT. BAT was collected in control (black bars) and metformin-treated (open bars) mice, as described in Fig. 4. The protein expression of HSL, ATGL, ADRP, eNOS, PGC-1α, and CS and of various mitochondrial respiratory chain subunits (C1, NDUFB8; C2, SDHB; C3, UQCRC2; C4, MTCO1 and COX4l1; C5, ATP5A) was assessed by Western blot (A), followed by densitometric quantification (BE). Tubulin expression was used as internal housekeeping protein. The ratio of mitochondrial respiratory chain complex 2 to complex 1 was calculated (F). Data are means ± SEM (n = 8 per group). *P < 0.05 vs. control. ETC, electron transport chain. p/T, phospho-to-total ratio.

Metformin Induces AMPK Activation and Lipolysis in T37i Brown Adipocytes

To investigate whether metformin can directly affect AMPK activity and lipid metabolism in BAT, we used T37i cells, a well-established in vitro model of differentiated brown adipocytes (26). Interestingly, we showed that therapeutic concentrations of metformin dose-dependently increased phosphorylation states of AMPK, ACC, and HSL (Fig. 6AF), as well as glycerol release into the medium, pointing toward enhanced intracellular lipolysis (Fig. 6G). Taken together, our results show that metformin not only promotes VLDL-TG uptake by BAT but also enhances both intracellular lipolytic and mitochondrial FA β-oxidation capacity in this tissue (Fig. 7).

Figure 6

Effect of metformin on AMPK signaling and lipolysis in T37i brown adipocytes. T37i cells were cultured and differentiated in brown adipocytes as described in RESEARCH DESIGN AND METHODS and next treated with increasing concentrations of metformin for 8 h. The protein expression and phosphorylation states of AMPKα, ACC, and HSL on various residues were assessed by Western blot (A) and the phospho-to-total ratios were calculated after densitometric quantification (BF). Tubulin expression was used as internal housekeeping protein. The lipolysis rate was assessed by measuring the time-dependent accumulation of glycerol in the culture medium (G) and expressed as μmol/h/mg protein. Data are means ± SEM (n = 3). *P < 0.05 vs. control. p/T, phospho-to-total ratio.

Figure 7

Proposed mechanism for the BAT-mediated TG-lowering effect of metformin. Metformin exerts a beneficial effect on circulating lipids by lowering plasma TG through a selective increase in TG-derived FA uptake by BAT. In addition, metformin also improves intracellular lipolytic capacity by increasing HSL expression and activity, thereby enhancing FA release from TG stored in lipid droplets. We propose that metformin next promotes FA oxidation in BAT by multiple (path)ways. First, metformin activates AMPK, leading to the subsequent phosphorylation and inactivation of its downstream target ACC. This relieves the inhibition exerted by malonyl-CoA on CPT1, ultimately promoting mitochondrial FA transport and oxidation. Second, metformin increases the tissue mitochondrial content, an effect that might be due to AMPK-mediated stimulation of mitochondrial biogenesis, as reflected by higher expression of eNOS and PGC-1α. Finally, by changing the qualitative composition of the mitochondrial respiratory chain, metformin can enhance respiratory chain complex 2–mediated FA oxidation and UCP1-independent metabolic uncoupling of oxidative phosphorylation. SA, serum albumin.


Metformin not only improves glycemic control in type 2 diabetic patients but also exerts beneficial effects on plasma lipid profiles (4) by a mechanism that has remained, so far, poorly understood. In the current study, we have therefore investigated the molecular mechanism(s) underlying this lipid-lowering property of metformin using E3L.CETP mice, a well-characterized transgenic model displaying a human-like lipoprotein metabolism and human-like responses to lipid-modulating drugs when fed a Western-type diet (1923). Our results show that chronic treatment of E3L.CETP mice with metformin recapitulates the effects on circulating lipoproteins observed in patients treated with the drug, i.e., reduction in plasma TG associated with significant reduction in VLDL (31). We next demonstrated that metformin does not affect hepatic VLDL-TG production but instead selectively promotes VLDL-TG clearance by BAT, an effect associated with elevated components of intracellular lipolytic and mitochondrial FA oxidation machinery in this highly active metabolic tissue. To the best of our knowledge, this study is the first one reporting that BAT is involved in the lipid-lowering effect of metformin and therefore constitutes an important target tissue for the drug.

Plasma TG levels are determined by the balance between production of chylomicron-TG and VLDL-TG in intestine and liver, respectively, and their LPL-mediated TG clearance in peripheral tissues. In our study, all the experiments were performed in fasted mice, thereby excluding any significant contribution of intestine-derived chylomicrons to the change observed in circulating TG concentrations. Furthermore, metformin treatment did not affect the postprandial response to an oral lipid load (Supplementary Fig. 3), suggesting that impaired intestinal TG absorption is not involved in the TG-lowering effect of the drug. Besides its central role in glucose homeostasis, the liver plays a key role in lipid metabolism, notably by regulating synthesis and secretion of apoB-containing VLDL-TG particles (32). Hepatic VLDL-TG production is mostly driven by intracellular substrate availability resulting from both FA uptake from the circulation and the balance between de novo lipogenesis and mitochondrial FA β-oxidation in the liver (33). In our study, we found that metformin did not significantly affect plasma FA levels; hepatic lipid content; AMPK activity; expression of genes involved in FA/TG uptake, synthesis, and oxidation; VLDL-TG and VLDL-apoB secretion rates; and composition of the excreted VLDL particles. Although we did not find an apparent contribution of the liver to the TG-reducing effect of metformin, we cannot completely exclude that some of its hepatic effects were lowered or masked due to our experimental conditions, e.g., fasting state, and the pharmacokinetic features of the drug. Of note, we found that expression of some genes involved in hepatic HDL uptake (Lrp1 and Scarb1) and remodeling (Abca1 and Pltp) was decreased by metformin, suggesting that part of the mild HDL-raising effect of the drug might be partly due to subtle changes in cholesterol metabolism in the liver. Future studies are required for clarifying the exact underlying molecular mechanism.

Plasma VLDL-TG clearance is driven by LPL-mediated lipolysis in the capillaries of peripheral tissues (34). The most striking result of our present study was that metformin induced a potent and selective increase in VLDL-glycerol tri[3H]oleate-derived [3H]oleate retention in BAT without affecting VLDL-TG uptake by heart, muscle, and various white adipose tissues. Recently, Bartelt et al. (35) were the first to identify BAT as a major organ involved in plasma VLDL-TG clearance in rodents. In this elegant study confirming previous observations (36), they reported that BAT constitutes a quantitatively relevant lipid-clearing organ displaying very high rates of VLDL-TG uptake (35) by a mechanism that still remains to be fully characterized. In the current study, our observation that metformin promotes VLDL-[3H]TG–derived FA but not albumin-bound [14C]FA retention in BAT suggests that the TG-lowering effect of the drug is mediated by a tissue-specific increase in LPL-mediated VLDL-TG hydrolysis with subsequent retention of the liberated FA in BAT. At the molecular level, it remains to be clarified whether increases in endothelial LPL expression and/or subtle changes in apolipoproteins and angiopoietin-like proteins regulating local LPL activity (37) are involved in the BAT-specific VLDL-derived TG hydrolysis induced by metformin.

Owing to its high mitochondrial and oxidative enzyme content, BAT has a marked ability to oxidize both glucose and FA, the latter being derived from either LPL-mediated hydrolysis of VLDL-TG or intracellular TG that is stored in lipid droplets. Once released, FAs are rapidly re-esterified in TG or directed to mitochondria for oxidation or activation of UCP1, leading to dissipation of the proton gradient across the inner mitochondrial membrane and heat production (38). At the molecular level, we found that metformin, both in vivo and in vitro, increased AMPK activity and Ser79-ACC phosphorylation, an effect that is expected to promote mitochondrial FA transport and oxidation by relieving the inhibition of CPT1α by malonyl-CoA (9). AMPK activation is known to trigger mitochondrial biogenesis, at least in skeletal muscle (39) and liver (40). Interestingly, the expression of key proteins of the mitochondrial respiratory chain complexes and CS was increased by metformin in BAT, indicating enhanced mitochondrial content in this tissue. Mechanistically, the expression of PGC-1α and eNOS, which are both recognized as important regulators of mitochondrial biogenesis (41,42), was found to be higher in BAT from metformin-treated mice, suggesting activation of the AMPK-PGC1α-eNOS pathway by metformin in this tissue. Finally, we found that metformin affected the qualitative composition of the mitochondrial respiratory chain in BAT, leading to an increase in complex 2 relative to complex 1. This effect might also contribute to enhanced FA oxidation by promoting electron supply to the respiratory chain complex 2. Interestingly, modulating the ratio of FADH2 to NADH oxidation will also affect the stoichiometry of oxidative phosphorylation and promote UCP1-independent metabolic uncoupling, with the yield of ATP synthesis being lowered by ∼40% when FADH2 is oxidized as compared with NADH (43). Taken together, we propose that secondary to its tissue-specific increase in VLDL-TG uptake, metformin promotes FA oxidation in BAT by enhancing both intracellular lipolytic capacity and mitochondrial oxidative machinery. It is important to underline that we do not exclude that another AMPK-independent mechanism(s) in BAT might also contribute to the TG-lowering effect of metformin in vivo.

The recent discovery of metabolically active BAT in adult humans (4446) has caused a revival of interest in this potential new therapeutic target for the treatment of obesity and metabolic disorders (47,48). Although the precise role of BAT in TG metabolism remains to be established, it has been recently shown that FA uptake and oxidation in BAT significantly contributes to energy expenditure in humans (49). Thus, it is tempting to speculate that part of the weight-lowering property of metformin might be secondary to enhanced lipid oxidation and energy dissipation in BAT. Further studies allowing imaging of lipid metabolism in BAT from metformin-treated patients, for instance using 18F-labeled FA incorporated into VLDL-TG coupled to position emission tomography scanning (50), would be crucial to specifically address this point.

In summary, we show that metformin exerts a beneficial effect on circulating lipids by lowering plasma TG, through a selective BAT-mediated increase in VLDL-TG uptake/lipolysis (Fig. 7). The current study is the first identifying BAT as a new important mechanistic player in the lipid-lowering action of metformin, suggesting that targeting this tissue, on top of being interesting for body weight management, might also be of therapeutic importance in the treatment of dyslipidemia.

Article Information

Acknowledgments. The authors are grateful to Elsbet Pieterman, Chris van der Bent, Linda Switzar, and Amanda Pronk (Leiden University Medical Center) for their valuable technical assistance.

Funding. M.R.B. is supported by the Board of Directors of the Leiden University Medical Center. This work was supported by a research grant from the Netherlands Diabetes Foundation (DFN2007.00.010 to P.C.N.R.). P.C.N.R. is an Established Investigator of the Netherlands Heart Foundation (Grant 2009T038).

Duality of Interest. This work was supported by a Société Francophone du Diabète (SFD)–Roche Diagnostics grant from the SFD to B.G. No other potential conflicts of interest relevant to this article were reported.

Author Contributions. J.J.G. and M.R.B. performed experiments, analyzed data, and drafted the manuscript. G.C.v.d.Z., S.A.A.v.d.B., and A.M.v.d.H. performed experiments, analyzed data, and/or critically reviewed the manuscript. M.L. provided the T37i cell line and critically reviewed the manuscript. H.M.G.P. and L.M.H. contributed to discussion and critically reviewed the manuscript. P.C.N.R. performed experiments, supervised the project, contributed to discussion, and critically reviewed the manuscript. B.G. conceptualized the project, performed experiments, analyzed data, and wrote and edited the manuscript. B.G. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.


  • Received February 4, 2013.
  • Accepted November 17, 2013.

Readers may use this article as long as the work is properly cited, the use is educational and not for profit, and the work is not altered. See for details.


| Table of Contents