AMP-activated protein kinase (AMPK) acts as a cellular energy sensor: it responds to an increase in AMP concentration ([AMP]) or the AMP-to-ATP ratio (AMP/ATP). Metformin and phenformin, which are biguanides, have been reported to increase AMPK activity without increasing AMP/ATP. This study tests the hypothesis that these biguanides increase AMPK activity in the heart by increasing cytosolic [AMP]. Groups of isolated rat hearts (n = 5–7 each) were perfused with Krebs-Henseleit buffer with or without 0.2 mM phenformin or 10 mM metformin, and 31P-NMR-measured phosphocreatine, ATP, and intracellular pH were used to calculate cytosolic [AMP]. At various times, hearts were freeze-clamped and assayed for AMPK activity, phosphorylation of Thr172 on AMPK-α, and phosphorylation of Ser79 on acetyl-CoA carboxylase, an AMPK target. In hearts treated with phenformin for 18 min and then perfused for 20 min with Krebs-Henseleit buffer, [AMP] began to increase at 26 min and AMPK activity was elevated at 36 min. In hearts treated with metformin, [AMP] was increased at 50 min and AMPK activity, phosphorylated AMPK, and phosphorylated acetyl-CoA carboxylase were elevated at 61 min. In metformin-treated hearts, HPLC-measured total AMP content and total AMP/ATP did not increase. In summary, phenformin and metformin increase AMPK activity and phosphorylation in the isolated heart. The increase in AMPK activity was always preceded by and correlated with increased cytosolic [AMP]. Total AMP content and total AMP/ATP did not change. Cytosolic [AMP] reported metabolically active AMP, which triggered increased AMPK activity, but measures of total AMP did not.
- 31P-NMR spectroscopy
- total AMP content
amp-activated protein kinase (AMPK) acts as an ultrasensitive cellular energy sensor system (15). Increases in AMP concentration ([AMP]) result in an increase in Thr172 phosphorylation of AMPK-α and, thereby, an increase in AMPK activity. We have demonstrated that AMPK activity responds to cytosolic [AMP] in the isolated rat heart (11, 12) and to the AMP analog ZMP in mouse hearts treated with 5-aminoimidazole-4-carboxamide-riboside (38).
Activated AMPK functions to conserve ATP or promote ATP generation. AMPK stimulates ATP synthesis in the heart by Ser79 phosphorylation of acetyl-CoA carboxylase (ACC), which inhibits ACC activity and, in turn, lowers malonyl-CoA formation. Malonyl-CoA inhibits carnitine palmitoyl transferase 1, the primary transporter and control point for fatty acyl-CoA entry into the mitochondrion. Thus lower malonyl-CoA results in an increase in the transport of fatty acids into the mitochondrion, acceleration of β-oxidation of fatty acids, and increased generation of ATP. Fatty acid oxidation by AMPK increases during reperfusion after ischemia. Increased fatty acid oxidation in this setting can contribute to ischemic injury by inhibiting glucose oxidation (9).
AMPK also promotes ATP generation by stimulating glucose uptake in the heart (31). Increased ATP synthesis resulting from AMPK activity in skeletal muscle also appears to include an increase in glucose uptake (14, 20).
Metformin (N,N-dimethyl biguanide), which is used in the treatment of Type 2 diabetes, decreases hepatic glucose production via gluconeogenesis and increases peripheral glucose utilization (1). In insulin-sensitive and insulin-resistant adult cardiomyocytes, the relationship between AMPK activation and glucose uptake stimulation by biguanide has been reported (4). In the absence of insulin, biguanides restored stimulation of glucose uptake in insulin-resistant cardiomyocytes via an additive activation of AMPK and protein kinase B. Treatment of cardiomyocytes with metformin for 18 h increased glucose transport by increasing the residence time of GLUT4 in the plasma membrane due to an AMPK-dependent reduction in endocytosis (37). Metformin also decreases fatty acid oxidation by 10–20% (30). Use of phenformin (N-phenylethyl biguanide), a more lipophilic biguanide, by humans has been discontinued because of frequent occurrence of lactic acidosis (1).
Zhou et al. (39) reported that AMPK is a target for metformin in hepatocytes and skeletal muscle. Subsequent studies have shown that metformin activates AMPK in cardiomyocytes (7), skeletal muscle (28), and heart and aorta (40). Other studies have shown that phenformin activates AMPK in pancreatic islets (25) and transfected HeLa cells (16).
The mechanism by which the biguanides increase AMPK activity remains uncertain. Fryer et al. (13) showed in muscle cells that metformin activated AMPK in the absence of any increase in the AMP-to-ATP ratio (AMP/ATP) (13). Hawley et al. (17) reported that metformin increased AMPK activity in the absence of an increased AMP/ATP in Chinese hamster ovary fibroblasts and rat hepatoma cells.
This study tested the hypothesis that phenformin and metformin activate AMPK by increasing cytosolic [AMP] in the perfused heart. Treatment of rats with perfusate containing phenformin or metformin increased AMPK activity and Thr172 phosphorylation of AMPK-α. This biguanide-induced increase in AMPK activity was preceded by and correlated with an increase in cytosolic [AMP]. The total AMP content of metformin-treated hearts did not change.
MATERIALS AND METHODS
Preparation of Isolated Perfused Rat Hearts
Hearts of male Sprague-Dawley rats (280–320 g) were isolated and perfused in the isovolumic Langendorff model. Krebs-Henseleit buffer (KH; in mM: 118 NaCl, 5.9 KCl, 1.2 MgSO4, 25 NaHCO3, 1.75 CaCl2, 0.5 Na-EDTA, 10 d-glucose, and 0.5 pyruvate) was equilibrated with 95% O2-5% CO2 to achieve pH 7.4. Phenformin (Sigma-Aldrich) and metformin (Sigma-Aldrich) were added to KH.
The experimental protocols were approved by the Harvard Medical Area Standing Committee on Animals and followed the recommendations of the National Institutes of Health and the American Physiological Society's guiding principles in the care and use of animals.
Study Design and Experimental Protocols
All rat hearts were paced at 300 beats/min. In the phenformin experiments, all hearts were perfused with KH during a 12-min baseline period. Seven groups were created by alteration of the perfusion conditions as follows for the remainder of the protocol: 1) perfusion with KH for 38 min (KH, n = 6); 2) perfusion with KH + 0.2 mM phenformin for 18 min (Phen, n = 5); 3) perfusion with KH + 0.2 mM phenformin for 18 min followed by KH for 20 min (Phen-KH, n = 6); 4) perfusion with KH + 0.04 mM phenformin for 18 min followed by KH for 20 min (Phen 0.04-KH, n = 5); 5) perfusion with KH + 0.2 mM phenformin for 38 min (Phen 38, n = 3); 6) perfusion with KH + 0.2 mM phenformin for 38 min followed by KH for 20 min (Phen 38-KH, n = 1); and 7) perfusion with KH + 25 mM KCl + 0.2 mM phenformin for 60 min (Phen-KCl 60, n = 4). In the metformin experiments, hearts were perfused with KH during a 12-min baseline period. Four groups of hearts were created by alteration of the perfusion conditions as follows for the remainder of the protocol: 1) perfusion with KH for 61 min (KH, n = 5); 2) perfusion with KH + 10 mM metformin for 53 min (Met 53, n = 5); 3) perfusion with KH + 10 mM metformin for 61 min (Met 61, n = 5); and 4) perfusion with KH + 10 mM metformin for 83 min (Met 83, n = 3). Two additional groups of hearts were perfused for measurement of total adenine nucleotide by HPLC, as well as cytosolic [AMP] and AMPK activity: 1) perfusion with KH (n = 8) for 26 min and 2) perfusion with KH + 10 mM metformin for 70 min (n = 12). At the end of the protocol, hearts were freeze-clamped with Wollenberger tongs cooled in liquid nitrogen and stored at −80°C until further measurements. Freeze-clamping occurred 1 min after removal from the magnet, which was 3 min from the midpoint of the final 31P-NMR spectrum.
31P-NMR spectroscopy of isolated perfused hearts.
31P-NMR free induction decays (FIDs) were acquired at 161.8 MHz using a spectrometer (Inova, Varian, Palo Alto, CA). For heart 31P spectra, 96 FIDs were averaged over 4 min using 60° pulses and a recycle time of 2.5 s. The resonance areas of phosphocreatine (PCr) and ATP and the chemical shifts of Pi were quantified using Bayesian analysis software (G. L. Bretthorst, Washington University, St. Louis, MO), which uses a direct statistical analysis of the time-domain data (FID) based on Bayesian probability theory (6). Saturation factors for resonances were determined from fully relaxed spectra; recycle time was 15 s.
Heart 31P metabolite content.
ATP was used as an internal standard to derive the concentrations of the 31P-NMR-visible metabolites of hearts: the average of the area of the β-resonance of ATP during the baseline period for each heart was set to 10.8 mM (2), and the areas of other resonances, corrected for variable saturation, were calculated on this basis. Intracellular pH (pHi) was calculated from the chemical shift of intracellular Pi (3). Metabolite measurements are an average of the 4 min of the 31P-NMR data acquisition, with the midpoint of the spectrum reported as the measurement time.
Calculation of cytosolic AMP.
31P-NMR measurements of PCr concentration ([PCr]), ATP concentration ([ATP]), and H+ concentration ([H+]) were entered into the creatine (Cr) kinase reaction (PCr + ADP + H+ ↔ Cr + ATP) equilibrium expression to calculate cytosolic ADP concentration ([ADP]) using an equilibrium constant (Keq) of 1.66 × 109 M−1 (23) (1) Cr concentration ([Cr]) is calculated as the difference between PCr content and total Cr content (3). The 31P-NMR-measured pHi was used to estimate [H+]. The near equilibrium of the adenylate kinase reaction (2ADP ↔ AMP + ATP) converts increased cytosolic [ADP] to an increase in cytosolic [AMP]. Thus cytosolic [AMP] is set by the adenylate kinase equilibrium expression using a Keq of 1.05 (23) (2)
Methods described by Lazzarino et al. (24) were used to measure adenine nucleotide (ATP, ADP, and AMP) contents (nmol/mg protein) in the extracts of freeze-clamped hearts. Protein was determined using the method of Lowry et al. (26).
Measurement of AMP-activated protein kinase activity.
Total AMPK activity was measured using the method of Dagher et al. (8) modified by Frederich et al. (12). AMPK activity was quantified in the resuspended pellet as incorporation of 32P from [γ-32P]ATP (10 GBq/mmol; NEN, Boston, MA) into a synthetic peptide with the specific target sequence for AMPK, the SAMS peptide (HMRSAMSGLHLVKRR, American Peptides, Sunnyvale, CA). Radioactivity was measured using a liquid scintillation counter (Tri-Carb 2100TR, Packard Biosciences, Meriden, CT). Protein content in the solution containing the resupended (NH4)2SO4 pellet was determined using the Bradford method (5).
Measurement of phosphorylation of AMPK-α at Thr172 and ACC at Ser79.
Phosphorylation was measured using Western immunoblots as described by Frederich et al. (12). Membranes were incubated with an anti-phosphorylated AMPK-α polyclonal antibody (Thr172; Cell Signaling Technology, Beverly, MA) or an anti-phosphorylated ACC polyclonal antibody (Ser79; Upstate Biotechnology, Lake Placid, NY) overnight at 4°C and then with horseradish peroxidase-conjugated secondary antibody (Southern Biotech, Birmingham, AL) for 2 h at room temperature. Signal detection was facilitated with LumiGlo Reagent and peroxide (Cell Signaling Technology). The membrane immunosignals were quantitated using an image-scanning densitometer (model GS-700 Imaging Densitometer and Quantity One version 4.2.1, Bio-Rad).
Values are means (SD). Statistical computations were performed with Statistica (version 6.1, StatSoft, Tulsa, OK). ANOVA was used to compare measurements among all groups. For the 31P-NMR measurements of PCr, ATP, and AMP, a repeated-measures ANOVA was used for within-group analysis. Fisher's post hoc protected least significant difference was used for comparison of the means. Differences were declared statistically significant if P < 0.05. Prism for Windows (version 4.0, GraphPad Software, San Diego, CA) was used for graphs.
Study Design and Experimental Protocols
The experimental protocols of this study were designed to test the hypothesis that phenformin and metformin activate AMPK by increasing cytosolic [AMP] in the perfused heart. Consequently, the main protocols that treat hearts with phenformin (see ⇓Fig. 2) and metformin (see Fig. 3) were designed to identify the threshold increases in AMPK activity during measurement of cytosolic [AMP].
Hemodynamics of Rat Hearts Perfused With Phenformin and Metformin
At 6 min after perfusion was switched to KH containing phenformin, the LV systolic pressure (SP) and coronary flow (CF) began to decrease (Fig. 1, A and B). End-diastolic pressure (EDP) of the phenformin-treated hearts decreased transiently near the end of the treatment period. By the end of phenformin treatment, SP had declined to 45% of baseline, CF to ∼80% of baseline, and EDP to ∼69% of baseline. The effect of phenformin on myocardial function (Fig. 1A) preceded changes in PCr or AMP (Fig. 2, A and C).
To examine the concentration dependence of phenformin's effects on the heart, another group of hearts was treated with 0.04 mM phenformin for 18 min and perfused with KH for an additional 20 min. [AMP], AMPK activity, phosphorylated AMPK, and phosphorylated ACC were unchanged (data not shown), but SP declined to 66% of baseline, EDP to 70% of baseline, and CF to 87% of baseline (data not shown). Thus a fivefold reduction of phenformin concentration alters heart function but does not increase [AMP].
Hearts were also perfused with KH containing 25 mM KCl (KH-KCl), which arrests contraction and effectively silences the functional effect of phenformin and also reduces ATP demand. These hearts were treated with 0.2 mM phenformin (Phen-KH-KCl, n = 5) and compared with hearts perfused with KH-KCl (n = 5). Beginning at 42 min, cytosolic [AMP] increased in Phen-KH-KCl hearts compared with their own baseline or KH-KCl hearts; no difference was observed in the CF of the Phen-KH-KCl hearts (not shown).
The effect of phenformin on myocardial function is quite rapid and precedes its effect on metabolism. The effect on myocardial function also occurs at a lower phenformin concentration than the metabolic effect. Finally, the increase in [AMP] and AMPK activity in the phenformin-treated K+-arrested heart occurred without changes in CF. Taken together, these results suggest that the effects of phenformin on function and metabolism are independent.
The functional response of hearts perfused with KH + 10 mM metformin was quite different from that of phenformin-treated hearts. SP, EDP, and CF were unchanged during the first 46 min of the treatment period (Fig. 1, C and D). At 50 min, SP and CF increased, while EDP was unchanged. These changes in LVSP and CF occur concomitantly with the increase in cytosolic [AMP] (Fig. 3C). These data demonstrate that SP was not compromised in metformin-treated hearts, even with substantial decreases in [PCr] and increases in cytosolic [AMP]. The increase in CF may result from the elevated [AMP], which will result in increased adenosine production (19) and vasodilation (22).
31P-NMR-measured metabolite content of the isolated rat heart: phenformin and metformin decreased [PCr] and increased [AMP].
31P-NMR was used to measure PCr and ATP contents of the isolated heart continuously during the experimental protocols. In hearts treated for 18 min with phenformin, [PCr] decreased at 22 min relative to baseline and was different from [PCr] of KH-perfused hearts at 26 min (Fig. 2A). In phenformin-treated hearts, [ATP] decreased after 6 min of treatment relative to baseline but was never different from that of the KH-perfused group (Fig. 2B). In phenformin-treated hearts, cytosolic [AMP] increased at 26 min (Fig. 2C) relative to baseline and KH-perfused hearts.
In metformin-perfused hearts, [PCr] decreased relative to baseline at 10 min and was different from that in KH-perfused hearts at 30 min (Fig. 3A). In metformin-perfused hearts, myocardial [ATP] exhibited a sustained decrease at 30 min relative to baseline but was only different from that in KH-perfused hearts at several time points (Fig. 3B). Cytosolic [AMP] in metformin-perfused hearts increased at 42 min relative to baseline and at 50 min relative to KH-perfused hearts (Fig. 3C).
AMPK activity and AMPK and ACC phosphorylation in isolated rat heart: effects of phenformin and metformin.
AMPK activity and phosphorylation were measured in hearts that were freeze-clamped just after the end of phenformin treatment at 21 min (Phen) and at the end of the posttreatment period at 37 min (arrows in Fig. 2C). After the treatment period, [AMP], AMPK activity, phosphorylated AMPK, and phosphorylated ACC of the KH- and phenformin-perfused hearts were equal (Fig. 4). In hearts perfused for another 20 min with KH after 18 min of phenformin treatment, [AMP], AMPK activity, phosphorylated AMPK, and phosphorylated ACC were increased (Fig. 4).
In metformin-treated hearts, AMPK activity and phosphorylation were measured at 53 and 61 min (arrows in Fig. 3C). In hearts perfused for 53 min with metformin, [AMP] (Fig. 5A, 50 min), AMPK activity (Fig. 5B), phosphorylated AMPK (Fig. 5C), and phosphorylated ACC (Fig. 5D) were equal to those in KH-perfused hearts. In hearts perfused for another 8 min with metformin (61 min of perfusion) [AMP], AMPK activity, phosphorylated AMPK, and phosphorylated ACC were increased. In metformin-treated hearts, cytosolic [AMP] was significantly different than that in KH-perfused hearts at 50 min. Measurement of [AMP] at 50 min includes all 10 hearts. [AMP] measured at 53 min (Fig. 5A) includes the five hearts clamped at 53 min.
The relationship between AMPK activity and cytosolic [AMP] is shown in Fig. 6 for hearts treated with phenformin and metformin. Data from all KH-perfused hearts and hearts treated with phenformin, phenformin-treated and KH-perfused hearts, and hearts treated with metformin for 53, 61, and 70 min (Figs. 2–5 and 7) are included. Measurements from hearts treated with phenformin and metformin under conditions that surpass the threshold increase in AMPK activity are included as well.
The cause of increased AMPK activity resulting from phenformin and metformin exposure has not been defined. There are two known pathways for AMPK regulation in mammalian cells. The first and best defined pathway involves increased AMP binding to the cystathione β-synthase domains of AMPK-γ. This increases Thr172 phosphorylation of AMPK-α by AMPK kinase, which is required for increased AMPK activity. In the heart, LKB1 is the AMPK kinase that phosphorylates AMPK-α2, but not AMPK-α1, heterotrimers during ischemia and hypoxia, both of which increase AMP (32). We have demonstrated in the isolated rat heart that AMPK activity responds to cytosolic [AMP] (11), but in a cytosolic [AMP]-independent manner to hypoxia (12). The pathway for regulation by LKB1 after an increase in [AMP] corresponds to the cellular energy sensor function of AMPK. The second pathway for AMPK regulation is directed by Ca2+/calmodulin-dependent protein kinase kinase (CaMKK) activity (18, 21, 35). CaMKK also phosphorylates AMPK-α at Thr172. AMP binding to AMPK is not required. CaMKK activity increases in response to increased Ca2+. This pathway for AMPK regulation by CaMKK, which is dependent on intracellular Ca2+, corresponds to a metabolic regulation function for AMPK at the organism level (34).
A third AMPK kinase has been postulated and may indicate a third pathway for AMPK regulation. TGF-β-activated kinase-1 (TAK1) has been reported to function as an AMPK kinase in yeast and HeLa cells (27). In cultured fibroblasts, TAK1 was activated by a number of stimuli, including metformin, that are known to increase AMPK activity (36). Loss of TAK1 prevented the increase in Thr172 phosphorylation of AMPK-α induced by metformin and interfered with activation of LKB1. These results suggest that TAK1 increases LKB1 activity and does not directly phosphorylate AMPK. It appears that TAK1 modulates the AMPK energy sensor pathway.
All studies that have attempted to define the regulation of AMPK activity by phenformin and metformin have measured total AMP content or the ratio of total AMP content to total ATP content (total AMP/ATP). In muscle cells (13) and Chinese hamster ovary fibroblasts and rat hepatoma cells (17), metformin activated AMPK in the absence of any increase in AMP/ATP. There are a few exceptions to these reports. In human skeletal muscle, ATP and PCr content decreased with chronic metformin administration (28). Decreases in PCr typically result in increased AMP. High concentrations of phenformin (1–5 mM) increased AMP/ATP in rat brain slices (18).
In the heart, total AMP content is much greater than cytosolic AMP content. For example, conversion of HPLC-measured total AMP content of the well-perfused rat heart to [AMP] would result in ∼350 μM AMP; yet 31P-NMR-estimated cytosolic [AMP] is ∼0.4 μM. Cytosolic [AMP] is estimated from measured [PCr], [ATP], and pHi: the creatine kinase equilibrium expression (Eq. 1) is used to calculate cytosolic [ADP], and cytosolic [ADP] and measured [ATP] are used to calculate cytosolic [AMP] using the adenylate kinase equilibrium expression (Eq. 2). These calculations are valid, because the creatine kinase and adenylate kinase reactions are in states of near equilibrium in vivo and both enzymes are abundant. Cytosolic [ADP] and [AMP] represent the metabolically active fractions of the total content of these two nucleotides. Hence, AMPK detects cytosolic [AMP], not total AMP or total AMP/ATP.
To test this prediction, cytosolic [AMP] and total AMP were measured (Fig. 7) in metformin-treated and KH-perfused hearts. Cytosolic [AMP] and AMPK activity were increased in the metformin-treated hearts. Total AMP and total AMP/ATP were not different between the metformin-treated and KH-perfused hearts. Thus total [AMP] and total AMP/ATP did not correlate with AMPK activity, whereas cytosolic [AMP] correlated with increased AMPK activity.
In the heart, acute decreases in [PCr] typically indicate a reduction in ATP synthesis (imbalance of synthesis and use). The decreased [PCr] and increased cytosolic [AMP] are consistent with biguanides acting as metabolic inhibitors. Owen et al. (29) reported that metformin inhibited complex 1 of the respiratory chain in liver mitochondria. El-Mir et al. (10) also reported that metformin treatment inhibited complex 1 but that the inhibition was the result of an indirect action by metformin. Myocardial O2 consumption was similar in isolated rat hearts treated with metformin and control hearts (not shown). Thus we were unable to confirm an inhibition of oxidative phosphorylation by metformin in the isolated heart.
The concentrations of phenformin (0.2 mM) and metformin (10 mM) used in the present study were selected to be compatible with those used in many in vitro studies to increase AMPK activity. Oral metformin is widely used to treat Type 2 diabetes. In vivo metformin concentrations reported in human plasma are ∼10 μM (1). Thus the results of this study may not necessarily extrapolate to metformin's effect on [AMP] in vivo. Shaw et al. (33) reported, however, that AMPK phosphorylation was increased in livers from wild-type mice injected with metformin, but not in livers deficient in LKB1. Liver LKB1 was also necessary for the blood glucose-lowering effect of metformin. Since LKB1 is required for Thr172 phosphorylation of AMPK-α resulting from increased [AMP] (see above), the results of Shaw et al. are consistent with the concept that metformin increases AMPK activity in vivo by increasing [AMP].
This study demonstrates in the perfused heart that the increase in AMPK activity caused by phenformin and metformin results from an increase in cytosolic [AMP]. Metformin did not alter total AMP content or total AMP/ATP. The failure to detect biguanide-induced changes in total AMP content or total AMP/ATP results from the insensitivity of these measures, which do not quantify the metabolically active cytosolic [AMP] that is detected by AMPK.
This work was supported by National Heart, Lung, and Blood Institute Grant HL-46033 (to J. A. Balschi).
The costs of publication of this article were defrayed in part by the payment of page charges. The article must therefore be hereby marked “advertisement” in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
- Copyright © 2007 by the American Physiological Society