AMP-Activated Protein Kinase α2 Deficiency Affects Cardiac Cardiolipin Homeostasis and Mitochondrial Function

The generation of AMPKα2^−/− mice has been described elsewhere (15). Ten-month-old male AMPKα2^−/− (n = 27) and control (n = 25) littermate mice were used. Animals were housed under temperature-controlled conditions (21°C) and had free access to water and to a standard mouse chow. All procedures were performed in accordance with the principles and guidelines established by the European Convention for the Protection of Laboratory Animals. Mice were killed by lethal intraperitoneal injection of pentothal (150 mg/kg) for mitochondrial respiration experiments and tissue storage. Left ventricular tissue was isolated, part of which was immediately used for mitochondrial function measurement and part of which was rapidly frozen and kept at −80°C.

Additional mice were anesthetized with urethane (2 g/kg). Hearts were quickly removed and retrogradely perfused at a constant flow of 2.5 ml/min in the isovolumic Langendorff perfused mode without pacing. They were first equilibrated with 11 mmol/l glucose for 30 min and then with 5 mmol/l glucose and 0.4 mmol/l oleate prebound to 1% BSA as substrates for 20 more min. Left ventricular pressure was monitored from a water-filled balloon introduced in the left ventricle. Oxygen consumption (QO₂) was calculated from the difference in oxygen content between incoming (aortic) and outcoming (pulmonary artery) perfusates (17).

Left ventricular wall and papillary muscles were quickly isolated in oxygenated buffered Krebs solution without calcium to avoid ischemia and fixed with 2% glutaraldehyde as previously described (18). For stereological analysis, papillary muscles were used to ensure longitudinal sections, and three randomly selected levels separated by >50 μm were used. From randomly selected cardiomyocytes (12–14 myocytes from each animal), the volume density of organelles was estimated by the point counting method (18). The volume density V_v of the organelles was estimated as V_v = p/P (p is the number of the test points hitting the image of the cellular components, P is the number of all points falling on the cardiomyocytes).

Oxygen consumption measurements of saponin-skinned fibers from left ventricle have been described previously (19,20). Rates of respiration are given in micromoles O₂ per minute per gram dry weight.

Two different experimental protocols were used based on substrate utilization pathways (Fig. 1). The first protocol determined the sensitivity of mitochondrial respiration to various substrates in the presence of 2 mmol/l ADP, by cumulative substrate addition as described previously (21). The second protocol was aimed at determining the dependency of respiration on external [ADP] and [creatine] (22), with glutamate plus malate as substrates. Respiration through complex III was designed according to Ray et al. (23).

Frozen tissue samples were weighed and homogenized in ice-cold buffer, and enzyme activities were determined as described previously (24). Complex I activity was measured in heart homogenized in ice-cold buffer containing 10 mmol/l Tris base (pH 7.2), 75 mmol/l sucrose, 225 mmol/l mannitol, 100 μmol/l EDTA, and 0.1% Triton-X 100. To measure NADH–coenzyme Q reductase activity with decyl-ubiquinone as electron acceptor, samples were incubated in 25 mmol/l phosphate buffer (pH 7.5), 2.5 mg/ml BSA, and 100 μmol/l decyl-ubiquinone at 30°C. Activity was reported as rotenone-insensitive decrease in NADH absorbance at 340 nm. NADH-ferricyanide reductase activity was measured with ferricyanide as electron acceptor, and the decrease in absorbance was followed at 410 nm.

Total muscle RNA was extracted using standard procedures. Real-time RT-PCR was performed using the SYBR Green method on a LightCycler rapid thermal cycler (Roche Diagnostics) as previously described (25). Values of each gene were normalized to cyclophilin A mRNA content and then corrected for the amount of RNA relative to muscle weight. Primers are listed in Table 1.

Specific antibodies were used to measure the protein content of the oxoglutarate/malate carrier (OMC) (26) (gift from Thomas Scholz and Stacia Koppenhafer, University of Iowa), AMPKα1, and AMPKα2 (Upstate Biotechnology, Lake Placid, NY) in control and AMPKα2^−/− mice.

Mitochondria were isolated from cardiac ventricles of control and AMPKα2^−/− mice according to Moreno-Sanchez et al. (27). Cardiolipin was quantified by the spectrophotometric method of Petit et al. (28), using the high affinity of 10N-nonyl acridine orange (NAO) for cardiolipin of freshly isolated mitochondria.

Lipids were extracted from heart, isolated cardiac mitochondria, liver, and plasma in 2:1 chloroform:methanol. Phospholipids were separated from nonphosphorous lipids on silica acid cartridges, and fatty acids were trans-methylated with BF₃-methanol. Methyl esters were analyzed by gas chromatography on an Econo-cap EC-WAX capillary column (0.32 × 30 m; Alltech Associates) coupled to a flame ionization detector using C17:0 as the internal standard as previously described (29).

Data are expressed as means ± SE. Student’s t test was used to determine the statistical significance of differences between group means. Statistical significance was defined as P < 0.05.

AMPKα2^−/− mice had normal body (32 ± 1 in control vs. 37 ± 3 g) and heart (187 ± 8 in control vs. 197 ± 16 mg) weights, indicating that cardiac atrophy or hypertrophy were not present. Neither left ventricular pressure or heart rate nor oxygen consumption of isolated hearts differed between control and AMPKα2^−/− mice whether glucose or glucose plus oleate were used as substrates, suggesting that AMKPα2 deficiency does not alter cardiac work during normal perfusion (Table 2).

AMPKα1 protein content (as measured by Western blotting; not shown) was the same in control and AMPKα2^−/− mice (1.24 ± 0.12 in control vs. 1.49 ± 0.21 arbitrary units [AU]). Thus the deficiency in AMPKα2 was not compensated for by AMPKα1 overexpression.

In control cardiomyocytes, mitochondria formed longitudinal rows under sarcolemma and between myofibrils (Fig. 2A and B). In contrast, in AMPKα2^−/− cardiomyocytes from both papillary and ventricular muscles, splitting of myofibrils was frequently observed (Fig. 2C–F). No change in total mitochondrial volume was observed (Table 3), but mitochondria lost their ellipsoid shape and became irregular in size (Fig. 2C–F). They lost their arrangement in longitudinal rows and formed large and spread clusters especially under the sarcolemma (Fig. 2D and F), so that the volume density of subsarcolemmal mitochondria was increased while that of intermyofibrillar mitochondria was decreased (Table 3). No change in the volume of lipid droplets was observed (0.97 ± 0.22 in control vs. 0.79 ± 0.61%).

Respiration rates for almost all substrates were significantly lower in cardiac fibers of AMPKα2^−/− mice (Fig. 3A). Compared with controls, respiration was decreased by 35% with malate, 34% after addition of octanoyl-carnitine, 38% after addition of pyruvate, and 31% after final addition of glutamate in AMPKα2^−/− mice, but it did not change with glycerol-3-phosphate (G3P).

Respiration in the presence of phosphate acceptors ADP and creatine was studied with glutamate plus malate (V_glu+mal) as substrates (Table 3). Basal respiration rate (V₀) and acceptor control ratio (ACR = V_glu+mal/V₀) were similar in control and AMPKα2^−/− cardiac fibers. No difference was observed between the two groups for the affinity for ADP with or without creatine. The maximal oxygen consumption rate was again 24% lower with glutamate plus malate in the AMPKα2^−/− mice (Table 3; Fig. 3). To determine at which point the respiratory chain was altered in AMPKα2^−/− mice, complex I was inhibited by amobarbital, and respiration through complex II was measured with succinate (V_suc). Under succinate, no significant difference between the two groups was observed, whereas the ratio between complex I–and complex II–stimulated respiration (V_suc/V_glu+mal) was increased by 42% in AMPKα2^−/− mice (Table 3; Fig. 3B). No difference was observed for complex III–activated respiration with duroquinol (V_duroq). Thus, AMPKα2 deficiency induces a defect in mitochondrial respiration that appears limited to complex I.

To understand the origin of mitochondrial defects, activities of key enzymes of energy metabolism were assessed (Table 3). Activity of citrate synthase, an index of mitochondrial mass, was similar in the two groups of mice. Other enzymes of the energy metabolism, total creatine kinase (CK), mitochondrial CK (mi-CK), adenylate kinase, total lactate dehydrogenase, and cytochrome c oxidase (COX) activities also remained unchanged in AMPKα2^−/−. Despite decreased respiration through complex I, maximal complex I and complex II activities determined in tissue extracts were the same in both groups. To ensure that maximal complex I activity was preserved in AMPKα2^−/−, we measured its activity using different electron acceptors, but no difference between the two groups was observed, suggesting that AMPK deficiency did not affect mitochondrial enzyme content.

In an attempt to understand the origin of the decreased respiration through complex I, malate-aspartate shuttle enzymes were measured (Table 3). Total malate dehydrogenase (MDH), mitochondrial MDH, and the amount of OMC (6.8 ± 1.2 [n = 6] in control vs. 5.1 ± 0.5 AU in AMPKα2^−/− [n = 6]; not shown) were unchanged in AMPKα2-deficient mice. Moreover, the mRNA levels of PGC-1α, the main regulator of mitochondrial biogenesis, were similar in AMPKα2^−/− and control hearts (Fig. 4B), consistent with the preserved mitochondrial enzyme activities.

Complex I activity in situ is critically dependent on the cardiolipin environment (30,31). In AMPKα2^−/− mice, the amount of cardiolipin was 25% lower than in controls (Fig. 4A). In addition, as evidenced in Table 4, the fatty acid composition of whole-heart phospholipids and of isolated cardiac mitochondria in the AMPKα2^−/− mice showed a significant decrease in linoleic acid proportion (∼80% of cardiolipin fatty acids). Conversely, linoleic acid was significantly increased in the nonphosphorous lipid fraction (which mainly contains storage triacylglycerol) of the whole-heart homogenate. This decrease in linoleic acid content was specific to the heart because it decreased neither in liver phospholipids (19.4 ± 0.8 and 19.6 ± 1.1%, control vs. AMPKα2^−/−) nor in plasma lipids (31.3 ± 0.4 and 27.8 ± 2.2%, respectively).

The level of expression of enzymes involved in cardiolipin homeostasis was determined (Fig. 4B). Levels of CTP:PA cytidylyltransferase (CDS2), the enzyme that catalyzes the initial key step of cardiolipin synthesis, i.e., the conversion of phosphatidic acid to CDP-diacylglycerol, and of acyl-CoA:lysocardiolipin acyltransferase 1 (ALCAT1), which is implicated in rapid cardiolipin remodeling (32,33), were significantly decreased in AMPKα2^−/− mice, whereas mRNA content of taffazin, another phospholipid acyltransferase, was unchanged. Interestingly, expression of CPT-1, the mitochondrial middle- and long-chain fatty acid transporter, was also significantly decreased in AMPK-deficient mice.

The main results of this study can be summarized as follows. 1) Specific deficiency in AMPKα2 catalytic subunit did not induce cardiac atrophy or hypertrophy. 2) Oxygen consumption of isolated perfused heart was normal, but mitochondrial ultrastructure was altered. 3) AMPK deficiency induced a decrease in maximal oxidative capacity of the cardiac muscle whether lipids, pyruvate, or glutamate plus malate were used as substrates. 4) This was not accompanied by changes in mitochondrial enzyme activities, the malate/aspartate shuttle, or isolated complex I or II in vitro activities. 5) When succinate was used as substrate, mitochondrial respiration was normal in AMPKα2^−/− cardiac fibers, suggesting a defect in complex I function in situ. 6) Cardiolipin content of AMPKα2^−/− cardiac mitochondria was decreased by 25%, consistent with the altered ultrastructure and the functional defect of mitochondrial respiration by complex I. 7) This defect was accompanied by a significant decrease in linoleic acid content of mitochondrial phospholipids and 8) could be explained by the decrease in the expression of key enzymes of cardiolipin biosynthesis and remodeling. Altogether, these results suggest that AMPKα2 is involved in the control of mitochondrial respiration through cardiolipin homeostasis.

AMPKα2^−/− mice exhibit perturbation in whole-body insulin sensitivity, probably modulated by the sympathetic nervous system (15). These mice show normal cardiac content in AMPKα1 isoform, suggesting that remnant AMPKα1 is unable to compensate for the lack of AMPKα2 (16).

A role for AMPK in controlling cardiac weight is still controversial. Although AMPK was suggested to play a role in pressure overload hypertrophy (9), it was recently reported that adiponectin blocks cardiac hypertrophy by an AMPK-dependent mechanism (11). Hearts of mice expressing a dominant-negative mutant of AMPKα2 exhibiting a residual AMPKα2 activity have preserved weight and baseline function (13). In another model overexpressing a kinase-dead rat α2 isoform, the α2 protein content and activity were absent, and α1 content and activity were also decreased, with the remnant α1 activity being attributed to endothelial cells (12). These mice exhibit a slightly decreased cardiac weight and contractility. However, we show here that AMPKα2^−/− mice had normal cardiac weight. This suggests that the α2 subunit by itself is not essential for cardiac growth.

Oxygen consumption and contractile function were not different under basal conditions either with glucose or glucose plus oleate as substrates, as already described with glucose and pyruvate (16) in this mouse line. In KD mutant mice, cardiac function is slightly depressed at baseline (12). This difference could be due to the decreased AMPKα1 activity in this line. However, the workload of hearts perfused in the Langendorff mode is significantly less than in the working mode or in vivo, and oxygen consumption does not reach maximal capacity.

To assess maximal mitochondrial function, respiration rates were measured in skinned fibers with saturating amounts of substrates, oxygen, and phosphate acceptor. Depending on available substrates and cardiac demand, mitochondria are able to use acetyl-CoA produced from pyruvate by pyruvate dehydrogenase downward of glycolysis and/or acetyl-CoA produced by the β-oxidation of fatty acids, which enters the Krebs cycle (Fig. 1). Mitochondria can also slightly use G3P by mitochondrial glycerophosphate dehydrogenase, which produces reduced flavine adenine dinucleotide (FADH₂) that enters the respiratory chain directly at the level of complex II (34). We found a clear decrease in the respiration rate of AMPKα2^−/− cardiac fibers, whether carbohydrate- or lipid-derived substrates were used. This suggests that AMPKα2 is involved in the control of cardiac oxidative capacity and in mitochondrial substrate utilization. Decreased energy availability in AMPKα2^−/− mice would probably occur only at maximal workloads, so that the mitochondrial alterations observed here might be more evident during heavy exercise or under pathologic stress.

To understand the decrease in respiration of cardiac fibers, we looked at indicators of mitochondrial activity. In AMPKα2^−/− mouse heart, levels of citrate synthase, a marker of mitochondrial mass; of COX, an enzyme of the respiratory chain; of mi-CK, a phosphotransfer kinase; and of markers of the malate/aspartate shuttle, which are involved in glutamate plus malate utilization, were normal. Additionally, PGC-1α, which induces mitochondrial biogenesis (35), and total mitochondrial volume density were unchanged as was also observed in skeletal muscles of mice with dominant-negative mutant of AMPK (36). All of this suggests that AMPKα2 is not essential for determining and maintaining mitochondrial mass in skeletal and cardiac muscle. However, it may play a role in cardiac mitochondrial biogenesis induced by external stimulus, as was observed in skeletal muscle in response to exercise or chronic energy depletion (36).

Mitochondrial function also depends on the sensitivity to ADP and to creatine, the final acceptor of high-energy phosphates in cardiac mitochondria (37). No change in the sensitivity of respiration to ADP and creatine was observed, showing that the efficiency of cardiac mitochondria in phosphorylating ADP and producing phosphocreatine was not affected by AMPKα2 loss.

To explain the decreased respiration with preserved enzyme activities, we compared the respiration rates during activation from complex I, II, or III (Figs. 1 and 3; Table 2). Respiration with glutamate plus malate mainly produces NADH, which activates respiration through complex I, whereas succinate oxidation mainly produces FADH₂, which is further oxidized by complex II. Compared with controls, respiration was lower in AMPKα2^−/− mice with glutamate plus malate but was normal with succinate, revealing an inhibition of complex I in these animals. Interestingly, respiration from G3P, which produces only FADH₂, was normal in AMPKα2^−/− mice, whereas it was decreased for NADH-producing substrates: pyruvate and octanoyl-carnitine. Moreover, respiration through complex III was unchanged in AMPKα2^−/− mice.

Despite a 24% decrease in respiration rate through complex I, complex I activity measured in total heart extracts was normal in AMPKα2^−/− mice, independent of the electron acceptor. This indicated that the activity of isolated complex I was not affected by AMPKα2 deficiency, suggesting an alteration of the in situ regulation of complex I by cardiolipin. Activity of the complexes of the respiratory chain, and particularly of complex I, strongly depends on the surrounding phospholipid environment of the inner mitochondrial membrane. Cardiolipin is the main functional phospholipid of the mitochondrial inner membrane, being present almost exclusively in mitochondria and representing 8–15% of the entire cardiac phospholipid mass. A 25% lower mitochondrial cardiolipin content was observed in AMPKα2^−/− mice. Cardiolipin content is critical for the adaptation of energy metabolism to demand. It rises with increased metabolic rate or muscle performance (38) and plays a key role in the activity of several inner membrane proteins including complex I (rev. in 39). Both complex I and III were shown to require cardiolipin, but although necessary for full complex I activity (40), cardiolipin plays a structural rather than catalytic role for complex III (41). Accordingly, respiration through complex I but not complex III was decreased in AMPKα2^−/− mice. Interestingly, in CHO cells, similar decrease in cardiolipin content induces similar alteration in complex I activity in the respiratory chain without changes in isolated, complex II, complex III and IV, and NADH reductase activity (31). mi-CK binding in the vicinity of translocase is dependent on cardiolipin environment (42). It is possible that the rather small decrease in cardiolipin observed here was not sufficient to affect mi-CK function and activity.

Examination of cardiac ultrastructure of AMPKα2^−/− mice revealed abnormal structure of mitochondria of various sizes arranged in clusters under the sarcolemma and less in regular rows between myofilaments as in control hearts. Although unchanged total mitochondrial volume was in accordance with preserved mitochondrial enzyme activity and PGC-1α expression, partial relocation of mitochondria from myofibrillar to subsarcolemmal space occurred. No evidence of ultrastructural changes was observed in AMPK KD mutant (12). However, in the present study, these morphological changes were observed at higher magnification and could be more pronounced in older animals. Similarly, in CHO cells, an alteration of mitochondrial ultrastructure is specifically associated with reduction in cardiolipin content (31). Moreover, although to a lower extent, this resembles the ultrastructural abnormalities of mitochondria in the Barth syndrome, an X-linked disease. This disease, due to mutations in the tafazzin gene that belongs to the superfamily of phospholipid acyltransferase, is characterized by a dramatic reduction in cardiolipin content and by abnormal mitochondrial ultrastructure (43).

Cardiolipin is the only phospholipid with four acyl chains, ∼80% of which are composed of linoleic acid (18:2ω6), a polyunsaturated fatty acid of the ω6 series. The decrease in cardiolipin content in AMPKα2^−/− heart was accompanied by a significant decrease in linoleic acid content of mitochondrial and whole-heart phospholipids. The decrease in mitochondrial membrane linoleic acid observed in this study is due to neither nutritional differences, because animals received the same chow diet, nor a systemic effect because the plasma and liver lipid composition was unchanged. The fact that 1) the linoleic acid decrease parallels the cardiolipin decrease, 2) cardiolipin contains ∼80% linoleic acid, and 3) the linoleic acid content in storage lipids was higher suggests that the decrease in linoleic acid is a consequence rather than a cause of the cardiolipin decrease in AMPKα2-deficient mice. The lower linoleic acid content in heart compared with liver of AMPKα2^−/− mice could be related to the higher proportion of AMPKα1 in liver (4). These animals are thus characterized by a dysfunction of cardiac cardiolipin homeostasis that affects the inner mitochondrial membrane and that could be the key effector of mitochondrial dysfunction.

Similar decreases in cardiolipin content and complex I activity were reported after partial inhibition of cardiolipin synthesis in CHO cells (31). One important rate-limiting step of de novo cardiolipin synthesis is the initial reaction catalyzed by CDS (33). In AMPKα2^−/− mice, expression of CDS2, the only cardiac isoform (44), is decreased, possibly explaining the decrease in cardiolipin content. A second mechanism of cardiolipin biosynthesis is through deacylation-reacylation, which is catalyzed by phospholipases and acyltransferases, which are regarded as the principal enzymes involved in phospholipid remodeling in mammalian tissues (33). The acyltransferase ALCAT1 is regulated in concert with the level of cardiolipin and cardiolipin biosynthesis in mammalian heart (32,45). ALCAT1 mRNA expression was decreased in AMPKα2^−/− mice like both CDS2 mRNA and cardiolipin content. All of this suggests that AMPK is involved in cardiolipin homeostasis at least at the transcriptional level. Finally, expression of the mitochondrial fatty acid transporter CPT-1 was also decreased in AMPKα2-deficient mice. Decreased activation of CPT-1 because of the lack of AMPK-modulated regulation of malonyl CoA content, together with the downregulation of CPT-1, would have an additive inhibitory effect on fatty acid oxidation in AMPKα2^−/− mice.

In summary, selective deficiency of AMPKα2 causes a significant decrease in maximal mitochondrial respiration, whether carbohydrates or lipids were used as substrates. This is due to a defect in the function of complex I of the respiratory chain within the inner mitochondrial membrane, probably because of the decrease in cardiolipin content. Interestingly, it has been recently reported that the diabetic heart, characterized by altered lipid homeostasis and mitochondrial dysfunction, exhibits a dramatic decrease in cardiolipin content (46). These findings suggest that AMPK plays a critical role in the control of cardiac phospholipid homeostasis, possibly by modulating their metabolism through transcriptional machinery. More work is needed to elucidate the precise mechanisms by which AMPK may enhance energy production by favoring cardiolipin synthesis and thus improving the efficiency of the respiratory chain.

R.V.-C. is supported by the Centre National de la Recherche Scientifique. This work was supported by the Association Française contre les Myopathies, the Fondation de France, the European Commission FP6 program (EXGENESIS Grant QLG1-CT-2001-01488), and the European Union Contract (LSHM-CT-2005-018833/EUGeneHeart). The Franco-Slovak collaboration was funded by a French STEFANIK grant and by Slovak VEGA 2/6079/26 and APVT-51-31104.

We thank Dr. R. Fischmeister for continuous support. We thank J. Degrouard and D. Jaillard from the Centre Commun de Microscopie Electronique, Université Paris XI, Orsay, France, and D. Fortin for skillful technical assistance.