The AMPK signaling cascade in metabolic regulation: view from the chair

The AMP-activated protein kinase (AMPK) system is a key player in the regulation of cellular energy levels. By switching on catabolic pathways that generate ATP, this molecular switch triggers substrate oxidation processes during physical exercise and mediates the metabolic actions of several hormones and adipokines known to control whole-body energy balance. In this symposium, we learned about the discovery of AMPK, from the details of its complex molecular structure to its molecular mechanisms of activation by metabolic stresses, exercise and humoral factors. As AMPK is also activated by two classes of antidiabetic drugs (for example, biguanides and thiazolidinediones) it represents an attractive target for the treatment of obesity and its associated metabolic abnormalities. The potential upstream and downstream signals that may be involved in relaying AMPK's metabolic actions were also discussed by several speakers of the symposium as summarized below.

The mammalian AMPK system was discovered two decades ago by the first speaker of this session, Grahame Hardie, who initially reported that AMPK phosphorylates and inactivates key enzymes of lipid biosynthesis, that is acetyl-CoA carboxylase and 3-hydroxy-3-methylglutaryl-CoA reductase.¹ Hardie showed that AMPK exists as heterotrimeric complexes consisting of catalytic α-subunits and regulatory β- and γ-subunits encoded by distinct genes. Evidence was also presented that the AMPK α-subunits are phosphorylated and activated by the tumor suppressor LKB1 kinase complex and this likely explains much of the tumor suppressor effect of AMPK activation. In addition, we also learned that in some cells AMPK activation can be induced by increases in cytoplasmic Ca²⁺ and consequent activation of calmodulin-dependent kinase kinase-β (CaMKKβ) (see summary of David Carling's presentation for more details on upstream AMPK kinases). Furthermore, it was proposed that AMPK can also sense some aspect of glycogen structure and provide feedback regulation on glycogen synthesis. It is thought that this might occur through an interaction of glycogen particles with the AMPK β-subunit glycogen-binding domain. Whether this interaction plays a role in the phosphorylation and inactivation of glycogen synthase by AMPK remains to be clarified. Interestingly, the AMPK β-subunits appear to form the regulatory binding sites for AMP and ATP through specific modules (termed Bateman domains) and mutations in these domains cause heart diseases associated with excessive glycogen storage.

The LKB1/AMPK signaling pathway is activated by any metabolic stress that disturbs energy balance by interfering with ATP synthesis, such as glucose deprivation, hypoxia or ischemia. Some interesting examples were discussed. The activation of AMPK by glucose deprivation is particularly critical in specialized glucose-sensing cells such as the pancreatic β cells and specific neurons in the hypothalamus, where AMPK is a key signal for inhibiting insulin secretion or increasing food intake, respectively. AMPK also mediates the effect of hypoxia in oxygen-sensing cells such as glomus cells in the carotid body and pulmonary artery smooth muscle cells, both of which are key to the precise control of oxygen supply to the brain and lung, respectively. AMPK can also be activated by metabolic stressors that increase ATP consumption, the most notable example being muscle contraction during exercise. The role of AMPK in the metabolic effects of exercise will be discussed in greater detail below.

AMPK is also modulated by many adipose-derived factors (adipokines) that regulate whole-body energy metabolism, such as leptin, adiponectin and resistin.² Leptin and adiponectin both induce muscle glucose uptake and fatty acid oxidation by AMPK activation whereas resistin, which inhibits AMPK, appears to have the opposite effects. In an interesting contrast to its peripheral effect, leptin inhibits AMPK in the hypothalamus as a means to repress food intake, an effect that is also triggered by other anorexigenic factors (for example, insulin, melanocortin receptor agonists). Conversely, orexigenic agents (for example, agouti-related protein, ghrelin and cannabinoids) activate the kinase in the hypothalamus of fed animals.³ Thus, AMPK signals to regulate whole-body energy balance in both the periphery and the brain.

It is thought that activators of the kinase might be effective agents for treatment of obesity-linked type 2 diabetes and its co-morbidities because AMPK can be activated by both increased levels of physical exercise and by pharmacological means. Accordingly the pharmacological AMPK activator 5-aminoimidazole-4-carboxamide ribonucleoside (AICAR) reduces insulin resistance and glucose intolerance, with concomitant lipid-lowering effects in several animal models.¹ Furthermore, two classes of antidiabetic drugs, namely the biguanides (for example, metformin) and the thiazolidinediones (for example, rosiglitazone), both activate AMPK. The latter drug likely induces its therapeutic actions through adiponectin release and activation of AMPK in liver and peripheral tissues, whereas the glucose-lowering action of metformin is ablated in liver-specific LKB1 knockout mice in which AMPK can no longer be activated by the drug. A possible antiobesity effect of AMPK activation is also suggested by the observations that AMPK is more active in various animal models that are resistant to diet-induced obesity (for example, SCD1 knockouts, UCP1 overexpression in white fat and UCP3 overexpression in skeletal muscle).

We next learned from the second speaker of this session, Jørgen Wojtaszewski, that although all subunits of AMPK seem to be expressed in human skeletal muscle tissue, only three AMPK heterotrimers (α1β2γ1, α2β2γ1 and α2β2γ3) appear to account for the AMPK activity present in this tissue.⁴ In situ electrical stimulation-induced AMPK phosphorylation/activity is abolished, to a large extent, in the absences of LKB1, and thus at least under these conditions LKB1 seems to be the principal regulator of AMPK in mouse skeletal muscle.^{5, 6} The ability of contraction to induce AMPK is thought to be a local and tissue autonomous event as the enzyme can be activated in isolated muscle in vitro and because AMPK activation in human skeletal muscle during one-legged exercise occurs only in the exercising muscle.

A relatively high intensity of exercise (that is, 60–75% of VO) appears to be necessary for AMPK activation in humans, which is mostly accounted for by the activity of the α2β2γ3 heterotrimer.⁴ Interestingly, we learned that different heterotrimers may be recruited at different times during prolonged low-intensity exercise with the activation of α2β2γ1 AMPK following that of α2β2γ3 AMPK. Moreover, the activation pattern of α2β2γ3 AMPK appeared to correlate with phosphorylation of ACCβ whereas the time course of α2β2γ1 activation was associated with the phosphorylation of another AMPK target, namely Akt substrate of 160 kDa (AS160).

Mice lacking α2-AMPK have perturbed muscle energy balance during exercise, as reflected by reduced ATP content and an increase in AMP/IMP levels during both short-term high-intensity and long-term moderate-intensity exercises.⁷ It was speculated that oxidative metabolism is impaired in the muscle of these mice—either as a result of impaired substrate delivery or reduced mitochondrial oxidative capacity, which may compromise fatty acid oxidation or mitochondrial respiration. This would induce a higher reliance on exogenous glucose for anaerobic or aerobic metabolism in the working muscle, and in this way accelerate the development of fatigue. AMPK's importance for aerobic metabolism is also supported by earlier studies suggesting that it plays a role in mitochondrial biogenesis within skeletal muscle. Accordingly, mitochondrial biogenesis following chronic AICAR or β-guanadinopropionic acid treatment was severely compromised in transgenic mouse models of α2-AMPK deficiency. This appears to be unrelated to peroxisome proliferator-activated receptor γ co-activator-1α, a key transcriptional activator of mitochondriogenesis.

Wojtaszewski also discussed the potential role of AMPK in mediating the insulin-sensitizing action of an acute bout of exercise. A role for acute AMPK activation in this phenomenon is supported by the findings that both in vivo and ex vivo AICAR administration can acutely improve muscle insulin sensitivity. Some studies suggest that this effect cannot be solely explained by altered glycogen levels or changes in the classical IR/IRS-1/PI3K/Akt signaling cascade. However, it was proposed that this AMPK-mediated insulin sensitization may be related to activation of the Akt substrate AS160, a key regulator of GLUT4 translocation that is activated by both insulin and muscular contraction (see below for details). It was proposed that AS160 phosphorylation is increased and sustained after an exercise bout, which could then potentiate insulin action on glucose transport.

The complexity of the signaling cascades and their cross talk was discussed in even greater details by the next speaker, Juleen Zierath, who illustrated the concept of ‘critical nodes’ in signaling pathways, whereby many factors (for example, hormones, peptides) orchestrate metabolic and gene regulatory events that are integrated to initiate a biological response. She noted that contrary to the insulin-signaling cascade, the pathway(s) by which exercise or muscle contraction increase glucose uptake and metabolism remain poorly resolved. The identification of ‘critical nodes’ required for eliciting the highly regulated metabolic responses to exercise is still a challenging task nearly 40 years after the discovery that muscle glucose uptake and glycogen resynthesis is improved after exercise.

The effect of insulin and muscle contraction/cellular stress on glucose transport are additive, providing evidence for independent pathways in the regulation of glucose metabolism. However, both signals eventually lead to the mobilization of the GLUT4 glucose transporters to the cell surface, which suggests that a point of convergence exists within these two distinct signaling pathways. The point of this converging signaling event(s) remained unknown until it was recently realized that the Akt susbtrate AS160, a Rab-GTPase activating protein and a key regulator of GLUT4 traffic, lies in a critical node and is regulated by both Akt and AMPK, key signaling mediators of the insulin and contraction effects, respectively.

Zierath first reviewed the evidence linking insulin signals to AS160 and its role in insulin-mediated GLUT4 translocation and glucose transport. When AS160 is active Rab proteins are maintained in an inactive GDP-bound conformation and GLUT4 is retained intracellularly. The phosphorylation of AS160 by Akt upon insulin stimulation inhibits its Rab-GTPase-activating protein activity, and the activated GTP-bound Rab can then promote GLUT4 translocation. A major role for insulin-induced Akt activation in regulating AS160 is supported by the observation that AS160 phosphorylation and glucose uptake are severely affected in skeletal muscle from Akt2 knockout mice and mutation of key Akt-targeted phosphorylation sites within AS160 blunts insulin-induced GLUT4 translocation and glucose transport in both adipocytes and skeletal muscle.^{8, 9} Accordingly, siRNA-mediated gene silencing of AS160 increases GLUT4 translocation and basal glucose uptake in adipocytes. Finally, AS160 phosphorylation is also impaired in insulin-resistant states as well as in skeletal muscle from type 2 diabetic patients, which further supports an important role of AS160 in insulin-mediated glucose uptake in skeletal muscle.

It was first realized that AS160 may be a converging signaling element leading to both insulin and contraction-induced glucose uptake when it was found that AS160 could also be phosphorylated by the AMPK activator AICAR, and in response to in vitro contraction,¹⁰ or after endurance exercise.¹¹ Interestingly, it has been shown that the addition of active recombinant α1β1γ1 or α2β2γ1 AMPK complexes to muscle lysates can phosphorylate AS160. However, it is still unclear whether this occurs through a direct interaction between AS160 and AMPK. More specifically, it remains to be tested whether the recombinant AMPK complexes may have phosphorylated AS160 through activation of other kinase(s) present in the lysates. Importantly, AICAR-induced phosphorylation of AS160 is completely ablated in skeletal muscle from mice lacking either the α2 or γ3 AMPK subunits as well as in mice expressing a kinase-dead AMPKα2 subunit. Contraction-induced AS160 phosphorylation is also completely suppressed in mice lacking AMPKα2 or expressing a kinase-dead AMPKα2 in muscle, thus providing genetic evidence that AS160 is a downstream target of AMPK in contracting skeletal muscle.

Zierath concluded her presentation on the concept that multiple pathways likely contribute to AS160 phosphorylation in contracted muscle, and ultimately the glucose metabolic effects of exercise. This is supported by the observations that multiple phosphoproteins are readily observed in human skeletal muscle after an acute bout of aerobic exercise.¹¹ Multiple Ser/Thr residues are present in a conserved ‘RXRXXS/T’ motif known to be recognized by a large number of related kinases that include Akt and AMPK but also many other kinases (for example, protein kinase C and p70 S6 kinase) that can both be modulated by insulin and contraction signals. Thus, further studies will be required to decipher the molecular pathways and the critical signaling nodes involved in AS160 phosphorylation upon insulin stimulation and contraction.

The last speaker of this session, David Carling, reviewed current knowledge on the role of various upstream kinases in the control of AMPK and energy metabolism. Indeed, the identification of key upstream kinases in the AMPK cascade only recently emerged from studies conducted in yeast, where the AMPK equivalent SNF1 is activated by phosphorylation on Thr210 (equivalent to Thr172 in mammalian AMPK). Using mass spectrometry of yeast protein complexes, two protein kinases, Tos3 and Sak1 (originally termed Pak1), were identified and later found to interact and activate the SNF1 complex. These yeast kinases are members of the CaMKK family and their discovery was followed by that of a protein kinase termed LKB1. At the time, several pieces of evidence pointed toward LKB1 as the best candidate for the upstream kinase in the AMPK cascade and the field almost unanimously turned its attention to LKB1.

Early studies rapidly confirmed that LKB1 is an AMPK kinase. Indeed, purified recombinant LKB1 phosphorylates (Thr172) and activates AMPK whereas cells lacking LKB1 do not respond to AICAR or to other stimuli that increase the AMP/ATP ratio.¹² Genetic studies in animals also support a role for LKB1 as an upstream AMPK kinase. LKB1 deletion in skeletal or cardiac muscle severely abolishes AMPKα2, but not AMPKα1, activity in response to multiple stimuli (for example, contraction, ischemia, phenformin and AICAR), whereas LKB1 disruption in adult liver was found to markedly reduce total AMPK activity.

The mechanism by which AMP may activate the LKB1/AMPK signaling cascade has been the subject of debate. Early studies suggested that the LKB1 complex isolated from rat liver could be activated directly by AMP and that it could promote AMPK phosphorylation on Thr172. However, it was recently reported that AMP does not promote phosphorylation of AMPK by recombinant LKB1. Carling proposed that the apparent discrepancy may be linked to the fact that the preparations of AMPK and LKB1 from rat liver may have contained protein phosphatase 2C.¹³ Indeed, it has been reported that AMP inhibits dephosphorylation of AMPK and consequently, inhibition of dephosphorylation, rather than activation of phosphorylation, may account for the AMP effect.

Although LKB1 is recognized as a key upstream AMPK kinase, several lines of evidence suggest that other upstream kinases also activate AMPK in certain cell types. One such kinase is CaMKK, in particular the CaMKKβ isoform, and it has been recently suggested that an increase in intracellular Ca²⁺ is involved in the activation of this upstream kinase by thrombin in endothelial cells as well as following T-cell antigen receptor activation in T lymphocytes.

Are there other AMPK kinases beside LKB1 and CaMKK in mammalian cells? Carling mentioned that another unknown AMPK kinase might exist based on the identification of at least three kinases upstream of SNF1 in yeast. One such candidate is TGF-β-activated kinase-1 (TAK1), a member of the mitogen-activated protein kinase kinase kinase family, which was recently reported to phosphorylate and activate AMPK in vitro.¹⁴ The physiological significance of AMPK activation by TAK1 was recently demonstrated by the observation of severely impaired AMPK phosphorylation and activity after disruption of TAK1 in mouse embryos and embryonic fibroblasts by Cre-mediated recombination, or after ablation of floxed TAK1 in culture by viral delivery of Cre.¹⁵ In the latter case, TAK1 ablation also interfered with LKB1 activation, thus demonstrating a novel role for TAK1 in the LKB1/AMPK signaling axis.

Concluding remarks

AMPK is a fascinating enzyme that continues to inspire many research groups in the scientific community. We learned at this symposium that AMPK is a major energy checkpoint in several cells and a key mediator of the metabolic effects of muscular contraction. We are now beginning to understand the upstream activators of this energy gauge as well as the complex signaling nodes that are responsible for perpetuating its biological actions. A key goal in the future will be to determine the specific roles of the upstream kinases (for example, LKB1, CaMKK and TAK1) in activating distinct or overlapping AMPK complexes in different tissues. Another challenging task will be to improve our knowledge of the molecular mechanisms of AMPK action in key metabolic tissues. It is hoped that the identification of new downstream targets of AMPK complexes may help in the design of novel strategies to combat obesity and its related metabolic complications.