The hydrolysis of ATP drives virtually all of the energy-requiring processes in living cells. A prerequisite of living cells is that the concentration of ATP needs to be maintained at sufficiently high levels to sustain essential cellular functions. In eukaryotic cells, the AMPK (AMP-activated protein kinase) cascade is one of the systems that have evolved to ensure that energy homoeostasis is maintained. AMPK is activated in response to a fall in ATP, and recent studies have suggested that ADP plays an important role in regulating AMPK. Once activated, AMPK phosphorylates a broad range of downstream targets, resulting in the overall effect of increasing ATP-producing pathways whilst decreasing ATP-utilizing pathways. Disturbances in energy homoeostasis underlie a number of disease states in humans, e.g. Type 2 diabetes, obesity and cancer. Reflecting its key role in energy metabolism, AMPK has emerged as a potential therapeutic target. In the present review we examine the recent progress aimed at understanding the regulation of AMPK and discuss some of the latest developments that have emerged in key areas of human physiology where AMPK is thought to play an important role.
- AMP-activated protein kinase (AMPK)
- cell growth
- energy homoeostasis
Living cells use ATP as the immediate source of energy and in order to survive they need to ensure that they maintain a relatively high level of ATP . Within an individual cell, the demand for ATP must be matched by its supply. In order to achieve this, the cell requires an ATP sensor coupled in some way to an output that allows a response to be initiated if a fall in ATP is detected. One such system that has been identified in eukaryotic cells is the AMPK (AMP-activated protein kinase) cascade. AMPK is a heterotrimeric protein kinase complex, orthologues of which are expressed in virtually all eukaryotes. The primary function of AMPK is to monitor changes in the intracellular level of ATP and to couple this to phosphorylation of downstream substrates leading to an increase in the rate of ATP-producing pathways and/or a decrease in the rate of ATP-utilizing pathways.
Interest in AMPK has grown at a considerable rate over the last 5–10 years, and this has been matched by the number of groups working on projects directly related to AMPK. According to the Web of Science (http://apps.webofknowledge.com/), by 2001 nearly 250 papers had been published with AMPK in their title or abstract. By 2011 this had risen to nearly 4000. In the last three years (2009–2011) 2000 papers were published on, or relating to, AMPK. On average, this equates to almost two papers per day! Clearly, it would not be possible for a single review to provide comprehensive coverage of every aspect of AMPK biology so in an attempt to highlight what we consider to be some of the most important breakthroughs, we have drawn a ‘roadmap’ charting various key discoveries (Figure 1). There have been many excellent reviews on a wide variety of areas involving AMPK in the last few years (e.g. [2–8]) and the reader is referred to these for additional information. In the present review we have chosen to focus on a few areas where there have been exciting breakthroughs or significant new developments in the previous 2–3 years.
SUBUNIT STRUCTURE OF AMPK
AMPK is composed of three subunits, α, β and γ (Figure 2), and in mammalian cells there are two isoforms of the α subunit, two isoforms of the β subunit and 3 isoforms of the γ subunit . The α subunit contains a typical serine/threonine protein kinase domain at the N-terminus and a C-terminal regulatory domain. Within the β subunit there is a domain that has been termed the glycogen-binding domain or CBM (carbohydrate-binding module), since its primary sequence is similar to sequences conserved in a family of proteins that bind to oligosaccharides [10,11]. The C-terminal region of the β subunit interacts with the α and γ subunits, acting as a scaffold for the interaction of the heterotrimeric complex . In mammals, both β1 and β2 undergo N-terminal myristoylation, and the α and β subunits are subject to phosphorylation. The γ subunit isoforms contain four copies of a motif that was first identified in CBS (cystathionine-β-synthase) . CBS domains are found in a number of proteins and are almost always expressed as tandem repeats, forming what has been termed a Bateman domain . In AMPK, the CBS domains form the adenine nucleotide-binding regions (see below).
AMPK has been highly conserved throughout evolution and homologues are present in virtually all eukaryotes . In the budding yeast Saccharomyces cerevisiae, the homologue of AMPK is the SNF1 kinase complex, which has been studied extensively because of its role in glucose derepression [15,16]. In the presence of an abundant source of glucose, SNF1 is maintained in a low-activity state. However, depletion of glucose, or growth on an alternative carbon source, such as sucrose or galactose, leads to a rapid increase in SNF1 activity [17,18]. Under these conditions, the expression of genes required for metabolism of non-glucose carbon sources is increased.
REGULATION OF AMPK
As its name suggest, mammalian AMPK is allosterically activated by AMP. Allosteric activation appears to be specific for AMP and closely related AMP analogues , such as ZMP [AICAribotide (5-amino-4-imidazolecarboxamide) monophosphate] [20,21]. The degree of allosteric activation depends on the nature of the γ isoform present in the AMPK complex , but the maximum activation achieved is approximately 5-fold, which is very modest compared with the effect of phosphorylation on AMPK activity (see below). Notably, SNF1 is not allosterically activated by AMP.
The primary mode of regulation of AMPK is by reversible phosphorylation. AMPK is activated by phosphorylation of a threonine residue within the activation loop segment of the α subunit (Thr172 in rat AMPK) . Recombinant AMPK expressed in Escherichia coli is not phosphorylated on Thr172 and has negligible kinase activity . Phosphorylation of Thr172 acts therefore as an on–off switch. Although a number of kinases have been reported to phosphorylate Thr172 in vitro , CaMKKβ (Ca2+/calmodulin-dependent protein kinase kinase β) [26,27] and the tumour suppressor kinase, LKB1 (liver kinase B1) [28,29] are the predominant upstream kinases responsible for phosphorylation of Thr172 in mammalian cells. In addition to Thr172 a number of other phosphorylation sites within the α and β subunits of AMPK have been identified. Some of these sites are autophosphorylation sites, but in other cases the identity of the kinase leading to phosphorylation is unknown . Moreover, for most of these sites, there is no evidence that phosphorylation has a direct effect on AMPK activity and further studies are required to determine the physiological relevance of phosphorylation at these sites. One site for which there is some evidence for a functional role is Ser108 in β1. Ser108 undergoes autophosphorylation, and mutation of this site to an alanine residue completely blocked activation of AMPK by a small molecule activator (A769662 ), although the mechanism for this regulation is not yet understood .
The identity of the protein phosphatases acting on Thr172 remains to be determined. Members of the PP1 and PP2A [PPP (phosphoprotein phosphatase) family] and PP2C [PPM (metal-dependent protein phosphatase) family] have been shown to dephosphorylate Thr172 efficiently in vitro [19,32]. Okadaic acid, a toxin produced by dinoflagellates that is a potent inhibitor of the PPP, but not the PPM, family , had no effect on AMPK activity in rat hepatocytes , suggesting that AMPK dephosphorylation is catalysed primarily by one or more members of the PPM family. In a recent study, siRNA (small interfering RNA)-mediated knockdown of PPM1E/F increased Thr172 phosphorylation of AMPK in HEK (human embryonic kidney)-293 cells, and a weak interaction between AMPK and PPM1E and PPM1F was observed . In a separate study, siRNA knockdown of PP1α and PP1β catalytic subunits reduced the dephosphorylation of Thr172 seen in MIN6 cells following a switch from low to high glucose . In addition, knockdown of R6, a regulatory subunit of PP1, was also shown to reduce Thr172 dephosphorylation . The R6 subunit was reported to interact with the AMPKβ subunits, as determined by co-immunoprecipitation and yeast two-hybrid experiments . The conclusions of these two studies point to completely different outcomes, but perhaps this is not so surprising. There is no a priori reason to presume that a single phosphatase controls AMPK dephosphorylation and it seems reasonable to hypothesize that the identity of the phosphatases involved in dephosphorylation of AMPK may be dependent on the cell type and/or the stimulus to which the cell is responding. Further studies aimed at the identification of the physiologically relevant AMPK phosphatases in vivo are required to address these issues.
The level of Thr172 phosphorylation is determined by the relative activities of the upstream kinases (CaMKKβ and LKB1) and the protein phosphatases acting on AMPK. CaMKKβ is activated by calcium and calmodulin  and conditions which increase intracellular calcium lead to an increased phosphorylation of Thr172 via activation of CaMKKβ [26,38–40]. In contrast, LKB1 appears to be constitutively active, since its activity is unaffected by treatments that cause a marked increase in AMPK activity [29,41], raising the issue of how LKB1 mediates changes in Thr172 phosphorylation. Early studies suggested that AMP promoted phosphorylation of AMPK by three independent mechanisms [32,42]. The first of these, direct allosteric activation of the upstream kinase, has subsequently been dismissed [26,27,29,43]. It was also proposed that binding of AMP to AMPK caused a conformational change in the complex, promoting phosphorylation of Thr172 whilst at the same time protecting against dephosphorylation. The role of AMP to promote phosphorylation has proved controversial. Our own studies indicated that the stimulatory effect of AMP on phosphorylation of Thr172 was due to contamination of the upstream kinase preparation with PP2C . More recently, Kemp and colleagues reported that AMP can promote phosphorylation of Thr172, but only when the β subunit is myristoylated at the N-terminus . However, another study using native AMPK purified from rat liver in which the β subunit is N-terminally myristoylated  failed to detect a stimulatory effect of AMP on Thr172 phosphorylation . Further studies seem warranted in order to resolve these conflicting reports. A third independent mechanism by which AMP was proposed to increase Thr172 phosphorylation is by protecting the kinase from dephosphorylation [32,43,46]. In this case, there appears to be no dispute and recent data from structural studies provides strong evidence underlying a mechanistic basis for the protective effect  (see below).
Regulation by ADP
An important discovery in AMPK research is the recent finding that ADP, as well as AMP, protects against the dephosphorylation of Thr172 . Indeed, at first sight it might appear odd that this has only recently emerged, but unlike AMP, ADP does not allosterically activate AMPK . As allosteric activation of AMPK by AMP provided an additional criterion in determining AMPK activity, the obvious lack of allosteric activation by ADP meant that the role of ADP in the regulation of AMPK was overlooked. ADP has also been reported to promote phosphorylation of Thr172, and as shown for AMP, this effect requires N-terminal myristoylation of the β subunit . We reported recently that ADP protects SNF1 against dephosphorylation , providing an answer to a long-standing conundrum. Although the obvious similarities between the regulation of AMPK and SNF1 have been recognized for many years, identification of the metabolic signal leading to activation of SNF1 in response to glucose limitation proved elusive. Using AMPK as the paradigm, the obvious candidate, AMP, turned out to have no effect on SNF1 activity [17,18,43,50]. The finding that ADP protects against dephosphorylation of SNF1 raises the intriguing possibility that ADP represents a unifying ligand for activation of AMPK orthologues in all eukaryotic species.
INSIGHTS INTO THE MECHANISMS REGULATING AMPK
Significant progress has been made during the last few years regarding the molecular mechanisms underlying the activation of AMPK. The crystal structure of a regulatory core of AMPK, containing the C-terminal regions of α1 and β2 together with full-length γ1 subunits revealed that there are four potential nucleotide-binding sites within the γ subunit . However, only three of the four sites were found to be occupied by nucleotide in the AMPK crystal structure, with two molecules binding between CBS3 and CBS4 and one molecule binding between CBS1 and CBS2 . The finding that AMPK binds three nucleotide molecules was also unexpected in light of binding data that clearly showed that phosphorylated full-length AMPK bound only two exchangeable nucleotides , consistent with a previous study showing that the isolated γ2 subunit of AMPK binds two adenine nucleotide molecules . These apparently discordant results are explained by the finding that one of the three nucleotide-binding sites identified in the crystal structure binds AMP so tightly that it is non-exchangeable under normal conditions, and so is not detected in binding assays. This non-exchangeable AMP binds at site 4 (according to the nomenclature proposed by Kemp et al. ; see Figure 3), whereas the two exchangeable nucleotides bind at sites 1 and 3. The two exchangeable sites bind nucleotide with significantly different affinities. The tight binding site (site 1) binds approximately 30-fold stronger than the weak binding site (site 3). In contrast with sites 1, 3 and 4, site 2 in mammalian AMPK does not bind nucleotide, perhaps because conserved aspartate residues in CBS1, CBS3 and CBS4 are replaced in CBS2 by arginine . The nucleotide-binding properties of AMPKγ1 are summarized in Figure 3.
The finding that there are two exchangeable nucleotide-binding sites and that nucleotides affect AMPK activity in two different ways (allosteric and regulation of Thr172 phosphorylation) raises the question of whether nucleotide binding at distinct sites might mediate the different effects on activity. Unfortunately, attempts to abolish nucleotide binding by mutation of the aspartic acid residues in CBS1, CBS3 and CBS4, or mutation of other candidate amino acids within the canonical binding sites, have been unsuccessful , and so it has not been possible to address this issue directly. Despite this, there are some clues that suggest that individual nucleotide-binding sites are responsible for distinct functional effects. One clue comes from the finding that the binding affinity at the tighter site is consistent with the concentration required for allosteric activation by AMP, whereas the affinity at the weaker site correlates with the concentration range of AMP and ADP for protection against dephosphorylation . Another clue stems from the observation that NADH binds specifically to AMPK at the tighter binding site, and although NADH has no direct effect on AMPK activity it does compete with AMP for allosteric activation, but not with the effect of AMP or ADP on protection against dephosphorylation . These results suggest that it is binding of AMP at the tighter of the two exchangeable sites that leads to allosteric activation, and binding of AMP or ADP at the weaker site that protects against dephosphorylation. AMPK crystals soaked in the presence of one molar equivalent of ADP identified site 1 as the tighter of the two exchangeable binding sites . Taken together, these results support the conclusion that AMP binding at site 1 mediates allosteric activation, whereas binding of AMP or ADP at site 3 mediates protection against dephosphorylation.
Role of site 3 in protection against dephosphorylation
The structure of a phosphorylated form of AMPK, containing the kinase domain and the regulatory core of the trimeric complex has been reported . The activation loop, including phosphorylated Thr172, is well defined and makes extensive contacts with the C-terminal regions of the α and β subunits. In this conformation access to Thr172 by PPPs would be blocked [47,54]. Another notable feature of this structure involves a region, termed the α-hook, that is well conserved in the α1 and α2 subunits throughout vertebrates and which interacts with site 3. Structural studies suggest a model in which association of the α-hook with site 3 in the presence of AMP/ADP favours formation of the kinase/regulatory interaction protecting against dephosphorylation of Thr172. However, on the basis of the superimposition of previous structures with MgATP , when MgATP is bound the α-hook cannot interact at site 3 due to a predicted steric clash with an arginine residue (Arg69). This promotes a conformation of the kinase in which the kinase/regulatory interaction is weakened, allowing access of phosphorylated Thr172 to PPPs. The key aspects of this model are depicted in Figure 4. Consistent with this model, mutations within the α-hook abolished AMP- or ADP-mediated protection against dephosphorylation .
Whilst addressing some of the key aspects of AMPK regulation, this model does not provide all of the answers. No insight into how AMP or ADP might promote phosphorylation of Thr172 is afforded, nor is there any clue regarding the mechanism for allosteric activation of AMPK by binding of AMP at site 1. Previous studies using a truncated form of the α subunit revealed that allosteric activation could involve a region of the α subunit termed the AID (autoinhibitory domain) [55,56]. In the structure of the phosphorylated AMPK complex, the region containing the AID was mainly unstructured, in contrast with the ordered conformation adopted in the truncated kinase structure [47,57]. Whether or not the AID plays a role in further inhibition of the unphosphorylated form of AMPK remains to be determined.
Regulation of AMPK by adenine nucleotides in vivo
The binding studies also revealed that although there was a marked difference in affinity between the tight and weak sites, at an individual site AMP, ADP and ATP all bound with essentially the same affinity. Since the concentration of ADP in mammalian cells is 10–100-fold higher than that of free AMP , it is likely that ADP, not AMP, is the main regulator of AMPK activity under normal cellular conditions. An important result from the binding studies was the finding that magnesium caused a substantial reduction in the affinity of ATP at both the tight and the weak sites. This has significant implications in terms of understanding the regulation of AMPK in vivo, because in mammalian cells most of the ATP (~95%) is in complex with magnesium . The weakening effect of magnesium on ATP binding acts to counteract the higher physiological concentration of MgATP relative to ADP, helping to explain how ADP is able to compete with ATP and MgATP for binding to the γ subunit and regulate the rate of Thr172 dephosphorylation. A major problem facing research into AMPK in vivo is that AMPK activity is measured following isolation from tissue, and this raises the possibility that the activity determined is not the same as the activity in situ. A fluorescent probe that can be used to monitor AMPK activity in living cells has been reported , and it is possible that this probe, or something similar, could be used to measure AMPK activity in vivo. In addition, fluorescent probes are available that can monitor ATP levels in mammalian cells , so it is conceivable that a multiplexing approach could be developed that would allow simultaneous monitoring of AMPK activity and ATP levels in living animals. Developments in this area in the future may provide important insights into the in vivo regulation of AMPK in response to changes in energy homoeostasis.
Role of glycogen in the regulation of AMPK
The β subunit of AMPK contains a region that is related to CBMs that are found in a number of enzymes that metabolize polysaccharides, such as glycogen or starch [10,11]. Consistent with the presence of a CBM, AMPK binds to glycogen in vitro [10,11] and has also been shown to associate with glycogen particles in the rat liver . Although the physiological significance of the association of AMPK with glycogen remains uncertain, it has been proposed that AMPK might act as a glycogen sensor , analogous to its role as an ATP/ADP sensor. In a previous study, glycogen was reported to inhibit AMPK , although an older study had failed to see any effect of glycogen on AMPK activity . One possible explanation for these discordant results may be related to the finding that the degree of inhibition varied markedly depending on the preparation of glycogen used in the assay . This finding also raises the interesting possibility that AMPK could act not only as a sensor of total glycogen levels, but also monitor the degree of branching within the glycogen. Of note, the presence of the CBM within the β1 subunit was reported to be required for activation by a small molecule activator (A-769662) of AMPK . The β2 isoform binds carbohydrate more tightly than β1, and a recent study provided evidence that an additional threonine residue, present within a loop in β2, but missing in β1, might be an important determinant for this difference in binding affinity .
Acetylation of AMPK
A recent study examining the global role of lysine acetylation identified the catalytic subunit of AMPK as a potentially important target for this modification . Lys31, Lys33 and Lys71 in human α1 (corresponding to Lys29, Lys31 and Lys69 in rat α1, and conserved in α2) were found to be major sites for acetylation by p300 acetyltransferase. Using a short hairpin RNA-based knockdown system in HCT116 and HepG2 cells, HDAC1 (histone deacetylase 1) was identified as the major deacetylase acting on AMPKα1 . Changes in the degree of acetylation of AMPK isolated from cells treated under conditions known to alter AMPK activity, e.g. glucose starvation, were detected and an inverse relationship between acetylation and Thr172 phosphorylation was suggested. Interestingly, in HepG2 cells, knockdown of p300 increased basal Thr172 phosphorylation, whereas knockdown of HDAC1 decreased basal phosphorylation. These changes were not seen in cells in which LKB1 expression was down-regulated, suggesting that LKB1 is required for this effect . In the same study, the acetylation of AMPKα was reported to influence the association of AMPK with LKB1. Conditions favouring low acetylation enhanced the interaction of AMPK and LKB1, whereas high acetylation weakened the interaction. These findings would support a model whereby acetylation of AMPKα regulates the interaction of AMPK with LKB1, thus providing a direct mechanism for controlling the phosphorylation and activation of AMPK mediated by LKB1. One issue stemming from the potential role of acetylation in the regulation of AMPK is that of the subcellular localization of the key components involved. Although there are reports of nuclear localization, under most conditions, the majority of AMPKα1 complexes are localized in the cytoplasm . Both p300 and HDAC1 are predominantly localized within the nucleus, raising the question of how they gain access to their substrate. Moreover, how is a change in cellular energy coupled to a change in the acetylation of AMPK? It will be interesting to determine whether nucleotide binding to AMPK directly affects acetylation/deacetylation and whether the activity of p300 and/or HDAC1 is altered by cellular energy status.
THE ROLE OF AMPK IN AUTOPHAGY
Autophagy is the mechanism by which cells dispose of and recycle damaged organelles under basal conditions and involves the degradation of proteins and whole organelles by lysosomes. However, if energy demands become extreme, due to lack of nutrients, autophagy is increased in order to meet the requirement for amino acids and other nutrients. As a result of this process the sequestered contents of the lysosomes are degraded and recycled as a source of energy. Autophagy is subject to stringent control and allows cells to protect themselves against multiple stresses, which in turn provides protection to the organism against many disease states, including inflammation and neurodegeneration . It is not surprising that AMPK has a role to play in this mechanism of energy regulation to keep the balance between nutrient homoeostasis and excessive internal degradation.
In mammals autophagy is initiated by ULK (Unc-51-like kinase) 1, the mammalian homologue of the yeast ATG (autophagy) 1 kinase. Within the cell, ULK1 exists as a complex with ULK2, ATG13, ATG101 and FIP200 (focal adhesion kinase family-interacting protein of 200 kDa) [67,68]. Under conditions of adequate nutrient supply, ULK1 interacts with RAPTOR [regulatory associated protein of mTOR (mammalian target of rapamycin)] and is phosphorylated on Ser758 by mTORC (mTOR complex) 1 to maintain it in an inactive state . Upon starvation, this site becomes dephosphorylated allowing ULK1 to form a stable association with AMPK . AMPK was reported to phosphorylate ULK1 at a number of sites: Ser467, Ser555, Thr574 and Ser637 . In primary mouse hepatocytes lacking AMPK, p62 was found to accumulate together with increased mitochondrial content indicating a lack of autophagy. This phenotype is similar to that observed in ULK1-deficient cells. Reconstitution of ULK1-deficient cells with a mutant form of ULK1 that could not be phosphorylated by AMPK revealed that phosphorylation of ULK1 was essential for cell survival during starvation . Another study  found that AMPK activation resulted in phosphorylation of ULK1 at different sites (Ser317 and Ser777) to those reported by Egan et al. . Although there is disagreement on the identity of the AMPK phosphorylation sites in ULK1, there is consensus that AMPK activation enhances autophagy via ULK1. In addition to direct phosphorylation of ULK1, AMPK can regulate its activity by suppression of mTORC1, thus relieving the inhibitory phosphorylation of ULK1 (Figure 5).
Recent studies have reported that exercise induces autophagy in skeletal and heart muscle [73,74]. Exercise-induced autophagy was shown to require phosphorylation of BCL2 (a regulator of apoptosis), although the kinase(s) responsible for increased phosphorylation of BCL2 in response to exercise have not yet been identified . Using a transgenic knockin mouse model in which native BCL2 was replaced with a mutant form harbouring substitution of three phosphorylatable residues (Thr69, Ser70 and Ser84) to alanine, it was found that autophagy was no longer induced by exercise . One of the most well-studied physiological effects of exercise is to increase insulin sensitivity and this is thought to be responsible for many of the beneficial health effects of exercise. Remarkably, mice expressing the mutant BCL2 protein had impaired exercise-induced increases in insulin sensitivity . In an attempt to elucidate the mechanism underlying this impairment, He et al.  investigated AMPK signalling, since it is well established that exercise activates AMPK in skeletal muscle. In the BCL2 mutant mice, phosphorylation and activation of AMPK in response to exercise was blunted, suggesting that increased autophagy is required for normal AMPK activation in muscle. How this might be achieved is unclear, and seemingly places AMPK downstream of the autophagic response, which at a simple level seems paradoxical. Increased signalling through CaMKKβ has been suggested to play a role in the activation of AMPK and induction of autophagy . However, no difference in calcium homoeostasis was detected in muscle from mice expressing the mutant BCL2 protein . It would be interesting to examine whether there is any change in metabolism of ATP in the mutant mice in response to exercise as this was not reported in the study. Determining the molecular link between BCL2 and AMPK, and in turn understanding how autophagy signals to AMPK, represents an exciting challenge for the field.
AMPK AND INFLAMMATION
Inflammation is a complex biological process that the body initiates in response to tissue damage or infection in order to protect itself against harmful stimuli and initiate the healing process after insult. The initial stage is the production of cytokines, cell signalling proteins that regulate immunity and inflammatory responses. Without the inflammatory response wounds would not heal and there would be progressive destruction of tissue, which would compromise the survival of the organism. Inflammatory disorders include atherosclerosis, allergies and cancer. The production of inflammatory cytokines from key metabolic tissues, including adipose tissue, liver and macrophages, has also been shown to increase in obesity and precedes development of insulin resistance  providing a clear connection between inflammatory and metabolic pathways.
There is a growing body of evidence suggesting that AMPK plays an important role in inflammation. Several studies have shown that treatment of inflammatory conditions with AMPK activators, such as AICAriboside, is advantageous, e.g. autoimmune encephalomyelitis  and colitis . However, some reports claim that the anti-inflammatory response of AICAriboside is AMPK-independent [79,80]. It has also been found that treatment of cells with LPS (lipopolysaccharide), a membrane component of bacteria, leads to dephosphorylation and inactivation of AMPK, whereas the anti-inflammatory cytokines IL (interleukin)-10 and TGF (transforming growth factor)-β result in AMPK activation . Yang et al.  showed that in macrophages AMPKα1 negatively regulates the inflammatory response caused by exposure to free fatty acids, LPS or diet-induced obesity by inhibiting NF-κB (nuclear factor κ-B) signalling. A study using mice lacking haematopoietic expression of AMPKβ1 revealed an essential role for AMPK in protecting against adipose tissue inflammation and insulin sensitivity in obesity . In a clever strategy involving transplanting bone marrow from AMPKβ1-deficient mice into wild-type mice the researchers showed that mice receiving β1-deficient bone marrow were more susceptible to macrophage adipose tissue inflammation and liver insulin resistance in response to a high-fat diet than wild-type mice . Another recent study showed that under calorie restriction and exercise AMPK acts in conjunction with PGC1α (peroxisome proliferator-activated receptor γ coactivator 1α) to control the expression of hepatic inflammatory response mediators and induce expression of anti-inflammatory IL-1 receptor agonist . Together, these studies suggest a link between the control of dietary state, lipid metabolism and inflammation (see Figure 6). This raises the possibility that the insulin-sensitizing effects of metformin and exercise (both of which have been shown to reduce inflammation [85,86]) could be mediated, at least in part, by activation of AMPK in macrophages.
One of the innate immune responses to infection is the phagocytosis of bacteria by macrophages and neutrophils, which plays an essential role in the control of inflammation. This process involves a rapid reorganization of the cytoskeleton , which has been shown to involve the small GTPases, RAC1 (ras-related C3 botulinum toxin substrate 1) and CDC42 (cell-division cycle 42) . AMPK has been reported to enhance the phagocytic ability of macrophages and neutrophils by activation of PAK [p21 protein (Cdc42/Rac)-activated kinase] 1/2 and WAVE2 [WASP (Wiskott–Aldrich syndrome protein) verprolin homologous 2], which are downstream effectors of RAC1 . Various activators of AMPK were shown to increase phagocytic activity of macrophages in culture and increase the phosphorylation of CLIP170 (CAP-Gly domain-containing linker protein 1), which has been shown previously to be a target of AMPK . This ability to increase the efficiency of the host to clear microbes or apoptotic cells by phagocytosis is important in the resolution of inflammation, and activation of AMPK could therefore reduce inflammation by increasing phagocytosis.
AMPK AND BILE ACID METABOLISM
Bile acids are made in the liver from cholesterol and form one of the main constituents of bile, which is secreted into the gut, via the gall bladder, where it plays a critical role in absorption of fats and fat-soluble nutrients from the diet. Several recent studies have implicated AMPK in the regulation of bile acid metabolism. In one study, AMPK was shown to phosphorylate and activate the transcriptional activator SRC-2 (steroid receptor coactivator-2) . In mice, liver-specific deletion of SRC-2 causes impaired intestinal fat absorption, but this can be completely reversed by the addition of exogeneous bile acids to their diet . Consistent with these observations, SRC-2 was found to regulate the expression of BSEP (bile salt export pump), and the lack of SRC-2 expression in liver led to a dramatic reduction in BSEP expression . BSEP plays a key role in the transport of bile acids from the liver into the bile ducts, and so reduced expression would lead to decreased secretion of bile acids into the gut, exactly as found in mice lacking hepatic expression of SRC-2. Activation of AMPK in hepatocytes was shown to increase BSEP mRNA expression via phosphorylation of SRC-2, which causes an increase in the intrinsic transcriptional activity of SRC-2. Taken together, these findings suggest a pathway whereby activation of AMPK in response to decreased energy supply leads to increased bile acid secretion into the gut allowing dietary fat absorption to replenish energy supplies.
A previous study had reported that inhibition of the LKB1–AMPK pathway impaired bile canalicular formation in collagen sandwich cultures of rat hepatocytes . In another study, liver-specific deletion of LKB1 was found to impair canalicular membrane integrity and bile duct formation, resulting in defective bile-acid clearance and subsequent accumulation of bile acids in the serum and liver . Interestingly, in the latter study no change in the expression of BSEP was detected. However, loss of LKB1 was associated with mislocalization of BSEP from a predominantly canalicular membrane localization in wild-type mice to a mainly cytoplasmic localization. It is noteworthy that lack of AMPK activity in the liver does not lead to a significant alteration in bile acid metabolism, or change in BSEP expression [92,94], suggesting that other kinases downstream of LKB1 are also involved in bile canalicular formation and bile acid clearance.
AMPK IN NEURODEGENERATIVE DISEASE
In comparison with peripheral tissues, delineating the function of AMPK in the central nervous system is still in its infancy and as yet, it is unclear whether activation of AMPK plays a protective or causative role in neurodegeneration. In the brain, AMPK is localized primarily in neurons although there is some expression in astrocytes, oligodendrocytes, microglia and neural precursor cells [95–97]. To date, the most comprehensive studies examining the physiological role of AMPK in brain have taken place in the arcuate nucleus of the hypothalamus. These studies implicate AMPK in the regulation of appetite and whole body energy homoeostasis [98,99]. However, more recently, there has been increased interest in the role of AMPK in neurodegenerative disease with results suggesting both a protective and degenerative role (Figure 7).
Beneficial effects of AMPK in the brain
AMPK is activated in response to glucose deprivation and hypoxic–ischaemic insult , both of which are potentially catastrophic for the brain. Transient activation of AMPK in neurons was found to offer short-term protection against these insults, in part by regulating the expression of glucose transporter 3 [100,101]. In a Drosophila model in which expression of the AMPKβ orthologue (referred to as alicorn) is deleted, early onset of retinal cone cell degeneration is observed, which is exacerbated by exposure to light suggesting that AMPK protects mature neurons from excitotoxicity . In primary neurons, AMPK was shown to interact with and phosphorylate the metabotropic GABAB (γ-aminobutyric acid type B) receptor, enhancing the activation of inwardly rectifying K+ channels. This mechanism was potentiated by increased intracellular AMP that might occur during pathological stimuli such as hypoxia or ischaemia and ultimately acts to suppress neuronal excitation, preventing cellular damage . Indeed, a subsequent study showed that prolonged receptor activation with NMDA (N-methyl-D-aspartate) resulted in decreased AMPK phosphorylation of GABAB receptors and targeting for lysosomal degradation . More recently, resveratrol was reported to be protective in cell models of Alzheimer's disease by negatively regulating the production of Aβ (amyloid β) peptide, a major component of amyloid plaques characteristic of Alzheimer's disease, through an AMPK-mediated pathway .
Evidence linking AMPK to neurodegenerative disease
One of the first studies to suggest that neural AMPK could have deleterious effects used a mouse model of stroke. In this study it was found that knockdown of AMPKα2, but not α1, reduced infarct volume  suggesting that the presence of active AMPK can promote brain lesions in stroke. More recently, it has been suggested that AMPK may play a role in mediating apoptosis. In primary neurons, activation of glutamate receptors by excitotoxic insult results in up-regulation of pro-apoptotic BIM (bisindolylmaleimide) expression in an AMPK-dependent manner . In contrast with the protective effect of resveratrol described earlier, activation of AMPK has been implicated in aspects of Alzheimer's disease. In response to metformin treatment, activation of AMPK promotes the formation of both intracellular and extracellular Aβ peptide . In a recent immunohistochemical study of Alzheimer's disease brain, Vingtdeux et al.  found that activated AMPK localized to areas in which pretangle or tangle formations of hyperphosphorylated tau were evident . This supports data from our laboratory showing that AMPK phosphorylated tau at key residues known to promote tangle formation in primary cortical neurons . In addition, we found that AMPK could be activated in a Ca2+-dependent manner when these neurons were exposed to aggregated Aβ peptide.
AMPK is also implicated in the progression of Huntington's disease. The original observation was made in a mouse model of Huntington's disease [R2/6; containing the characteristic CAG trinucleotide repeat expansion in exon 1 of the Htt (Huntingtin) gene], which identified aberrant AMPK activity and impaired glucose import in the striatum. Chronic treatment with CGS21680, an adenosine-A2A receptor agonist, not only reduced AMPK activation and improved glucose uptake, but also markedly improved animal performance in behavioural paradigms . Very recently, the same group that made the original finding reported that AMPKα1 translocated from the nucleus to the cytosol in cell models overexpressing the mutant Htt gene . Selective activation of AMPKα1 and nuclear localization was also found in the striatal neurons of both human and mouse brain Huntington's disease samples. In response to this activation, expression of the anti-apoptotic protein BCL2 was down-regulated, resulting in degeneration of the striatal neurons . Interestingly, within the last year, a study characterizing the levels of high-energy phosphates in striatum, hippocampus and forebrain of the R6/2 model found substantial deficits, correlating with an increase in activated AMPK and surprisingly an increase in phosphocreatine levels. These perturbations occurred prior to the onset of behavioural phenotypes, suggesting that high-energy phosphate imbalance is an early event in the progression of the disease .
Finally, there are initial data emerging to suggest that AMPK activation may be deleterious in motor neuron disease . Transgenic mice expressing a mutant form of superoxide dismutase replicate a number of key features of motor neurone disease, including energy imbalance. Lim et al.  found that AMPK was activated in both dissociated spinal cord neurons and striatal lysates from transgenic mice expressing mutant superoxide dismutase and that pharmacological reduction of AMPK activity in dissociated neurons was protective. To confirm the involvement of AMPK, a Caenorhabditis elegans model of mutant superoxide dismutase was used which was capable of replicating the motor neuron dysfunction. Placing these animals on an AMPKα2-null background, or knocking down AMPKα2 expression by RNA interference, restored a number of the locomotor defects, suggesting a beneficial effect of AMPK inhibition .
Reconciling the disparate data
A key question that arises from the different studies investigating AMPK in neuronal function is how to reconcile its apparently disparate protective and neurodegenerative roles. One possibility is that AMPK functions as a double-edged sword, having both beneficial and deleterious effects on neuronal function. In this model, short bursts of AMPK activation are neuroprotective, allowing the energy balance of the cell to be maintained, such as during normal excitatory stimulus. In neurons this would suggest that the CaMKKβ/AMPK pathway is primarily activated, and we have found some support for this argument [110,115]. On the other hand, it may be through prolonged stimulus that the LKB1/AMPK pathway is brought into play, contributing to the switch from physiological to pathological activation .
AMPK IN CANCER: FRIEND OR FOE?
It is well documented that inactivating mutations in the LKB1 gene results in tumour formation in humans as well as animal models [116,117]. However, it is less clear as to whether the tumour-suppressor functions of LKB1 are mediated via activation of AMPK. LKB1 acts as a master kinase, phosphorylating and regulating a number of AMPK-related kinases [118,119]. Although our understanding of the physiological role of the AMPK-related kinases is incomplete, they have been implicated in the control of cell polarity and cell growth , and so it is possible that some of the tumour-suppressor properties of LKB1 are mediated by members of the AMPK-related kinases. In this section we will concentrate on the evidence supporting a role for AMPK in cancer.
AMPK as a tumour suppressor
Cancer cells have a huge demand for energy to allow for rapid growth and division. A large body of evidence has led to the consensus that during energy-limiting conditions, activation of AMPK causes a reduction in anabolic pathways, ultimately leading to reduced cell growth (see  for a review on this topic). In this scenario, activation of AMPK would be predicted to antagonize cancer cell growth. In order to understand the role of AMPK in regulating cell growth, we need to consider another aspect of this pathway, specifically the network of proteins stimulating growth, such as mTOR. In mammalian cells mTOR exists in two functionally distinct protein complexes: mTORC1 and mTORC2 [120,121], and under most conditions, only mTORC1 is inhibited by rapamycin . When the nutrient supply is not limiting, or in the presence of growth factors, mTORC1 is active and stimulates pathways associated with cell growth, e.g. protein and lipid synthesis, ribosome biogenesis and nutrient import, and consequently dysregulation of the mTORC1 pathway has been implicated in many types of cancer. mTORC1 can be activated downstream of the PI3K (phosphoinositide 3-kinase)–Akt and Ras–Raf–MEK [MAPK (mitogen-activated protein kinase)/ERK (extracellular-signal-regulated kinase) kinase]–ERK signalling pathways . These pathways converge on TSC1 (tuberous sclerosis 1/hamartin)–TSC2 (tuberous sclerosis 2/tuberin), a GAP (GTPase-activating protein) complex that inhibits the activity of the small GTPase, RHEB (RAS homologue enriched in brain) . Both Akt and ERK1/2 phosphorylate TSC2 at multiple sites, inhibiting the GAP activity of the TSC1–TSC2 complex [122,123]. This in turn leads to accumulation of the GTP-bound form of RHEB, a potent activator of mTORC1. In terms of energy status, the activity of mTORC1 follows an inverse relationship with that of AMPK: high-energy/nutrient replete conditions favour mTORC1 activity, whereas low-energy/starvation conditions signal AMPK activation. Given this relationship it is perhaps not surprising that many studies have shown a role for AMPK in negatively regulating mTORC1 activity. AMPK can directly phosphorylate mTOR at Thr2446, inhibiting phosphorylation at Ser2448 by Akt during nutrient deprivation . In addition, AMPK phosphorylates RAPTOR at Ser792 promoting binding of 14-3-3 and suppression of mTORC1 activity . AMPK also acts upstream of mTORC1 to regulate its activity. AMPK phosphorylates TSC2 at Ser1345, enhancing its GAP activity, leading to decreased RHEB function and lower mTORC1 activity . The role of AMPK in regulating mTORC1 provides the cell with a mechanism to integrate energy supply with cell growth. Low energy supply leads to increased AMPK activity reducing mTORC1 function and inhibiting cell growth (see Figure 8).
There are several reports that AMPK activity is necessary for cell-cycle arrest at the G1-phase under conditions of limited nutrient supply, via phosphorylation of the tumour suppressors p53 [127,128] and p27 . However, convincing evidence that either p53 or p27 are direct targets for AMPK in vivo has not emerged. There are also reports of AMPK-depleted mammalian and Drosophila cells undergoing cell-cycle arrest [130,131] implying that in these systems AMPK is not necessary for cell-cycle arrest.
Another pathway by which AMPK can affect tumour growth is via its impact on lipid synthesis. Cancer cells exhibit high rates of de novo fatty acid synthesis irrespective of circulating lipid levels and independently of physiological hormones normally involved in regulating fatty-acid synthesis. As a consequence, in many cancers there is increased expression of FAS (fatty acid synthase) [132–134], the enzyme catalysing the terminal reactions in de novo fatty-acid synthesis. Although the exact function of de novo fatty acids in the cancer cell is unknown, mechanisms to inhibit fatty-acid synthesis have been suggested as possible therapeutic tools to slow cancer progression. A number of studies have shown that inhibition of FAS results in significant anti-tumour activity, including growth inhibition and increased apoptosis [135–138]. In the liver, FAS gene expression is negatively regulated by AMPK [139,140] via suppression of SREBP1 (sterol regulatory element-binding protein 1) . Activation of AMPK in cancer cells, therefore, would be predicted to decrease FAS expression and reduce tumour growth. Consistent with this hypothesis, treatment of a prostate-cancer-derived cell line (LNCaP) with AICAriboside decreased FAS protein expression induced by the synthetic androgen, methyltrienolone (R1881) . In parallel, cell growth was significantly reduced following incubation with AICAriboside. In addition to FAS, ACC (acetyl-CoA carboxylase) provides another target for AMPK in the regulation of fatty-acid synthesis in cancer cells. In human breast cancer cell lines it was found that down-regulation of ACC1 (also referred to as ACCα) caused a reduction in fatty-acid synthesis and led to a concomitant increase in apoptosis . AMPK phosphorylates and inhibits ACC1 and in lipogenic tissues, such as the liver, this provides an important mechanism for regulating the rate of fatty-acid synthesis .
The evidence presented so far would imply that AMPK acts as a tumour suppressor and that activation of AMPK would be advantageous in treating cancers (see Figure 8). Indeed although the effects of metformin to lower hepatic glucose output now appear to be independent of AMPK activity , its ability to activate AMPK and restrict cell proliferation may contribute towards the beneficial effect of metformin on reducing the lifetime risk of developing cancer [146,147]. More recent studies have also shown that metformin has anti-tumour effects in breast cancer . However, it has also been reported recently that metformin can inhibit mTORC1 signalling by an AMPK-independent mechanism and that this could account for some of the beneficial effects of metformin in cancer progression .
AMPK as a tumour promoter
Most of the studies carried out to date lend weight to the idea that activation of AMPK leads to a reduction in cell growth and proliferation. Coupled with its inhibitory role in fatty acid synthesis, activation of AMPK appears as a good candidate for anti-tumour therapy. However, challenging this view, a number of studies have emerged that lead to the opposite conclusion, namely that inhibition of AMPK may reduce cancer cell growth. It was reported that AMPK activity, assessed by monitoring ACC phosphorylation, was increased in malignant human prostate tissue compared with normal tissue cells . Moreover, down-regulation of AMPK in prostate cancer cell lines using siRNA reduced cell proliferation, and treatment with a non-selective AMPK inhibitor led to increased apoptosis . These findings provided the first evidence that AMPK activity might be required for prostate cancer cell growth and survival. In another study, also using a prostate cancer line, inhibition of AMPK was shown to sensitize cells to FAS-induced apoptosis by a mechanism involving activation of caspases and down-regulation of cFLIP (CASP8 and FADD-like apoptosis regulator) . Interestingly, in this study the authors showed that activation of AMPK induced by Fas was blocked by STO-609, an inhibitor of CaMKKβ. This finding may link to recent reports implicating CaMKKβ in prostate cancer. Gene expression profiling studies revealed that CaMKKβ expression is elevated in prostate cancers and that CaMKKβ mRNA is increased in response to androgen receptor stimulation [152,153]. Consistent with these findings, AMPK activity was increased in prostate cancer cell lines in response to R1881, via an androgen receptor- and CaMKKβ-dependent mechanism [152,153]. Moreover, inhibiting the CaMKKβ/AMPK pathway was shown to block androgen-stimulated cell migration and invasion  and cell growth . Taken together, these studies suggest that, at least in prostate cancer, inhibition of AMPK would be an advantage in treating tumour growth. How these studies can be reconciled with those of the study reporting that AICAriboside reduced cell growth and decreased FAS protein expression induced by R1881  is unclear.
In a recent study aimed at identifying novel downstream targets of AMPKα2 complexes, using an ingenious genetic chemical screen, several new targets were identified that are involved in mitotic cell division . These included PAK2, a kinase involved in cytoskeletal reorganization, and PPP1R12C (PPP1 regulatory subunit 12C). A potential convergence point for PAK2 and PPP1R12C is at the regulation of MLC (myosin regulatory light chain) phosphorylation. Prompted by this possibility, the authors studied the significance of their phosphorylation by AMPK, and in particular whether this had any bearing on MLC phosphorylation. Previous studies have demonstrated that PAK2 phosphorylates MLC at Ser19 . AMPK phosphorylated PAK2 on Ser20, and although there was no direct effect of AMPK phosphorylation on PAK2 activity in vitro, mutation of Ser20 to an alanine residue reduced phosphorylation of MLC in U2OS cells in response to treatment with a direct small molecule activator of AMPK, A769662 . PPP1R12C (also referred to as MBS85) is a member of the MYPT (myosin phosphatase-targeting) family that interacts with the catalytic subunit of PP1β (also known as PP1δ) to confer substrate specificity and subcellular localization of the myosin phosphatase complex [156,157]. Although little is known about the physiological role of PPP1R12C, AMPK was found to phosphorylate it on Ser452. PPP1R12C interacts with 14-3-3 and mutation of Ser452 to an alanine residue abolished this interaction, but had no effect on interaction of PPP1R12C with PP1β . Activation of AMPK in HEK-293T cells increased the interaction between PPP1R12C and 14-3-3, whereas the combined knockdown of AMPKα1 and α2 subunits decreased this interaction . A previous study reported phosphorylation of PPP1R12A (also known as MYPT1) by NUAK1 (AMP-related kinase 5) . Phosphorylation of PPP1R12A at Ser445, Ser472 and Ser910 by NUAK1 promoted its interaction with 14-3-3. In this case it was shown that binding of 14-3-3 to the PPP1R12A–PP1β complex inhibited its ability to dephosphorylate MLC2 in vitro . Interestingly, however, although AMPK was capable of phosphorylating PPP1R12A in vitro at the same sites as NUAK1, the authors were unable to find any evidence to support a role of AMPK in regulating PPP1R12A–PP1β in cells . It will be important to determine whether NUAK1, or other AMPK-related kinases, can phosphorylate PPP1R12C and if so, to determine the physiological relevance of the different upstream signalling pathways in regulation of PP1β activity.
Previous studies have demonstrated that phosphorylation of MLC plays a role in the regulation of many diverse functions, including cell shape and polarity, cell migration, cell adhesion and cell cycle [156,157,159]. Phosphorylation of MLC on Ser19, and in some cases Thr18, leads to activation of myosin II and increased contractility . The finding that AMPK has the potential to co-ordinately regulate both a kinase and a phosphatase involved in MLC phosphorylation led Banko et al.  to investigate the possibility that AMPK might play a role in mitosis. Altering AMPK activity, or expression of a mutant form of PPP1R12C harbouring a S452A mutation, in synchronized cells led to an increase in multinucleated cells. Expression of PAK2 harbouring a S20A mutation, however, had no effect. The idea that AMPK might be involved in mitosis is not new, as previous lines of evidence had implicated a role for AMPK in cell-cycle regulation. Drosophila embryos lacking AMPK show defects in both mitosis and polarity, and these effects appear to be due to decreased phosphorylation of MLC . In this case, however, AMPK was proposed to directly phosphorylate MLC . Active AMPK has been proposed to localize to the centrosome and the central spindle midzone during mitosis . However, the evidence for this was entirely on the basis of immunological techniques, using antibodies directed against phosphorylated Thr172. These antibodies can cross-react with some of the AMPK-related kinases, and so without control staining in cells lacking AMPK these results should be interpreted with caution.
A number of unresolved issues remain regarding any putative role for AMPK in mitosis. It seems counterintuitive to imagine that AMPK would lead to increased mitosis during conditions of low energy supply, and indeed glucose starvation, which is associated with activation of AMPK, did not alter mitotic progression . Perhaps AMPK only plays a role in mitosis in certain cells under certain conditions. There are several isoforms of the myosin phosphatase targeting subunits, and there is evidence that in intact cells AMPK does not phosphorylate all of the isoforms . It is possible, therefore, that AMPK only plays a role in mitosis in cells expressing specific targeting subunits e.g. PPP1R12C. On this point, it is worthwhile noting that in smooth muscle cells, activation of AMPK has been reported to phosphorylate and inhibit myosin light chain kinase leading to attenuation of MLC phosphorylation . Furthermore, it is clear from the available knockout mouse models that tissue-specific deletion of AMPK activity does not lead to obvious cell-cycle defects.
At this juncture, it is difficult to reconcile the apparently opposing effects of AMPK on cell growth other than to speculate that the downstream consequences of AMPK activation could alter dramatically depending on the cellular context under which it operates. In normal non-cancerous cells, activation of AMPK increases energy supply, but at the same time has a negative effect on cell growth and proliferation, e.g. by antagonizing mTOR activity. In certain cancer cells, e.g. prostate, increased AMPK activity could provide the cells with an advantage by increasing the rate of glucose uptake and glycolysis , increasing mitosis , and increasing cell migration  (see Figure 9). Unlike normal cells, however, the cancer cell has somehow adapted to dampen the AMPK signalling pathways that would normally lead to reduced cell growth. How this adaptation might occur is not clear, but it should be possible to devise studies to test the hypothesis. It is intriguing that in many tumours LKB1 is inactivated, whereas in prostate cancer, CaMKKβ expression is increased. Is it possible that in some cancers there is a switch from LKB1- to CaMKKβ-mediated activation of AMPK? If so, could this switch to CaMKKβ signalling be important for altering the downstream response of AMPK? Despite considerable effort over the last few years, it is clear that more research is required to determine the role of AMPK in cancer. It will be important to examine the role of AMPK in different types of cancer and at different stages of cancer progression. Only then will it be possible to contemplate future cancer therapies based around modulation of AMPK activity.
AMPK AS A DIRECT THERAPEUTIC TARGET: EVIDENCE FROM IN VIVO STUDIES
Given the wide range of metabolic targets downstream of AMPK, it is perhaps not surprising that AMPK has been considered a promising therapeutic target. Although there is no direct evidence that AMPK will prove to be a beneficial target in humans, the number of patents describing potential AMPK activators is growing rapidly .
Pharmacological activators of AMPK can be divided into two groups: indirect AMPK activators [biguanides such as metformin and phenformin, TZDs (thiazolidinediones) such as rosiglitazone and plant-derived compounds such as resveratrol] and direct AMPK activators (A-769662, ZMP, PT-1 and OSU-53) . A plant-derived compound, Honokiol, has been described as an in vivo activator of AMPK, protecting neurons from injury in an animal model of cerebral ischaemia, but as yet there is no information on whether this compound acts as a direct or indirect activator of AMPK . Apart from ZMP, which is an AMP mimetic, the mode of action of other direct AMPK activators remains enigmatic. Any compound that increases the ADP/ATP and/or the AMP/ATP ratio within a cell will cause activation of AMPK, and the majority of the indirect activators of AMPK appear to work via this mechanism, e.g. phenformin. In addition to AMPK, there are likely to be many other enzymes whose activity is altered by a reduction in cellular energy charge, and so determining whether the effect of a particular indirect AMPK activator is mediated by AMPK may not be straightforward. For instance, it now seems clear that the ability of metformin to reduce hepatic glucose output does not require AMPK, even though metformin causes a robust activation of AMPK in hepatocytes . Ideally, in order to demonstrate that AMPK is required for specific downstream effects of a potential AMPK activator, studies showing that the effect of the compound is abolished in cells lacking AMPK activity should be included. Without this evidence, we would urge caution when interpreting a role for AMPK in the effects of indirect activators.
Another mechanism by which indirect activators of AMPK could operate is via an increase in intracellular calcium and subsequent activation of CaMKKβ. In a recent study, low concentrations of resveratrol were reported to inhibit PDE (phosphodiesterase) activity , leading to an accumulation of cAMP. The increase in cAMP led to an increase in the activity of Epac1 (exchange protein directly activated by cAMP 1), a cAMP-regulated guanine nucleotide-exchange factor [167,168]. Previous independent studies have shown that both activation of Epac1 [169,170] and resveratrol [171,172] lead to an increase in cytosolic calcium. An attractive possibility, therefore, is that resveratrol increases cytosolic calcium via inhibition of PDE, accumulation of cAMP and activation of Epac1 . One of the expected consequences of increased calcium would be activation of CaMKKβ leading to phosphorylation and activation of AMPK. Interestingly, in another study resveratrol was shown to activate AMPK as a result of ATP depletion . In this case, however, the concentrations of resveratrol used were higher than those used in the study reporting inhibition of PDE, suggesting that indirect activators may lead to AMPK activation through multiple pathways.
In a recent genome-wide association study, a single nucleotide polymorphism, rs11212617, was identified that was associated with the efficacy of metformin treatment in Type 2 diabetes . One of the genes in this locus is ATM (ataxia telangiectasia mutated), a protein kinase activated by DNA damage which is mutated in ataxia telangiectasia, a neurodegenerative disease linked with predisposition to cancer . A selective ATM inhibitor (KU55933) reduced the activation of AMPK by metformin, suggesting a role for ATM in the activation of AMPK . However, this view has now been challenged [175,176]. KU55933 was shown to reduce cellular uptake of metformin independently of ATM, providing an explanation for its effect on AMPK activation by metformin, independent of the requirement for ATM [175,176].
CONCLUSIONS AND PERSPECTIVES
AMPK plays a role in sustaining cellular energy levels, and recent studies suggest that ADP may be the universal trigger that signals activation of AMPK in eukaryotic cells. Given that virtually every energy-requiring reaction in the cell uses the energy derived from ATP hydrolysis, it is not so surprising that AMPK has been implicated in a wide range of biochemical pathways. In fact, at times it can seem like AMPK is involved in regulating every biochemical pathway conceivable! As a consequence of the wide range of pathways in which AMPK has been implicated, it is perhaps inevitable that in some instances apparent conflicts and paradoxes arise due to our incomplete understanding of the biology. In these cases, further studies are required to resolve the controversies. We urge caution when using pharmacological approaches to manipulate AMPK; development of well-characterized direct activators and specific inhibitors of AMPK will facilitate future studies. Transgenic mouse models are already available that allow generation of tissue- and isoform-specific AMPK knockouts and these will help with our understanding of the physiological role of AMPK. A goal for the future is to translate more of the knowledge gained from in vitro studies and animal models to human physiology and pathology. A major challenge is to identify specific modulators of AMPK activity and only then will the true potential of AMPK as a therapeutic target be realized.
Work in the authors' laboratories is funded by the MRC (Medical Research Council). C.T. is also supported by a grant from the Wellcome Trust.
Abbreviations: Aβ, amyloid β; ACC, acetyl-CoA carboxylase; AID, autoinhibitory domain; AMPK, AMP-activated protein kinase; ATM, ataxia telangiectasia mutated; BSEP, bile salt export pump; CaMKKβ, Ca2+/calmodulin-dependent protein kinase kinase β; CBM, carbohydrate-binding module; CBS, cystathionine-β-synthase; Epac1, exchange protein directly activated by cAMP 1; ERK, extracellular-signal-regulated kinase; FAS, fatty acid synthase; FIP200, focal adhesion kinase family-interacting protein of 200 kDa; GABAB, γ-aminobutyric acid type B; GAP, GTPase-activating protein; HDAC1, histone deacetylase 1; HEK, human embryonic kidney; Htt, Huntingtin; IL, interleukin; LKB1, liver kinase B1; LPS, lipopolysaccharide; MLC, myosin regulatory light chain; mTOR, mammalian target of rapamycin; mTORC, mTOR complex; MYPT, myosin phosphatase-targeting; NUAK1, AMP-related kinase 5; PAK, p21 protein (Cdc42/Rac)-activated kinase; PDE, phosphodiesterase; PPM, metal-dependent protein phosphatase; PPP, phosphoprotein phosphatase; PPP1R12C, PPP1 regulatory subunit 12C; RAC1, ras-related C3 botulinum toxin substrate 1; RAPTOR, regulatory associated protein of mTOR; RHEB, RAS homologue enriched in brain; siRNA, small interfering RNA; SRC-2, steroid receptor coactivator-2; TSC, tuberous sclerosis; ULK, Unc-51-like kinase; ZMP, AICAribotide monophosphate
- © The Authors Journal compilation © 2012 Biochemical Society