Circulation Research. 2007;100:328-341
doi: 10.1161/01.RES.0000256090.42690.05
(Circulation Research. 2007;100:328.)
© 2007 American Heart Association, Inc.
AMP-Activated Protein Kinase in Metabolic Control and Insulin Signaling
Mhairi C. Towler,
D. Grahame Hardie
From the Division of Molecular Physiology, College of Life Sciences, University of Dundee, Scotland, UK.
Correspondence to Prof D. G. Hardie, Division of Molecular Physiology, College of Life Sciences, University of Dundee, Sir James Black Centre, Dow St, Dundee, DD1 5EH, Scotland, UK. E-mail d.g.hardie{at}dundee.ac.uk
This Review is part of a thematic series on AMP Kinase, which includes the following articles:
AMP-Activated Protein Kinase in Metabolic Control and Insulin Signaling
Cardiac AMP-Activated Protein Kinase in Health and Disease
Bruce Kemp Guest Editor
 |
Abstract
|
|---|
The AMP-activated protein kinase (AMPK) system acts as a sensor
of cellular energy status that is conserved in all eukaryotic
cells. It is activated by increases in the cellular AMP:ATP
ratio caused by metabolic stresses that either interfere with
ATP production (eg, deprivation for glucose or oxygen) or that
accelerate ATP consumption (eg, muscle contraction). Activation
in response to increases in AMP involves phosphorylation by
an upstream kinase, the tumor suppressor LKB1. In certain cells
(eg, neurones, endothelial cells, and lymphocytes), AMPK can
also be activated by a Ca
2+-dependent and AMP-independent process
involving phosphorylation by an alternate upstream kinase, CaMKKß.
Once activated, AMPK switches on catabolic pathways that generate
ATP, while switching off ATP-consuming processes such as biosynthesis
and cell growth and proliferation. The AMPK complex contains
3 subunits, with the

subunit being catalytic, the ß
subunit containing a glycogen-sensing domain, and the

subunits
containing 2 regulatory sites that bind the activating and inhibitory
nucleotides AMP and ATP. Although it may have evolved to respond
to metabolic stress at the cellular level, hormones and cytokines
such as insulin, leptin, and adiponectin can interact with the
system, and it now appears to play a key role in maintaining
energy balance at the whole body level. The AMPK system may
be partly responsible for the health benefits of exercise and
is the target for the antidiabetic drug metformin. It is a key
player in the development of new treatments for obesity, type
2 diabetes, and the metabolic syndrome.
Key Words: calcium signaling diabetes insulin metabolism signaling pathways
 |
Introduction
|
|---|
Genes encoding the

, ß, and

subunits of the AMP-activated
protein kinase (AMPK) are highly conserved in all eukaryotic
species in which genome sequences have been completed, including
vertebrates and invertebrates, plants, fungi, and protozoa.
1 Genetic studies show that in the yeast
Saccharomyces cerevisiae,
the genes encoding these subunits are required for the response
to glucose starvation.
2,3 In a primitive green plant, the moss
Physcomitrella patens, genes encoding the catalytic subunits
are required for growth in alternate light/dark cycles, as opposed
to continuous light.
4 Darkness represents a period of starvation
for a green plant because it is unable to produce carbohydrate
by photosynthesis. The AMPK system therefore appears to have
initially evolved to execute responses to carbon starvation.
Because of the sophisticated endocrine systems that control
glucose homeostasis in mammals (including the effects of insulin),
starvation for glucose is not a normal physiological event for
mammalian cells in vivo. Nevertheless, AMPK is still ubiquitously
expressed in mammalian cells, where it is involved in the response
to a variety of metabolic stresses that disturb cellular energy
homeostasis. More recently, it has been realized that hormones
and other extracellular signals have acquired the ability to
modulate the AMPK system, and it is now known to be involved
in regulating energy homeostasis at the whole body, as well
as the cellular, levels.
 |
Regulation of the AMPK Complex by 5'-AMP and Calcium
|
|---|
With hindsight, the first reports describing the AMPK system
were 2 independent articles published in 1973 involving poorly
defined protein fractions that, in the presence of ATP, inactivated
2 key metabolic enzymes involved in lipid synthesis, ie, acetyl-CoA
carboxylase (ACC) (involved in fatty acid synthesis)
5 and 3-hydroxy-3-methylglutaryl-CoA
(HMG-CoA) reductase (involved in isoprenoid/cholesterol synthesis).
6 The protein fractions were correctly surmised to contain protein
kinases, and subsequent studies revealed that both the ACC kinase
and the HMG-CoA reductase kinase were stimulated by 5'-AMP.
7,8 However, it was not realized that both functions were performed
by the same protein kinase until our laboratory provided evidence
in favor of that hypothesis in 1987
9; we renamed the activity
AMP-activated protein kinase the following year.
10
As well as allosterically activating the enzyme by up to 5-fold,9 AMP also promotes its phosphorylation11 at a specific threonine residue on the
subunit (Thr17212) by an upstream kinase that has recently been identified as a complex between the tumor suppressor protein LKB1 and 2 accessory subunits, termed STRAD and MO25.13,14 Phosphorylation of Thr172 produces at least 100-fold activation,15 so that it is quantitatively much more important than the allosteric activation. The LKB1 complex is not itself activated by AMP, and the effect of the nucleotide is to make AMPK a better substrate for LKB1,13 while at the same time making it a worse substrate for protein phosphatases that dephosphorylate Thr172.16 The 3 effects of AMP make the system very sensitive to small increases in AMP.17 All 3 effects are also antagonized by high concentrations of ATP. Because all eukaryotic cells express very active adenylate kinases, which maintains their reaction (2ADP
AMP+ATP) close to equilibrium at all times, the cellular AMP:ATP ratio varies approximately as the square of the ADP:ATP ratio,18 making it a very sensitive indicator of cellular energy status.
In cells lacking LKB1, such as HeLa cells (tumor cells in which LKB1 is not expressed), there is still some basal phosphorylation of Thr172 and AMPK activity,13 and both can be dramatically increased by addition of a Ca2+ ionophore. Three groups have now identified the upstream kinase responsible for phosphorylation of Thr172 under these circumstances as Ca2+/calmodulin-dependent protein kinase kinase (CaMKK), especially the CaMKKß (CaMKK2) isoform.1921 CaMKKs were originally identified as protein kinases that acted upstream of calmodulin-dependent protein kinases I and IV.22 Our laboratory showed that a purified CaMKK could phosphorylate and activate AMPK in cell-free assays as long ago as 1995,11 although at that time, we did not consider that this was likely to have any physiological relevance. CaMKKß is expressed primarily in the brain but is also expressed in testis, thymus, and T cells.23 Our laboratory has shown that when Ca2+ enters neurones in rat brain slices following K+-induced depolarization, there is a marked phosphorylation of Thr172 and activation of AMPK that is catalyzed by a CaMKK.
Given the limited tissue distribution of the CaMKKs compared with LKB1, the Ca2+-mediated pathway may be restricted to certain cell types such as neurones, although additional examples are discussed below. However, it is interesting to speculate about the function of the CaMKK
AMPK pathway in those cell types in which it occurs. Any treatment that causes an increase in cytoplasmic Ca2+ will create a subsequent demand for ATP, if only because the Ca2+ is immediately pumped out of the cytoplasm using ATP-driven pumps in the plasma membrane and endoplasmic reticulum. Activation of AMPK under these circumstances may represent a mechanism to anticipate the demand for ATP created by Ca2+ entry.
 |
Subunit Structure of the AMPK Complex
|
|---|
AMPK is a heterotrimer comprising

, ß, and

subunits.
There are 2 or 3 genes encoding each subunit (
Figure 1 and
Table 1),
giving rise to 12 possible heterotrimeric combinations, with
splice variants further increasing the potential diversity.
The 2 isoforms of the

subunit,
124 and
2,
25 contain the kinase
domain in their N-terminal half, with the C-terminal regions
being required to form a complex with the ß and

subunits.
26 They appear to have rather similar substrate specificities,
27 but the
2 isoform is enriched in the nucleus of several cell
types, including pancreatic ß cells,
28 neurones,
29 and skeletal muscle,
30 whereas
1 is predominantly cytoplasmic.
The
1 isoform is associated with the plasma membrane in carotid
body type 1 cells
31 and airway epithelial cells,
32 in the latter
case particularly with the apical membrane.

View larger version (25K):
[in this window]
[in a new window]
|
Figure 1. Domain structure of AMPK subunit isoforms and splice variants. Regions shown in the same color are related, and their functions, where known, are indicated.
|
|
The ß subunits from different eukaryotic species contain 2 conserved regions, located in the central and C-terminal regions (Figure 1).33 It is now clear that the C-terminal domain is all that is required to form a functional
ß
complex that is regulated by AMP,34 whereas the central conserved region is recognized to be a glycogen-binding domain.34,35 The crystal structure of the glycogen-binding domain of ß1 was recently determined,36 but the functions of this domain remain unclear, although it is present in the ß subunits of all eukaryotic species.
The
subunits (
1,
2, and
3) contain variable N-terminal regions followed by 4 tandem repeats of a 60-aa sequence termed by Bateman as a CBS motif.37 It is now known that these act in pairs to form 2 domains (now termed Bateman domains), each of which binds one molecule of AMP (Figure 1).38 Mutations in these domains reduce both AMP binding and AMP activation, proving that these are the regulatory binding sites for the nucleotides.38 The Bateman domains also antagonistically bind ATP with a lower affinity than AMP,38 consistent with previous findings that high concentrations of ATP oppose activation of the AMPK complex by AMP.39 The discovery that there were 2 binding domains for AMP or ATP was unexpected. The tandem domains bind AMP with a high degree of cooperativity38 suggesting that the second site is inaccessible until AMP has bound at the first. This represents yet another potential mechanism to increase the sensitivity of the AMPK system to small changes in AMP.
 |
Regulation of AMPK by Metabolic Stress
|
|---|
Any metabolic stress that inhibits ATP production, or that accelerates
ATP consumption, will tend to increase the cellular ADP:ATP
ratio, which will be amplified by adenylate kinase into a much
larger increase in the AMP:ATP ratio, with consequent activation
of AMPK. 2-Deoxyglucose, which depletes ATP by being converted
to the nonmetabolizable 2-deoxyglucose-6-phosphate, has sometimes
been used to activate AMPK.
40 Not surprisingly, AMPK can also
be activated by metabolic poisons that inhibit ATP production
via oxidative phosphorylation, such as arsenite, oligomycin,
antimycin A, azide, or dinitrophenol,
4143 or by ischemia
44 or hypoxia.
45 Although the effect of ischemia is relevant to
cardiovascular disease, these are all rather nonphysiological
stresses. However, a more physiological stress that activates
AMPK in skeletal muscle is exercise or contraction,
4648 and, as discussed further below, it seems likely that AMPK activation
underlies many of the health benefits of regular exercise. Although
there is currently no satisfactory method to measure changes
in AMP during exercise in vivo, electrical stimulation of mouse
hindlimb muscle in situ does cause small increases in the cellular
AMP:ATP ratio.
49 This is almost certainly responsible for the
accompanying activation of AMPK, because no activation occurs
if the same experiment is performed in mice with a muscle-specific
knockout of the upstream kinase LKB1,
49 which is known in other
cells to be required for the AMPK system to respond to changes
in AMP.
13 Intriguingly, the changes in AMP:ATP ratio in response
to electrical stimulation are significantly larger in the muscles
from the knockout mice,
49 confirming the proposed role of AMPK
in maintenance of cellular energy homeostasis.
In many tissues and organs, such as the heart, the AMPK system may be significantly activated by hypoxia only during pathological ischemic episodes, but there is evidence that, in certain specialized oxygen-sensing cells, the system is important in the response to more physiological variations in oxygen tension. These include the glomus cells in the carotid body, which sense the oxygen level in arterial blood supplying the brain, and also pulmonary artery smooth muscle cells.50 In glomus cells, hypoxia causes opening of voltage-gated Ca2+ channels and consequent entry of extracellular Ca2+, leading to release of neurotransmitters that cause firing of afferent neurones, triggering increased breathing. In pulmonary artery smooth muscle, hypoxia triggers release of Ca2+ from intracellular stores, causing contraction of the muscle. This contrasts with the behavior of the smooth muscle in other arteries, which relax in response to hypoxia to supply more blood to the hypoxic tissue. However, this unusual behavior makes physiological sense in the pulmonary circulation, because it ensures that blood flow is diverted away from poorly oxygenated areas of the lung in favor of well-oxygenated areas. In collaboration with the groups of Evans and Peers, our laboratory has recently shown that AMPK is activated by changes in AMP:ATP that occur in response to hypoxia in pulmonary artery smooth muscle.31 Unfortunately, glomus cells are too small to allow bulk measurements of AMP and ATP using the same methodology. However, AMPK-activating drugs (5-aminoimidazole-4-carboxamide riboside [AICAR] or phenformin; see below) precisely mimic all of the effects of hypoxia in both the carotid body glomus cell and in pulmonary artery smooth muscle, suggesting that hypoxia is detected by AMPK as it senses the changes in AMP:ATP ratio caused by inhibition of oxidative phosphorylation in these cells.31
As mentioned in the introduction, in lower eukaryotes such as the yeast S cerevisiae, the primary role of the AMPK homolog seems to be in the response to glucose starvation. Most mammalian cells express glucose transporters and hexokinase isoforms that have a very low Km for glucose, so that the rate of ATP synthesis from glucose only drops when blood glucose falls to pathologically low levels. However, just as there are specialized oxygen-sensing cells in the carotid body and pulmonary arteries, so too are there specialized glucose-sensing cells, including the ß cells in pancreatic islets that release insulin and neurones in the hypothalamus of the brain that control feeding behavior. These cells express isoforms of glucose transporter (GLUT2) and hexokinase (hexokinase IV, glucokinase) with a much higher Km for glucose, so that the rate of ATP synthesis falls in response to decreases in blood glucose within the physiological range. Consequently, AMPK in these cells is regulated by physiological variations in blood glucose. Thus, in cell lines derived from rodent pancreatic ß cells, AMPK is activated by low glucose and inhibited by high glucose,51,52 whereas, in fasted mice, intracerebroventricular injection of glucose, or refeeding of the mice, inhibits the
2 isoform of AMPK in the hypothalamus.53 Moreover, the known downstream consequences of lowering external glucose in both cell types, ie, decreased insulin secretion (ß cells) and increased feeding behavior (hypothalamus), can be mimicked by activation of AMPK at these sites, either by pharmacological or molecular biological interventions.5155 Thus, an ancient signaling pathway involved in the response to glucose starvation in single-celled eukaryotes responds to the same signal in multicellular eukaryotes, but its role has been adapted to control complex processes such as hormone release and feeding behavior.
 |
Regulation of AMPK by Hormones, Cytokines, and Other Extracellular Ligands
|
|---|
The AMPK system is found in all present day single-celled eukaryotes
and is therefore likely to have evolved long before multicellular
organisms, and the hormones and cytokines that help to integrate
their functions, had developed. Nevertheless, recently it has
become clear that hormones and cytokines have learnt how to
exploit the cell signaling capabilities of the AMPK system.
The first hints of this came with findings that AMPK complexes
in CHO cells stably expressing receptors coupled via the G protein
Gq, such as those for platelet-activating factor, epinephrine
(
1b), or bradykinin (B
2), were activated by stimulation of the
receptor with the appropriate ligand.
56 Because Gq couples these
receptors to phospholipase C, triggering production of inositol-1,4,5-trisphosphate
and Ca
2+ release, it seems likely that the activation of AMPK
in these systems is triggered by phosphorylation by CaMKKß,
although this has not yet been demonstrated conclusively. However,
in human umbilical cord vein endothelial cells Heller and colleagues
57 have provided evidence that thrombin (also known to act through
a Gq-coupled receptor) stimulates AMPK through the Ca
2+-mediated
pathway. This pathway also appears to operate in other cells
of the endothelial/hematopoietic lineage, such as T lymphocytes.
With Cantrell and colleagues, we have recently shown that AMPK
is dramatically but transiently activated when quiescent T cells
are stimulated via the antigen receptor, peaking at 1 minute
and returning to baseline by 60 minutes. A variety of evidence
suggest that this activation is mediated by the Ca
2+/CaMKKß
pathway.
58 Quiescent T cells have a very limited energy turnover
but when stimulated, they need to generate a considerable amount
of ATP for the rapid growth and proliferation that follows.
We speculate that activation of AMPK allows them to rapidly
switch on catabolic, ATP-generating processes to anticipate
this demand. However, there is now much evidence that a high
AMPK activity also represses cell growth and proliferation (see
below); therefore, it may be necessary for AMPK activity to
return to baseline values before T cell proliferation can commence.
Considerable excitement was generated by the discovery in the 1990s of adipokines (cytokines released by adipocytes) that control whole body energy homeostasis, especially leptin and adiponectin. Leptin, the product of the obese gene (ob), appears to represent a signal that stores of fat are adequate, and exerts a feedback inhibition of food intake via effects on hypothalamic neurones that express the leptin receptor.59 However, as well as inhibiting energy intake, leptin also stimulates energy expenditure by promoting uptake and oxidation of glucose60 and fatty acids61 in skeletal muscle. Kahn and colleagues62 reported that the
2 isoform of AMPK was activated in muscle by leptin injection in vivo in rats, accounting for its ability to stimulate these pathways. There appeared to be 2 effects: a rapid effect that was a direct effect on the muscle itself; and a longer-term effect that appeared to be a primary action of leptin on the hypothalamus, with a secondary effect on muscle being mediated by the sympathetic nervous system. Remarkably, the same group subsequently provided evidence53 that leptin inhibited AMPK-
2 activity in the arcuate and paraventricular regions of the hypothalamus of fasted mice. Along with leptin, other known anorexigenic agents, including insulin and a melanocortin receptor agonist, inhibit AMPK-
2 in the paraventricular region, whereas agouti-related protein, an orexigenic agent, increased
2 activity in the same region in fed mice. Other orexigenic agents, such as the gut hormone ghrelin55 and the cannabinoids,63 have also been reported to activate AMPK in the hypothalamus. Inhibition of AMPK appears to be necessary for the anorexigenic effects of leptin, because adenoviral expression of an activated form of AMPK in the hypothalamus blocked the effects of leptin.53
AMPK also appears to mediate many of the effects of adiponectin. This adipokine (also known as ACRP3064) was discovered as a protein secreted into the blood by adipocytes, although, paradoxically, its plasma concentration is reduced in obese rodents and humans.65,66 AMPK has been reported to be activated by adiponectin in skeletal muscle, where it stimulates glucose uptake and fatty acid oxidation, and in liver, where it inhibits expression of gluconeogenic genes.67,68 Experiments with liver-specific expression of dominant negative mutants67 and knockouts69 show that liver AMPK, and especially the
2 isoform, is necessary for the ability of adiponectin to lower blood glucose.
 |
AMPK-Activating Drugs and Identification of Metabolic Pathways Regulated by AMPK
|
|---|
Many of the metabolic pathways regulated by AMPK were originally
identified by incubating cells, or injecting whole animals,
with the adenosine analog AICAR. AICAR is taken up into cells
by adenosine transporters
70 and converted by adenosine kinase
71 to the phosphorylated nucleotide ZMP (5-aminoimidazole-4-carboxamide
ribonucleoside), which mimics all 3 effects of AMP on the AMPK
system.
39 Another pharmacological tool for AMPK activation became
available with the discovery that the antihyperglycemic drug
metformin, currently prescribed to more than 120 million people
with type 2 diabetes worldwide, activates AMPK both in intact
cells and in vivo.
72 A recent mouse gene targeting study involving
a liver-specific knockout of LKB1, which prevents activation
of AMPK by metformin in that tissue, suggested that activation
of liver AMPK is required for the plasma glucose-lowering effects
of metformin.
73 Thus, the major effect of metformin seems to
be inhibition of liver gluconeogenesis. Although one might expect
that metformin would also activate AMPK and stimulate glucose
uptake in skeletal muscle and other tissues, this may not occur
in vivo because the concentrations the drug reaches in the peripheral
circulation (as opposed to the portal vein, which supplies the
liver direct from the gut) may not be sufficiently high to yield
significant AMPK activation in those tissues. However, metformin
can be used to activate AMPK in muscle and other cells in culture
or ex vivo. The related biguanide drug phenformin, which is
no longer used for treatment in humans because of complications
resulting from lactic acidosis, is a more rapid and potent activator
of AMPK in cultured cells.
13
The use of AICAR and the biguanides has been extremely useful in the AMPK field, but like any pharmacological tools, results obtained using them must be interpreted with care. For example, the active derivative of AICAR, ZMP, also modulates other AMP-sensitive enzymes, such as glycogen phosphorylase74 and fructose-1,6-bisphosphatase.71 AICAR, itself, can compete with released adenosine for reuptake into cells, and if this allows adenosine to accumulate in the medium this can exert effects via adenosine receptors.70 In the case of biguanides, they do not directly activate AMPK in cell-free assays42 but instead appear to inhibit complex I of the respiratory chain,75,76 leading to an increase in cellular AMP:ATP ratio.19 Effects of biguanides could therefore be secondary effects of ATP depletion, rather than effects of AMPK activation per se. As an excellent example of these problems, Hue and colleagues have recently shown that effects of AICAR and metformin on the function of glucokinase in hepatocytes are still observed in cells from knockout mice lacking both AMPK catalytic subunits.77 Another example is provided by findings that, although AICAR was originally reported to inhibit autophagy in isolated hepatocytes,78 experiments involving expression of dominant negative mutants, or the use of compound C, now suggest that AMPK activation stimulates this process.79 Compound C, an AMPK inhibitor developed by Merck72 that has recently become commercially available, can be used as an AMPK inhibitor in intact cells, although its selectivity for AMPK remains uncertain.
These considerations highlight the increasing importance of confirming that the effects of pharmacological agents are indeed mediated by AMPK using molecular biological methods, such as the expression of constitutively active or dominant negative AMPK mutants,80,81 small interference RNAs targeted at AMPK or, best of all, AMPK or LKB1 knockouts.77 Another important approach is to identify the direct target for AMPK that mediates the effect on the pathway or process, identify the sites phosphorylated on that target, and confirm that phosphorylation of those sites correlates with AMPK activation. The use of Western blotting with phospho-specific antibodies is the usual method to achieve this, and phospho-specific antibodies against Thr172 on the AMPK-
subunits are often now used as a surrogate for measuring AMPK activity. An even more convincing approach is to show that the effect is lost in knock-in mice or cells expressing the target protein mutated at the AMPK sites. This method was first demonstrated in a landmark study of HMG-CoA reductase phosphorylation published in 1993.40
Although there are very few pathways thought to be regulated by AMPK for which all of these criteria have been met, those with reasonably convincing evidence for an effect of AMPK are listed in Table 2. In general, AMPK switches on catabolic pathways that generate ATP (eg, glucose uptake, glycolysis) while switching off ATP-consuming anabolic pathways (fatty acid, cholesterol, glycogen, and protein synthesis). Figure 2 summarizes the major effects of AMPK activation on glucose and lipid metabolism in liver, muscle, and adipose tissue. As well as its acute and long-term effects on metabolism, for which it is best known, it is becoming increasingly clear that AMPK is involved in regulating other cellular processes. For example, if cells are running short of ATP, it does not make sense for them to grow and proliferate. AMPK inhibits cell growth and proliferation by inhibiting both the activity and expression of biosynthetic enzymes involved in lipid, carbohydrate, and protein synthesis, and also by switching off the target-of-rapamycin (TOR)
S6 kinase I pathway,8284 which appears to be crucial in the regulation of cell growth and cell size. AMPK appears to achieve this by phosphorylation of tuberous sclerosis complex-2 (TSC2) (tuberin), an upstream regulator of TOR.85 AMPK activation also halts progress through the cell cycle at the G1
S phase transition, which is associated with accumulation of the tumor suppressor p53 and of the cyclin-dependent kinase inhibitors p21 and p27, which act downstream of p53.8689 Under these conditions, p53 becomes phosphorylated on Ser15,86,88 although it is not clear that this is a direct phosphorylation. Another mechanism potentially contributing to cell cycle arrest is that AMPK activation reduces the cytoplasmic: nuclear ratio of the RNA-binding protein HuR, reducing its ability to stabilize mRNAs encoding critical cell cycle regulators, including cyclins.90 AMPK activation also has complex effects on apoptosis. In some situations, it prevents the process,9194 whereas in others it induces it.95100 Successful apoptosis requires ATP, and an explanation for these conflicting findings may be that AMPK activation sometimes allows cells to recover from an energy crisis that would otherwise have triggered cell death. However, if the cells go past the "point of no return," AMPK may become proapoptotic.

View larger version (22K):
[in this window]
[in a new window]
|
Figure 2. Major effects of AMPK activation on glucose and lipid metabolism in liver, muscle, and adipose tissue. Pathways stimulated by AMPK are shown with thick arrows, those inhibited by thin arrows with thick bars across them. Effects of AMPK on pyruvate oxidation are mediated by upregulation of mitochondrial biogenesis, whereas effects on fatty acid oxidation are mediated by both phosphorylation of ACC2 and activation of fatty acid entry into mitochondria, as well as upregulation of mitochondrial biogenesis.
|
|
AMPK also has a role in cell senescence. Increased cellular AMP:ATP and AMPK activity appear to contribute to the senescence observed in human fibroblasts after a high number of passages in culture.101 Intriguingly, this is consistent with studies of life span in the nematode worm, Caenorhabditis elegans. Senescence in the worms is associated with increases in cellular AMP:ATP, and environmental stresses such as starvation or heat shock that increase cellular AMP:ATP, or mutations that inactivate genes involved in insulin signaling such as decay accelerating factor-2 (encoding the ortholog of the insulin/insulin-like growth factor-1 receptor), extend life span only in worms expressing AAK-2, which is 1 of the 2 catalytic subunits of AMPK in C elegans.102 This raises the intriguing possibility that AMPK may also be implicated in the increase in life span induced by caloric restriction in mammals.
 |
Direct Downstream Targets for AMPK and Target Recognition
|
|---|
The number of direct physiological targets identified for AMPK
is now approaching 20 (
Table 3). However, given the critical
importance of maintaining energy balance at both the cellular
and whole body levels, it seems likely that this may eventually
run into hundreds. How does AMPK recognize its downstream targets?
Figure 3 shows an alignment of 14 sequences containing sites
phosphorylated by AMPK in cell-free assays, all of which are
also believed to be targets in vivo. In every case, it is a
serine residue that is phosphorylated, although threonine can
be phosphorylated by AMPK in synthetic peptide substrates.
103 In most cases, there are bulky hydrophobic residues (shown in
bold) at positions P-5 and
P+4 (ie, 5 residues N terminal and
4 residues C terminal to the phosphorylated serine), and at
least 1 basic residue (underlined) at either P-4 or P-3. The
importance of this motif (

-[ß,X]-X-X-S/T-X-X-X-

, where

is hydrophobic and ß is basic) was initially confirmed
using variant synthetic peptides substrates.
103,104 Understanding
of substrate recognition was further extended by modeling the
binding of the sequence around Ser79 on rat ACC1 (still probably
the most efficient substrate) to a 3D model of the
1 kinase
domain, based on crystal structures of related protein kinases
(
Figure 4). This model was thoroughly tested by making complementary
mutations of the substrate and the kinase.
15 In the model, the
protein substrate binds in a long groove that runs from bottom
left to top right in
Figure 4, with the serine to be phosphorylated
(P-Ser) toward the right-hand end. The

phosphate of ATP (which
is buried by the "glycine-rich loop" at top left in
Figure 4)
is well positioned to transfer its

phosphate to this serine.
The side chains of the critical hydrophobic residues at P-5
(Met) and
P+4 (Leu) bind in hydrophobic pockets within this
substrate-binding groove, whereas the critical P-4 basic residue
(Arg) binds to an acidic patch (shown in red in
Figure 4, on
the left-hand side of the binding groove). In addition to the
basic residue at P-4 or P-3, many physiological targets also
have a basic residue at P-6 (
Figure 3), and mutagenesis of this
residue (His in ACC1) showed that it is an additional positive
determinant. In our model, this P-6 basic side chain binds to
another acidic patch located on the opposite side of the substrate-binding
groove from the patch that binds the P-4 side chain (
Figure 4).

View larger version (55K):
[in this window]
[in a new window]
|
Figure 3. Alignment of the consensus recognition motif for AMPK and sequences around sites believed to be phosphorylated by AMPK on physiological substrates. Sequences are from the human, although in many cases the sites were identified with enzymes from other species. The serine residues phosphorylated are shown by the arrow labeled "P" at the top, and "P-x or P+x" refers to residues x residues N or C terminal to the phosphorylated serine. In the consensus recognition motif, refers to a hydrophobic residue (M, L, I, F, or V) and ß to a basic residue (R, K, or H). Hydrophobic and basic residues believed to be important in recognition of the target sequences are in bold or underlined respectively.
|
|

View larger version (108K):
[in this window]
[in a new window]
|
Figure 4. View of a model of the polypeptide around Ser79 on acetyl-CoA carboxylase-1 (ACC1) phosphorylated by AMPK, bound to the 1 kinase domain of AMPK.38 The model was visualized and drawn using PyMOL version 0.97, with the kinase domain represented as spheres and hydrophobic residues in orange, acidic residues in red. The ACC1 peptide is shown in "cartoon" view with the side chains of selected amino acids in "stick" representation, with nitrogen atoms in blue, carbon in green and oxygen in red. Hydrogen atoms are omitted throughout. "P-x or P+x" refers to residues x residues N or C terminal to the phosphorylated serine.
|
|
As well as the crucial hydrophobic residue (Met) at P-5, ACC1 has additional hydrophobic side chains on the N-terminal side occurring every third or fourth residue, ie, at P-9, P-13, and P-16. This pattern of hydrophobic residues is also discernible in a few of the other substrates, including HMG-CoA reductase, endothelial NO synthase, TSC2, and TORC2 (transducer of regulated CREB-2) (Figure 3). In the case of HMG-CoA reductase, a crystal structure of the human protein has been determined105 and the residues from P-12 to P-5 do indeed form an
helix, with the hydrophobic residues pointing outward into the solvent. According to our model, these residues form an amphipathic
helix (labeled "P-5:P-16 helix" in Figure 4), with the hydrophobic residues that line 1 face fitting into the substrate-binding groove. Although this
helix almost certainly increases the affinity of binding between AMPK and the substrate, it is clearly not essential, because in muscle glycogen synthase, the leucine at the P-5 position is only the second amino acid in the whole protein, following cleavage of the N-terminal methionine (Figure 3). Finally, inspection of Figure 3 shows that 2 of the 13 substrates (cardiac PFK2 and endothelial NO synthase [eNOS]) have the basichydrophobicbasic motif that usually occurs at the P-6, P-5, and P-4 or P-3 positions displaced by 1 residue, ie, at P-5, P-4, and P-3. It may be significant that in both these cases the 2 basic residues at P-5 and P-3 are arginine. The side chain of arginine can probably position its positively charged nitrogen atoms with more flexibility than either lysine or histidine. It is conceivable that this, coupled with the large size of the 2 acidic patches on the kinase domain that bind the basic side chains, provides sufficient flexibility to allow a different spacing of the basic residues from the phosphoacceptor. This flexibility means that predicting phosphorylation sites for AMPK from raw sequence data are not straightforward.
Many of the direct targets for AMPK identified in Table 3 and Figure 3 are transcription factors (hepatocyte nuclear factor-4 [HNF4-
], ChREBP) or transcriptional coactivators (p300, TORC2), and it is now clear that mammalian AMPK has many effects on transcription, as has always been known to be the case for the yeast homolog. In most of the cases listed in Table 3, phosphorylation by AMPK inhibits transcription, either by inhibiting DNA binding (ChREBP,106 HNF4-
107), triggering degradation (HNF4-
107), or promoting binding of 14-3-3 proteins, thus causing cytoplasmic sequestration (TORC2108). However, AMPK activation can also increase gene expression, particularly in skeletal muscle. In a microarray study in mice, transgenic expression of a dominant negative AMPK mutant in muscle reduced AMPK activity and caused increased expression of 234 genes and decreased expression of 130 genes.109 AMPK activation has also been shown to downregulate expression of the transcription factor SREBP-1c (sterol response element binding protein-1c) in liver,72 and upregulate expression of the transcription factor peroxisome proliferatoractivated receptor-
(PPAR-
) and the coactivator PPAR-
coactivator-1
(PGC-1
)110 in muscle, although in these cases, this does not appear to be attributable to direct phosphorylation.
 |
Role of AMPK in the Vasculature
|
|---|
The role of AMPK in the heart is discussed in another article
in this issue, so will not be discussed in detail here. In endothelial
cells, AMPK activation has the same effects on metabolism as
it does in other cells. In human umbilical vein endothelial
cells, AMPK can be activated using AICAR or glucose deprivation,
and when activated it, stimulates acetyl-CoA carboxylase phosphorylation
and fatty acid oxidation.
111,112 One AMPK target that may be
particularly important in the circulatory system is the endothelial
isoform of NO synthase (eNOS). Phosphorylation of Ser1177 on
eNOS increases the
Vmax of the enzyme and its sensitivity to
Ca
2+ and calmodulin.
113 Because the classical effect of NO is
to relax the smooth muscle lining blood vessels, it is conceivable
that this is a mechanism to increase blood flow to hypoxic tissues.
Consistent with this, Salt and colleagues
114 have shown that
activation of AMPK in human aorta endothelial cells causes increased
phosphorylation of eNOS at Ser1177 and increased NO production.
AMPK is also activated by hypoxia in human umbilical vein endothelial
cells (HUVECs), and adenoviral expression of a dominant negative
AMPK mutant in those cells blocked increased eNOS phosphorylation
at Ser1177 in response to hypoxia.
115 However, AMPK activation
does not seem to account for eNOS phosphorylation in HUVECs
treated with thrombin.
57
AMPK may also be important in the process of angiogenesis. Mice with an embryonic knockout of the upstream kinase LKB1 die in mid-gestation with vascular abnormalities and abnormal expression of vascular endothelial cell growth factor.116 In addition, adenoviral expression of a dominant negative AMPK mutant in HUVECs inhibited their migration toward vascular endothelial cell growth factor and inhibited angiogenesis using in vitro and in vivo assays.115
 |
Interactions Between the AMPK and Insulin/Insulin-Like Growth Factor-1 Signaling Pathways
|
|---|
The insulin/insulin-like growth factor-1 signaling pathway is
activated when nutrients are available, whereas the AMPK pathway
is activated when cells are starved for a carbon source. One
would therefore expect these 2 pathways to oppose each other,
and this is often the case. In mammals, insulin promotes lipid,
protein, and glycogen synthesis, whereas AMPK inhibits these
biosynthetic pathways. The effect of insulin on protein synthesis
is mediated in part by activation of the TOR pathway via phosphorylation
of TSC2, whereas activation of AMPK causes phosphorylation of
different sites on TSC2 and inhibits TOR.
85,117 There also appears
to be direct crosstalk between the 2 pathways. In some tissues,
such as cardiac muscle, insulin antagonizes activation of AMPK,
118,119 and this appears to involve activation of the protein kinase
Akt (also termed protein kinase B [PKB]).
120 It has recently
been proposed that this is attributable to phosphorylation by
Akt of the AMPK

subunits on Ser485 or Ser491 (equivalent residues
on the
1 or
2 isoforms), which antagonizes AMPK activation via
phosphorylation at Thr172 by LKB1.
121 Whether this mechanism
operates in other tissues remains unclear.
In other cases, the insulin and AMPK signaling pathways work in the same direction, particularly in processes that regulate plasma glucose levels. In skeletal muscle, both insulin and AMPK activation stimulate glucose uptake by increased translocation of GLUT4 to the plasma membrane, although the subsequent fate of the glucose is different (glycogen synthesis in the case of insulin, which is anabolic; glycolysis/oxidation in the case of AMPK, which is catabolic). The 2 pathways appear to converge on the phosphorylation of AS160,122,123 a protein with a Rab-GTPaseactivating protein (Rab-GAP) domain that is involved in regulation of GLUT4 translocation. Activation of AMPK plays a major part in the well-known ability of muscle contraction to stimulate glucose uptake, because in mice overexpressing a dominant negative AMPK mutant,80 and in a skeletal musclespecific knockout of the upstream kinase LKB1,49 the effects of contraction on glucose uptake are reduced. There is also evidence that AMPK activation may be involved in the ability of a single bout of acute exercise to increase the insulin sensitivity of glucose uptake.124 The mechanism for this effect remains unclear, although a possibility is that it is attributable to the ability of AMPK to inhibit the TOR pathway because the latter, which is activated by insulin, is believed to exert a feedback regulation on insulin signaling by downregulating IRS1.125127 AMPK activation may also be involved in the effects of repeated exercise (ie, training) to improve insulin-sensitive glucose uptake, because of its ability to increase expression of GLUT4128 and perhaps other effects.
A second case in which insulin and AMPK act in the same direction occurs in the liver, in which both repress the expression of enzymes of gluconeogenesis, such as phosphoenolpyruvate carboxykinase and glucose-6-phosphatase.129 It makes obvious sense that insulin, a hormone released in response to high blood glucose, should repress hepatic glucose production, whereas in the case of AMPK it may perhaps have evolved as among its antianabolic actions. Gene repression by AMPK is thought to occur via phosphorylation and consequent cytoplasmic sequestration of the transcriptional coactivator TORC2.108 Repression of gluconeogenesis appears to be the major cause of the plasma glucose-lowering effects of the adipokine, adiponectin, and the antidiabetic drug metformin on plasma glucose levels, and in both cases, there is evidence that the mechanism involves activation of AMPK.67,73
A final case in which insulin and AMPK act in the same direction occurs in adipocytes, where both suppress activation of hormone-sensitive lipase, and hence lipolysis, by cAMP-elevating agonists.39,130,131 The 2 pathways act by different mechanisms, with insulin causing phosphorylation and activation of phosphodiesterase 3B by Akt, thus lowering cAMP,132 whereas AMPK phosphorylates hormone-sensitive lipase at a site (Ser565) that antagonizes activation by cAMP-dependent protein kinase.133 In the case of insulin, this can be regarded as among its anticatabolic actions, but the reason why AMPK should suppress lipolysis requires more explanation. If fatty acids released by lipolysis are not removed from the cell rapidly enough, they are known to recycle into triglyceride, thus consuming ATP.134 Inhibition of lipolysis by AMPK has been proposed135 as a mechanism to limit this recycling, ensuring that the rate of lipolysis does not exceed the rate at which fatty acids can be removed or metabolized by other routes, such as fatty acid oxidation.
The parallel effects of insulin and AMPK on muscle glucose uptake and hepatic glucose production formed part of the basis for the idea first proposed in 1999136 that activators of AMPK could be used to treat type 2 diabetes and perhaps obesity. This was initially tested successfully using AICA riboside in animal models such as genetically obese or fat-fed rodents,137140 and findings that activation of AMPK underlies the glucose-lowering effects of metformin72,73 provided a "proof-of-concept" for the idea in humans. Intriguingly, in at least 3 animal models that are resistant to diet-induced obesity, ie, mice overexpressing the uncoupling proteins UCP1141 or UCP3,142 or mice with a knockout of stearoyl-CoA desaturase-1,143 there appears to be a persistent activation of AMPK in the affected cell types.
The ability of AMPK to inhibit adipocyte lipolysis and to stimulate fatty acid oxidation in many tissues, both acutely via phosphorylation of ACC2144 and chronically by upregulating PGC-1
and mitochondrial biogenesis,110,145 is also likely to be beneficial in insulin resistance and type 2 diabetes. There is much evidence that these conditions are partly caused by abnormal accumulation of triglyceride in muscle and liver and that this can be at least partly caused by a defect in mitochondrial function.146 Interestingly, incubation of cultured cells with high levels of the saturated fatty acid palmitate affected the integrity of the endoplasmic reticulum and led to cell death, and this was ameliorated using AICAR to activate AMPK and stimulate fatty acid oxidation.147 One caveat is that it has also been proposed that high rates of fatty acid oxidation can be damaging because of oxidative stress, for example, in ischemic heart following reperfusion44 and in endothelial cells.148 It will be important to establish whether any adverse effects of AMPK activation in certain cell types are outweighed by the overall improvements in metabolic status.
 |
Conclusions and Perspectives
|
|---|
Based on the known effects of AMPK on carbohydrate and lipid
metabolism, it is already clear that the system is a major player
in the development and/or treatment of obesity, diabetes, and
the metabolic syndrome, which have reached epidemic proportions
in modern industrialized society. AMPK is probably involved
in the beneficial effects of regular exercise on these conditions
and now appears to mediate many of the effects of the adipokines
leptin and adiponectin, as well as the antidiabetic drug metformin.
AMPK also appears to be involved in cancer because of the ability
of the LKB1

AMPK pathway to act as a tumor suppressor, together
with potential roles of this pathway in angiogenesis, apoptosis,
and senescence of tumors. There are also now indications the
AMPK system is involved in the effects of caloric restriction
on life span. All of these represent fruitful areas for further
investigation.
 |
Acknowledgments
|
|---|
Sources of Funding
Supported by a Programme Grant from the Wellcome Trust and the EXGENESIS Integrated project (LSHM-CT-2004 to 005272) from the European Commission.
Disclosures
None.
 |
Footnotes
|
|---|
Original received October 26, 2006; revision received November
30, 2006; accepted December 12, 2006.
 |
References
|
|---|
- Hardie DG, Scott JW, Pan DA, Hudson ER. Management of cellular energy by the AMP-activated protein kinase system. FEBS Lett. 2003; 546: 113120.[CrossRef][Medline]
[Order article via Infotrieve]
- Hardie DG, Carling D, Carlson M. The AMP-activated/SNF1 protein kinase subfamily: metabolic sensors of the eukaryotic cell? Ann Rev Biochem. 1998; 67: 821855.[CrossRef][Medline]
[Order article via Infotrieve]
- Schmidt MC, McCartney RR. ß-Subunits of Snf1 kinase are required for kinase function and substrate definition. EMBO J. 2000; 19: 49364943.[CrossRef][Medline]
[Order article via Infotrieve]
- Thelander M, Olsson T, Ronne H. Snf1-related protein kinase 1 is needed for growth in a normal day-night light cycle. EMBO J. 2004; 23: 19001910.[CrossRef][Medline]