J. Winderickx, Laboratory for Functional Biology, Kasteelpark Arenberg 31, PO Box 2433, B-3001 Heverlee, Belgium Fax: +32 16 321967 Tel: +32 16 321516 E-mail: firstname.lastname@example.org
All life forms on earth require a continuous input and monitoring of carbon and energy supplies. The AMP-activated kinase (AMPK)/sucrose nonfermenting1 (SNF1)/Snf1-related kinase1 (SnRK1) protein kinases are evolutionarily conserved metabolic sensors found in all eukaryotic organisms from simple unicellular fungi (yeast SNF1) to animals (AMPK) and plants (SnRK1). Activated by starvation and energy-depleting stress conditions, they enable energy homeostasis and survival by up-regulating energy-conserving and energy-producing catabolic processes, and by limiting energy-consuming anabolic metabolism. In addition, they control normal growth and development as well as metabolic homeostasis at the organismal level. As such, the AMPK/SNF1/SnRK1 kinases act in concert with other central signaling components to control carbohydrate uptake and metabolism, fatty acid and lipid biosynthesis and the storage of carbon energy reserves. Moreover, they have a tremendous impact on developmental processes that are triggered by environmental changes such as nutrient depletion or stress. Although intensive research by many groups has partly unveiled the factors that regulate AMPK/SNF1/SnRK1 kinase activity as well as the pathways and substrates they control, several fundamental issues still await to be clarified. In this review, we will highlight these issues and focus on the structure, function and regulation of the AMPK/SNF1/SnRK1 kinases.
Convergence of energy and stress signaling on the AMPK/SNF1/SnRK1 kinases
The AMP-activated kinase (AMPK)/sucrose nonfermenting1 (SNF1)/Snf1-related kinase1 (SnRK1)-related protein kinase complexes are a family of highly conserved heterotrimeric serine/threonine kinases with orthologs being found in all eukaryotes ranging from yeast (SNF1), roundworms [AMP-activated kinase (AAK)] and insects (AMPK) to mammals (AMPK) and plants (SnRK1). Their primary role lies in the integration of nutrient availability, environmental stress signals and energy expenditure in order to induce the required adaptations to maintain energy homeostasis and cell survival.
In Saccharomyces cerevisiae, SNF1 is primarily required for adaptation to glucose limitation, allowing yeast cells to utilize alternative carbon sources, such as sucrose or ethanol [1,2] (Fig. 1). It is also involved in controlling the biosynthesis of reserve carbohydrates and the recycling of macromolecules and organelles by means of autophagy . It has an important role in developmental processes, such as meiosis and sporulation , filamentous growth and biofilm formation  as well as in the control of ageing and longevity [6,7]. In addition, SNF1 was found to be responsive to a variety of stressors, such as sodium ion stress, oxidative stress, alkaline pH and antimycin A, an inhibitor of the respiratory chain , which underscores its role in enabling protective mechanisms and processes that confer maximal stress tolerance to yeast cells [9,10].
The animal SNF1 orthologs exert very similar functions. In Caenorhabditis elegans roundworms, AAK functions as a metabolic sensor that couples lifespan to information about energy levels and insulin-like signals . Larvae that lack AAK signaling enter dauer stage (diapause) normally, but then rapidly consume their stored energy and die prematurely following vital organ failure . In Drosophilamelanogaster flies, AMPK has profound effects on epithelial integrity and cell division throughout development and in response to energetic stress [13,14]. During adult life, AMPK is also required for maintenance of the nervous system [15,16]. Mammalian AMPK is a metabolic master switch that is activated under conditions that increase the AMP/ATP ratio, including glucose deprivation, hypoxia, ischemia, oxidative stress or pharmacologic inhibition of oxidative phosphorylation. Once activated, the kinase mediates the up-regulation of energy-producing catabolic processes, such as glycolysis and fatty acid oxidation, induces autophagy and down-regulates energy-consuming anabolic metabolism such as protein synthesis or the synthesis of sterols and fatty acids [17,18] (Fig. 1). In addition to its role as a cellular energy sensor, AMPK acts as regulator of whole-body energy metabolism by integrating nutritional and hormonal signals in the hypothalamus to control food intake and body weight  or by inhibiting islet β-cell insulin production and secretion in response to low blood glucose levels . Consistently, AMPK has a tremendous impact on survival, growth and development at the organismal level and in humans its impairment is associated with a variety of disorders, including metabolic syndrome, insulin resistance, obesity, cardiovascular diseases, cancer, dementia and stroke [18,21].
Interestingly, the plant SNF1/AMPK ortholog, SnRK1, similarly senses energy deficit caused by nutrient deprivation, environmental stress or alternate light–dark cycles [22,23]. Although its physiological role is only starting to be uncovered, it is already clear that SnRK1 is a global regulator of plant metabolism, controlling energy-conserving processes and the remobilization of alternative energy sources, including sucrose, starch, cell wall compounds, amino acids and lipids (Fig. 1). In addition, SnRK1 coordinates stress-induced responses, including antiviral defense, and fundamental developmental processes, from germination and sprouting to reproduction and senescence [24–28].
Downstream of AMPK/SNF1/SnRK1
The AMPK/SNF1/SnRK1-related protein kinases exert a dual control over cellular metabolism. First, they regulate key metabolic enzymes through direct phosphorylation. For instance, AMPK/SNF1 phosphorylates and inactivates acetyl-CoA carboxylase both in yeast and in mammals, thereby inhibiting the formation of malonyl-CoA from acetyl-CoA and hence fatty acid biosynthesis [18,29,30]. Furthermore, both AMPK and plant SnRK1 inhibit the activity of HMG-CoA reductase, the rate-limiting enzyme in sterol synthesis [31–34] (Fig. 1). More direct phosphorylation targets have been identified and many more are predicted based on conserved phosphorylation sites.
Second, the AMPK/SNF1/SnRK1 kinases trigger massive transcriptional reprogramming [28,35,36], which is achieved by the activation, inactivation and control of recruitment and localization of various transcription factors [37–41]. The most important direct yeast Snf1 target is Mig1, a transcriptional repressor of glucose-repressed genes . When glucose becomes limiting, Snf1 promotes nuclear export of Mig1, thereby relieving glucose repression [43–46]. Additional effects of Snf1 are mediated by the transcriptional activators of gluconeogenic genes, Cat8 and Sip4, both transcriptionally inhibited by Mig1 and activated through direct phosphorylation by Snf1 [47–50] (Fig. 1). In mammals, AMPK inhibition of gluconeogenesis involves regulation of the transcriptional activators CRTC2 and HNF4α [18,51–53]. Finally, in Arabidopsisthaliana, the extensive SnRK1-dependent transcriptional response is mediated in part by a class of bZIP transcription factors . The transcriptional reprogramming by the AMPK/SNF1/SnRK1 complex has already been extensively reviewed [18,41,54,55] and will therefore not be further discussed here. In addition, it should be noted that the AMPK/SNF1/SnRK1 kinase complexes also affect transcription through interacting with different factors of the transcriptional machinery, including modulation of RNA polymerase activity via the SRB/mediator complex , direct association with chromatin, the phosphorylation and acetylation of histones [57–59] and the phosphorylation and nuclear export of histone deacetylases .
Hence, the AMPK/SNF1/SnRK1-mediated responses across species show remarkable similarities, indicating that the cross-species structural conservation of the kinase complex reflects the remarkable functional conservation [61,62]. This is further strengthened by the fact that also the upstream protein kinases, controlling AMPK/SNF1/SnRK1 activity, show significant conservation [63,64] (Fig. 1 and see below).
Structure of the AMPK/SNF1/SnRK1 kinases
The AMPK/SNF1/SnRK1 protein kinases typically function as heterotrimeric complexes that require a catalytic α-subunit and regulatory β- and γ-subunits for protein stability and kinase activity. The number of complexes that can be formed varies significantly between organisms (Fig. 2). For instance, the genome of S. cerevisiae encodes one catalytic subunit (Snf1), three β-subunits (Sip1, Sip2 and Gal83) and one γ-subunit (Snf4). Hence, three alternative Snf1 complexes can be formed in this yeast. C. elegans, on the other hand, has the potential for 20 heterotrimeric complexes , while in D.melanogaster, the α-, β- and γ-subunits are encoded by single genes: dSnf1A/dAMPKα, alicorn and dSnf4A/loechrig, respectively. However, the latter can presumably form a number of different complexes because there is evidence for alternative splicing and differential transcription initiation [15,16]. Humans express two isoforms of the α- and β-subunit each (α1, α2 and β1, β2) and three isoforms of the γ-subunits (γ1, γ2 and γ3), all encoded by different genes, and thus can form at least 12 different complexes. However, this number is again probably higher because of alternative splicing and transcription initiation .
In the plant, A. thaliana, the proteins that display most similarity to Snf1 and AMPKα are KIN10 and KIN11, together with the likely pseudogene KIN12, making up the SnRK1 subfamily. In addition, plants encode two larger plant-specific subfamilies with sequence similarity to the yeast and mammalian catalytic α-subunits (SnRK2 and SnRK3), but these have clearly diverged more and do not complement the yeast snf1Δ deletion mutant growth phenotype . Two Arabidopsis genes, KINβ1 and KINβ2, encode proteins that have the typical characteristics of yeast and mammalian β-subunits, and one protein (KINγ) was dubbed the A. thalianaγ-subunit. However, as described in more detail below, plants also express structurally atypical and plant-specific β- and γ-subunits and proteins that were found to bind to the catalytic subunits but with only very weak sequence similarity to yeast Snf4 and AMPKγ and were unable to complement the yeast snf4Δ deletion mutant . Hence, plants appear to display plant-specific, alternate complexes besides the ‘classical’ heterotrimeric complexes and this may relate to their sessile, autotrophic lifestyle and dependency on a more stringent control of energy homeostasis to cope with continuous nutritional and environmental stress conditions.
The catalytic α-subunit
As indicated above, the remarkable evolutionary conservation applies to all three subunits and allowed identification of several subunits based on functional complementation screening. The catalytic subunit displays the highest degree of cross-species conservation (Fig. 3A), especially in the kinase domain located in the N-terminal half of the protein [25,68]. A conserved threonine residue in the activation loop of the kinase domain (Thr210 in Snf1, Thr172 in AMPK and Thr175 in SnRK1.1/KIN10) requires phosphorylation by upstream kinases to confer kinase activity [9,18,23,69]. The C-terminal half of the proteins typically harbors an auto-inhibitory regulatory sequence (AIS) right next to the kinase domain, and a β-subunit interaction domain (β-SID), which is found near the C-terminal end and is followed by a very recently identified conserved leptomycin-sensitive nuclear export sequence (NES) . Under activating conditions, an additional interaction can occur between the γ-subunit and the AIS domain [71,72] (see below). Interestingly, both the yeast and mammalian catalytic subunits can form homodimers, but its physiological significance is still unclear [73,74].
The regulatory β-subunit
The β-subunits contain two characteristic and distinct domains. The first domain, located in the middle of the protein, was initially identified as the kinase interaction sequence (KIS) domain , but is now rather referred to as the largely, but not completely overlapping, glycogen-binding domain (GBD) (Fig. 3A). It is still a matter of debate whether glycogen binding has regulatory consequences . In S. cerevisiae, recent publications suggest that deletion of the GBD of Gal83 results in constitutive activation of SNF1, independent of glucose availability, while deletion of the GBD in Sip1 and Sip2 had no significant effect [75,76]. However, these effects were not caused by the inability of glycogen to bind to the β-subunits, as normal regulation of SNF1 was observed in mutants lacking glycogen synthase activity . While the GBD/KIS domain is required for the interaction with the regulatory domain of the catalytic α-subunit of yeast SNF1 and plant SnRK1, it is apparently not essential for the formation of stable heterotrimeric AMPK complexes in mammals. The second domain, the so-called association with SNF1 complex (ASC), is found at the C-terminal end of the proteins and mediates the interaction with the γ-subunit of yeast SNF1 and plant SnRK1  and both the α- and γ-subunits of mammalian AMPK . Interestingly, plants express an atypical β-subunit (i.e. the KINβ3-type), which lacks the entire GBD/KIS domain and, in this case, the C-terminal ASC domain interacts with both α- and γ-subunits (Fig. 3B). In spite of this feature, KINβ3 from A. thaliana still complements a yeast mutant lacking all β-subunits .
Besides these conserved domains, the β-subunits have more variable N-termini that control the subcellular localization of the kinase complexes through sequences important for nucleo–cytoplasmic translocation  or N-myristoylation, important for membrane targeting and binding [9,18]. For instance, in yeast, all three β-subunits are cytoplasmic under high-glucose conditions, while under glucose-limiting conditions, Gal83 is translocated to the nucleus, while Sip1 relocalizes to the vacuole and Sip2 remains cytoplasmic . Furthermore, Sip2 also acts as a negative regulator of nuclear SNF1 in young cells by sequestering Snf4 at the plasma membrane, a process which depends on N-terminal myristoylation of Sip2 . An age-associated shift of Sip2 from the plasma membrane to the cytoplasm, however, results in nuclear translocation of Snf4 and induction of histone kinase activation and chromatin remodeling , the derepression and regulation of sterol biosynthesis , as well as the control of yeast cell longevity . The β-subunit N-terminus also contains serine residues that are known to be ‘auto’-phosphorylated in mammalian AMPK β1 . The exact mechanisms that regulate the intracellular localization of the kinase complexes are, however, still poorly understood and may depend on the phosphorylation of different sites in the α- and/or β-subunits [84–86].
The regulatory γ-subunit
The γ-subunits are characterized by divergent N-termini and two pairs of cystathionine-beta-synthase (CBS) repeats, called Bateman domains, that bind adenosine derivates [9,18] (Fig. 3A). Plants, in addition, also express an atypical γ-subunit (i.e. the KINβγ-type), that apparently resulted from the fusion of a γ-subunit with sequences related to the GBD (Fig. 3B). This GBD-related domain can also interact with proteins unrelated to the SnRK1 complexes. Nevertheless, this βγ-subunit is able to complement the yeast snf4Δ mutant phenotype [87,88].
Initial studies on SNF1 suggested that, upon glucose-deprivation, the yeast Snf4 binds to the AIS sequence in the C-terminal regulatory part of Snf1 and thereby relieves auto-inhibition of the kinase . More recently, a pseudo-substrate sequence was found to be conserved in the γ-subunits of different species, leading to a model where this sequence interacts with the active-site groove of the α-subunit to prevent phosphorylation and activation by upstream kinases. Binding of AMP to the γ-subunit would then result in a conformational change in the AMPK complex that disrupts its interaction with the α-subunit and relieves the inhibition . However, the specific signal that brings about the activation of the yeast and plant kinases is still a matter of debate (see below).
Regulation of the AMPK/SNF1/SnRK1 kinases
Regulation by phosphorylation
Phosphorylation of the catalytic subunit is essential for AMPK/SNF1/SnRK1 activation (Fig. 1). Three upstream kinases have been identified that phosphorylate the conserved threonine in the activation loop of Snf1 in yeast, in other words Sak1, Elm1 and Tos3. These kinases are highly similar and partially functionally redundant [63,90,91]. AMPK in mammals is activated by two kinases: one that is constitutively active (i.e. LKB1) and one that is activated by Ca2+ ionophores (i.e. CaMKKβ) [17,18]. Interestingly, both are functionally conserved as they activate SNF1 in a yeast mutant lacking Sak1, Tos3 and Elm1. In addition, the mammalian TAK1 kinase also complements a yeast sak1Δ tos3Δ elm1Δ mutant and hence may activate AMPK in animals . Sequence comparison and complementation and phosphorylation assays have similarly implicated GRIK1/2 (SnAK1/2) as the upstream activating kinases of SnRK1 in plants [93–95]. However, in yeast the activities of the SNF1-activating kinases do not appear to be glucose-regulated. Instead, glucose promotes the dephosphorylation by changing the ability of the activation loop to act as a substrate for the PP1 phosphatase Glc7-Reg1 . In addition, the PP2A-like phosphatase, Sit4, is also able to dephosphorylate Thr210 of Snf1 in vivo . PP2A and PP2C phosphatases were reported to dephosphorylate and inactivate mammalian AMPK and plant SnRK1 in vitro [98–100], but specific phosphatases in vivo have not yet been identified.
Mammalian AMPK is subject to allosteric regulation by the AMP/ATP ratio, a sensitive indicator of cellular energy status (Fig. 1). Binding of AMP to the γ-subunit not only promotes the phosphorylation of the catalytic α-subunit by upstream kinases and protects it against dephosphorylation, it also results in allosteric activation of AMPK [98,101–103]. Moreover, ADP has been shown to protect AMPK against dephosphorylation, although this does not result in allosteric activation . In addition, fatty acids are also able to allosterically activate AMPK [18,105].
However, yeast SNF1 and plant SnRK1 are not subject to allosteric regulation by AMP/ATP (Fig. 1). Indeed, early studies in yeast reported that the SNF1 complex does not bind AMP in vitro , which was explained by the presence of a substitution (His151Gly) in Snf4 . Nonetheless, activation of SNF1 correlates with increased AMP/ATP ratios , and mutation of conserved Snf4 sites that contribute to AMP binding in AMPK do relieve glucose inhibition of Snf1 . Plant SnRK1 is also not allosterically regulated by the AMP/ATP ratio, although T-loop dephosphorylation and, in consequence, inactivation of the kinase is inhibited by physiological concentrations of AMP . Interestingly, sugar phosphates, such as glucose 6-phosphate (Glc6P), were found to inhibit SnRK1 in a cell-free system  and a more recent study reported an inverse correlation between KIN10/11-controlled gene expression and trehalose 6-phosphate (Tre6P) levels, identifying the SnRK1 protein kinase as an important Tre6P target . Tre6P is the intermediate of trehalose biosynthesis from Glc6P and UDP-glucose catalysed by Tre6P synthase (TPS) and located at the crossroads of primary carbon metabolism. In A. thaliana, Tre6P is required for efficient carbohydrate utilization and the levels correlate well with sugar supply [111,112]. In addition, trehalose (and Tre6P) metabolism in plants plays an important role in root meristem establishment and activity as well as in the transition to flowering [113,114]. Mutation of the TPS gene, TPS1, is embryo-lethal [115–117]. In yeast, Tre6P plays an essential role in glucose signaling, the control of the glucose influx into glycolysis [118,119], mitochondrial cytochrome content, respiratory chain activity  and oscillation of energy metabolism . Furthermore, trehalose synthesis is required for an optimal stress response [121–123] and survival of yeast cells . It is conceivable that also in S. cerevisiae, a negative correlation exists between Tre6P levels and SNF1 activity. Additional research is required to identify the (allosteric) regulator(s) of the SNF1 kinase complex.
In C. elegans, trehalose was recently shown to increase longevity , and overexpression of TPS was shown to increase hypoxia tolerance in flies [126,127]. Intriguingly, while mammals do not encode TPS, the overexpression of D. melanogaster dTPS1 in HEK-293 cells also increased their hypoxia tolerance . However, the latter effects could be attributed to the stress-protectant effects of trehalose. The main C. elegans catalytic α-subunit, AAK-2, also functions downstream of DAF-2 (receptor)-mediated insulin-like signaling (ILS) to positively regulate adult lifespan. The regulation of lifespan by AAK-2 probably occurs in parallel with the DAF-16 transcription factor that is also negatively regulated by ILS [11,12,129–131]. Trehalose metabolism has not been extensively studied in C. elegans, but exogenous trehalose was recently found to increase thermotolerance and longevity . While trehalose did not further extend lifespan in a long-lived daf-2 mutant, RNA interference (RNAi) silencing of the two C. elegans TPS genes (tps-1 and tps-2), reduced the daf-2 extended lifespan. The latter effect was mostly rescued by trehalose supplementation and expression of tps-1 and tps-2 was found to be upregulated in daf-2 mutants. These data are consistent with increased lifespan upon reduced ILS being at least in part dependent on the downstream biosynthesis of trehalose as a stress-protectant . It is not clear, however, whether and how Tre6P is involved in ILS. Given the striking overlap of phenotypes influenced by Tre6P or trehalose synthesis, and those controlled by AMPK/SNF1/SnRK1, it is tempting to speculate that Tre6P could be an allosteric regulator not only in plants but also in other eukaryotes.
In all eukaryotic systems, the AMPK/SNF1/SnRK1 kinases typically function as heterotrimers with a catalytic α-subunit and regulatory β- and γ-subunits that contribute to substrate specificity and determine subcellular localization. As most organisms encode a variety of subunit isoforms, different complexes can be formed, but the exact number is not known because of differential expression and alternative transcription initiation and/or splicing. In addition, the functional consequences of incorporating alternate subunit isoforms into the complex remains to be explored. Studies in various model systems also indicate that the activity of the AMPK/SNF1/SnRK1 kinase complex is tightly regulated by dynamic phosphorylation of its catalytic subunit, but while the upstream kinases are relatively well characterized, much less is known about the protein phosphatases that trigger dephosphorylation, especially of AMPK and SnRK1 in animals and plants. Increasing evidence also suggests that the function of the complex is also controlled by post-translational modification of the β-subunits, which, besides myristoylation, appears to involve phosphorylation by the AMPK α-subunit and as-yet-to-be-identified kinases. Finally, the AMPK/SNF1/SnRK1 kinases are subject to allosteric regulation, and while this directly relates to changes in the AMP/ATP ratio in mammals, the metabolic intermediates that control yeast SNF1 and plant SnRK1 activity are not known. A regulatory role, however, is suggested for specific sugar phosphates.
Disruption of the metabolic control function of AMPK/SNF1/SnRK1 kinases has been implicated in the aging process and cell death in all eukaryotic systems as well as in a number of highly prevalent human diseases, including type 2 diabetes, cardiovascular disease and cancer. The precise molecular mechanisms that underlie their cell-protective function and their involvement in patho-physiological processes are only starting to be uncovered. Unraveling all of the abovementioned issues will not only advance our understanding of conserved and diverged signaling mechanisms, but will also provide important insights into the initiation and progression of AMPK-associated human diseases and SnRK1-controlled developmental processes and stress tolerance in plants.
Work on AMPK/SNF1/SnRK1 in the K.U.Leuven laboratories is funded by Fonds voor Wetenschappelijk Onderzoek (FWO), Agentschap voor Innovatie door Wetenschap en Technologie in Vlaanderen (IWT) and K.U.Leuven (BOF, Bijzonder Onderzoeksfonds). Work on SnRK1 at MGH is funded by NSF grant IOS-0843244.