Small Molecule Antagonizes Autoinhibition and Activates AMP-activated Protein Kinase in Cells*

AMP-activated protein kinase (AMPK) serves as an energy sensor and is considered a promising drug target for treatment of type II diabetes and obesity. A previous report has shown that mammalian AMPK α1 catalytic subunit including autoinhibitory domain was inactive. To test the hypothesis that small molecules can activate AMPK through antagonizing the autoinhibition in α subunits, we screened a chemical library with inactive human α1394 (α1, residues 1-394) and found a novel small-molecule activator, PT1, which dose-dependently activated AMPK α1394, α1335, α2398, and even heterotrimer α1β1γ1. Based on PT1-docked AMPK α1 subunit structure model and different mutations, we found PT1 might interact with Glu-96 and Lys-156 residues near the autoinhibitory domain and directly relieve autoinhibition. Further studies using L6 myotubes showed that the phosphorylation of AMPK and its downstream substrate, acetyl-CoA carboxylase, were dose-dependently and time-dependently increased by PT1 with-out an increase in cellular AMP:ATP ratio. Moreover, in HeLa cells deficient in LKB1, PT1 enhanced AMPK phosphorylation, which can be inhibited by the calcium/calmodulin-dependent protein kinase kinases inhibitor STO-609 and AMPK inhibitor compound C. PT1 also lowered hepatic lipid content in a dose-dependent manner through AMPK activation in HepG2 cells, and this effect was diminished by compound C. Taken together, these data indicate that this small-molecule activator may directly activate AMPK via antagonizing the autoinhibition in vitro and in cells. This compound highlights the effort to discover novel AMPK activators and can be a useful tool for elucidating the mechanism responsible for conformational change and autoinhibitory regulation of AMPK.

The AMP-activated protein kinase (AMPK) 3 is a highly conserved serine/threonine protein kinase that is widely expressed in higher eukaryotes, yeast, and plants and plays a unique and central role in the responses of cells to metabolic stresses such as nutrient starvation, heat shock, ischemia/hypoxia, and vigorous muscular exercise by depleting cellular ATP and elevating AMP levels (1,2). Once activated, AMPK prevents depletion of ATP by increasing the rate of ATP generation, triggering changes in the rates of glucose transport, fatty acid oxidation, lipogenesis, sterol synthesis, and gluconeogenesis through direct regulation of key metabolic enzymes and transcriptional control of specific genes (1)(2)(3)(4). There is mounting evidence of the involvement of AMPK in human physiological and pathological processes, especially type 2 diabetes and obesity. Previous studies indicate that several of the beneficial effects of rosiglitazone and metformin, two widely used antidiabetic drugs, are mediated by indirect activation of AMPK, suggesting the potential role of the AMPK pathway in the treatment of type 2 diabetes (5,7). Two adipocyte-derived hormones, leptin and adiponectin, stimulate fatty acid oxidation and glucose uptake in peripheral tissues such as skeletal muscle and liver, which are also induced by AMPK activation (8 -13). Furthermore, total AMPK ␣2 knock-out mice displayed impaired glucose tolerance, reduced insulin-stimulated whole-body glucose utilization and skeletal muscle glycogen synthesis (14), and increased body weight and fat mass as compared with the wild-type mice after a high fat diet (15). Therefore, AMPK is considered as a promising target of drugs for treatment of type II diabetes and obesity (16,17).
The mammalian AMPK ␣1 subunit consists of a constitutively active catalytic domain, ␣1 312 (residues 1-312), when Thr-172 is phosphorylated, and a C-terminal domain (residues 313-548), which is responsible for the binding of ␤ and ␥ subunits. Binding of the C-terminal domain with the ␤ and ␥ subunits induces a conformational change in the ␣1 subunit, which results in Thr-172 phosphorylation by AMPK kinase and protects ␣ subunits from degradation. AMPK ␣1 subunit exhibits little activity in the absence of regulatory ␤/␥ subunits, and truncation of the ␣1 subunit from residues 1-548 to 1-392 results in loss of ␤/␥ binding and catalytic activity (22,25,35). We recently reported that the autoinhibitory domain (AID) of the AMPK ␣1 subunit is located between residues 313 and 335 (36) and influences exposure of the catalytic cleft or the Thr-172-activating phosphorylation site or both.
5-Amino-imidazole-4-carboxamide-1-␤-D-ribofuranoside (AICAR) was first reported for regulation of cellular metabolism (37) and is a well known cell-permeable activator (38) of AMPK (39,40). AICAR mediated AMPK activation by conversion into ZMP (AICAR monophosphate) by an adenosine kinase inside cell, which acts as an AMP analogue to activate AMPK and LKB1 without affecting the cellular AMP:ATP ratio (38). Very recently, Cool et al. (41) have first reported a small-molecule AMPK activator, A-769662, which stimulated partially purified rat liver AMPK with an EC 50 of 0.8 M and had potential metabolic effects in vivo.
In this study we identified a novel small-molecule AMPK activator, PT1, which directly activated the inactive forms of ␣1 335 , ␣1 394 , homologous ␣2 398 (␣2, residues 1-398) in a dosedependent manner. According to PT1-docked human AMPK ␣1 subunit structure model, PT1 may interact with Glu-96 and Lys-156 residues near the autoinhibitory domain in ␣1 subunit and directly relieve the autoinhibition. Further study revealed that PT1 also activated native AMPK in L6 myotubes or HeLa cells deficient in LKB1 by increasing the phosphorylation of AMPK ␣ Thr-172 and its downstream substrate, acetyl-CoA carboxylase (ACC), without an increase in the AMP:ATP ratio. Moreover, PT1 treatment lowers hepatic lipid content in a dosedependent manner in HepG2 cells. It suggested that a novel small-molecule activator can directly activate AMPK through conformational change to stimulate AMPK functions.
Recombinant Plasmid Construction-The coding sequences of human AMPK ␣1 or ␣2 subunit truncations and mutants were amplified from cDNA of human AMPK ␣1 or ␣2 subunit and cloned into pET28b vector. All mutants of AMPK-␣1-(1-394) were prepared using the QuikChange site-directed mutagenesis kit for generating the single point mutants from the plasmid pET28b/AMPK-␣1-(1-394). All recombinant plasmids containing wild-type truncations and mutations were verified by DNA sequencing and transformed into E. coli strain BL21 Codon Plus (DE3)-RIL for expression.
Phosphorylation and Activation of Recombinant AMPK Proteins-Recombinant AMPK proteins were fully phosphorylated by incubation with CaMKK␤ (200 nM) at 30°C for 4 h in 20 mM Tris-HCl, pH 7.5, 1 mM dithiothreitol, 5 mM MgCl 2 , 200 M ATP as described previously (36). Analyses of the activities of AMPK proteins were carried out in typical assay conditions of a 50-l reaction mixture containing 20 mM Tris-HCl, pH 7.5, 1 mM dithiothreitol, 5 mM MgCl 2 , 50 mM NaCl, 50 M ATP (0.4 Ci of [␥-33 P] ATP per reaction), and 50 M SAMS peptide. The reaction was initiated by the addition of AMPK proteins (100 nM), incubated at 30°C for 10 min, and terminated by the addition of 50 l of 1% H 3 PO 4 . Particulate matter was then transferred to P30 filter paper and washed 3 times with 0.1% H 3 PO 4 . Radioactivity that had been incorporated in the AMPK proteins was determined by liquid scintillation counting in a Wallac Microbeta plate counter. Background radioactivity estimated from reactions conducted without enzymes was subtracted from sample radioactivities. All reactions were repeated in three independent experiments.
Discovery of a Novel Small-molecule Activator and Its Effects on AMPK Activation-Until now there have been no reports of small-molecule activators that can be used for studying the autoregulatory function of the AMPK ␣ catalytic subunit in the absence of regulatory subunits. We speculate that small molecules might activate AMPK through inducing a conformational change, which antagonizes the autoinhibition. Using the inactive ␣1 394 form (100 nM), we randomly screened 3600 compounds using a diverse compound library under typical assay conditions containing 2 l of compound (40 g/ml) dissolved in dimethyl sulfoxide (DMSO) and discovered a novel smallmolecule activator, PT1. The effects of PT1 on ␣1 truncations (100 nM) and ␣2 398 (400 nM) were studied by varying the concentration of PT1 from 0 to 60 M. The EC 50 (concentration of the compound at which enzymatic activity equals 50% of maximum activity) was calculated from a nonlinear regression curve fitted using GraphPad software (San Diego, CA).
Homology Modeling and Molecular Docking-Homology modeling, molecular dynamics simulations, and energy minimization were calculated by using the molecular operating environment (MOE) 2006.08 (Chemical Computing Group Inc., Montreal, Quebec, Canada) on a HP xw8200 Linux work station. The molecular docking simulation was made by using the Moloc molecular modeling package (43) on a HP NC6220 laptop running Windows XP SP 2. Energy minimization was calculated again with the Moloc package by using the MAB force field.
Homology Modeling-The protein sequence of human AMPK ␣1 was retrieved from the ExPASy server (Q13131, 550 amino acids; Expert Protein Analysis System proteomics server of the Swiss Institute of Bioinformatics). Based on the sequence and structure analysis as well as our previous work (36), the template proteins were determined to be the activated AURORA-A (Protein Data Bank (PDB) code 1OL5) and MARK2 (PDB code 1Y8G), which were used to build up the kinase domain (residues Met-1-Cys-312) and AID (residues Leu-313-Asn-332) of human AMPK ␣1, respectively. The chain alignment was performed with the MOE-Align module, which implements a modified version of the alignment algorithm originally introduced by Needleman and Wunsch (45) using the Blosum50 matrix with a penalty gap of 3 and a penalty for extending a gap of 1 and combining with manual examination. The construction of the three-dimensional model was carried out with the MOE-Homology module (46,47) by means of calculation of 20 intermediate models that were coarsely minimized by using 1OL5 and 1Y8G crystal structure as templates, and the final model was taken as the Cartesian average of all the intermediate models and was further refined by molecular dynamic simulations and energy minimizations, during which only the side chains of all residues were allowed to move. The molecular dynamics simulations were set to 5 ps of heating to 300 K, 10 ps of equilibrium at 300 K, and 5 ps of cooling to 0 K. Energy minimization were made in 100 steps of steepest decent with a root mean square (r.m.s.) gradient test of 100 and 500 truncated Newton steps with a r.m.s. gradient test of 0.01. The Amber94 force field was used. The polypeptide backbone and side chains were then evaluated by MOE-Ramachandran plots (48), a program used to check the stereochemical quality of protein structures (dihedrals, bond angles, etc). This model was used to identify the active site and for docking of activator with the kinase.
Active Site Identification of Human AMPK ␣1-The active site of modeled human AMPK ␣1 was identified using the Alpha Site Finder in MOE, a methodology based upon Alpha Shapes which are a generalization of convex hulls developed by Edelsbrunner (49). In brief, a collection of three-dimensional points is triangulated using a modified Delaunay triangulation. For each resulting simplex there is an associated sphere, i.e. ␣ sphere. These spheres have different radii including infinite radii (corresponding to the planes of the convex hull of the point set). The collection of ␣ spheres is pruned by eliminating those that correspond to inaccessible regions of the receptor as well as those that are too exposed to solvent. In addition, only the small ␣ spheres are retained since these correspond to locations of tight atomic packing in the receptor. Next, each ␣ sphere is classified as either "hydrophobic" or "hydrophilic" depending on whether the sphere is in a good hydrogen bonding spot in the receptor. Hydrophilic spheres not close to a hydrophobic sphere are eliminated (since these generally correspond to water sites). Finally, the ␣ spheres are clustered using a single-linkage clustering algorithm to produce a collection of sites. Each site consists of one or more ␣ spheres at least one of which is hydrophobic.
Molecular Docking Studies-Initial geometric optimizations of the activator, PT1, was carried out using the MAB all atom force field of Moloc (43) with a final root mean square gradient of 0.1 kcal/mol/Å. The optimization parameters and thresholds were adopted with their default values. Mdck of Moloc was used for batch run docking human AMPK ␣1 60 times. The force entry used by Mdck was generated based on the active site determined by Alpha Site Finder as described above. An ensemble of conformers of PT1 was generated by a systematic search as implemented in the Mdck with the dihedral variation of 1 and force margin of 1.5. The resulting complexes were visually inspected in order to discard the unacceptable ones. All the complexes were finally minimized by using the MAB force field again with all the residues of the active site fixed during the optimization. The estimated binding energy was calculated by the score function in the Moloc implementation, which is an empirical formula of the Boehm type (50).
Cell Culture and PT1 Treatment in L6 Myotubes, HeLa Cells, and HepG2 Cells-L6 myoblasts were routinely maintained in ␣-minimum Eagle's medium supplemented with 10% FBS, 4.5 g/liter glucose, 100 units/ml penicillin, and 100 g/ml streptomycin in a humidified incubator in an atmosphere of 5% CO 2 at 37°C. For L6 myoblasts differentiation in 6-well plates, the concentration of FBS was decreased from 10 to 2% over 5 days. Cells were deprived of serum for 3 h before the subsequent compound incubation, which lasted for 1 h. PT1 was prepared as 40 mM stock solutions in DMSO and diluted in medium before application to the cells. Before Western blotting analysis, medium was aspirated, and the cells were rinsed with phosphate-buffered saline and lysed using SDS-PAGE loading buffer.
HeLa cells were maintained in HG-Dulbecco's modified Eagle's medium in 6-well plates supplemented with 10% FBS and antibiotics. Before compounds treatments for 6 h, HeLa cells were starved overnight free of serum, and then specific AMPK inhibitor (40 M compound C) or specific CaMKK␤ inhibitor (10 g/ml STO-609) was added, and cells were incubated for a further 30 min. Before Western blotting analysis, medium was aspirated, and the cells were rinsed with phosphate-buffered saline and lysed using SDS-PAGE loading buffer.
Human hepatoma HepG2 cells were cultured in Dulbecco's modified Eagle's medium containing 10% FBS, antibiotics, and 5.5 mM D-glucose. For experiments, HepG2 cells were incubated in complete medium with 10% FBS in 100-mm diameter dishes, grown to 70% confluence, and maintained in serum-free Dulbecco's modified Eagle's medium overnight. Cells were treated with compounds as indicated in figure legends using metformin as a positive control. Before activator PT1 or metformin treatment, 40 M of compound C was added, and cells were incubated for another 60 min.
Other Methods-The cellular adenine nucleotides were analyzed and detected as we previously described (51). Preparation of the cell lysates and determination of triacylglycerol and total cholesterol content in cell lysates were performed as described previously (6). Triacylglycerol and total cholesterol content were determined in cell lysates using a colorimetric assay and expressed as g of lipid per mg of cellular protein.
Statistical Analysis-Results are expressed as the means Ϯ S.E. Significance was analyzed using a twotailed unpaired Student t test. p Ͻ 0.05 was considered significant.

RESULTS
Discovery of a Novel Small-molecule Activator PT1 for AMPK ␣ Subunit-Until now there have been no reports of small-molecule activators that can be used for studying the autoregulatory function of the AMPK ␣ catalytic subunit in the absence of regulatory subunits. We considered it worthwhile to search for small molecules that might affect the conformation of AMPK ␣ subunit and, thus, antagonize the autoinhibition. Using inactive form human AMPK ␣1 394 , we carried out a random screening of a small organic-compound library for low molecular weight activators. The 3600 compounds were initially screened at concentrations of 40 g/ml using typical assay protocols. One hit (PT1) was identified (Fig.  1A), and its activation effect was confirmed by an EC 50 of about 8 M. Various concentrations of PT1 did not induce significant change in the activity of ␣1 312 but increased ␣1 394 activity in a dose-dependent manner (Fig. 1B). ␣1 394 activity reached a maximum at 20 M PT1, which is about 8-fold higher than that obtained in the absence of PT1.
Because the autoinhibitory mechanism of the AMPK ␣2 subunit is similar to that of the AMPK ␣1 subunit, the effect of PT1 on inactive AMPK ␣2 subunit ␣2 398 was studied. As expected, PT1 activated ␣2 398 in a dose-dependent manner (EC 50 Ϸ 12 M). ␣2 398 activity was maximal at 30 M PT1, which is more than 4-fold higher than that recorded in the absence of PT1 (Fig. 1C). Also, we investigated whether PT1 can directly activate AMPK heterotrimer in vitro as AMPK exists as three subunits heterotrimer in eukaryotic cell. Using recombinant AMPK(␣1␤1␥1) heterotrimer purified from bacterial cells, we found that PT1 did stimulate AMPK(␣1␤1␥1) heterotrimer activity in a dose-dependent manner (EC 50 Ϸ 0.3 M). AMPK(␣1␤1␥1) activity reached a maximum at 5 M PT1, which is approximate 1.5-fold that obtained in the absence of PT1 (Fig. 1D). However, PT1 could not stimulate the activities of AMPK-related protein kinases such as human MARK2, BRSK1, NUAK2, and MELK, which indicates that PT1 has a strong selectivity for AMPK ␣ catalytic subunit activation (data not shown).

PT1 Stimulates Truncated AMPK ␣1 Subunit Proteins Including Autoinhibitory Domain of Residues 313-335-We
have previously reported that autoinhibitory domain of human AMPK ␣1 subunit was located between residues 313 and 335 (36). To determine whether PT1 stimulates AMPK ␣ subunit through relieving the autoinhibitory conformation by interaction with autoinhibitory domain, we investigated the effects of variety concentrations of PT1 on five truncated ␣1 subunit proteins, ␣1 335 , ␣1 341 , ␣1 351 , ␣1 367 , ␣1 377 , each of which includes autoinhibitory domain of residues 313-335. We found that PT1 stimulates these truncations in a dose-dependent manner with similar EC 50 values of about 8 M (Fig. 2). Because ␣1 312 activity cannot be stimulated by PT1 suggests that the autoinhibitory con-formation of AMPK ␣1 catalytic subunit can be regulated by PT1 interaction with kinase domain or autoinhibitory domain.
Construction of PT1 Docking Model in Human AMPK ␣1 Subunit Structure-To determine which domain or which residues in ␣1 subunit interact with PT1, we constructed a PT1docked AMPK ␣1 subunit structural model, and the binding site for activator in the current study was estimated to lie in a cleft formed between the AID domain and N-lobe of AMPK ␣1 after comparing the modeled protein structure to the computational results of Alpha Site Finder. Docking of the activator, PT1, was performed using Mdck of Moloc. The algorithm exhaustively searches the entire rotational and translational space of the ligand with respect to the receptor. The flexibility of the ligand is given by dihedral angle variation, whereas the protein flexibility is not considered. Various solutions are evaluated by a Boehm-type empirical score function. The final solution was determined based on the lowest binding energy together with shape complement. As shown in Fig. 3, PT1 was inserted on the interface between the kinase domain and the autoinhibitory domain of ␣1 subunit. Because the residues surrounding PT1 should influence the binding of PT1 to AMPK ␣1 and the subsequent activation efficacy, around 14 residues, based on the interaction model, were chosen to conduct the site-directed mutagenesis experiment to investigate their role. It was found that the region supposed to interact with PT1 is very sensitive in maintaining the autoinhibition of ␣1 subunit, since most mutants increased the basal activity of ␣1 394 . Although 20 M of PT1 still stimulated mutants of the ␣1 subunit, except for E96A and K156A, the ratios of ␣1 394 enzymatic activity induced by PT1 were all decreased in mutants compared with wild type. Among them, Glu-96 and Lys-156 are the most significant, since PT1 eventually lost the ability to activate ␣1 394 when they were mutated to Ala (Fig.  4). However, when residue Glu-96 was mutated to aspartic acid or when residue Lys-156 was mutated to arginine in ␣1 394 , the E96D or K156R mutant was stimulated by PT1 in a dose-dependent manner (Fig. 4C), similar to wild type of ␣1 394 . Because aspartic acid or arginine has similar polarity and similar molecule size with glutamic acid or lysine, respectively, it is deliberately inferred that activator PT1 with carboxyl group has strong a polar interaction such as H-bond with similar polar residues such as glutamine acid or lysine (Fig. 3). The experimental phenomenon that PT1 removed carboxyl acid was the lack of activation efficacy could partly support this point (data not shown). AMPK Activation by PT1 in L6 Myotubes-Although PT1 dose-dependently activated AMPK ␣1 394

Novel Small-molecule Activator of AMPK
and AMPK ␣2 398 , which has an autoinhibitory domain in the C terminus and is originally inactive, it was unclear whether it could activate the full-length ␣ subunit or the holoenzyme ␣1/␤1/␥1 heterotrimer. To investigate whether AMPK is activated by PT1 in L6 cells, the phosphorylations of AMPK and its downstream target, ACC, an enzyme in the fatty acid synthesis pathway, were used as indicators of AMPK activation. AICAR was used as a positive control in this experiment. The L6 myotubes were treated with 80 M PT1 for 0.5, 1, 2, 3, and 6 h. The immunoblot result showed that PT1 stimulated the phosphorylation of Thr-172 of the AMPK ␣ subunit and the phosphorylation of Ser-79 of ACC in a time-dependent manner, and its effects reached a maximal level when at 2 h of treatment time (Fig. 5A). Then we investigate the dose-response of PT1 on AMPK activity in L6 myotubes. The cells were treated with 10 M, 20 M, 40 M, 80 M PT1, 1 mM AICAR, or DMSO for 1 h. We found that PT1 greatly increased the phosphorylations of AMPK and ACC in a dose-dependent manner (Fig. 5B). The phosphorylations of AMPK and ACC reached a maximal level by treatment of 40 M PT1, which were greatly higher than that induced by 1 mM AICAR.
Several reported compounds induced an increase in cellular AMP:ATP ratio to activate AMPK (53). To investigate whether PT1 stimulates AMPK activity by an increase in AMP:ATP ratio in L6 cells, the L6 myotubes were treated by 40 M PT1 for 2 h, and 100 M compound PD98059, which has been reported to activate AMPK by increasing the cellular AMP:ATP ratio, was used as a positive control (53). As shown in Fig. 5, compared with control, PT1 did not induce an increase in AMP: ATP ratio (p ϭ 0.699), but PD98059 can significantly increase cellular AMP:ATP ratio (p ϭ 0.004), which means PT1 stimulates cellular AMPK activity without an increase in AMP:ATP ratio.
AMPK Activation by PT1 in HeLa Cells Independent of LKB1-There are two kinases, LKB1 and CaMKK␤, identified as AMPK upstream kinases in cells and the major AMPK upstream kinase in L6 cells is LKB1. To investigate whether AMPK is activated by PT1 independent of LKB1, we chose HeLa cells deficient in LKB1 to be treated by PT1. We found that PT1 can also stimulate the phosphorylations of AMPK and ACC in HeLa cells after PT1 treatment for 6 h (Fig. 6). Because HeLa cells only express CaMKK␤ as an AMPK upstream kinase, we used specific CaMKK␤ inhibitor STO-609 and specific AMPK inhibitor compound C to determine the PT1 effects on AMPK phosphorylation and ACC phosphorylation. After 10 g/ml STO-609 or 40 M compound C pretreatment in HeLa cells as described under "Experimental Procedures," the phosphorylations of AMPK and ACC induced by 40 M PT1 were significantly inhibited, respectively (Fig. 6A). Moreover, the cellular AMP:ATP ratio was not changed after 40 M PT1 treatment for 6 h compared with control of DMSO treatment (Fig. 6B). These data demonstrate that PT1 is a novel direct activator of AMPK and the cellular AMPK pathway independent of LKB1.

PT1 Stimulates AMPK and ACC Phosphorylation, Decreases Lipid Content, and these Effects are Diminished by Pretreatment of Compound C in HepG2
Cells-To test whether PT1 has certain cellular function in lipid metabolism, we then determine the effect of PT1 treatment in human hepatoma HepG2 cells on lipid accumulation. We first examined the phosphorylation state of AMPK and ACC. In HepG2 cells exposed to increasing concentrations of PT1 for 24 h, the phosphorylation of AMPK and ACC was significantly stimulated, but the expression of endogenous AMPK␣ and ACC protein were not changed, respectively (Fig. 7A). As a positive control, 2 mM metformin also caused a significant increase in the phosphorylation of AMPK and ACC without an increase in total endogenous AMPK␣ and ACC protein. To determine whether PT1 has a cellular lipid-lowering effect, intracellular levels of triacylglycerol and cholesterol in HepG2 cells exposed to PT1 for 24 h also were measured using metformin as a positive control. Increasing concentrations of PT1 (5, 10, 20, 40, 80 M) decreased intracellular triacylglycerol and cholesterol content in a dose-dependent manner (Fig. 7B), which were in concert with these changes in phosphorylation of AMPK and ACC. The concentration of PT1 above 20 M lowers intracellular lipid content more than that of 2 mM metformin. To determine whether AMPK activity is required for PT1 to lower lipid, HepG2 cells were pretreated with AMPK inhibitor compound C (40 M) followed by incubation with or without PT1 (40 M) for 24 h using metformin (2 mM) as a positive control. The ability of PT1 or metformin to stimulate the phosphorylation of ACC was diminished by compound C (Fig. 8A). Moreover, compound C at least in part blocked the inhibitory effect of PT1 or metformin on triglyceride in HepG2 cells (Fig.  8B), which indicates that AMPK may mediate the effect of PT1 on cellular lipid content.

DISCUSSION
AMPK is a heterotrimeric complex comprising a catalytic subunit ␣ and two regulatory subunits ␤ and ␥. The mammalian AMPK ␣1 subunit consists of a constitutively active catalytic domain, ␣1 312 (residues 1-312), when Thr-172 is phosphorylated, and a C-terminal domain (residues 313-548), which is  responsible for the binding of ␤ and ␥ subunits. AMPK ␣ catalytic subunit exhibits little activity in the absence of regulatory ␤/␥ subunits. Binding of the C-terminal domain with the ␤ and ␥ subunits induces a conformational change in ␣ subunit, which results in an increase in AMPK activity. In addition to regulatory subunits ␤ and ␥, NAD was recently shown to have an activation effect on AMPK (54). However, the details of these mechanisms remain unclear because of a lack of information about the three-dimensional structure of AMPK ␣ catalytic subunit. It is possible that conformational change is a common requirement for AMPK activation. Until now, there have been no reports of small-molecule activators that can be used for studying the autoregulatory function of the AMPK ␣ catalytic subunit in the absence of regulatory subunits. We considered it worthwhile to search for small molecules that might affect the conformation of AMPK ␣ subunit and, thus, antagonize the autoinhibition.
AICAR, an activator of AMPK in vitro and in vivo, has been described previously (38). However, AICAR was incapable of direct activation of AMPK in an in vitro enzymatic assay, and it is unclear whether the various metabolic effects by AICAR treatment are mediated primarily through AMPK activation.
Recently, it has been reported that compound D942, a furancarboxylic acid derivative, increases glucose uptake in L6 myocytes by AMPK activation and is a potent AMPK activator. D942 does not activate AMPK directly at the molecular level in vitro but binds specifically to mitochondrial complex I and inhibits its activity, resulting in an increase in the ratio of AMP to ATP (55). Recently, Abbott first reported a small molecule AMPK activator, A-769662, that stimulated partially purified rat liver AMPK with an EC 50 of 0.8 M and had potential metabolic effects in vivo but did not show whether these effects at large dose of this activator were related on AMPK activation (41). Their research results suggested that activator A-769662 stimulates AMPK with an activation mechanism that is different from that of AMP, and the other two papers published very recently have indicated that A-769662 allosterically activates AMPK probably through interaction with the glycogen binding domain of the ␤ subunit but that it has no direct effect on the activity of the ␣ subunit kinase domain with or without autoinhibitory domain (44,56). Because regulatory subunits ␤ and ␥ can antagonize the autoinhibitory conformation of the ␣1 subunit, we hypothesized that a small-molecule activator that induced a similar conformational change would activate AMPK.
AMPK activity was determined by incorporation of [ 33 P]phosphate into the SAMS peptide. It has been previously reported that CaMKK␤ phosphorylated and activated AMPK  in vitro, and AMPK activity was increased with the presence of AMP (24). Both LKB1 and CaMKK␤ function as an upstream kinase of AMPK in different tissues and cell lines, respectively (26 -31), suggesting different roles for each in the regulation of AMPK activity in vivo. Because CaMKK␤ is an efficient AMPK kinase in vitro (24, 25, 29 -31), we used rat CaMKK␤ as one upstream kinase of AMPK to assess AMPK activity in all enzyme assay experiments. In addition, AMP did not appear to allosterically regulate the activities of truncated human AMPK ␣1 proteins even in the presence of compound PT1 (data not shown), so all kinases assay were performed without the addition of AMP. As expected, human AMPK ␣1 312 without the autoinhibitory sequence (residues 313-394) was phosphorylated by CaMKK␤ and was active; human AMPK ␣1 394 was also phosphorylated by CaMKK␤ but remained inactive, which is consistent with a previous report on mammalian AMPK ␣1 (25,36).
Through screening with inactive form human AMPK ␣1 394 , we for the first time discovered a novel small-molecule activator PT1 to antagonize the autoinhibitory conformation of AMPK ␣ catalytic subunit. PT1 activated ␣1 394 and ␣2 398 , which were originally inactive, but did not increase the activity of AMPK ␣1 truncation without autoinhibitory sequence, ␣1 312 , which was already constitutively active (Fig. 1). It is likely that the activation of ␣1 394 by PT1 is mediated through conformational change, possibly resulting in dissociation of the autoinhibitory domain from the AMPK kinase domain. Interestingly, we found that PT1 can directly stimulate AMPK(␣1␤1␥1) heterotrimer activity in vitro dose-dependently with an EC 50 of 0.3 M (Fig. 1). It is possible that AMPK heterotrimer may have basal activity without stimulus, for the AMPK ␣ catalytic subunit binding with ␤/␥ regulatory subunits may still have a partly open conformation. However, once binding with PT1, the properly catalytic conformation of AMPK ␣ catalytic subunit, which also binds with ␤/␥ regulatory subunits, might be further opened, and then AMPK will exhibit higher kinase activity. It has been demonstrated by Abbott that one of AMPK activators, A-769662, stimulated AMPK probably through interaction with the glycogen binding domain of the ␤ subunit, but the AMPK ␣ catalytic subunit was not involved in the activation effect (44,56). However, in this paper our study for the first time provides strong biochemical evidence that PT1 might have direct activation effect on the activity of AMPK through interaction with several key residues in the AMPK ␣ catalytic subunit.
To better understand the molecular mechanism of PT1-activated AMPK ␣ subunit, we established a docking model of PT1 in human AMPK ␣1 (1-335) subunit. Previously we have constructed a structural model for human AMPK ␣1 (1-335) subunit in which the autoinhibitory domain bound to the small lobe of the kinase domain on the opposite face to the substrate binding groove, and this model demonstrated that some mutation of critical residues of the autoinhibitory domain in AMPK ␣1 subunit have greatly relieved AMPK ␣1 subunit autoinhibition (36). From our present speculative docking model and mutagenesis experiments, we found the region assumed to interact with PT1 is very sensitive to maintain the autoinhibition of ␣1 subunit because of the increased the basal activity of ␣1 394 from most mutants. Among them, Glu-96 and Lys-156 are the most significant as stated before. Our results suggested that PT1 might interact with the region and, especially, residues Glu-96 and Lys-156 to relieve the autoinhibitory conformation of human AMPK ␣ subunit. Considering the importance of carboxyl acid of PT1 from both experimental and modeling data, it is inferred that activator PT1, with a carboxyl group, has a strong polar interaction such as an H-bond with polar residues such as glutamine acid or lysine (Fig. 3). Binding at the cleft formed between the AID domain and N-lobe of AMPK ␣1, unfolding the CD domain and then further elongating the linker and ultimately closing the catalytic site, demonstrating the activation efficacy, is only one of devised molecule mechanisms from our viewpoint, and the details of the interaction between PT1 and ␣ subunit as well as activation mechanism are still under further investigation in our group.
As we expected, native AMPK and the downstream pathway in L6 myotubes were phosphorylated and activated in a dosedependent manner and in a time-dependent manner by PT1 without changing cellular AMP:ATP ratio. Moreover, PT1 stimulates cellular AMPK activity independent of LKB1, which is predominant upstream kinase of AMPK (27), because the phosphorylations of AMPK and ACC were also induced by PT1 treatment in HeLa cells deficient in LKB1. It demonstrates that a PT1-induced increase in cellular AMPK activity is not by PT1stimulated LKB1 or CaMKK␤ but by direct PT1-activated AMPK. In fat metabolism, we assessed the role of AMPK activation by PT1 in the regulation of hepatocellular lipids. We found that PT1 not only dose-dependently increased phosphorylation of AMPK and ACC but also reduced intracellular triglyceride content in a dose-dependent manner in HepG2 cells (Fig. 7). Moreover, AMPK inhibitor compound C partly blocked both the increase in ACC phosphorylation and the decrease in triglyceride content of HepG2 cells caused by PT1 or metformin (Fig. 8). These studies suggest that the effects of PT1 on the lipid content of HepG2 cells are mediated predominantly by activation of AMPK. Previous studies have shown that AMPK activation by either AICAR or metformin stimulates fatty acid oxidation in rat hepatocytes (5,52). This suggests that the reduction in triglyceride levels by PT1 observed in HepG2 cells is explained by increased fatty acid oxidation and/or decreased fatty acid synthesis through activation of AMPK and stimulation of ACC phosphorylation.
Recently, attention has been drawn to AMPK because of its important roles in the regulation of carbohydrate and lipid metabolism, glucose transportation and glycolysis, tumor cell growth, gene transcription, and protein synthesis (42). Because abnormal AMPK activity is associated with several diseases, it is considered an important therapeutic target for the treatment of diabetes, obesity, and cancer. Small-molecule activators of AMPK will be invaluable for elucidating the functions of AMPK and validating the pharmaceutical importance of AMPK as a drug target. PT1 is a novel small-molecule activator that directly activates AMPK through regulation of ␣ catalytic subunit autoinhibitory conformation and will be very useful for evaluating the effects of AMPK under physiological and pathological conditions and studying its downstream signaling pathway. Further investigation of the effects of PT1 on AMPK activation and of its antidiabetic and anticancer effects in vivo could Novel Small-molecule Activator of AMPK lay the foundation for a new therapeutic agent. Complexes between PT1 and ␣1 394 could be used to determine their threedimensional structures and to elucidate the activation mechanism, which would facilitate the discovery and design of more potent AMPK small-molecule activators. Also, this compound highlights the effort to discover novel AMPK activators and offers an alternative means by activating AMPK directly for treatment of metabolic disorders.