AMPK Is a Direct Adenylate Charge-Regulated Protein Kinase

See allHide authors and affiliations

Science  17 Jun 2011:
Vol. 332, Issue 6036, pp. 1433-1435
DOI: 10.1126/science.1200094


The adenosine monophosphate (AMP)–activated protein kinase (AMPK) regulates whole-body and cellular energy balance in response to energy demand and supply. AMPK is an αβγ heterotrimer activated by decreasing concentrations of adenosine triphosphate (ATP) and increasing AMP concentrations. AMPK activation depends on phosphorylation of the α catalytic subunit on threonine-172 (Thr172) by kinases LKB1 or CaMKKβ, and this is promoted by AMP binding to the γ subunit. AMP sustains activity by inhibiting dephosphorylation of α-Thr172, whereas ATP promotes dephosphorylation. Adenosine diphosphate (ADP), like AMP, bound to γ sites 1 and 3 and stimulated α-Thr172 phosphorylation. However, in contrast to AMP, ADP did not directly activate phosphorylated AMPK. In this way, both ADP/ATP and AMP/ATP ratios contribute to AMPK regulation.

Matching energy supply with demand is essential for the survival and function of organisms. In eukaryotes, strict maintenance of the cellular energy status, as reflected in relative concentrations of adenosine tri-, di-, and monophosphate (ATP, ADP, and AMP, respectively), is of paramount importance for the control of all energy-requiring metabolic processes. As energy demands increase, ATP is depleted, and increasing concentrations of ADP result in an increase in the concentration of AMP due to the adenylate kinase equilibrium (1). Eukaryotes have a sensitive signaling mechanism that both monitors cellular energy status and acts to restore energy balance to protect the cell from metabolic stress caused by demand (e.g., muscle contraction) or deprivation (e.g., ischemia). At the center of this mechanism lies the AMP-activated protein kinase (AMPK), an αβγ heterotrimeric serine-threonine kinase that senses the concentrations of AMP. Once activated, AMPK phosphorylates and regulates key enzymes in all branches of metabolism, as well as transcription factors that regulate gene expression, to redirect cellular metabolism away from anabolic, ATP-consuming pathways to energy-generating catabolic pathways.

Mammalian AMPK contains three adenine nucleotide-binding sites on the γ subunit that are occupied under physiological conditions: AMP and ATP bind interchangeably to sites 1 and 3, whereas site 4 binds AMP in a nonexchangeable manner (2). In the Schizosaccharomyces pombe AMPK homolog, AMP and ATP bind interchangeably at site 4 (3). AMPK is inactive unless phosphorylated on Thr172 in the α-catalytic subunit activation loop by the kinases LKB1 or CaMKKβ (4). AMP binding stimulates phosphorylation of α-Thr172, and this stimulation depends on the β subunit being myristoylated (5). Once phosphorylated on Thr172, AMPK is further directly activated two- to fivefold by AMP, and AMP binding also suppresses inactivation due to dephosphorylation of phosphorylated Thr172 (pThr172) by protein phosphatases PP2a and PP2c (6). Once phosphorylated on Thr172, neither direct activation by AMP nor protection against dephosphorylation requires β-subunit myristoylation (5). Inactivation of AMPK occurs in response to rising concentrations of ATP. Exchange of AMP for ATP at γ-subunit sites 1 and 3 inhibits direct activation and promotes dephosphorylation of pThr172 (4).

In early studies ADP was reported to activate AMPK (711), but activation by ADP was attributed to the presence of contaminating AMP (12). Nevertheless, ADP appears to bind to γ-subunit sites 2 and 4 in the crystal structure of S. pombe AMPK homolog, although no regulatory function has been attributed to it (13). To test the effects of ADP on Thr172 phosphorylation in mammalian AMPK, we purified ADP by high-performance liquid chromatography (HPLC) (fig. S1A) to remove contaminating AMP (14). With glutathione Sepharose (GE Healthcare, Rydalmere, Australia)–purified, recombinant AMPK expressed in COS7 cells, purified ADP (200 μM) activated CaMKKβ-mediated phosphorylation of Thr172 1.9-fold (Fig. 1A and fig. S1B). Stimulation by AMP (200 μM) was 2.2-fold (15). Although contaminating AMP concentrations during the ADP-activated phosphorylation assay remained below the detection limit of the HPLC system (0.2 μM) (fig. S1A), we cannot exclude minor stimulatory contributions from AMP in the lower nanomolar range. ADP-activated phosphorylation was dose-dependent; at 120 mM NaCl, the concentration of ADP giving half-maximal activation (K0.5) was 56 ± 19 μM (fig. S1C), which is comparable to the K0.5 for AMP (73 ± 14 μM) (5). Because ADP is a competitive inhibitor of protein kinases at the ATP binding site (16), we examined the effect of AMP and ADP on phosphorylation of the isolated AMPK α1(1-392) catalytic fragment (Fig. 1A and fig. S1B). At 200 μM, AMP and ADP inhibited Thr172 phosphorylation by 13 and 47%, respectively, through their inhibition of CaMKKβ at its active site. Similarly, ADP (250 μM) significantly inhibited CaMKKβ-mediated phosphorylation of the synthetic peptide substrate LKBtide (human NUAK2, 196 to 215 plus three Arg residues) by ~35% (fig. S2). If AMP- or ADP-dependent inhibition of CaMKKβ is taken into account the corrected ADP stimulation of Thr172 phosphorylation would be 3.5-fold (from 1.9-fold), and the corrected AMP stimulation would be 2.5-fold (from 2.2-fold).

Fig. 1

ADP stimulates CaMKKβ-mediated phosphorylation of AMPK α-Thr172. (A and B) Error bars denote mean ± SEM of three to seven independent experiments. Statistical analyses were done with Student’s t test. Thr172 phosphorylation. PP2c-dephosphorylated purified AMPK (α1β1γ1, WT, or indicated mutants from COS7 cells) or purified Escherichia coli–expressed α1(1-392) fragments were phosphorylated by CaMKKβ ± AMP or ADP. 200 μM adenosine nucleotides were used unless otherwise stated. Phosphorylation of Thr172 and total amount of α subunit were measured by simultaneous immunoblot. (A) Fold increase in pThr172 of α1β1γ1 or α1(1-392) relative to basal (non–AMP- or –ADP-incubated) phosphorylation for each protein (see fig. S1B). ****P < 0.0001 versus basal phosphorylation. (B) Regulation of phosphorylation of Thr172 by adenylate energy charge. Nucleotide concentrations were calculated (31) using a total adenine nucleotide pool of 2 mM and assumed equilibrium of the adenylate kinase reaction. 5 mM MgCl2 was used to ensure ATP chelation. **P < 0.002, ***P < 0.001 versus phosphorylation at adenylate charge = 1. The immunoblot shown is a single representative experiment.

Characterizing the relative contributions of ADP and AMP to AMPK regulation in vivo is challenging because, even during periods of high-intensity exercise, concentrations of ADP and AMP remain below detection limits of 31P nuclear magnetic resonance. ADP has been observed, but only in adenylate kinase null muscle (17). Also, estimates of intracellular ADP concentrations based on HPLC methods (after perchlorate extraction) do not take into account that the majority of ADP is tightly bound to actin and metabolically unavailable (18). Within these limitations, free ADP concentrations in skeletal muscle have been measured from 1 to 10 μM in the resting state and from 50 to 300 μM under vigorous contraction conditions (1921), thereby occurring over the range that influenced phosphorylation of AMPK on Thr172 in our in vitro experiments. The relative proportions of ATP, ADP, and AMP within the cell is a reflection of cellular energy state and can be expressed as adenylate energy charge. (22, 23). Resting cells possess an adenylate charge close to one due to the high ATP/ADP ratio. We examined the effect of changes in adenylate energy charge, within a total adenine nucleotide pool of 2 mM, on CaMKKβ-mediated phosphorylation of Thr172 (Fig. 1B). A fall in adenylate charge from 1.0 to 0.9 produced a significant (1.6-fold) stimulation in phosphorylation of Thr172, with half-maximal activation occurring at an adenylate charge of 0.91 (corresponding to 1678 μM ATP, 284 μM ADP, and 38 μM AMP). Thus, AMPK is ideally poised to sense small fluctuations in adenylate charge, even at millimolar concentrations of ATP.

Three adenine nucleotide-binding sites (1, 3, and 4) on the γ subunit influence AMP stimulation of Thr172 phosphorylation (5). Mutation of the conserved Asp to Ala residues in the γ1 adenine nucleotide-binding sites 1, 3, or 4 each resulted in significantly reduced sensitivities to AMP-activated phosphorylation of Thr172, with site 3 the most sensitive to mutation. Only disruption of sites 1 and 3 reduced ADP activation of Thr172 phosphorylation (Fig. 2 and fig. S3). The net decrease in phosphorylation in the Asp mutants relative to that of wild-type (WT) enzyme reflects both the loss of ADP stimulation plus inhibition of CaMKKβ by ADP. When corrected for CaMKKβ inhibition, site-1 and -3 mutants retain some residual sensitivity to ADP-activated phosphorylation (≈1.8-fold increase in each case), but both sites are required for full activation.

Fig. 2

ADP stimulation of Thr172 phosphorylation is mediated via γ-nucleotide binding sites 1 and 3. (A) Contribution of individual γ adenine nucleotide-binding sites to AMPK regulation by ADP (see fig. S3). Error bars denote mean ± SEM of four independent experiments. Statistical analyses were done with Student’s t test. ***P ≤ 0.002, ****P ≤ 0.0001 versus WT AMPK. (B) Mammalian AMPK γ1 nucleotide-binding sites. Diagram shows AMP/ADP occupancy of individual γ (blue) adenine nucleotide-binding sites involved in mammalian AMPK regulation, derived from structural (3) and/or mutational analyses (5). Numbers represent nucleotide site/cystathionine beta synthase motif, and conserved γ- or β-flap (green) Asp residues are indicated.

ADP stimulation of Thr172 phosphorylation caused a 1.6-fold increase in AMPK activity when measured by a linked AMPK peptide substrate assay (Fig. 3A). Increased AMPK activity appeared to be caused entirely by enhanced phosphorylation of Thr172 because direct addition of ADP to phosphorylated AMPK αβγ heterotrimer or isolated α1 kinase domain (1 to 315) resulted in similar suppression of activities (Fig. 3B), confirming that ADP binding alone does not directly activate AMPK αβγ heterotrimers (24).

Fig. 3

ADP stimulation of AMPK activity occurs independently of direct allosteric activation. (A and B) Error bars denote mean ± SEM of three to five independent experiments. Statistical analyses were done with Student’s t test. (A) Linked assay: PP2c-dephosphorylated, purified AMPK (α1β1γ1) was phosphorylated by CaMKKβ ± ADP and activity measured in the presence of 10 μM ADP using the SAMS assay. **P < 0.007 versus basally phosphorylated (non–ADP-incubated) AMPK. The immunoblot shown is a single representative experiment. (B) Direct activation: activities of AMPK (α1β1γ1) expressed in COS7 cells or CaMKKβ-phosphorylated, α1(1-315) kinase domain expressed in E. coli were measured by SAMS assay ± AMP or ADP. *P < 0.02, **P < 0.002 versus basal (non–AMP- or –ADP-incubated) activity of each protein.

Myristoylation of the β subunit is essential for AMP stimulation of α-Thr172 phosphorylation, and substitution of β-Gly2 with alanine (G2A) to prevent myristoylation of AMPK blocks AMP-enhanced phosphorylation of Thr172 (5). ADP did not activate Thr172 phosphorylation of nonmyristoylated AMPK [α1β1(G2A)γ1] (fig. S4). The nucleotide-induced reductions in Thr172 phosphorylation compared to basal phosphorylation of the nonmyristoylated enzyme are consistent with suppression of CaMKKβ activity observed with AMPK α1(1-392) as the target substrate (Fig. 1A).

AMP additionally regulates AMPK by suppressing dephosphorylation of pThr172; however, at 200 μM ADP conferred significantly more protection against PP2c-mediated dephosphorylation than AMP (49 versus 37%, respectively) (Fig. 4). This difference was specific to the αβγ heterotrimer, because dephosphorylation of the catalytic fragment α1(1-392)-pThr172 by PP2c was not suppressed by either AMP or ADP (fig. S5). As for AMP, ADP-inhibited rates of PP2c-mediated pThr172 dephosphorylation were unaffected by the myristoylation state of the β subunit (Fig. 4).

Fig. 4

ADP protects against PP2c-mediated dephosphorylation of pThr172 independently of β-myristoylation. (A and B) Error bars denote mean ± SEM of three to seven independent experiments. Statistical analyses were done with Student’s t test. Dephosphorylation assay: CaMKKβ-phosphorylated, purified AMPK [α1β1γ1 or α1β1(G2A)γ1] was incubated with PP2c ± AMP or ADP. Residual pThr172 and total amount of α subunit were measured by simultaneous immunoblot. (A) Percent of residual pThr172 compared to untreated (non–PP2c-incubated) controls. **P < 0.007, ***P < 0.001 versus basal (non–nucleotide-incubated) dephosphorylation. The immunoblot shown is a single representative experiment. (B) Percent of protection against dephosphorylation conferred by addition of nucleotide, relative to basal dephosphorylation. *P < 0.03 versus AMP treatment for each protein.

A central argument underpinning the role of AMP as the key regulatory molecule for AMPK has been that the adenylate kinase reaction responsible for generating AMP and ATP from ADP is held at near equilibrium. As a result, AMP concentrations were proposed to fluctuate as a square function of changes in ADP, thereby lending increased sensitivity to the AMPK regulatory system (25, 26). However, this assumption neglects the contribution of AMP deaminase in muscle, which rapidly and irreversibly removes AMP from the system to form inosine monophosphate (IMP), a clearance mechanism that may drive the adenylate kinase reaction to reduce ADP that would otherwise inhibit protein kinase signaling (27). Concentrations of IMP increase more than 350-fold in skeletal muscle after stimulated contraction or exercise, with a corresponding drop in total adenine nucleotide content (2729). We found no AMPK regulatory function for IMP, and IMP did not antagonize AMP-enhanced phosphorylation of Thr172 (fig. S6, A to C). Presumably, IMP cannot form critical N6 hydrogen bonds to backbone residues in sites 1 and 3 (fig. S6D).

We show that mammalian AMPK Thr172 phosphorylation is regulated by both ADP and AMP, whereas only AMP directly activates AMPK. Given that control of Thr172 phosphorylation is the major mechanism for AMPK regulation, the enzyme is a direct adenylate charge-sensing protein kinase—along the lines envisaged by Atkinson in his “adenylate charge hypothesis” for regulating metabolic pathways (23)—rather than specifically an AMP-activated kinase.

Supporting Online Material

Materials and Methods

Figs. S1 to S6


References and Notes

  1. Yeh et al. (9) concluded that AMP stimulated acetyl coenzyme A carboxylase phosphorylation by acting on the substrate, not the upstream kinase (AMPK).
  2. Materials and methods are available as supporting material on Science Online.
  3. 200 μM AMP produces submaximal activation of the phosphorylation of AMPK αThr172; hence, the fold activation reported here is lower than that reported previously using 500 μM AMP (5).
  4. Adenylate energy charge is calculated by the equation ([ATP] + 1/2[ADP])/[ATP + ADP + AMP], producing a value between 1 (100% ATP) and 0 (100% AMP).
  5. Acknowledgments: We thank F. Katsis for the preparation of antibodies and L. Macaulay at Commonwealth Scientific and Industrial Research Organisation Parkville, Victoria, Australia, for CaMKKβ expression constructs. We are most grateful to S. Gamblin and D. Carling and their colleagues at Medical Research Council National Institute of Medical Research and Hammersmith Hospital, respectively, for generously sharing with us the observation that ADP inhibited the dephosphorylation of AMPK by PP2c (30), which prompted us to test whether ADP also controlled the kinase kinase reaction. This work was supported by grants from the Australian Research Council, the National Health and Medical Research Council (NHMRC), and the Victorian Government Operational Infrastructure Support Scheme. B.E.K. is an NHMRC Fellow.
View Abstract

Navigate This Article