Signal transduction

How cells sense energy

Maintaining an optimal cellular energy status requires sensing the levels of the adenine nucleotides ATP, ADP and AMP. Biochemical and structural studies of the enzyme AMPK provide insights into how this is achieved. See Letter p.230

Most cellular processes consume energy, which is provided by the hydrolysis of adenosine triphosphate (ATP) to adenosine diphosphate (ADP). Cells usually maintain ATP:ADP ratios that are orders of magnitude away from equilibrium, in much the same way that chemicals in an electrical battery store energy. Animal cells recharge this 'battery' by oxidizing fuels such as glucose, using some of the released energy to convert ADP back to ATP. Clearly, it is essential that the battery remains constantly charged, and this requires systems that monitor the cellular levels of ATP and ADP as well as AMP (adenosine monophosphate, derived from ADP). Pre-eminent among these systems is AMP-activated protein kinase (AMPK) — a control enzyme that triggers corrective changes in metabolism when energy status is compromised1. On page 230 of this issue, Xiao et al.2 provide intriguing biochemical and structural insights into the mechanism by which cellular energy status is sensed. Their data suggest that AMPK responds not only to AMP and ATP as previously thought, but also to ADP.

AMPK consists of a catalytic α-subunit and two regulatory subunits (β and γ)1. On attachment of a phosphate group to the side chain of T172 (an amino-acid residue on the α-subunit), AMPK's catalytic activity increases more than 100-fold. But as upstream kinase enzymes catalyse phosphate attachment, under normal conditions phosphatase enzymes immediately remove it. AMP binding inhibits phosphate removal3, providing a sensitive mechanism for switching on AMPK. AMP binding also promotes a further 10-fold increase in AMPK activity (yielding an overall activation of greater than 1,000-fold) by allosteric means — that is, by causing a change in the shape of AMPK that indirectly affects its catalytic activity4.

The AMPK γ-subunit contains four tandemly arranged sequence repeats that form the binding sites for AMP, ADP or ATP5. An earlier structure6 revealed that these repeats assemble symmetrically into a flattened disk, with one repeat in each quadrant; the disk contains four cavities in which the adenine nucleotides might bind. Of these, site 2 always seems to be unoccupied, whereas site 4 contains a permanently bound AMP of unknown function. As for sites 1 and 3, AMP binds to them reversibly, in competition with ATP and ADP.

Xiao et al.2 first investigated adenine nucleotide binding to AMPK. They found that AMP, ADP and ATP bind with similar affinity to site 1, which seems to mediate the allosteric effect, and that this site has a 50-fold higher affinity for all three nucleotides than site 3, which regulates T172 dephosphorylation. The authors also made the intriguing finding that ADP protects against phosphate removal just as well as does AMP, although only AMP causes allosteric activation as reported previously7.

In a tour de force of crystallography, Xiao et al. also provide the most complete AMPK structure to date: it contains almost the entire α-subunit, the carboxy-terminal domain of the β-subunit and the entire γ-subunit with AMP in sites 3 and 4. The structure shows that the catalytic (kinase) domain abuts a 'regulatory interface' (formed from the carboxy-terminal domains of the α- and β-subunits), with its substrate-binding site facing away from the rest of the complex (Fig. 1a). The activation loop, which houses T172, is well resolved in the structure, and interacts with the regulatory interface; this would restrict the access of phosphatases to phosphorylated T172 (Fig. 1b).

Figure 1: The new AMPK structure from two angles2.

a, The two lobes of the catalytic domain of the enzyme's α-subunit are shown in two shades of green. The arrow indicates the substrate-binding site at the back of the complex. The carboxy termini of the α-subunit (dark blue) and the β-subunit (light blue) together form the regulatory interface. The linker peptide of the α-subunit (red) contains a structure called the α-hook, which interacts with AMP (yellow) at the γ-subunit (purple), specifically at site 3. Although the structure of the auto-inhibitory domain (AID) in the α-subunit was not resolved in the crystallographic structure, its expected position is labelled. b, View from the left side of a. In the activation loop (orange) within the α-subunit, T172 (pink) lies in a cleft between the catalytic domain and the regulatory interface, where phosphatases would have restricted access.

Xiao and colleagues' structure also provides an elegant explanation for how AMP binding to site 3 prevents T172 dephosphorylation. In the α-subunit, a long linker peptide that connects the catalytic and carboxy-terminal domains includes a structure termed the α-hook; residues at the tip of the hook interact with AMP in site 3 (Fig. 1a). The authors propose that a favourable interaction between the α-hook and AMP (or ADP) in site 3 stabilizes the binding of the catalytic domain to the regulatory interface, thus blocking phosphatase access to T172. They also suggest that, when ATP is bound, this interaction would be disrupted, causing the catalytic domain to dissociate from the regulatory interface and exposing T172 for dephosphorylation. Their biochemical analysis supports this model.

Naturally, puzzles remain. For instance, the structure provides no clues to how AMP binding to site 1 might allosterically activate the catalytic domain. It has been proposed8 that this involves the auto-inhibitory domain (AID, residues 305–330), which follows the catalytic domain on the α-subunit. In the new structure no residues of the AID were resolved, suggesting that it was unfolded and mobile. Although it is possible that the AID folds up when AMP binds to site 1, they are quite remote from each other (Fig. 1b), making it difficult to see how the two events would be connected. What's more, a previously proposed idea9 that AMP promotes phosphate attachment to — as well as inhibiting phosphate removal from — T172 was recently resurrected10. The new structure gives no indication of how this might happen.

Perhaps the most interesting finding of the paper2 is that ADP prevents phosphate removal, which seems likely to be of physiological relevance. Normally, cellular concentrations of AMP are much lower than those of ADP, except in situations of severe energy deficit. During mild energy stress, displacement of ATP by ADP at site 3 would activate AMPK by preventing phosphate removal from T172; the allosteric effect of AMP binding at site 1, however, would amplify this response during more severe stress. This dual mechanism would allow AMPK to sense energy deficit progressively over a wide range of energy availability.


  1. 1

    Hardie, D. G. Nature Rev. Mol. Cell Biol. 8, 774–785 (2007).

  2. 2

    Xiao, B. et al. Nature 472, 230–233 (2011).

  3. 3

    Davies, S. P., Helps, N. R., Cohen, P. T. W. & Hardie, D. G. FEBS Lett. 377, 421–425 (1995).

  4. 4

    Suter, M. et al. J. Biol. Chem. 281, 32207–32216 (2006).

  5. 5

    Scott, J. W. et al. J. Clin. Invest. 113, 274–284 (2004).

  6. 6

    Xiao, B. et al. Nature 449, 496–500 (2007).

  7. 7

    Carling, D., Clarke, P. R., Zammit, V. A. & Hardie, D. G. Eur. J. Biochem. 186, 129–136 (1989).

  8. 8

    Chen, L. et al. Nature 459, 1146–1149 (2009).

  9. 9

    Hawley, S. A. et al. J. Biol. Chem. 270, 27186–27191 (1995).

  10. 10

    Oakhill, J. S. et al. Proc. Natl Acad. Sci. USA 107, 19237–19241 (2010).

Download references

Author information

Correspondence to D. Grahame Hardie.

Rights and permissions

Reprints and Permissions

About this article

Cite this article

Hardie, D. How cells sense energy. Nature 472, 176–177 (2011).

Download citation

Further reading


By submitting a comment you agree to abide by our Terms and Community Guidelines. If you find something abusive or that does not comply with our terms or guidelines please flag it as inappropriate.