Regular article
Protein kinase substrate recognition studied using the recombinant catalytic domain of AMP-activated protein kinase and a model substrate1

https://doi.org/10.1006/jmbi.2001.5316Get rights and content

Abstract

We have expressed a truncated form of the α1 kinase domain of AMP-activated protein kinase (AMPK) in Escherichia coli as a glutathione-S-transferase fusion (GST-KD). A T172D mutant version did not require prior phosphorylation and was utilized for most subsequent studies. We have also created a recombinant substrate (GST-ACC) by expressing 34 residues around the major phosphorylation site (Ser79) on rat acetyl-CoA carboxylase-1/α (ACC1) as a GST fusion. This was an excellent substrate that was phosphorylated with similar kinetic parameters to ACC1 by both native AMPK and the bacterially expressed kinase domain. We also constructed a structural model for the binding of the ACC1 sequence to the kinase domain, based on crystal structures for related protein kinases. The model was tested by making a total of 25 mutants of GST-ACC and seven mutants of GST-KD, and measuring kinetic parameters with different combinations. The results reveal that AMPK and ACC1 interact over a much wider region than previously realized (>20 residues). The features of the interaction can be summarised as follows: (i) an amphipathic helix from P − 16 to P − 5 on the substrate binds in a hydrophobic groove on the large lobe of the kinase; (ii) basic residues at P − 6 and P − 4 bind to two acidic patches (D215/D216/D217 and E103/D100/E143, respectively), on the large lobe; (iii) a histidine at P + 3 interacts with D56 on the small lobe; (iv) the side-chain of P + 4 leucine could not be precisely positioned, but a new finding was that asparagine or glutamine could replace a hydrophobic residue at this position. These interactions position the serine residue to be phosphorylated in close proximity to the γ-phosphate group of ATP. Although based on modelling rather than a determined structure, this represents one of the most detailed studies of the interaction between a kinase and its substrate achieved to date.

Introduction

Living cells can be regarded as information-processing devices containing a signalling network (analogous to the “central processor unit” of a computer) in which protein kinases and phosphatases play a dominant role. A variety of receptors and other input systems monitor extracellular and intracellular conditions, and this information is processed by the network and converted into appropriate output, i.e. changes in the activity or expression of individual proteins. A key part of this process is the mechanism by which protein kinases recognize their protein substrates. Some protein kinases phosphorylate serine or threonine residues on numerous different target proteins, but need to be able to distinguish these from the hundreds of thousands of other serine or threonine residues that are not in the correct context.

The main superfamily of eukaryotic protein kinases have conserved kinase domains of about 250–300 amino acid residues.1 Structures of several of these kinase domains derived from X-ray crystallography have been reported. In one case, i.e. phosphorylase kinase,2 the structure was determined in the presence of a non-hydrolysable ATP analogue and a peptide substrate. However the peptide substrate only contained seven residues and had a much higher Km than the natural substrate, phosphorylase. This model does not therefore provide a complete description of the manner in which the physiological substrate might bind. In other cases, e.g. the catalytic subunit of cyclic AMP-dependent protein kinase (PKA3) and calmodulin-dependent protein kinase I (CaMKI4) the structures were determined in the presence of inhibitory sequences that are thought to mimic the binding of substrates. These structures provide good models for protein substrate binding, but in no case have the interactions between a kinase domain and a natural protein substrate been studied in molecular detail.

The AMP-activated protein kinase (AMPK) is the downstream component of a protein kinase cascade that acts as a sensor of cellular energy charge by monitoring AMP and ATP.5, 6, 7, 8 ATP depletion can occur either by inhibition of ATP production (e.g. by deprivation of oxygen or a carbon source) or by factors that increase ATP consumption (e.g. exercise in skeletal muscle). Once activated, AMPK phosphorylates a number of targets and this has the overall effects of switching on ATP-producing catabolic pathways, and switching off ATP-consuming processes, both via direct phosphorylation of metabolic enzymes and via effects on gene expression.5, 6, 7, 8 There is considerable current interest in the AMPK system because pharmacological agents that activate the system are valuable in the treatment of Type 2 diabetes,7, 9 and also because mutations in genes encoding regulatory subunits can give rise to hereditary heart conditions.10, 11

Previous studies on substrate recognition by AMPK have utilized short (15-mer) synthetic peptide variants.12, 13 These identified the core recognition motif Φ(X,β)XXS/TXXXΦ, where Φ is a hydrophobic residue (M, L, F, I or V) at the P−5 and P+4 positions (i.e. five residues N-terminal, or four residues C-terminal, to the phosphorylated amino acid), and β is a basic residue (R, K or H) that can be at P−3 or P − 4.

With the publication of crystal structures for several protein kinases, especially that of CaMKI,4 it occurred to us that we could extend this analysis by structural modelling of the interaction between AMPK and a protein substrate. CaMKI is 40 % identical in amino acid sequence to AMPK (α1 isoform) within the kinase domain, and its core recognition motif (ΦXβXXS/TXXXΦ) is very similar to that of AMPK.13 In the crystal structure4 an autoinhibitory sequence binds in the active site cleft, thus blocking access to substrates. This sequence contains the motif FXKXXWXXXF, which fits the recognition motif above except that a non-phosphorylatable tryptophan residue occurs in the position where serine or threonine would be in a substrate. The autoinhibitory sequence is therefore a pseudosubstrate,14 which provides a model for the manner in which a true substrate might bind. For this study, we constructed a model for the kinase domain of AMPK bound to a substrate peptide containing the primary site (Ser79) phosphorylated by AMPK on rat acetyl-CoA carboxylase (ACC1, residues 60–94). Our model (see Figure 1) suggested that the interaction between AMPK and ACC1 might occur over at least 20 residues of the substrate, i.e. a much longer region than the ten residues implied by the previously established core recognition motif. To test the validity of the model, we have expressed the kinase domain of AMPK (α1 isoform) as a catalytically active glutathione-S-transferase (GST) fusion in Escherichia coli. We also cloned DNA encoding residues 60–94 of ACC1 and expressed it as a GST fusion in E. coli. After some modifications this was an excellent substrate for both native rat liver AMPK, and for the bacterially expressed kinase domain. We have tested our model for binding of ACC1 to the kinase domain by making complementary mutations of both kinase and substrate.

Section snippets

Development of the model for AMPK kinase domain with a bound substrate

Although this project was initiated following the publication of the crystal structure of CaMKI,4 our final model (Figure 1) was built as a structural chimera based on the atomic coordinates (1ATP.pdb)15 for a ternary complex between PKA, a peptide inhibitor, and ATP for the N-terminal lobe, and those of an autoinhibited form of twitchin kinase (1KOB.pdb)16 for the C-terminal lobe. The use of two structures rather than one was partly due to the lack of certain regions in the individual

Discussion

Although our approach was similar to those used previously to investigate the binding of autoinhibitory sequences to phosphorylase kinase22 and calmodulin-dependent protein kinase II,23 we believe that this study represents one of the most complete analyses of the interaction between a protein kinase and a protein substrate to date. Our approach was based on a model for the AMPK kinase domain derived from homology with other protein kinases, rather than on real structural data. We also utilized

Materials and bacterial strains

Hepes, Tris, imidazole, Triton X-100, Tween-20, formamide, β-mercaptoethanol, ampicillin, Amberlite mono bead resin and Protein-G horseradish peroxidase conjugate were from Sigma, UK. ATP, AMP, reduced glutathione, leupeptin, pepstatin, anti c-myc antibodies, and Complete EDTA Free Protease Inhibitor Cocktail were from Roche Diagnostics, Lewisham, UK. Glutathione-Sepharose, Chelating-Sepharose, Q-Sepharose, scintillants [γ-32P]ATP, Western blotting detection kits (enhanced chemi-luminescence)

Acknowledgements

This study was supported by a Programme Grant (to D.G.H.) and a Prize Studentship (to J.W.S.) from the Wellcome Trust. We are very grateful to David Carling for gifts of plasmids expressing AMPK and the CCL13 cells.

References (35)

  • P Cohen et al.

    Protein phosphatase-1 and protein phosphatase-2A from rabbit skeletal muscle

    Methods Enzymol.

    (1988)
  • M.M Bradford

    A rapid and sensitive method for the quantitation of microgram quantities of protein utilizing the principle of protein-dye binding

    Anal. Biochem.

    (1976)
  • A Sali et al.

    Comparative protein modelling by satisfaction of spatial restraints

    J. Mol. Biol.

    (1993)
  • S.K Hanks et al.

    Protein kinases 6. The eukaryotic protein kinase superfamilykinase (catalytic) domain structure and classification

    FASEB J.

    (1995)
  • E.D Lowe et al.

    The crystal structure of a phosphorylase kinase peptide substrate complexkinase substrate recognition

    EMBO J.

    (1997)
  • D.R Knighton et al.

    Structure of a peptide inhibitor bound to the catalytic subunit of cyclic adenosine monophosphate-dependent protein kinase

    Science

    (1991)
  • D.G Hardie et al.

    The AMP-activated protein kinasefuel gauge of the mammalian cell?

    Eur. J. Biochem.

    (1997)
  • Cited by (145)

    View all citing articles on Scopus
    1

    Edited by J. Karn

    View full text