AMP-activated protein kinase (AMPK) plays a major role in regulating cellular energy balance by sensing and responding to increases in AMP/ADP concentration relative to ATP. Binding of AMP causes allosteric activation of the enzyme and binding of either AMP or ADP promotes and maintains the phosphorylation of threonine 172 within the activation loop of the kinase. AMPK has attracted widespread interest as a potential therapeutic target for metabolic diseases including type 2 diabetes and, more recently, cancer. A number of direct AMPK activators have been reported as having beneficial effects in treating metabolic diseases, but there has been no structural basis for activator binding to AMPK. Here we present the crystal structure of human AMPK in complex with a small molecule activator that binds at a site between the kinase domain and the carbohydrate-binding module, stabilising the interaction between these two components. The nature of the activator-binding pocket suggests the involvement of an additional, as yet unidentified, metabolite in the physiological regulation of AMPK. Importantly, the structure offers new opportunities for the design of small molecule activators of AMPK for treatment of metabolic disorders.
AMP-activated protein kinase (AMPK) plays an important role in regulating energy homeostasis in eukaryotic cells1,2,3. In response to a decrease in cellular ATP levels, for instance following nutrient deprivation or muscle cell contraction, AMPK is activated by phosphorylation of a threonine residue (Thr-172) within the activation loop of the kinase domain4. Two upstream kinases, liver kinase B1 (LKB1)5,6,7,8 and calcium/calmodulin-dependent protein kinase kinase (CaMKKβ)9,10,11 catalyse phosphorylation of Thr-172 in mammalian cells. Increased AMPK activity leads to a concomitant increase in the phosphorylation of its downstream targets such as acetyl-CoA carboxylase12. AMPK phosphorylates a broad range of substrates involved in diverse cellular processes13,14,15,16,17,18,19,20, but in general terms activation of AMPK leads to a reduction in the rate of anabolic pathways (ATP-utilising) and an increase in the rate of catabolic pathways (ATP-producing)1,2,3. Maintaining a high concentration of ATP relative to ADP is a pre-requisite for eukaryotic cell survival and disturbances in energy homeostasis underlie a wide range of disease states in humans, such as type 2 diabetes and cancer. This key role of AMPK in energy homeostasis makes it an attractive target for the development of drugs aimed at preventing and/or alleviating the detrimental effects of metabolic diseases15.
AMPK is a heterotrimeric enzyme complex composed of a catalytic α-subunit, together with β- and γ-regulatory subunits21,22,23,24,25,26. The α-subunit contains an N-terminal protein kinase domain and a C-terminal regulatory domain. The γ-subunit contains four copies of a cystathionine-β-synthase domain27, and we previously showed that the γ-subunit binds three molecules of adenine nucleotide28. One molecule of AMP is bound tightly and is not exchangeable in solution (Site-4), although soaking studies have suggested that this AMP can be exchanged with ATP in crystalline AMPK29,30. Nucleotides can bind reversibly to Site-1 and Site-3 in solution, providing a mechanism by which AMPK can respond to changes in ATP levels28. Adenine nucleotide binding regulates AMPK activity by three mechanisms. First, AMP causes a two- to threefold allosteric activation31,32. Second, it has recently been reported that ADP and AMP promote phosphorylation of Thr-172 by LKB1 and CaMKKβ when the β-subunit is myristoylated at its N-terminus3,33,34. Third, ADP and AMP protect Thr-172 against dephosphorylation35,36,37,38, and we previously reported the structure of an active form of AMPK phosphorylated on Thr-172, which allowed us to suggest how this mechanism might occur38.
In addition to regulation by adenine nucleotides, a number of small molecules have been identified that directly activate AMPK39,40. The first of these to be reported was A-769662 (refs 39, 41, 42, 43). Two subsequent studies42,43 showed that activation by A-769662 required the presence of the carbohydrate-binding module (CBM, also known as the glycogen-binding domain) at the N-terminus of the β-subunit, since it shares sequence similarity with a domain found in a number of proteins that bind carbohydrates44,45. The structure of the isolated CBM from AMPK β1 in complex with β-cyclodextrin has been reported46 together with the corresponding region from a yeast analogue as part of a truncated heterotrimer47. However, the molecular architecture of drug binding and activation of AMPK has not been determined. The kinase domain of the α-subunit and the CBM of the β-subunit are connected to their C-terminal scaffold domains by flexible linkers. Here we report the crystal structure at 3 Å resolution of AMPK bound to a small molecule activator.
Role of carbohydrate-binding module in activator binding
Our previously published structure of an active AMPK complex contained full-length versions of the α- and γ-subunits but the N-terminal domain of the β-subunit, including the CBM, was engineered out of the construct to facilitate crystallization38. Although this construct is allosterically activated by AMP, it is not by A-769662 (ref. 36), and it does not show protection against dephosphorylation by AMP or A-769662 (Supplementary Fig. 1). We recently solved two different crystal forms of AMPK constructs that include the CBM but in neither case was there electron density present for this domain. Given that the CBM is required to achieve protection against dephosphorylation by small molecule activators, we reasoned that the presence of one of these compounds might stabilise the interaction between the CBM and the rest of the complex. Initial attempts at co-crystallisation with A-769662 were unsuccessful likely because of the relatively weak binding, poor solubility and the presence of additional, non-specific, binding sites that promote protein aggregation. We therefore looked for an alternative compound for co-crystallisation. From structures deposited in patent databases40, we selected and synthesised a cyclic benzimidazole derivative developed by Merck Sharp and Dohme Corporation and Metabasis Therapeutics , hereafter referred to as compound 991 (Fig. 1a and Supplementary Fig. 2). We found that 991 was 5- to 10-fold more potent than A-769662 in assays monitoring allosteric activation and protection against dephosphorylation (Fig. 1b and Table 1). We then determined the effectiveness of 991 in cell-based assays and found that it showed activation at much lower concentrations than required for A-769662 (Fig. 1c), in keeping with the enhanced dose response seen in vitro. Consistent with our previous studies with A-769662 (ref. 42), we found that 991 did not activate a complex lacking the CBM and that the dose response curve was shifted to the right upon mutation of Ser-108 (Fig. 1d).
Next, we compared binding of 991 and A-769662 to AMPK complexes using Biolayer Interferometry (BLI) and circular dichroism (CD) assays (see Fig. 2, Table 2 and Supplementary Methods). Several conclusions emerged:
Both the CBM and the kinase domain are required for effective compound binding.
Compound 991 binds about 10-fold tighter than A-769662.
The activator dose–response curves monitoring AMPK activity, both in allostery and protection against dephosphorylation formats (Fig. 1), show the same patterns as the binding studies (Tables 1 and 2).
The CD data also establish that there is a single tight site for 991 binding whose structure we describe in the next section. While there was no CD change upon binding of A-769662, this compound competed for the binding of 991 to AMPK. Together, these data establish that there is a single-binding site common to A-769662 and 991, and likely related compounds40. The site requires both the CBM and kinase domain, and is responsible for both the allosteric activation and protection against dephosphorylation by small molecule activators but not by nucleotides. Interestingly, activation of AMPK by salicylate has also been shown to be ablated by removal of the CBM, or the mutation of Ser-108, implying that it too binds at the same site48.
Structure of AMPK/activator complex
We solved the structure of full-length human α2β1γ1 AMPK bound to compound 991, and crystallographic statistics are presented in Table 3. The overall structure is shown in cartoon representation in Fig. 3a,b. The organization of the kinase domain with the regulatory complex (comprising the C-terminal domains of the α- and β-subunits, together with the γ subunit) is similar to our earlier report of the structure of the ΔCBM construct38 (the main-chain r.m.s.d. is 5.9 Å). In our new structure, the CBM is ordered and bound to the N-terminal lobe of the kinase domain. The interface between these two domains generates the binding cavity for 991 (Fig. 3 and Supplementary Fig. 3).
One end of the five stranded β-sheet that constitutes the core of the N-terminal domain of the kinase packs against a pair of anti-parallel β-strands from the CBM. 991 sits in a pocket generated at this interface and interacts with hydrophobic residues from each domain, mainly involving ring-1 and ring-2 of the activator (Figs 1a and 3c). Interestingly, the phosphorylated serine (pSer) at position 108 from the CBM is involved in a network of electrostatic interactions with Lys-31(Kinase) and potentially Thr-21(Kinase) and Asn-111(CBM) (Fig. 3d). The role of pSer-108(CBM) at the domain interface corroborates our earlier findings that this residue is important for the regulation and binding of compounds to AMPK42. We carried out site-directed mutagenesis experiments to test the importance of some of these polar interactions (Tables 1 and 2). Mutations of Lys-29(Kinase)/Lys-31(Kinase), Arg-83(CBM) and Ser-108(CBM) all lead to more than a 25-fold reduction in activator affinity, without altering the regulation of AMPK by AMP (Table 1).
Utilizing the crystallization conditions for 991, we were able to refine conditions that gave crystals of AMPK in complex with A-769662. Although the diffraction data were limited to 3.9 Å, unbiased electron density maps (Supplementary Fig. 4) clearly show that A-769662 is located in the same binding site as 991 and how it overlaps with ring-1 and ring-2 of 991 (Fig. 3e). The two activator complex structures therefore confirm the interpretation of our functional and binding data that A-769662 and 991 activate AMPK through binding to the same site at the interface of the kinase domain and CBM.
An α-helix immediately C-terminal to the CBM, absent in previous crystal structures of the CBM38 and just outside of the CBM consensus sequence, interacts with the αC helix of the kinase (hereafter, the C-interacting helix) (Supplementary Fig. 5). The two helices are oriented approximately antiparallel with three residues at the N-terminus of the latter (Val-162, Phe-163 and Leu-166) making hydrophobic contacts with the αC helix. Sequence alignment of this region demonstrates that the C-interacting helix is conserved in AMPKs across numerous species (Supplementary Fig. 6; note, there is a high sequence identity of the entire β1 sequence across all species (over 95%)). We designed a mutation in the C-interacting helix with the aim of blocking its interaction with αC (Leu-166->Glu). The resulting AMPK complex showed a reduction in allosteric activation by 991 while, importantly, allosteric activation by AMP was not affected (Supplementary Fig. 7). The role of the C-interacting helix in mediating allosteric activation by 991 has close parallels in other kinases49. For example, cyclin-dependent kinases (CDKs) are allosterically activated by binding of their regulators, the cyclins. An alpha helix from the cyclin packs against the αC helix of the CDK kinase domain promoting an active conformation (Supplementary Fig. 8).
Most structural features of the activator complex of AMPK presented here show the hallmarks of an activated kinase but one feature of the kinase domain does not (a detailed comparison of the CBM/activator/kinase structure with our earlier ΔCBM structure is presented in Supplementary Figs 9–12). In brief, the residues that contribute to the regulatory and catalytic spines are all in place, similar to our earlier structure, consistent with an active kinase structure. Similarly, the activation loop is fully ordered and continues to mediate the interaction of the kinase with the regulatory fragment. However, the top of the C-helix is rotated away from the active site, relative to our earlier structure, such that the Lys-45->Glu-64 interaction is broken (which is regarded as essential for efficient phosphoryl transfer). This interaction is still intact in the CDK–cyclin complex despite the similar structural packing of the CBM and cyclin against the αC helices of AMPKα and CDK, respectively. Importantly, we are able to activate AMPK in cells and in cell-free assays by 991 and A-769662, suggesting that this Lys-Glu interaction does form during the catalytic cycle. This ‘swung out’ αC conformation has been widely seen in other systems50, and it has been argued that the multiple observations of this conformation suggest that it highlights conformational changes required for an active kinase. Specifically, since ADP release is considered to be the rate-limiting step in the catalytic cycle, and not phosphoryl transfer, it is thought that a conformational flip in the DFG motif is necessary for this slow step and that movement of the C-helix facilitates this. Thus, by analogy with other kinases the conformation captured in our activator complex, in the absence of ATP and substrate, suggests that facilitation of movement in the αC-helix may be an important contribution to activation of AMPK.
α2-hook and the autoinhibitory domain (AID)
The complex used in this study contains the α2 isoform, compared with α1 in our previous structure38, and in both cases there is a segment in the α-subunit, between the N-terminal kinase domain and the C-terminal scaffolding domain, that interacts with one of the exchangeable nucleotide-binding sites (Site-3) on the γ-subunit (Fig. 4). In the current structure, we term this α2-hook residues (defined as residues α2 365–371 and referred to as α-RIM in Chen et al.51), and it is very similar for that seen for α1, suggesting that the hook plays a similar role in communicating the nucleotide status at AXP Site-3 (see below). The improved resolution of the X-ray data used in the present study makes the identification of the sequence register in the α2-hook region unambiguous, and it is different from the sequence register we built into our earlier α1 structure. Wu and colleagues51 have recently suggested an alternative way of building the hook sequence in our earlier electron density maps for the α1 complex and we agree with their reinterpretation of the sequence register (α1, 359–365; α2 365–371). We have reviewed our previous structure of α1 in the light of this and the current structure and found some additional areas that can now be interpreted more fully than was possible before (Supplementary Figs 13–15) and have deposited revised coordinates at the Protein Databank (PDB ID: 4CFH).
Comparison of the α1 and α2 structures shows that the hook regions make the same interactions over the five residues that are conserved between them (Pro-His-Pro-Glu-Arg) (Fig. 4b). Interestingly, Glu-368(α2-hook) makes salt-bridges with Lys-170(γ) and potentially Arg-70(γ) (Fig. 4b), and the latter residue we have argued mediates the initial signal that distinguishes between AMP/ADP and ATP being bound at the exchangeable adenine nucleotide Site-3 (ref. 38). Mutation of Glu-368(α2-hook) ablates nucleotide protection against dephosphorylation (Fig. 4c) consistent with our proposal that site-3 mediates this effect38. We also find that this mutation reduces allosteric activation by AMP (Fig. 4c) as reported previously51. How this mutation affects both aspects of nucleotide regulation is discussed below.
Chen et al.51 have also suggested how their isolated AID (autoinhibitory domain) crystal structure can be docked onto our earlier ΔCBM structure. In our current structure there is significantly better density for the α3 helix of the AID where the sequence register is now convincing, and reasonable main chain, but not side-chain, definition for α1 helix (Supplementary Fig. 14). Again there is not much electron density for other parts of the domain, but our new structure further supports the interpretation of Chen et al.51 that the α1 and α3 helices of the AID adopt a similar structure in the isolated domain and in full-length AMPK. Comparison of our structures with the yeast kinase/AID structure (PDB ID: 3H4J) shows that the AID has undergone a major rigid-body rotation such that it mostly interacts with the regulatory fragment and not the kinase. What does this mean for the autoinhibitory function? The AID can exist in a kinase-bound state and in a regulatory fragment-bound state. The former is understood to be inhibitory and the latter active. Given its location between the kinase domain and the α-hook, it is plausible that the switch in position of the AID occurs in response to which AXP is bound to the regulatory fragment, providing a signalling mechanism for nucleotide regulation52.
Protection against dephosphorylation
In our previous paper38, we suggested that AMPK is protected against dephosphorylation by AMP/ADP binding to the γ-subunit and stabilizing the interaction of the α-hook, and thus enhancing the recruitment of the kinase domain to the AMPK regulatory fragment. In the current structure, as in the ΔCBM structure, nearly all the contacts with the regulatory fragment are made by residues on the activation loop of the kinase, thus stabilizing the activation loop structure. While the phosphate group of Thr-172 is partially solvent exposed, it is not accessible to the active site of a phosphatase. To be dephosphorylated, the activation loop must undergo a conformational change that enables the phosphate of pThr-172 to be flipped out of the RD pocket of the kinase. The interactions of the regulatory fragment of AMPK with the activation loop thus leads to protection of pThr-172 (refs 1, 2, 38).
In our current structure, the CBM is recruited to the core complex by interaction with the kinase domain (Figs 3 and 4d). The interaction of the CBM with the kinase domain similarly makes it less likely that the kinase domain will dissociate from the regulatory fragment. Thus, by promoting this CBM/Kinase/regulatory–fragment interaction, 991 and A-769662 enhance protection of the activation loop against dephosphorylation. As mentioned above (and in Supplementary Fig. 1), protection against dephosphorylation by AMP requires the CBM thus implying that the CBM/kinase interface described here is important for protection against dephosphorylation by physiological regulators. The fact that nucleotide-mediated protection against dephosphorylation is a weaker effect than that shown by compounds is consistent with the interaction between CBM/kinase being weaker in the absence of 991.
As described above, mutation of Glu-368(α2-hook) reduces the extent of allosteric activation, as well as protection against dephosphorylation. Does this imply that is site-3 responsible for both aspects of activation? We previously presented data that demonstrated that the AMP dose dependence of the allosteric effect was consistent with it being mediated by binding to the tighter nucleotide binding site (site-1) while the protection effect had a weaker dose response consistent with site-3. These observations are reconciled on the assumption that allosteric activation by AMP occurs when the kinase is bound to the regulatory fragment. Indeed, it is difficult to envisage how conformational changes in the regulatory fragment could be propagated to the kinase domain unless it is bound to the regulatory fragment. Therefore, mutations that reduce the proportion of AMPK molecules with ‘bound kinase’, like Glu-368->Ala, should reduce the extent to which AMP can activate.
Our current structure and binding studies reveal the site on AMPK that mediates the activation by compounds, including A-769662, 991 and perhaps salicylate. Inspection of the CBM/kinase interface, with the activator removed (Supplementary Fig. 16), reveals that the protein subunits make contacts around the rim of the interface but leave a large, significantly hydrophobic, cavity where the activator binds. It is tempting to speculate, therefore, that this cavity has evolved to bind a natural ligand that acts to stimulate AMPK. Although we have not yet managed to identify this ligand we think that it is likely to exist and that it represents an important missing link in our understanding of the physiological regulation of AMPK.
The following antibodies were used in this study. The source and catalogue number are indicated in parentheses. anti-GFP (Clontech, Catalogue number 632381), anti-pACC (Cell Signalling, Catalogue number 3661), anti-pT172 (Cell signalling Catalogue number 2535), anti-AMPKβ1/β2 (Cell signalling Catalogue number 4150), Goat anti-rabbit HRP (ProteinSimple, Catalogue number 71598) and Goat anti-mouse IRDye 800CW (Li-Cor, Catalogue number 926-32210).
Human AMPK subunits (His-α1/2, β1/2 and γ1) were cloned into a pET-3d vector, and successive rounds of cloning were performed to generate a tricistronic vector similar to that previously reported (Neumann et al.,53). For BLI and crystallization experiments an AVI-tag (GLNDIFEAQKIEWHE) was engineered onto the N-terminus of the β-subunit and cloned into a pET-3d vector. The tricistronic vector was generated as previously reported (Neumann et al.,53). Biotin ligase (BirA) lacking a His-tag was amplified from His-BirA pET-47 vector (a kind gift from Vangelis Christodoulou) and cloned into pET-47 vector.
Recombinant AMPK complexes were expressed in E. coli BL21 cells (BL21-one shot DNA Star, Stratagene) and purified by nickel affinity chromatography (His-Trap, GE Healthcare) and gel filtration (Superdex 200, GE Healthcare). AMPK complexes were phosphorylated by incubation with CaMKKβ in the presence of 0.5 mM ATP, 2.5 mM MgCl2 and 0.5 mM AMP overnight at 18 °C as previously reported38. The phosphorylated AMPK complex was repurified by Nickel-sepharose and gel filtration. For the BLI experiments, the AVI-tagged AMPK complexes were coexpressed with a biotin ligase (BirA) to generate biotinylated AMPK that could be purified (as above) and immobilized onto superstreptavidin sensors on an Octet RED biolayer interferometer. The growth media for expressing the biotinylated complexes was supplemented with 50 μM D-Biotin. It was discovered later in this study that the endogenous bacterial BirA produced sufficient biotinylated AMPK complex without overexpressing the BirA in bacteria and supplementing the media with biotin.
AMPK functional assays
For allosteric effects, AMPK activity was determined by phosphorylation of the SAMS peptide in the presence or absence of varying concentrations of compound, as indicated in the appropriate figures. Results are plotted as fold activation relative to the activity in the absence of added compound and are the mean±s.e.m. from at least three independent experiments. For dephosphorylation studies, an aliquot of phosphorylated AMPK was incubated in 50 mM HEPES, pH 7.4, 100 mM NaCl, 2.5 mM MgCl2, in the presence or absence of recombinant PP2Cα (26 ng) and in the presence or absence of varying concentrations of compound for 20 min at 37 °C (or as indicated in the appropriate figure legends). For determination of AMPK activity, the reaction mixture was diluted 1:25 in 50 mM Hepes, pH 7.4 to prevent further dephosphorylation. The appropriate compounds were added to the diluted samples in order to equalize their final concentration in all samples and AMPK activity measured using the SAMS peptide assay. Results are plotted as a percentage of the activity measured in the absence of PP2C and are the mean±s.e.m. from at least three independent experiments.
Endogenous AMPK activity in HEK293 cells
HEK293 cells (from American Tissue Culture Collection (ATCC)) were grown at 37 °C with 5% CO2 in Dulbecco’s modified Eagle’s medium (4,500 mg l−1 glucose with sodium pyruvate and pyridoxine) supplemented with 10% fetal calf serum and 1 mM glutamine42. Cells were transferred into serum-free media for 2 h before treatment with varying concentrations of A-769662 or 991, as indicated in the figure legends. Cells were washed briefly with ice-cold phosphate-buffered saline before rapid lysis in ice-cold buffer (50 mM HEPES, pH 7.4, containing protease inhibitor mixture (Roche Applied Science), 1 mM EDTA, 10% glycerol (v/v), 50 mM NaF and 1% (v/v) Triton X-100). Insoluble material was removed by centrifugation at 10,000 g for 10 min at 4 °C, and the supernatant was used for subsequent analysis.
Transient transfection of HEK293 cells
Plasmid DNA was prepared using a Qiagen maxiprep kit according to the manufacturer’s instructions. Cells were transfected by Ca2+-phosphate precipitation with 10 μg of each plasmid encoding myc-α1 and FLAG-γ1 subunits, and either wild-type β1, β1 harbouring the S108A mutation or lacking the CBM (ΔCBM; β1, 186–270)42. Thirty hours post transfection, cells were collected as described above. cDNA encoding the β1 subunit was cloned into a vector to allow expression of a fusion protein with green fluorescent protein (N-terminal of β1). This allowed us to determine the expression of the β1 subunit using western blot analysis. The primary antibody was used at 1:1,000 dilution, and the secondary antibody was used at 1:10,000 dilution.
Immunoprecipitation of AMPK from mammalian cells
Endogenous AMPK was immunoprecipitated using a rabbit antipan-β antibody42 (1:10,000 dilution). Recombinant AMPK from transfected cells was immunoprecipitated using an anti-FLAG antibody. Immune complexes were washed extensively, and AMPK activity was determined by performing an SAMS assay.
Simon western blot analysis (ProteinSimple, California, USA)
In some cases, western blot analysis was performed using a capillary-based automated system (http://www.proteinsimple.com/simon.html). The western blot was performed using the standard manufacturer’s protocol using a primary antibody dilution of 1:200 and the secondary antibodies from ProteinSimple were used neat.
The binding of A-769662 and A-769662-like compounds to different AMPK constructs was measured on an Octet RED biolayer interferometer (Pall ForteBio Corp., Menlo Park, CA, USA). AMPK constructs biotinylated by addition of an AviTag sequence to the N-terminus of the β subunit were immobilized on superstreptavidin (SSA) biosensors (Pall ForteBio Corp., Menlo Park, CA, USA) at a concentration of 40 μg ml−1. Binding of compounds at concentrations in the range 0.75 to 10 μM was measured at 25 °C using 1–5 min association steps and 1–10 min dissociation steps. The buffer was Fortebio kinetics buffer (10 mM phosphate (pH 7.4), 150 mM NaCl, 0.1 mg ml−1 BSA and 0.005% Tween-20) supplemented with 1 mM TCEP and a final DMSO concentration of 2%. In the majority of cases, some non-specific binding and/or instrument drift meant that the association and dissociation phases could not be fit to a single-exponential function. The single observed rate (kOBS) employed in the analyses was the one that accounted for at least 85% of the total observed reaction amplitude (Supplementary Methods). Kinetic constants (kon and koff) were obtained from plots of the association phase kOBS values versus compound concentration. In some cases an accurate value for koff could not be determined from such plots, and in these cases this constant was determined independently from analysis of the dissociation phase (see Supplementary Information).
Near-UV CD spectra (340–255 nm) were recorded on a Jasco J-815 spectropolarimeter fitted with a cuvette holder thermostatted by a CDF-426S/15 Peltier unit. All measurements were made at 20 °C in 10-mm path length fused silica cuvettes (Hellma, Jena, Germany). Spectra were typically recorded with 0.1-nm resolution and baseline corrected by subtraction of the appropriate buffer spectrum (50 mM Tris, pH 8.0, 300 mM NaCl and 1 mM TCEP). Molar CD extinction coefficients (ΔεM) were calculated by dividing with the protein concentration.
Full-length AMPK, γ1(human, 1–331) AVI-tag-β1(human, 1-270) His-α2(human, 1–552), was cloned into a tricistronic vector and subsequently expressed in E. coli BL21 Star (DE3). Proteins were purified using a nickel affinity chromatography and gel filtration. The purified protein was phosphorylated using CAMKKβ kinase overnight at 18 °C then repurified by Nickel affinity chromatography and gel filtration. The complex stock solution was prepared at 7 mg ml−1 in 50 mM Tris, pH 8.0, 300 mM NaCl and 1 mM TCEP, mixed with a threefold molar excess of AMP and onefold of staurosporine and 991 compound. Crystals were grown by vapour diffusion technique at 4 °C in hanging drops. Drops were prepared by mixing equal volumes of protein complex with reservoir solution containing 13% PEG3350, 0.1 M MgCl2, 1% Glucose, 0.15% CAPB in 100 mM Imidazole (pH 6.2). Crystals were first transferred into mother liquor with an additional 30% ethylene glycol, before plunging into liquid nitrogen. Diffraction data were collected on a Pilatus 2 M detector (Dectris), Diamond Light Source, Oxford. Data were integrated using Denzo and scaled with Scalepack54. The structure was solved by molecular replacement using Phaser55 and standard refinement was carried out with Phenix56 using 2Y94.pdb and 2F15.pdb as the search models, with manual model building with COOT57. General crystallographic calculations were carried out using the CCP4 package58. Figures were created with Pymol (http://pymol.sourceforge.net/).
How to cite this article: Xiao, B. et al. Structural basis of AMPK regulation by small molecule activators. Nat. Commun. 4:3017 doi: 10.1038/ncomms4017 (2013).
Accession codes: Atomic coordinates and structure factors have been deposited in the Protein Data Bank under accession codes 4CFE (AMPK:991 complex), 4CFF (AMPK:A-769662 complex) and 4CFH (AMPK ΔCBM, revised co-ordinates).
Protein Data Bank
Work in both laboratories is supported by the MRC. B.X., M.J.S., L.F.H., E.U., R.B.H., P.A.W., S.J.G. were funded by the Medical Research Council through programme U117584222, S.M. by the programme U117570592 and D.C. N.J.B., B.R.P. and D. Carling by programme U1200275737. We gratefully acknowledge Diamond Light Source for synchrotron access on grant 7707. We thank Stephen Smerdon and John Skehel for advice and critical reading of the manuscript and Martin Noble and Jane Endicott for valuable discussions.
Supplementary Figures 1-22, Supplementary Table 1, Supplementary Methods and Supplementary Reference