Mechanisms of Activation and Inhibition of the Metabolic Regulator AMP-Activated Protein Kinase (AMPK)

Mechanisms of activation and inhibition of the metabolic regulator AMP-activated protein kinase (AMPK)

Toby Dite

Submitted in total fulfilment of the requirements of the degree of Doctor of Philosophy to the University of Melbourne

February 2018

Metabolic Signalling Unit St. Vincent’s Institute of Medical Research

Department of Medicine, St. Vincent’s Hospital The University of Melbourne

Abstract Biochemical energy in the form of ATP is crucial for normal metabolic function of all cells. Relative levels of AMP, ADP and ATP reflect the energy status of the cell, and are described in a single parameter called the adenylate energy charge. The master metabolic regulator of the cell, AMP-activated protein kinase (AMPK), maintains energy homeostasis by responding to fluctuations in adenylate energy charge, and adjusting ATP consuming and producing pathways.

AMPK activation can lower plasma glucose, reduce de novo cholesterol and fatty acid synthesis and increase fatty acid oxidation, which has encouraged the development of many small molecule activators that directly activate AMPK. AMPK activity can also play a deleterious role in cancer cell survival, neurodegeneration and stroke, which has alternatively highlighted AMPK inhibition as a potential therapeutic strategy, although small molecule antagonists are currently limited.

Contextualising AMPK regulation in an intracellular environment requires simultaneous monitoring of adenine nucleotide ratios. While this is not routine in the field, I have used mass spectrometry to highlight several scenarios where quantitation of adenine nucleotide ratios is critical.

I have shown that glucose starvation is able to decrease adenylate energy charge in some cells, but not others. I have also shown that high concentrations of any small molecule may indirectly activate AMPK by increasing [AMP]/[ATP]. Finally, whereas the adenylate kinase equilibrium predicts that [AMP]/[ATP] will increase as the square of [ADP]/[ATP] during stress, I have shown empirically that, under most circumstances, this is not the case. Together, these results demonstrate that measuring adenine nucleotide ratios is fundamental to distinguish novel mechanisms from canonical mechanisms of intracellular AMPK regulation.

I next investigated the regulation of AMPK by allosteric drugs. The allosteric drug and metabolite (ADaM) site on AMPK is the most commonly reported drug binding site in AMPK/drug crystal structures. Work conducted prior to my PhD has shown that phosphorylation of AMPK at β1-Ser108 stabilises the ADaM site and is critical for activation by many drugs, and enhances activation by others. I have identified and confirmed ULK1 as an upstream kinase of AMPK β1-Ser108. Phosphorylation of AMPK by ULK1 sensitized AMPK to activation by two allosteric drugs, A-769662 and salicylate. Pharmacological inhibition or genetic deletion of ULK1 in cells reduced β1-Ser108 phosphorylation on

iii

recombinant AMPK in cells, and increases in β1-Ser108 phosphorylation. In addition, mutation of Thr172 to Ala did not prevent AMPK pathway signalling under circumstances of increased β1-Ser108 phosphorylation in the presence of A-769662, demonstrating the first activation loop-independent AMPK signalling in cells.

Finally, I have identified and characterised SBI-0206965 as a novel type II inhibitor of AMPK.

SBI-0206965 was able to inhibit both α1- and α2-AMPK heterotrimers with nanomolar IC50 values. SBI-02069965 displayed preferable potency and selectivity profiles compared to compound C, and was able to inhibit cellular AMPK signalling across a range of AMPK- activating stimuli. Co-crystallisation of AMPK α2-kinase domain with SBI-0206965 revealed that it binds to AMPK at the active site, in a catalytically inactive, DFG-out conformation. Biochemical characterisation of SBI-0206965 showed that, with regard to ATP competition, it is a mixed-type inhibitor, and is able to maintain potency in activity assays when [ATP] is high. This discovery is the first example of a type II AMPK inhibitor, which is more selective and potent that currently available inhibitors. SBI-0206965 is a promising lead for small molecule AMPK inhibitor development against a backdrop of increasing applications for AMPK inhibition.

In conclusion, work in this thesis provides insights into mechanisms of AMPK activation and inhibition that may be useful for the development of AMPK drugs with clinical potential. Current small molecule activators of AMPK have not reached clinical trials, and the handful of existing small molecule inhibitors of AMPK are poor. Exploiting endogenous mechanisms that increase AMPK β1-Ser108 phosphorylation, such as activating ULK1, could be used as a potential strategy for increasing AMPK pathway activation by drugs. Using SBI-0206965 as a research tool, and as a platform for the development of even better antagonists will further enhance our understanding of the roles of AMPK, and may lay the foundation for a new class of therapeutic AMPK inhibitors. Essential to the understanding of both AMPK activation and inhibition will be the adoption of adenine nucleotide quantitation as standard practice in the field.

Declaration

This is to certify that:

i) This thesis comprises only my original work towards the degree Doctor of Philosophy except where indicated in the Preface,

ii) Due acknowledgement has been made in the text to all other material used, iii) The thesis is fewer than 100,000 words in length, exclusive of tables and bibliography.

Toby Dite

Preface

Pursuant to the regulations governing the degree of Doctor of Philosophy at the University of Melbourne, I hereby submit that:

I. This thesis contains no material that has been accepted for the award of any other degree or diploma in any university.

II. To the best of my knowledge and belief, this thesis contains no material previously published or written by another person, except where due reference has been made.

Chapter 3 of this thesis contains findings that were published in a peer-reviewed journal and result from the collaboration with others in the scientific community. I declare that I am the primary author of the publication, having contributed >50% of the work presented. The declaration for thesis with publication forms for the paper have been submitted with this thesis, along with the signed co-author collaboration authorisation documents confirming collaborator contributions and that my contributions were >50%.

Chapter 4 of this thesis contains findings that were published in a peer-reviewed journal and result from the collaboration with others in the scientific community. I declare that I am the primary author of the publication, having contributed >50% of the work presented. The declaration for thesis with publication forms for the paper have been submitted with this thesis, along with the signed co-author collaboration authorisation documents confirming collaborator contributions and that my contributions were >50%.

I would like to acknowledge the help of others in contributing to the work presented in this thesis:

Dr Sandra Galic performed primary hepatocyte and primary neuron isolation, and SBI- 0206965 cell treatments on SH-SY5Y cells.

Richard Rebello and Dr Luc Furic (Peter MacCallum Cancer Centre, Melbourne) performed proliferation assays using PC3 cells and SBI-0206965.

vii

Dr Kevin Ngoei performed ATP-agarose binding assays using SBI-0206965 and compound C.

Dr Christopher Langendorf co-crystallised α2 kinase domain and SBI-0206965, collected X- ray diffraction data and solved the crystal structure.

Dr Jon Oakhill performed the LKB1-Thr172 phosphorylation assay using SBI-0206965.

Dr John Scott analysed kinetic data relating to SBI-0206965.

Dr Naomi Ling performed lentiviral transfection of WT, AMPK knock-out and ULK1/2 knockout MEF cells.

Dr Ashfaqul Hoque performed phosphoproteomic experiments and analysis.

Dr Benjamin Parker performed MS/MS analysis on ULK1 phosphorylated AMPK.

viii

Publications

The following publications are a result of work generated for this thesis:

Scott, J. W., Galic, S., Graham, K. L, Foitzik, R, Oakhill, J. S., Ling, N. X. Y, Dite, T. A., Langendorf, C. G., Weng, Q.P., Thomas, H.E., Birnberg, N.C., Kemp, B.E., and Oakhill, J. S (2015) Inhibition of AMP-Activated Protein Kinase at the Allosteric Drug-binding site Promotes Islet Insulin Release, Chemistry and Biology.

Dite, T. A., Ling, N. X. Y, Scott, J. W., Galic, S., Parker, B. L., Ngoei, K.R.W, O’Brien, M. T., Hoque, A., Kundu, M., Steinberg, G.R, Sakamoto, K., Kemp, B.E. and Oakhill, J. S. (2017) The Autophagy Initiator ULK1 sensitizes AMPK to Allosteric Drugs, Nature Communications.

Lindqvist, L. M., McArthur, K., Dite, T. A., Lazarou, M., Oakhill, J. S., Kile, B. T., Vaux, D. L. (2017) Autophagy induced during Bax/Bak-mediated apoptosis degrades mitochondria and inhibits Type I interferon secretion, Cell Death and Differentiation.

Dite, T. A., Langendorf, C. G., Hoque, A., Galic, S., Rebello, R. J., Ovens, A. J., Lindqvist, L. M., Ngoei, K. R. W., Ling, N. X. Y., Furic, L., Kemp, B. E., Scott, J. W. and Oakhill, J. S. (2018) AMP-activated protein kinase selectivity inhibited by the type II inhibitor SBI-0206965. Journal of Biological Chemistry.

Acknowledgments

This study has only been made possible through the support and contributions of many people. First of all, I would like to thank my supervisors Dr. Jon Oakhill and Prof. Bruce Kemp. Thank you for picking me up as an honours student in 2012, and supporting and mentoring me through to the start of my first post doc position. Jon, my enthusiasm for science has been cemented under your supervision, and without your guidance I would never have made it to where I am today. Bruce, your dedication to science and vast knowledge of the field have guided me through the toughest problems in my PhD.

I was lucky to have worked and studied in the Protein Chemistry and Metabolism, and Metabolic Signalling labs at SVI. I’d like to thank all the past and present members of these labs, who have helped me with experiments, and shared a drink with me at the pub. John Scott, Matt O’Brien, Naomi Ling, Sandra Galic, Kevin Ngoei, Chris Langendorf, Kim Loh, Sam Issa, Frosa Katsis, Lisa Murray-Segal and Vy Hoang, thank you for your hard work, help and company over the last 5 years. The endless hours I have spent in the lab, outside of the lab, at conferences like Lorne and on Manila Tarmac with you have taught me invaluable lessons, and will shape the rest of my career.

I would also like to thank the staff and students at SVI, who have created an amazing environment to work and study in. I’m confident that the student community at SVI is the best in Melbourne, and I have made many friendships in the course of my studies that I know will carry on for the rest of our lives.

In particular I’d like to thank Christina Vrahnas, who has been my friend since I started at SVI. You have helped me through every stage of my time in science so far, and I am lucky that we will be taking the next steps of our lives together in Scotland.

Finally I’d like to thank my parents, who have supported me through every stage in my life. Without you I would never have been able to start or get through my PhD. I love you both.

List of Figures Box 1 The adenylate energy charge (AEC) of the cell is equal to half of the kinase-accessible phosphoanhydride bonds in the adenine nucleotide pool...... 2 Figure 1- 1 3D structure, subunit schematic and post-translational modifications of AMPK...... 6 Figure 1- 2 Effects of AMPK activation on cell metabolism ...... 17 Figure 1- 3 The Allosteric Drug and Metabolite (ADaM) site ...... 27 Figure 2- 1 Mass chromatogram and standard curves of AMP, ADP and ATP ...... 43 Figure 2- 2 Endogenous AMPK activity reduces stress induced changes to adenylate energy charge ...... 46 Figure 2- 3 The effect of glucose starvation on the adenylate energy charge of different cells .... 49 Figure 2- 4 High concentrations of small molecules can increase [AMP]/[ATP] ...... 51 Figure 2- 5 ...... 54 Figure 2- 6 Mechanisms of indirect activation of AMPK ...... 55 Figure 2- 7 The effect of metabolic stress on [AMP]/[ATP], [ADP]/[ATP] and predictions of [AMP]/[ATP] made using the adenylate kinase equilibrium ...... 58 Figure 3- 1 A-769662 activation of cellular AMPK signaling is dependent on β1-pSer108...... 66 Figure 3- 2 Quantitative global and phosphoproteomic analysis uncovers cellular roles for β1- pSer108 ...... 67 Figure 3- 3 β1-Ser108 trans-phosphorylation occurs via an AMPK independent mechanism .... 68 Figure 3- 4 ULK1 phosphorylates β1-Ser108 in vitro ...... 69 Figure 3- 5 ULK1 phosphorylation of Ser108 is specific for the AMPK β1 isoform ...... 70 Figure 3- 6 ULK phosphorylates β1-Ser108 in cells...... 71 Figure 3- 7 An AMP-myristoyl switch triggers ULK1 phosphorylation of β1-Ser108 ...... 72 Figure 3- 8 Cellular AMPK signaling occurs independently of α1-pThr172...... 73 Figure 3- 9 An integrated model for ULK1 regulation of β1-AMPK signaling ...... 74 Figure 4- 1 Biochemical characterization of SBI-0206965 ...... 83 Figure 4- 2 Kinase selectivity profile for SBI-0206965 (0.25 μM) and compound C (2.5 μM). .... 84 Figure 4- 3 SBI-0206965 is an AMPK type II inhibitor ...... 85 Figure 4- 4 Crystal structure of α2 kinase domain complexed to SBI-0206965 ...... 86 Figure 4- 5 SBI-0206965 suppresses AMPK signalling in diverse cultured cell lines ...... 87 Supplementary Figure 4- 1 Kinase inhibitor profiles for (A) SBI-0206965 (250 nM) and (B) compound C (2.5 μM) ...... 95 Supplementary Figure 4- 2 SBI-0206965 is a mixed-competitive AMPK inhibitor ...... 96 Supplementary Figure 4- 3 Structural features of SBI-0206965-bound a2 kinase domain ...... 97 Supplementary Figure 4- 4 Compatibility of SBI-0206965 to investigate cellular AMPK signalling ...... 98

List of Abbreviations 6965 SBI-0206965 ACC Acetyl-CoA carboxylase ADP Adenosine diphosphate AgRP Agouti-related peptide AICAR 5-Aminoimidazole-4-carboxamide ribonucleotide AID Auto-inhibitory domain AMP Adenosine monophosphate AMPK AMP-activated protein kinase AR Androgen receptor ARC Arcuate nucleus of the hypothalamus ATG Autophagy related genes ATP Adenosine triphosphate BAT Brown adipose tissue BMAL1 Brain and muscle ARNT-like protein 1 CaMKK2 Calcium/calmodulin-dependent kinase kinase 2 CBM Carbohydrate binding module CBS cystathionine-β-synthase CLOCK Circadian locomotor output cycles kaput CPT1 Carnitine palmitoyltransferase 1 DKO Double knock-out FAS Fatty acid synthase GS Glycogen synthase GSK3 Glycogen synthase kinase 3 GP Glycogen phosphorylase GPAT Glycerol-3-phosphate acyl-transferase HDX-MS Hydrogen-deuterium exchange mass spectrometry HMGCR HMG-CoA reductase IGF-1 Insulin-like growth factor 1 KO Knock-out LC3 Microtubule-associated proteins 1A/1B light chain 3 LC-MS Liquid chromatography – mass spectrometry LKB1 Liver kinase B1

xii

MEF Mouse embryonic fibroblast mTOR Mechanistic target of rapamycin mTORC1 Mechanistic target of rapamycin complex 1 NPY Neuropeptide Y OCT1 Organic cation transporter 1 PAS Phagaphore assembly site PE Phosphatidylethanolamine PFKFB 6-phosphofructo-2-kinase/fructose-2,6-biphosphatase PI3K Phosphatidylinositol-3-kinase PKA Protein kinase A POMC Proopiomelanocortin PP2A Protein phosphatase 2A PP2C Protein phosphatase 2C PtIns3P Phosphoinositide 3-phosphate RIM Regulatory subunit interacting motif SCN Suprachiasmatic nucleus SDS-PAGE Sodium dodecyl sulfate polyacrylamide gel electrophoresis Snf1 Sucrose non-fermenting 1 SnRK1 SNF1-Related protein kinase SREBP Sterol regulatory element binding protein T2D Type 2 diabetes UCP1 Uncoupling protein 1 ULK1 Unc-51-like kinase 1

xiii

Table of Contents Abstract ...... iii Declaration ...... vi Preface ...... vii Publications ...... ix Acknowledgments ...... x List of Figures ...... xi List of Abbreviations ...... xii Chapter 1 Literature Review ...... 2 1.1 An Energy Sensing Kinase ...... 2 1.1.1 Adenosine nucleotides and cell metabolism ...... 2 1.1.2 The AMP-activated protein kinase ...... 3 1.2 Structural organisation of AMPK ...... 5 1.2.1 Catalytic α-subunit ...... 8 1.2.2 Regulatory β-subunit...... 9 1.2.3 Regulatory γ-subunit ...... 10 1.3 Crystal structures of AMPK ...... 11 1.4 Regulation of AMPK ...... 12 1.4.1 Binding of nucleotides to the γ-subunit ...... 12 1.4.2 Regulation by phosphorylation ...... 13 1.4.2 Regulation by dephosphorylation ...... 15 1.5 AMPK regulation of metabolism ...... 15 1.5.1 Carbohydrate metabolism ...... 18 1.5.2 Lipid metabolism ...... 19 1.5.3 Protein Synthesis and Cell Cycle ...... 20 1.5.4 Autophagy ...... 21 1.5.5 Appetite ...... 23 1.5.6 Thermogenesis ...... 23 1.5.7 Circadian rhythm ...... 23 1.6 Therapeutic activation of AMPK ...... 24 1.6.1 Metformin ...... 24 1.6.2 Allosteric Drug and Metabolite (ADaM) site ...... 24 1.6.3 Drugs that target the γ-subunit ...... 28 1.7 Inhibition of AMPK ...... 29 1.7.1 Physiological inhibition of AMPK ...... 29 1.7.2 Opportunities for therapeutic inhibition of AMPK ...... 30

xiv

1.7.3 Pharmacological inhibition of AMPK...... 32 1.8 Concluding Remarks ...... 36 1.9 Aims and Hypotheses ...... 36 CHAPTER 2: Investigating intracellular adenine nucleotide ratios during metabolic stress using mass spectrometry ...... 38 2.1 INTRODUCTION ...... 38 2.2 MATERIALS AND METHODS ...... 39 2.3 RESULTS...... 44 2.3.1 Endogenous AMPK activity reduces stress induced changes to adenylate energy charge . 44 2.3.2 Measuring changes in adenylate energy charge in response to glucose starvation ...... 47 2.3.3 High concentrations of small molecules can increase [AMP]/[ATP] ...... 50 2.3.4 The degree to which [AMP]/[ATP] increases in response to metabolic stress is stimulus- dependent and cannot be predicted by the adenylate kinase equilibrium ...... 52 2.4 DISCUSSION ...... 59 CHAPTER 3: The autophagy initiator ULK1 sensitises AMPK to allosteric drugs ...... 63 3.1 Introduction ...... 63 CHAPTER 4: AMPK is selectively inhibited by the type II inhibitor SBI-0206965 ...... 79 4.1 Introduction ...... 79 CHAPTER 5: General Discussion ...... 101

Chapter 1 Literature Review 1.1 An Energy Sensing Kinase 1.1.1 Adenosine nucleotides and cell metabolism The ability for living cells to sense their intra- and extracellular environments for available nutrients and sources of energy, and adapt their metabolism accordingly is a homeostatic mechanism that exist at all levels of life. In yeast, the regulation of gene transcription, transporter translocation and kinase activation and inhibition allows for fermentation of a range of sugars, allowing continued growth if the available carbon source changes. In animals, a homologous mechanism results in a huge number of physiological and behavioural responses to metabolic stress.

Organisms have a number of mechanisms for converting nutrients into a form that can be used for cellular processes that require energy. In all organisms, the principle molecule that is generated during these processes is adenosine triphosphate (ATP). The high energy phosphoanhydride bonds in ATP allow energy to be released during hydrolysis, as well as phosphate or pyrophosphate, converting ATP to ADP (adenosine diphosphate) or AMP (adenosine monophosphate) respectively.

The idea that relative levels of the energy storing nucleotide ATP and its dephosphorylated metabolites ADP and AMP are a direct input for the regulation of metabolic pathways was first proposed by Atkinson and colleagues in 1965. They noted that a number of enzymes - phosphofructokinase, isocitrate dehydrogenase and citrate synthase - appear to be activated by AMP or inactivated by ATP (1). As all three of these enzymes are involved in the metabolism of carbohydrate for the production of ATP, they proposed that AMP and ATP were functioning as regulators of metabolism, to overall maintain ATP homeostasis (1).

In 1967, Atkinson et al. proposed a “fundamental metabolic control parameter” derived from intracellular concentrations of ATP, ADP and AMP. This was called the “energy charge” of the adenylate system, and is defined by the equation shown in box 1. They proposed that the

Box 1 The adenylate energy charge (AEC) of the cell is equal to half of the kinase-accessible phosphoanhydride bonds in the adenine nucleotide pool.

reaction velocity of enzymes that consume ATP and are inhibited by ADP and/or AMP, more specifically the synthesis of fatty acids, will be regulated by energy charge (2).

Many decades later, a master metabolic regulator that responds directly to AMP, ADP and ATP was discovered, linking adenylate energy charge to many more pathways than just carbohydrate and lipid metabolism.

1.1.2 The AMP-activated protein kinase All eukaryotes are able to adapt their metabolism to their environment through orthologues of the AMP-activated protein kinase (AMPK).

1.1.2.1 Snf1 The budding yeast Saccharomyces cerevisiae has been used in winemaking, beer brewing and baking since ancient times. It primarily uses glucose as a carbon source, which it uses for ATP production through fermentation, the process which also produces ethanol in wine and beer. But S. cerevisiae is able to metabolise a number of different carbon sources if glucose is unavailable. By studying mutants of S. cerevisiae that were unable to use sucrose as a carbon source, Carlson et al. identified mutations in novel locus they named sucrose-nonfermenting (SNF1)(3). The SNF1 gene encodes the catalytic subunit of the protein kinase complex Snf1, which senses the signals of glucose unavailability and alters the transcription of genes involved in gluconeogenesis, respiration and the metabolism of carbon sources other than glucose (4, 5). Many of these genes are transcribed when Snf1 phosphorylates Mig1, a transcriptional repressor, leading to its cytosolic translocation out of the nucleus (6). The transcription of a number of these glucose-repressed genes then allows for catabolism of other carbon sources, such as galactose via GAL1 transcription (galactokinase) and sucrose via SUC2 transcription (invertase) (7, 8). Interestingly, the Snf1 kinase complex is crucial the regulation of other pathways not involved in metabolism. For example, Snf1 also regulates elements of normal cell cycle, and Snf1 mutants with reduced catalytic activity have impaired transcription of mitotic genes, and defects in spindle assembly leading to cell cycle arrest (9-11).

The Snf1 kinase complex is a heterotimer. The SNF1 (α-) subunit of the complex contains the catalytic kinase domain. The SNF4 (γ-) subunit is a regulatory subunit and is required for the expression of many glucose-repressed genes (12). One of three other proteins, SIP1, SIP2 or GAL83 can act as the other regulatory (β-) subunit of the complex (13).

1.1.2.2 SnRK1 Plants are also able to adapt their metabolism to the availability of nutrients in their environment via a family of Snf homologues. For example, the presence of sugar leads to the repression of photosynthetic genes in corn, tobacco and tomato plants (14-16). The presence of sugar also represses the expression of isocitrate lyase and malate synthase which are involved in glyoxylate cycle (17), and sugar starvation induces the expression of α-amylase which breaks down starch, the carbohydrate storage molecule of plants (18).

Plants respond to carbohydrates in their environment via the SNF1-Related Protein Kinase (SnRK1). SnRK1 was first sequenced in 1991 from Rye and shown to be a plant homologue of Snf1 (19), and members of the SnRK1 family have since been found in a wide range of plants (20). The first substrates of SnRK1 to be discovered were HMG-CoA reductase in Arabidopsis (21), sucrose phosphate synthase (SPS) (22) and nitrate reductase (NR) (23) involved in the synthesis of isoprenoid, sucrose and nitrogen containing molecules respectively. SnRK1 is activated by phosphorylation in the absence of sugars and inhibits all of these enzymes as a carbon conserving mechanism. SnRK1 is also able to regulate gene expression, such as sucrose synthase in the presence of sugar, and α-amylase in the absence of sugar (24, 25).

While both Snf1 and SnRK1 can regulate yeast and plant metabolism respectively by sensing carbon availability, neither are activated directly by AMP or ADP, although AMP is able to inhibit inactivation of SnRK1 by phosphatases (25), and ADP prevents dephosphorylation of Snf1 (26). The discovery of a master regulator of cell metabolism that responds directly to AMP, ADP and ATP came from identifying regulators of animal cell metabolism.

1.1.2.3 AMPK In animals, the rate limiting enzymes responsible for cholesterol and fatty acid synthesis are, respectively, HMG-CoA reductase (HMGCR) and acetyl-CoA carboxylase (ACC). Early independent investigations into the regulation of these enzymes reported regulatory upstream kinases whose activity negativity regulated HMGCR and ACC, and whose activity was in turn regulated by AMP (27-29). It was later confirmed that the same ‘AMP-stimulated kinase’ was responsible for phosphorylating both HMGCR and ACC, and the name AMP-activated protein kinase (AMPK) was adopted (30-32). Similar to Snf1 and SnRK1, AMPK has been shown to phosphorylate an extensive range of substrates, controlling a vast number of cellular processes. These are described in detail in section 1.5.

1.2 Structural organisation of AMPK Like Snf1 and SnRK1, AMPK is a heterotrimer, with a catalytic α-subunit and regulatory β- and γ-subunits. Each subunit contains regulatory phosphorylation sites, metabolite binding sites as well as drug binding sites. There are multiple isoforms of each subunit – α1, α2, β1, β2, γ1, γ2 and γ3 – allowing for 12 heterotrimer combinations. The α-AMPK sequence is homologous to the yeast Snf1 sequence as well as its plant homologs (33). In addition, the β- subunit is homologous to yeast SIP1/SIP2/GAL83 proteins and the γ- subunit shares high sequence identity with SNF4 (34), evolutionarily linking these enzymes to a conserved system of energy homeostasis across eukaryotes.

A crystal structure, subunit schematic and table of post translational modifications of AMPK is shown in Figure 1-1, and are described in detail below.

Figure 1- 1 3D structure, subunit schematic and post-translational modifications of AMPK In each panel, AMPK α-subunits are coloured green, β-subunits are coloured blue and γ- subunits are coloured pink. (A) Crystal structure of full length, human α2β1γ1 AMPK (PDB: 4CFE). (B) Subunit schematic of α1, α2, β1, β2, γ1, γ2 and γ3 subunits, showing domain organisation and important post-translational modification sites. (C) Table of known post- translational modifications on AMPK subunits, the amino acid residue that is modified, the modifying enzyme and reported functional consequence of each post-translation modification.

Amino acid numbering based on human AMPK splice variants α1.3, α2.1, β1.1, β2.1, γ1.1, γ2.1, γ3.1.

1.2.1 Catalytic α-subunit The α-subunit is the catalytic subunit of the AMPK heterotrimer, containing the AMPK kinase domain. There are two distinct isoforms of the α-subunit, (α1 and α2) and although the catalytic cores of the subunits are 90% identical, their amino acid sequences diverge on other, C-terminal parts of the protein (35).

While the length of the α1 and α2 isoforms are almost identical, (550 and 552 residues respectively in humans)(33), their relative expression across tissues is very different, although most cell types express both isoforms (36). The α1 isoform is widely expressed in heart, liver, kidney, brain, spleen, lungs, as well as skeletal muscle (35). The α2 isoform is strongly expressed in skeletal muscle but less so in brain, liver and heart tissues, and is not expressed at all in haematopoietic cell lines (35). AMPK dependent autophagy plays a crucial role in erythrocyte development, and as a result global knockouts of α1 have a severe anaemic phenotype (37, 38). α2 global knockout mice have an insulin-resistant and glucose intolerant phenotype, possibly due to loss of hypothalamic AMPK activity (36, 39). Simultaneous knockout of α1 and α2 are embryonic lethal (40).

Activation loop phosphorylation and dephosphorylation is the most common regulatory mechanism of kinase activity (41). Phosphorylation of the activation loop allows for electrostatic interaction between the phosphoresidue and a conserved basic pocket. This interaction rearranges the activation loop in a way that allows for substrate binding and efficient catalysis (42). Both AMPK α isoforms contain a Thr172 residue (amino acid position based on the rat α2 sequence, Thr174 on human α1) that was first hypothesized to be the major activating phosphorylation site based on its location within the activation segment of AMPK – between the DFG and APE motifs (35, 43). Phosphorylation of Thr172 by an upstream kinase increases AMPK activity >100-fold, and was thought to be an absolute requirement for catalytic activity, however some small molecule activators of AMPK are able to activate AMPK without Thr172 phosphorylation (44).

The α-subunit also contains an autoinhibitory domain (AID) which lies between residues 313- 392 (45). Expression of a truncated α1(1-392) yields low activity, while the shorter sequence with the AID missing, α1(1-312) yields AMPK with high activity, although still dependent of Thr172 phosphorylation (45). Within the auto inhibitory domain is the α-regulatory subunit interacting motif (α-RIM, 358-362) which makes contact with the β- and γ-subunits and is critical for AMP-dependent activation of AMPK (46, 47). Two conserved residues within the

α-RIM, Arg363 and Glu362 make contacts with a loop in the β-subunit (219-235) and basic residues on the γ-subunit (Arg69 and Lys169) that are also in contact with AMP (47). The originally identified α-RIM was later renamed RIM2, when an additional N-terminally located segment, named RIM1, was found to pack against another site on the γ-subunit (48). The C- terminal region of the α-subunit has a scaffold function, interacting with the β-subunit to allow for heterotrimer assembly (45, 49).

Another region of the α-subunit called the “ST loop” is rich in serine and threonine residues and lies between residues 472-525. Several phosphorylation sites within the ST loop have been reported to have a regulatory role by impacting on the rate of Thr172 phosphorylation, although the mechanism is unknown (50, 51). These are discussed in more detail in section 1.4.2.

1.2.2 Regulatory β-subunit The β-subunit provides the scaffolding that allows AMPK heterotrimer formation – both the α- and γ-subunit interact with β (34, 52). The two isoforms of the β-subunit (β1 and β2), are similar in length but their amino acid sequences diverge in their N-terminal regions (53), and they have different expression profiles across different tissues – the β1-subunit is the predominant isoform in heart, kidney and rodent liver, but not human liver; the β2-subunit is the predominant isoform expressed in human liver and skeletal muscle (54-56). In addition, β-subunit isoforms have been shown to affect AMPK sensitivity to drugs, nucleotides, and capacity for glycogen binding. For example, A-769662, the first synthetic, direct activator of AMPK is unable to activate β2 AMPK (57), whereas the β2 subunit binds glycogen more strongly (36).

1.2.2.1 Carbohydrate binding module The carbohydrate binding module (CBM) of the β-subunit was revealed following sequencing of the β-subunits of AMPK. It was observed that a non-catalytic domain between residues 68- 163 in rat β1 AMPK is related to isoamylase domains found in plant enzymes that metabolise glycogen (58, 59). Confocal microscopy then confirmed that AMPK co-localises with glycogen and glycogen synthase, and NMR studies have shown that carbohydrates with more than 5 glucose units, like maltohexose, are able to bind with Kd values less than 1 mM, while shorter carbohydrates bind less strongly (58, 60). 3D homology modelling with three known structures of isoamylase domains revealed that the β1 CBM contains two tryptophan residues (Trp100 and Trp133) that mediate binding of cyclodextrin (59). The importance of the CBM in mediating drug binding has been demonstrated with structural and biochemical studies,

however the role that carbohydrates play in regulating AMPK activity is less clear, and will be discussed later.

1.2.2.2 Myristoylation of the beta subunit The covalent linking of lipids to proteins is a common post translational modification that allows for membrane targeting and compartmentalisation of proteins (61). Myristoylation involves linking a myristoyl group, derived from myristic acid, to an N-terminal glycine residue by the enzyme N-myristoyl transferase. Both β-subunit isoforms of AMPK are myristoylated at Gly2, which determines the ability for the AMPK heterotrimer to bind to membranes (62), but also confers additional levels of regulation to the AMPK complex - AMP is able to promote phosphorylation of Thr172 by upstream kinases in a myristoyl-dependent mechanism (63). In addition, AMP and nutrient stress lead to a “myristoyl-switching” mechanism, whereby AMPK is cytosolic under nutrient replete conditions, and is targeted to membranes during metabolic stress (63). Under basal conditions, the myristoyl group is hypothesised to be buried in a binding pocket on the AMPK molecule - following [AMP] increase, the myristoyl group is released from the binding pocket and is able to interact with cell membranes. The N-terminus of the β-subunit is flexible, and as a result is not resolved in crystal structures, however hydrogen-deuterium exchange mass spectrometry (HDX-MS) data suggest that the myristoyl binding pocket lies near the interface of the DFG motif and catalytic loop on the α-subunit (64). This may explain why the non-myristoylated mutant of AMPK (G2A) not only remains cytosolic, but has a higher activity as a result of increased pThr172.

There is also evidence to suggest that the β2 subunit, but not the β1 subunit, can be sumoylated by the E3-SUMO ligase PIASy, enhancing its overall activity and preventing it from being modified by ubiquitination and subsequently degraded (65).

The β1-subunit contains phosphorylation sites at Ser24/25, Ser108, Ser148 and Ser182 (numbering based on human β1 sequence) with Ser24/25, Ser108 and Ser148 identified as autophosphorylation sites (62, 66). In addition, the β2 subunit has 4 phosphorylation sites identified as being phosphorylated by the autophagy initiator Unc-51-like kinase 1 (ULK1), Ser38, Thr39, Ser68 and Ser173 (67).

1.2.3 Regulatory γ-subunit There are three γ-subunit isoforms, with each varying in length - the γ1 subunit is 331 residues in length, while the γ2 and γ3 isoforms have N-terminal extensions of 240 and 150 residues respectively (68). While the γ1 subunit is ubiquitously expressed, the γ2 subunit is expressed

in the heart, with lower expression in the brain, placenta and skeletal muscle, and the γ3 subunit is only expressed in skeletal muscle (69).

The ability to sense intracellular energy status in the form of adenine nucleotides is conferred to AMPK through cystathionine-β-synthase (CBS) motifs on the γ-subunits (70). CBS motifs are short motifs (~60 amino acids long) that exist in a wide range of proteins, and harbour binding sites for adenosine derivatives, allowing AMPK to sense energy availability in the form of AMP/ADP/ATP ratios. CBS motifs are always found in pairs, and all three AMPK γ isoforms contain two pairs of CBS motifs. Each pair of CBS motifs form two cavities, comprised of hydrophobic and charged residues, that constitute the binding pockets for adenine nucleotides. On AMPK, these binding sites are numbered according to the conserved aspartate, (or substitute residue) that interact with hydroxyls of the nucleotide ribose (CBS sites 1-4 contain Asp90, Arg171, Asp245, and Asp317 respectively, numbering based on human γ1 sequence) (71).

Interestingly, the different γ isoforms of AMPK respond differently to changes in AMP, ADP and ATP (72). Most notably, γ3 AMPK has very little allosteric activation by AMP, which may reduce AMPK activation by AMP in skeletal muscle which largely expresses γ3 containing AMPK complexes.

Many phosphorylation sites have been identified on the N-terminus of the γ2 subunit using phosphoproeomic analysis, suggesting heavy post-translational regulation (73-75). In addition, ULK1 can phosphorylate the γ1-subunit at Ser261, Thr263 and Ser270 (67). Whether or not these phosphorylation sites have any functional significance is unknown, however it is proposed that ULK1 phosphorylation of AMPK decreases Thr172 phosphorylation (67).

1.3 Crystal structures of AMPK In 2007, the heterotrimer core of the AMPK homolog from the fission yeast Schizosaccharomyces pombe was solved, providing the first crystal structure of AMPK subunit arrangement as well as AMP and ATP bound to the same site in two different structures (76) (PDB: 2OOY ATP, 2OOX AMP). Later structures yielded complexes with ADP and ZMP (77) (PDB: 2QRE ZMP, 2QRC and 2QRD ADP). The first structure of the heterotrimer core of the Snf1 complex was solved in 2007, providing additional detail of the subunit interactions, as well as positioning of the glycogen binding domain (GBD, CBM in mammalian AMPK)(PDB 2QLV) (78). The first mammalian trimer fragment of AMPK was also published in 2007, revealing three of the four CBS sites on the γ-subunit bound by AMP or Mg.ATP (52).

In 2010, the first crystal structure of the AMPK kinase domain was solved using a truncated α2 construct (6-279)(PDB: 2H6D)(79). This truncated α-subunit is missing the AID, but provides evidence that the kinase domain on its own (without the AID) is auto-inhibited via a restructuring in the activation loop and catalytic loop, interfering with both ATP and substrate binding. The activation loop of kinases contains a conserved N-terminal DFG motif and C terminal APE motif, critical for positioning Mg.ATP for kinase activity. In this structure, the Asp from the DFG motif is in a position that is unable to coordinate Mg2+ ions for ATP catalysis, and the aromatic ring of the Phe is buried into adenyl-binding pocket of ATP.

In 2011, a construct using rat α1(1-469; 524-548), human β2 (187-272) and rat γ1 was solved in complex with ADP (PDB: 2Y8L) (46). A region of the α subunit (373-382) was solved showing the first interaction of the α subunit with the nucleotide binding site, and was named the α-hook (46), however the structure was rebuilt, removing the α-hook and a new motif, the α-RIM, was discovered (47).

Structures of AMPK have now been solved with full-length human α1β1γ1 and α2β1γ1 heterotrimers, revealing allosteric drug binding sites between the α- and β-subunits (PDB: 4CFE, 4CFF, 5KQ5)(80-82) and on the γ-subunit (4ZHX) (83). Several flexible regions on these structures remain unresolved including in the α-subunit (ST-loop residues [α1/2 472- 525], β-subunit (N-terminal residues [1-78], residues between CBM and scaffolding [β1 173- 203]), and γ-subunit (γ1 N-terminal residues [1-26]).

1.4 Regulation of AMPK 1.4.1 Binding of nucleotides to the γ-subunit Since the initial identification of an AMP-activated protein kinase in 1980, understanding of the complex relationship between all three adenosine nucleotides and their regulation of AMPK has been vastly improved, however remains incomplete.

Despite AMPK having four CBS domains, only sites 1, 3 and 4 are able to bind nucleotides. Sites 1 and 3 bind AMP, ADP and ATP interchangeably, whereas site 4 appears to only bind AMP, however is capable of exchanging with ATP under some circumstances (52, 84, 85). Aspartic acid residues CBS sites 1, 3 and 4 (Asp90, Asp245 and Asp317 respectively) interact with the 2’ and 3’ hydroxyl groups of the ribose groups of the binding adenine nucleotide, and mutations of these residues leads to loss of the stimulating effects of AMP and ADP (52, 84). At CBS site 2, an arginine residue in place of an aspartate disrupts nucleotide binding (52).

Binding of AMP to the γ-subunit increases AMPK activity by three mechanisms: 1) stimulates Thr172 phosphorylation by upstream kinases; 2) prevents dephosphorylation of Thr172 by phosphatases; 3) directly stimulates AMPK activity of Thr172 phosphorylated AMPK. Mg.ATP binding to the AMP binding sites competitively blocks AMPK activation by AMP. ADP binding to CBS sites 1 and 3 protects AMPK from Thr172 dephosphorylation by phosphatases, and has been reported to stimulate phosphorylation by an upstream kinase (84), however the ability of ADP to stimulate phosphorylation has been debated – while Oakhill et al. observed a 1.9-fold stimulation of Thr172 phosphorylation by CaMKK2 by ADP, Gowens et al. did not observe any significant increase under the same conditions (86). As a result, the relative contribution of ADP to the regulation of AMPK remains contentious.

1.4.2 Regulation by phosphorylation The canonical mechanism for AMPK activation is via phosphorylation of Thr172 on the α- subunit activation loop. Three kinases have been identified as the upstream activators of this site in mammals: Liver Kinase B1 (LKB1), Calcium/calmodulin dependent kinase kinase 2 (CaMKK2), and TGF-β-activated kinase 1 (TAK1).

1.4.2.1 LKB1 LKB1 was identified as an upstream kinase of AMPK as an orthologue of Snf1 kinases, as well as identifying its substrate specificity sequence at the AMPK activation loop. The catalytic domains of the three upstream kinases of the SNF1 primary phosphorylation site - Tos3, Pak1 and Elm1 – share sequence identity with LKB1, leading to the hypothesis that it was an upstream kinase for AMPK (87). This hypothesis was consistent with an “AMPK-kinase” (AMPKK) purified from rat liver which was estimated to match the mass of LKB1, 58kDa, as measured by SDS-PAGE staining (43). In 2003, purified LKB1 was shown to phosphorylate AMPK in vitro, and “AMPKK” purified from rat liver was shown to consist of a complex LKB1, STRADα and MO25α by probing with specific antibodies (88). LKB1 is a tumour suppressing kinase, and germline mutations of LKB1 result in Peutz-Jeghers syndrome, with other inactivating mutations found in sporadic human cancers, in particular lung cancers (89). Interestingly, metformin, a treatment of type 2 diabetes and an indirect activator of AMPK, is associated with 30% lower cancer incidence (90), however this is in part due to LKB1-AMPK independent pathways (91).

1.4.2.2 CaMKK2 CaMKK2 was identified before LKB1 as an upstream kinase of AMPK (92), but was initially dismissed as a physiologically relevant upstream kinase. Interestingly, the kinase domains of the three upstream kinases of SNF1 show closest similarity to CaMKK2 (87). Work using LKB1 double knock-out cells showed that AMPK was still able to be phosphorylated on Thr172 in response to treatment with the calcium ionophore ionomycin, and that this phosphorylation could be reduced using the CaMKK inhibitor STO-609 (93). In addition, Hawley et al. showed that treatment of LKB1-deficient HeLa cells with the calcium ionophore A23187 activated AMPK without changing [ATP]/[ADP] ratios (94), and Woods et al. showed that activation of AMPK in HeLa cells by ionomycin was CaMKK dependent, and could be reduced by STO-609 (95). This group also showed that CaMKK2 (previously CaMKKβ) was the predominant CaMKK responsible for AMPK Thr172 phosphorylation, with CaMKKα showing less preference for AMPK phosphorylation. Many signalling molecules communicate with cells by increasing intracellular Ca2+. When this Ca2+ then couples with calmodulin (CaM) it is able to activate CaMKK2, which can then activate AMPK. This pathway allows AMPK to respond to hormones such as ghrelin, which activate G-coupled protein receptors.

1.4.2.3 TAK1 A genetic screen of using a mouse cDNA library revealed that Tos3/Pak1/Elm1 deficient yeast maintained Snf1 activity when expressing TAK1, and biochemical analysis showed that TAK1 could phosphorylate Snf1 and AMPK in vitro (96). Furthermore, TAK1 deficient cells have reduced AMPK activation in response to a number of indirect AMPK activators (97). However, as mice carrying loss-of-function TAK1 mutations have phenotypes similar to mice with gain- of-function mutations in AMPK γ2, the physiological role that TAK1 could be playing in regulating AMPK activity is unclear (33).

Several other phosphorylation sites on the α-subunit have been reported to regulate the phosphorylation status of Thr172 by affecting the rate of phosphorylation by LKB1 or CaMKK2, or the rate of dephosphorylation by phosphatases.

One such site is located immediately C-terminally to Thr172. α1-Ser173 (Ser175 on the human α1 sequence) and has been reported to be phosphorylated by PKA in response to glucose starvation (98). Phosphorylation at this site appears to reduce the degree to which AMPK is activated during glucose starvation by limiting Thr172 phosphorylation.

Additional sites located in the ST-loop of the α-subunit have been reported to influence Thr172 phosphorylation by different mechanisms. Glycogen synthase kinase 3 (GSK3) has been shown to phosphorylate AMPK in the ST-loop at Thr481 and Ser477 (50). Individually mutating these residues to alanines revealed that GSK3 phosphorylates Thr481, and then subsequently phosphorylates Ser477, and higher activity of GSK3 in HEK293T cells resulted in higher Thr481 phosphorylation, and lower Thr172 phosphorylation and AMPK activity. While GSK3 phosphorylation of AMPK did not appear to reduce LKB1 phosphorylation of Thr172, it did appear to increase dephosphorylation of Thr172 by PP2C. It is proposed that this site promotes ST-loop association with the kinase domain when dephosphorylated, protecting Thr172 from access by phosphatases. When phosphorylated by GSK3, the ST loop is hypothesised to move away from the kinase domain and expose Thr172 to phosphatases (50). Another site within the ST-loop, Ser487, is phosphorylated by Akt. Phosphorylation at this site prevents phosphorylation of Thr172 by LKB1 in cell free assays and by both LKB1 and CaMKK2 in cells. In cells expressing α1 AMPK and treated with IGF-1 to activate Akt, pSer487 was increased, which could be blocked using the Akt inhibitor MK2206. It has been proposed that phosphorylation of Ser487 allows interaction of the ST-loop with basic residues on the kinase domain αC helix, physically blocking access to Thr172 by upstream kinases, and that this mechanism reduces AMPK activation in cancer cells (51). Additionally, p70S6 kinase (p70S6K) can phosphorylate hypothalamic α2 AMPK at Ser491 as a result of leptin signalling. Phosphorylation of AMPK at this site results in a decrease of AMPK activity, and subsequent reduction of appetite and body weight in mice (99).

1.4.2 Regulation by dephosphorylation Thr172 dephosphorylation is an important mechanism for AMPK deactivation, and Thr172 is rapidly dephosphorylated in cells following removal of metabolic stress (44). Two phosphatases have been identified as being able to dephosphorylate AMPK on Thr172: PP2C and PP2A, with PP2C being the dominant AMPK phosphatase identified in rat liver extract (100, 101).

1.5 AMPK regulation of metabolism The outcome of AMPK activation is an overall increase in catabolic, ATP producing pathways in the cell and decrease in anabolic, ATP consuming pathways. These outcomes are dictated by the direct and indirect regulation of enzymes controlling transcription, macromolecule synthesis and oxidation, membrane transport, autophagy and more. AMPK’s central role in regulating metabolism is illustrated in figure 2, and is reviewed in detail below.

Figure 1- 2 Effects of AMPK activation on cell metabolism The direct phosphorylation targets of AMPK are shown, as well as the metabolic outcomes of their regulation by AMPK. Green arrows indicate an increase in target enzyme activity, and red arrows indicate a decrease in target enzyme activity, as a result of phosphorylation by AMPK. Catabolic pathways that are activated by AMPK include fatty acid oxidation, glycolysis, glucose uptake and autophagy. Anabolic pathways that are inhibit by AMPK include cholesterol, triglyceride and fatty acid synthesis, thermogenesis, glycogen synthesis and protein synthesis.

1.5.1 Carbohydrate metabolism 1.5.1.1 Glucose uptake Large increases in blood glucose concentration following feeding are predominantly controlled by insulin-stimulated glucose uptake into skeletal muscle. Here, glucose is stored as glycogen or oxidised to produce ATP, accounting for 60%-80% of the increase in glucose metabolism in response to insulin (102). Like insulin, exercise/muscle contraction stimulates glucose uptake. And while both mechanisms involve the regulation of two Rab GTPAses TBC1D1 and TBC1D4, the two mechanisms are distinct.

Insulin binding to its receptor triggers a class I phosphatidylinositol-3-kinase (PI3K) dependent signalling cascade, leading to recruitment and activation of Akt. Akt phosphorylation of TBC1 (tre-2/USP6, BUB2, cdc16) domain family, member 4 (TBC1D4) increases the binding of 14- 3-3 proteins (103), inhibiting its activity and ultimately releasing the glucose transporter (GLUT4) from intracellular compartments to the plasma membrane (104). TBC1D1, a closely related paralog of TB1D4 which contains Akt, PKA and AMPK phosphorylation sites is also regulated in response to insulin signalling, although the role that AMPK plays in insulin stimulated glucose uptake in muscle is unclear as insulin does not increase pT172 on AMPK or pS79/pS212 on ACC (105).

In contrast, AMPK rather than Akt has been proposed to be the major regulator of TBC1D4 and TBC1D1 in response to exercise/muscle contraction, as Akt phosphorylation is not increased during exercise and AICAR stimulates glucose uptake independently of insulin (104). Both TBC1D1 and TBC1D4 are phosphorylated in response to pharmacological AMPK activation and muscle contraction (105-107), and AMPK muscle knockout mice have reduced TBC1D1 phosphorylation and glucose uptake (108).

1.5.1.2 Glycogen metabolism The production of glycogen from glucose allows energy to be stored in the liver and skeletal muscle to be used later during fasting and exercise. The synthesis of glycogen and its breakdown are controlled by glycogen synthase (GS) and glycogen phosphorylase (GP) respectively. AMPK is able to inhibit GS directly through phosphorylation of Ser7 (109). Chronic AMPK activation however, has shown to increase glycogen content in skeletal muscle, rather than reduce it. This is a result of increased glucose uptake, and subsequently increased glucose-6-phosphate (g-6-P), which allosterically activates GS, suggesting that AMPK inhibition of GS may only be significant under basal conditions (33, 110).

1.5.1.3 Glycolysis In addition to increasing glucose uptake, AMPK is able to promote the catabolism of glucose through the activation of the glycolytic enzyme 6-phosphofructo-2-kinase/fructose-2,6- biphosphatase (PFKFB). PFKFB is a bifunctional enzyme that catalyses the synthesis and degradation of fructose-2,6-bisphosphate, controlling the rate of glycolysis. Although there are four isoforms of PFKFB in vertebrates (PFKFB1-4) only PFKFB2 and PFKFB3 are known to be phosphorylated and regulated by AMPK (111). Following ischaemia and a drop in oxygen supply, cells are able to switch from oxidative respiration to anaerobic metabolism, with an increase in glycolysis. During myocardial ischaemia, an increase in [AMP]/[ATP] activates AMPK, which then phosphorylates PFKFB2 at Ser466 increasing its activity, fructose-2,6- bisphosphate concentration, and glycolysis (112).

AMPK is also able to phosphorylate PFKFB3 in monocytes and macrophages, and during mitosis. During prolonged mitotic arrest as a result of some cancer therapeutics, cells lose mitochondrial content as a direct result of mitophagy. AMPK is activated following an increase in [AMP]/[ATP], and by phosphorylating PFKFB3, is able to increase glycolysis and ATP production, allowing cell survival (113).

1.5.2 Lipid metabolism Acetyl-Coenzyme A carboxylase (ACC) catalyses the carboxylation of acetyl-CoA to malonyl- CoA. Mammals have two isoforms of ACC, ACC1 and ACC2. ACC1 is mostly expressed in lipogenic tissues, such as liver and adipose, and catalyses the rate-limiting step in the synthesis of long chain fatty acids. In these lipogenic tissues, malonyl-CoA synthesised by ACC1 is used by fatty acid synthase (FAS) to generate long chain fatty acids for triacylglyceride and phospholipid synthesis.

ACC2 is mostly expressed in the heart and muscle. In these tissues, the synthesis of malonyl- CoA by ACC2 inhibits fatty acid oxidation. To be able to cross the mitochondrial membrane for β-oxidation, long chain fatty acids must first be converted into acylcarnitines by the enzyme carnitine palmitoyltransferase 1 (CPTI). Malonyl Co-A produced by ACC2 is a potent inhibitor of CPT1, and ACC2 double knockout mice have continuous fatty acid oxidation and reduced body fat mass and body weight, and are protected against high fat diet induced obesity and diabetes (114).

AMPK phosphorylates and inactivates both isoforms of ACC, and is therefore able to inhibit fatty acid synthesis (via ACC1 inhibition) and increase fatty acid oxidation (via ACC2 inhibition) (115).

The initial identification of AMPK phosphorylation sites on ACC came from amino acid sequencing of ACC1, purified from rat mammary gland. Of the three sites identified, (Ser79, Ser1209 and Ser1215), Ser79 was the most rapidly phosphorylated by AMPK, and was shown to be responsible for the inhibition of ACC1 (31). Importantly, no other kinase has been discovered as an upstream kinase for this site, and so ACC1 Ser79 phosphorylation is a commonly used indicator of AMPK activity in cells (115).

AMPK also inhibits the synthesis of cholesterol and triglycerides by inhibiting the activity of key rate limiting enzymes. The conversion of 3-hydroxy-3-methylglutaryl-Co enzyme A (HMG-CoA) to mevalonate by HMG-CoA reductase (HMGCR) is the rate limiting step in the synthesis of cholesterol. Statins, the most commonly prescribed drugs to treat hypercholesterolaemia, work by inhibiting HMGCR and have successfully increased the survival rate in patients with cardiovascular disease (116). AMPK directly phosphorylates HMGCR at Ser872 to inhibit its activity, and activation of AMPK in macrophages decreases cholesterol content and the synthesis of fatty acids and sterols (117).

Glycerol-3-phosphate acyl-transferase (GPAT) catalyses the rate limiting step in triglyceride synthesis from carbohydrate metabolism. Pharmacological activation of AMPK in cultured hepatocytes leads to a reduction in GPAT activity (118), and endurance exercise reduces GPAT activity in liver and adipose tissue, but not in skeletal muscle (119, 120). It remains unclear whether AMPK directly phosphorylates GPAT or not (33).

Finally, AMPK is able to directly phosphorylate sterol regulatory element binding protein (SREBP), preventing its translocation to the nucleus. SREBP is a transcription factor that controls the expression fatty acid synthase (FAS) and GPAT in lipogenic tissues like liver and adipose. Pharmacological activation of AMPK reduces SREBP gene expression, and activation by the synthetic polyphenol S17834 protects against hepatic steatosis, hyperlipedemia and atherosclerosis in mouse livers (121, 122).

1.5.3 Protein Synthesis and Cell Cycle Protein synthesis accounts for a large proportion of energy consumption in the cell, and is also regulated by fluctuations in adenylate energy charge via AMPK. This is achieved in large part through inhibition of the mechanistic target of rapamycin complex (mTORC) pathway, which

regulates cell growth and proliferation, motility and survival, and protein transcription and translation (123). Tuberous sclerosis 2 (TSC2) is a GTPase-activating protein (GAP) that is able to hydrolyse GTP in the Rheb-GTP complex to form Rheb-GDP. Rheb-GTP is able to activate mTORC, therefore increasing the GAP activity of TSC2 leads to an increase in Rheb- GDP, and subsequent decrease in overall mTORC activation. Growth hormones like insulin are able to inactivate TSC2 via activation of Akt. Akt can phosphorylate TSC2 and inhibit its activity, increasing mTOR activity and protein synthesis (124, 125). AMPK is able to activate TSC2 by phosphorylating Ser1345, increasing its activity and subsequently decreasing mTORC pathway activation (126). AMPK is also able to inhibit mTOR through phosphorylation of RAPTOR, a subunit of mTORC1. When AMPK phosphorylates RAPTOR at Ser722 and Ser792 it promotes 14-3-3 binding and inhibition of mTORC1 signalling (127). Both mechanisms of mTORC1 signalling inhibition lead to a decrease in mRNA transcription and translation. In addition, AMPK inhibition of mTORC1 acts as a metabolic checkpoint, arresting cell cycle when nutrients are low, and preventing apoptosis (33, 127).

1.5.4 Autophagy Autophagy involves the delivery of cytosolic components to lysosomes for degradation. In times of nutrient starvation, autophagy can regenerate carbohydrates, amino acids and lipids for metabolism and ATP production, but can also recycle other nutrients like ferritin. Autophagy is active even under nutrient-rich conditions, and is important for the removal of protein aggregates (128), and has many additional roles in controlling immune response (129), development (130), tumour suppression (131) and cell death (132).

Autophagy was first characterised morphologically in the 60s, but the genes involved in autophagy initiation, and the complex network of signalling pathways and feedback loops have only recently begun to be unravelled. The precise roles of individual proteins are currently under intense investigation. Autophagy is evolutionarily conserved between all eukaryotes, from yeast to mammals. In yeast, there are over 30 autophagy-related genes (ATGs) that contribute to the initiation of the autophagy signalling pathway, recruitment of additional proteins, and the formation and expansion of the membrane bound structure that delivers components to the lysosome, called the autophagosome (133).

In yeast, Atg8 was the first of the ATGs that was found to localise at the phagophore and autophagosome. It was observed that while yeast were in a nutrient rich environment, Atg8 is dispersed through the cytoplasm, but forms large puncta near the vacuole when switched to

starvation conditions. This site has been named the phagophore assembly site (PAS) and is a localisation point for most of the Atg proteins (134).

Identification of the genes that constitute the core machinery of autophagy has led to a better understanding of autophagy. This core machinery can now be divided into different subgroups.

The Atg1/ULK1 complex (Atg1, Atg11, Atg13, Atg17, Atg29 and Atg31) make up the initial complex formed at the PAS and controls the induction of autophagosome formation and autophagy. Atg9, Atg2 and Atg18 are involved in membrane delivery to the phagophore. The phosphoinositide 3-kinase (PtdIns3K) complex (Vps34, Vps15, Vps30/Atg6, and Atg14) regulates the recruitment of phosphoinositide 3-phosphate (PtdIns3P)-binding proteins to the PAS. Finally, the Atg12 (Atg5, Atg7, Atg10, Atg12 and Atg16) and Atg8 (Atg3, Atg4, Atg7, Atg8) conjugation systems are involved in vesicle expansion. Notably, Atg8 is now a commonly used core marker of autophagy. In mammals, there are 9 orthologues of Atg8, including microtubule-associated proteins 1A/1B light chain 3B (LC3B). LC3B is cleaved by Atg4 within minutes of being synthesised. The cleaved product, named LC3-I, can then be lipidated with phosphatidylethanolamine (PE) to form LC3-II. The exact function of LC3-II on the surface of the autophagosome is still being investigated, but likely involves the expansion of the autophagosome, and determination of the cytosolic cargo that it carries (134). Whatever its role, the ratio of LC3-I to LC3-II is a common tool used to quantify autophagy flux in cells (135).

The Atg1 homolog in mammals has two isoforms, Unc-51-like kinase 1 and 2 (ULK1 and ULK2), which are able to form complexes with ATG13 (homolog of yeast Atg13), and focal adhesion kinase family interacting protein of 200kDa (FIP200) as a part of the core complex involved in autophagy initiation. Both mTOR and AMPK are able control autophagy induction by directly phosphorylating and regulating ULK1, which together, link autophagy to nutrient stress sensing (136).

Amino acid deprivation has been known to increase autophagy in cells since 1977, and one of the key effectors of amino acid and growth factor signalling is mTOR (137-139). mTOR is able to phosphorylate and repress the activity of ULK1 at Ser757 in the presence of amino acids, and nitrogen stress relieves mTOR repression of ULK1 activity (140). In contrast, AMPK activates ULK1 by directly phosphorylating residues Ser555 and Ser317 when nutrients are low enough to increase [AMP]/[ATP] (140, 141). ULK1 has also been shown to directly

phosphorylate AMPK, and increased ULK1 activity reduces Thr172 phosphorylation in cells (67).

1.5.5 Appetite In the arcuate nucleus of the hypothalamus (ARC), anorexigenic proopiomelanocortin (POMC) neurons and the orexigenic agouti-related peptide/neuropeptide Y (AgRP/NPY) neurons are able to regulate food intake (142). Reduced AMPK activity in the brain results in lower mRNA expression of AgRP and NPY in the ARC, whereas increased AMPK activity results in higher AgRP and NPY (143). While global AMPK α2 knock-out (KO) mice show larger increases in body weight than WT mice under high fat diets (144), mice with a conditional deletion of AMPK α2 in AgRP neurons of the ARC developed an age-dependent lean phenotype (145).

1.5.6 Thermogenesis Brown adipose tissue (BAT) is able to burn glucose and fatty acids to generate heat. It does this through uncoupling protein 1 (UCP1), which allows protons that have been pumped out of the mitochondrion by the respiratory chain to flow back, bypassing ATP synthase (146). BAT therefore contributes to temperature homeostasis by generating heat after exposure to cold. Thyroid hormone (T3) is able to increase energy expenditure and weight loss, and the disorder hyperthyroidism involves excessive production of T3 (147). T3 has been shown to activate hypothalamic AMPK, leading to inhibition of ACC and increase BAT thermogenesis (148).

1.5.7 Circadian rhythm Circadian rhythm is closely related to metabolic processes, and disruptions to circadian rhythm are associated with cardiovascular disease, increased body mass, and elevated blood glucose and lipid levels (149). The master regulator of behavioural and physiological responses to day and night cycles is the hypothalamic suprachiasmatic nucleus (SCN), which responds to light. In peripheral, non-light sensitive organs, feeding regulates the rhythmic expression of genes to optimise the timing of metabolic processes with the day/night cycle and patterns of food intake (150). The rhythmic expression of genes in these tissues involves feedback to transcriptional regulators brain and muscle ARNT-like protein 1 (BMAL1) and circadian locomotor output cycles kaput (CLOCK). When BMAL1 and CLOCK are active, clock component cryptochrome 1 (CRY1) is expressed which in turn inhibits their activity. AMPK phosphorylates CRY1 in response to low glucose, leading to its destabilisation and degradation, and resetting circadian clock in the liver after fasting (151).

1.6 Therapeutic activation of AMPK 1.6.1 Metformin The drug metformin was introduced into the clinic as an anti-diabetic, glucose lowering drug in the 1950s. Its ability to lower plasma glucose without causing hypoglycaemia, as well as increase insulin sensitivity has made it the frontline prescription for patients with type 2 diabetes (T2D)(152). While metformin’s mechanism of action is still incompletely understood, its main therapeutic function is the reduction of hepatic glucose production by inhibiting gluconeogenesis in hepatocytes (153-155). The ability for metformin to accumulate in the liver is due to metformin reaching its highest concentration in the hepatic portal vein, as well as the predominant expression of the organic cation transporter 1 (OCT1) on the hepatocyte cell surface. Metformin is positively charged in plasma, but OCT1 allows for high concentrations to form in hepatocytes (hundreds of μM) via secondary active transport (156, 157). Once in hepatocytes, metformin inhibits glucose production, largely via activating AMPK (121). Rather than directly activating AMPK, however, metformin inhibits complex I of the mitochondrial respiratory chain and leads to an increase in [AMP]/[ATP], and indirectly leads to AMPK activation by LKB1 (157-159). In addition to inhibiting gluconeogenesis, metformin inhibits lipogenesis and increases fatty acid oxidation through AMPK activation and subsequent ACC inhibition (121, 160, 161). Fatty liver disease is associated with insulin resistance, but can be reversed by metformin induced activation of AMPK, which contributes to its therapeutic effect in T2D (152, 162).

1.6.2 Allosteric Drug and Metabolite (ADaM) site The realisation that AMPK activation in the liver was largely responsible for the benefits of metformin in T2D patients encouraged the search for direct AMPK activators. The first synthetic allosteric activator of AMPK was identified by Abbott laboratories in 2006. A non- nucleoside thienopyridone called A592107 was identified as a hit after high throughput screening of over 700,000 compounds, activating AMPK isolated from rat liver with an EC50 of 38 μM. This was then optimised to the compound A-769662, which has an EC50 of 116 nM. A-769662 is able to inhibit fatty-acid synthesis in primary rat hepatocytes with an IC50 of 3.2 μM, and reduced body weight and decreased plasma glucose by 40% in mice (163).

Biochemical characterisation of A-769662 revealed that truncation of bacterially expressed α1β1γ1 AMPK to remove the β-subunit N-terminal region and CBM (1-185) completely abolished its allosteric activation (164, 165). Phosphorylation of β1-Ser108 was found to be required for AMPK activation by A-769662 (165). Interestingly, A-769662 was also found to

exclusively activate β1 containing AMPK heterotrimers, rendering it unable to stimulate glucose uptake in muscle, where the β2 isoform is predominantly expressed (57).

Among other synthetic small molecules, a naturally occurring compound salicylate has also been shown to directly activate AMPK. Salicylate occurs naturally in the leaves of willow trees and has been used for thousands of years for its health benefits. The drug Aspirin (acetylsalicylic acid) is rapidly converted into salicylate once ingested (166).

When HEK293 cells were treated with salicylate at concentrations 1 mM or higher AMPK activity increased as measured by both pT172 and pACC (167). In cell-free assays using rat- liver AMPK, salicylate increased AMPK activity directly, and protected pT172 from dephosphorylation by PP2C. Interestingly, the S108A mutation abolished activation by salicylate, and addition of salicylate antagonised direct activation of AMPK by A-7696622, suggesting that salicylate is a partial agonist of AMPK and binds to the same site as A-769662.

While the CBM and specifically the Ser108 residue were known to be important for allosteric activation by A-769662 and salicylate, the binding site of these small molecules was not elucidated until years later. After extensive efforts to co-crystallise AMPK with the most well characterised allosteric activator A-769662 were unsuccessful, Xiao et al. synthesised a small molecule that had been published in patent databases, 991 (80). 991 shares characteristics of activation by A-769662, in that deletion of the CBM or mutation of Ser108 to Alanine disrupt its ability to allosterically activate AMPK and protect against dephosphorylation. Xiao et al. successfully co-crystallised AMPK and 991, and showed that the CBM of the β1-subunit was bound to the N-lobe of the kinase domain, forming the drug binding site, named the Allosteric Drug and Metabolite (ADaM) site at the interface (figure 1-3) (80, 168).

Importantly, this structure elucidated the importance of Ser108 phosphorylation in sensitizing AMPK to allosteric drugs. The phosphorylated β1-Ser108 residue forms electrostatic interactions with α2-Lys31 and β1-Asn111 on the CBM stabilising the drug binding site. By generating the mutants α2-Lys31Ala, β1-Arg83Ala or β1-Ser108Ala, 991 activation is largely abolished, while AMP stimulation of AMPK is maintained (80).

In addition to the polar contacts made by pSer108, hydrophobic interactions between the αC- helix of the kinase domain and a β-subunit α-helix located C-terminally of the CBM, named the C-interacting helix. Three residues in the C-interacting helix have been identified as hydrophobically interacting with the αC-helix (Val-162, Phe-163 and Leu-166). Mutation of Leu-166 to Ala reduces activation by 991 but maintains activation by AMP (80). Using the

crystallisation conditions for 991, Xiao et al. were also able to co-crystallise A-769662 and show that both compounds shared the ADaM binding site (80).

Figure 1- 3 The Allosteric Drug and Metabolite (ADaM) site Crystal structure of the α2 (green) β1 (blue) and γ1 (magenta) AMPK heterotrimer, with a close up of the Allosteric Drug and Metabolite (ADaM) site. The ADaM site is a hydrophobic, drug binding pocket formed between the β-subunit carbohydrate binding module and small, N-lobe of the α-subunit kinase domain. Phosphorylation of the residue β1-Ser108 allows for electrostatic interaction with α-K31 and β-N111 to stabilise the ADaM site and increase AMPK activation by ADaM site ligands. Shown here is the full length α2β1γ1 human heterotrimer in complex with the compound 991 (PDB: 4CFE).

The importance of Ser108 phosphorylation was emphasised further with the discovery that drug activation of β1-AMPK only required Ser108 phosphorylation, and could completely bypass the requirement for Thr172 phosphorylation on the activation loop. This surprising mechanism of activation was serendipitously discovered by our group using bacterially expressed α1β1γ1 AMPK, whose expression profile uniquely shows high Ser108 phosphorylation, and no Thr172 phosphorylation (44).

The expression of WT but not kinase-dead (KD) AMPK in bacteria yielded AMPK that was autophosphorylated at Ser108, as shown by mass spectrometry (44). This Ser108 phosphorylated AMPK preparation was catalytically inactive under basal conditions as it was not phosphorylated at Thr172. However, with the addition of A-769662, AMPK activity was stimulated >65 fold. By dephosphorylating the WT AMPK with λ phosphatase, A-769662 stimulation was lost. In addition, a Thr172Ala mutant of AMPK was able to be activated >100 fold by A-769662, confirming that activation loop phosphorylation was not required for drug activation of AMPK (44).

Several studies have shown that β1-Ser108 is an AMPK autophosphorylation site. This explains why, although endogenous enzymes are unable to phosphorylate AMPK during bacterial overexpression, gradual accumulation of Ser108 occurs via low basal activity of the AMPK itself. However, our group discovered that the autophosphorylation of AMPK is limited to cis-autophosphorylation, and cannot be trans-autophosphorylated (44).

In this study, we also showed that AMP and A-769962 can synergistically activate unphosphorylated AMPK. Activation of AMPK by AMP or A-769662 is critically dependent on phosphorylation of α-Thr712 or β1-Ser108 respectively. However, we showed that when both of these sites are unphosphorylated, AMPK activity can be stimulated >1000-fold by AMP and A-769662 co-incubation (44). This synergistic activation of AMPK through ADaM and γ- subunit ligands may be useful for future combinatorial therapies that target AMPK activation, and several studies have observed synergistic AMPK activation by simultaneously targeting both binding sites (83, 169-172).

1.6.3 Drugs that target the γ-subunit 5-Aminoimidazole-4-carboxamide ribonucleotide (AICAR) is an intermediate in the synthesis of inosine monophosphate (IMP) in the cell. The phosphorylated form of AICAR, called ZMP, is able to mimic the effects of AMP on a number of enzymes, including AMPK (173). AICAR is taken into cells via the adenosine transporter where it is phosphorylated by adenosine kinase,

allowing ZMP to accumulate in the cytosol (174). However, AICAR is not an ideal drug for AMPK activation, as it affects other AMP-sensitive enzymes, such as glycogen phosphorylase and fructose-1,6-bisphosphatase (175, 176). As a result, many of the effects of AICAR are AMPK independent (176). In addition, AICAR has poor bioavailability and a short half-life, and can cause lactic acidosis and uric acid production as unwanted side-effects (177).

The ability to activate AMPK by compounds that mimic AMP binding to the γ-subunit of AMPK led to the discovery of 5-(5-hydroxyl-isoxazol-3-yl)-furan-2-phosphonic acid, called compound 2 (C2) (178). C2 was identified from a screen of 1200 AMP mimetic compounds and is able to activate AMPK with more potency than AMP (EC50 = 6 nM and 6 μM for C2 and AMP respectively). As C2 is anionic it is unable to pass through cell membranes, however an esterified prodrug of C2 called C13 is able to activate AMPK in hepatocytes, and decrease de novo lipogenesis in vivo by 30% (178). The γ2 R531G mutant, which is insensitive to activation by AMP, is not activated by C2, suggesting that the mechanism of activation of AMP and C2 is shared (179). Structural studies revealed however that the binding sites of C2 are distinct from AMP (83).

Today, many more compounds have been developed that are able to directly activate AMPK (180). While a large portion of these compounds are selective for β1 AMPK activation, the β2 isoform dominates in terms of expression in other tissues, such as in liver and muscle, and offers an interesting target for drugs. Recent breakthroughs with β2 complex activating drugs have shown that by activating AMPK in skeletal muscle, plasma glucose concentrations fall without any effect on hepatic glucose production (54). Interestingly, phosphorylation of β- Ser108 is critical for drug activation of β1-, but not β2-AMPK (181).

1.7 Inhibition of AMPK 1.7.1 Physiological inhibition of AMPK The focus on AMPK as a drug target has predominantly involved the development and characterisation of activating agents, both direct and indirect, with the emphasis on its ability to lower glucose and increase insulin sensitivity (182). Inversely, low AMPK activity is associated with the pathology of metabolic disorders. Animal models where AMPK is inactive show increased sensitivity to high nutrient diets, and AMPK is inhibited under conditions of high nutrient intake, leading to increased disease states associated with metabolic disorder (183, 184). Mice that are fed a high fat diet have been reported to have low AMPK activity and expression in several tissues, and it has been shown that elevated levels of palmitate leads to

reduced AMPK activity as a result of PP2A activation (185). Furthermore, it has been shown that lipotoxicity in the myocardium of obese rodents occurs with an increase in PP2C expression and decrease in AMPK activation, an effect that can be reversed with the drug troglitazone (186).

High carbohydrates also have a negative effect on AMPK activation. A decrease in AMPK activity as a result of raising glucose above normal physiological levels has been observed in cultured cells included HepG2 cells (160), pancreatic islet β cells (187-189) and umbilical vein endothelial cells (190). While low glucose has been attributed to an increase in [AMP]/[ATP] leading to activation of AMPK (189, 191), high glucose concentrations have not been associated with a decrease in [AMP]/[ATP](192), and it has recently been reported that low glucose concentrations are not associated with a change in [AMP]/[ATP] in cultured MEF, but are able to activate AMPK in a novel fructose-1,6-bisphosphate/aldolase dependent mechanism (193).

Glycogen is also able to inhibit AMPK, although some data has shown to contradict this (183). AMPK activation in glycogen depleted muscle is larger than glycogen replete muscle, and some studies have shown that high glycogen content repressed AMPK activity in skeletal muscle (39, 194, 195). However, in human studies it has been found that high glycogen content is not associated with lower AMPK activation (196, 197). It proposed that the branch points of the glycogen molecule determine its inhibitory effects on AMPK, which may explain that it is not the content itself that is the determining factor for AMPK inhibition (198).

AMPK activity is also reduced in the presence of high amino acid availability (183). β-cells with high amino acids in the culture medium have decreased AMPK Thr172 phosphorylation and decreased AMPK activity, which correlated with an increase in mTOR activity (188, 199). AMPK and mTOR activity is inversely correlated, and AMPK is able to inhibit mTOR indirectly via phosphorylation of TSC2 and raptor (126, 127). Further studies are required to determine if mTOR itself is able to inhibit AMPK activity.

1.7.2 Opportunities for therapeutic inhibition of AMPK Although AMPK inhibition is largely associated with exacerbated disease states, several examples have been shown whereby inhibition of AMPK may offer therapeutic opportunity.

1.7.2.1 AMPK inhibition in stroke Ischaemic injury and stroke can lead to AMPK activation in neurons as a result of hypoxia or glucose starvation, with AMPK reported as having a neuroprotective role and preventing cell

death (200, 201). However it has also been reported that AMPK over-activation following ischaemic injury is detrimental to neuronal survival, and AMPK inhibition reduced infarct volume after stroke was induced in mice, compared to control treated mice (202). Similarly, WT mice demonstrate worse outcomes from induced stroke than AMPK α2 knock-out mice (203).

1.7.2.2 AMPK inhibition in cancer The role that AMPK plays in tumour cell survival and death is complex, as AMPK activation is associated with cell survival during times of nutrient stress, but the metabolic functions of AMPK also limit cell proliferation, and the AMPK signalling network involves a number of well-known tumour suppressors, including LKB1 and p53 (204, 205).

Studies have revealed that patients with type 2 diabetes that have been prescribed metformin have lower incidence of cancer than untreated patients (90, 206). Subsequent studies have shown that metformin induces apoptosis in p53 deficient tumours, but not in tumours with active p53 (207).

Alternatively, AMPK activation has also been observed to play a pro-survival role in cancer cells. Aggressive tumour growth often surpasses vascularisation, and cancer cells are often subject to low oxygen, glucose and other nutrients(208-211). AMPK has been shown to provide a protective role in this environment, preventing apoptosis in pancreatic tumours(212). Another metabolic function of AMPK hypothesised to contribute to cancer cell survival is its ability to increase glucose uptake via translocation of GLUT transporters to the cell membrane. Metabolic reprogramming of malignant cells involves a characteristic shift away from oxidative phosphorylation in favour of glycolysis, coupled with massive increases in glucose uptake, known as the Warburg effect. AMPK may play a role in this process, and in fact has been shown to protect breast cancer cells during metastasis by increasing glucose uptake, generating ATP and reducing oxidative stress (213).

An interesting example of a non-metabolic function of AMPK that promotes tumour survival involves cell proliferation in prostate cancers. Prostate cancers have been shown to rely on androgens for growth and survival, and while androgen ablation therapy can be useful, most patients experience a relapse of the disease and no longer respond to androgen deprivation therapy (214). Interestingly, the androgen receptor (AR)-regulated signalling pathways remain active (215). It has been shown that as a result of AR activation, functionally active splice variants 2 and 7 of CaMKK2 are overexpressed, leading to an increase in Thr172

phosphorylation on AMPK (216). Interestingly, inhibition of CaMKK2 with STO-609 or siRNA, or AMPK with an inhibitor or siRNA, is able to prevent proliferation and migration of LNCaP and VCaP cells (216). This may be a result of AMPK’s role in microtubule polymerisation through CLIP-170 phosphorylation, or via other critical mitotic functions (217, 218).

1.7.2.3 AMPK inhibition in neurodegenerative disease Neurodegenerative diseases such as Alzheimer’s disease, Parkinson’s disease and Huntington’s disease are all characterised by the presence of protein aggregates that lead to nerve cell degeneration and dementia (219). The positive role that AMPK plays in the removal of defective protein and organelles may suggest a neuroprotective role for AMPK. Indeed, Amyloid β production was decreased in primary neurons that were treated with AICAR, and were increased in primary neurons with α2 AMPK knocked-out (220). However, AMPK was also found to be upregulated in brains with Alzheimer’s disease (221, 222) where it associated with hyper-phosphorylated tau, a microtubule associated protein that is involved in the pathology of Alzheimer’s disease as well as a number of other neurodegenerative disorders (219).

1.7.3 Pharmacological inhibition of AMPK There are over 500 kinases in the human kinome, which vary greatly in their domain structure and regulatory subunits (223). The architecture of the kinase core, however, is highly conserved across the entire kinome, and contain many features that are able to be exploited for the development of new inhibitors (42).

The conserved catalytic core of protein kinases consists of two lobes. The N-terminal small lobe (N-lobe) consists of five β-strands (β-strand 1-5) and an α-helix (αC helix). The C-terminal large lobe (C-lobe) is made up of 7 α-helices (αD to αI) and four short β-strands. Together, these two lobes form a deep cleft – the catalytic site of the kinase – where a molecule of ATP bound to magnesium or manganese is coordinated for phosphoryl transfer to a substrate (42). The conformation of the kinase domain affects substrate binding and catalytic activity, and is determined by the interaction of conserved motifs.

The peptide or protein substrate binds to a groove in the C-lobe of the kinase between the α-D, -F and –G helices. The residue immediately following the substrate phosphorylation site buries into the “P+1” loop of the kinase, located in the C-terminal part of the activation segment (42).

This activation segment determines substrate binding and catalytic efficiency of the kinase. In most kinases, including AMPK α1 and α2, this segment begins with DFG (Asp-Phe-Gly) and ends with APE (Ala-Pro-Glu), separated by 21 residues. The Asp residue of the DFG motif binds to Mg2+ which in turn coordinates the three phosphates of ATP for catalysis. Usually, residues within the activation segment are required to activate kinase activity. For AMPK, this is the Thr172 residue (Thr174 in the human α1 sequence). In the tertiary structure, this “primary phosphorylation site” is located close to the Mg2+ binding DFG, the N-terminus of the αC-helix and a conserved HRD (His-Arg-Asp) motif, and coordinates the kinase domain for optimal catalysis and substrate binding.

Other conserved structures of kinases are the regulatory (R)- and catalytic (C)- spines. These are formed by the alignment of non-contiguous hydrophobic residues in the active structure to position both the protein substrate and ATP for catalysis. There are 8 residues that form a C- spine and 4 residues that form an R-spine (42). The C-spine consists of residues from both the large and small lobes of the kinase domain, and is completed by the adenine ring of the bound ATP molecule. The R-spine consists of a residue from the small lobe β4-strand, from the αC- helix, the phenylalanine from the DFG motif and histidine from the HRD motif (42). On AMPK, these residues correspond to Leu68, Leu79, His137 and Phe158, and are aligned in crystal structures of active AMPK, and are out of alignment in crystal structures of inactive AMPK (224).

While most protein kinases have similar tertiary structures when in the active state, crystal structures of the inactive protein kinases have shown that distinct conformations can be adopted. Most notably, the DFG motif on the activation segment can adopt a “DFG in” or “DFG out” conformation, which corresponds to the active and inactive states of the kinase respectively. In the “DFG in” conformation, the DFG Asp is pointed toward the ATP binding pocket of the catalytic cleft, while the DFG Phe is pointed outward and aligned with the other residues of the R-spine (225). In the “DFG out” conformation, the DFG Asp is flipped 180o relative to the “in” position, and can no longer assist in the catalytic coordination of Mg2+ ATP (225, 226). This conformation was first observed in the unliganded insulin receptor kinase (227), but has since been observed in a number of different kinase crystal structures, including AMPK (79, 228).

Whereas the majority of kinase inhibitors bind to the active site, directly competing with ATP and blocking catalysis, other inhibitors bind to a site distant from the active site. Furthermore,

some inhibitors are able to exploit distinct “inactive” conformations of kinases, such as those in the DFG out conformation, and keep them in this state. The various binding modes of inhibitors have been comprehensively categorised into “types”.

In 2011, Dar and Shokat defined three binding modes of kinase inhibitors: Type I inhibitors that bind competitively with ATP to the active site, and can bind to active and inactive forms of the enzyme; Type II inhibitors that selectively bind to kinases in the inactive, “DFG-out” confirmation; and Type III, that bind outside of the ATP binding pocket at an allosteric site (226). The allosteric inhibitors were later further divided into Type III and Type IV by Gavrin and Saiah in 2013 (229). They defined Type III inhibitors that bind within the catalytic cleft formed between the small and large lobes of the kinase domain, but adjacent to, and not overlapping with, the ATP binding site. They defined Type IV inhibitors as those that bind outside of the cleft. Bivalent inhibitors that can bind to the active site as well as another region of the kinase domain are described as Type V inhibitors (230). Finally, molecules that can inhibit kinases by forming covalent bonds are classified as Type VI (225).

Further subdivision of inhibitors relates to how the molecule binds to the catalytic cleft region between the small and large lobes(231). Regions in the catalytic cleft have been described as the “front cleft” and “back cleft”. The “front cleft” comprises residues that make contact with adenine, ribose and phosphates of the ATP molecule. The “back cleft” comprises the non-ATP binding residues – hydrophobic pockets that can be exploited for increased kinase selectivity, such as residues of the αC-helix. Type II inhibitors that extend into the “back cleft” have been described as Type IIA inhibitors, while those that don’t have been described as Type IIB (225).

Below are some examples of different types of inhibitors.

Type I Staurosporine is a type I inhibitor that was initially purified from the bacterium Streptomyces staurosporeus in 1977 (232). In early studies of kinase inhibitors, staurosporine was shown to

be a highly potent inhibitor of protein kinase C (PKC) (IC50 = 2.7 nM) (233). Although staurosporine is much larger than ATP, it is able to bind in the ATP binding pocket, and can achieve higher affinity than ATP by binding in pockets of the catalytic cleft not utilised by ATP (234). As a result, even at high concentrations of ATP, staurosporine can bind to and inhibit kinases. However, staurosporine is also highly promiscuous, and able to inhibit over 90% of kinases, which leads to apoptosis if introduced to cells (235). Other Type I inhibitors

can be very selective, however, such as the epidermal growth factor receptor (EGFR) inhibitor lapatinib (236).

Compound C, also called dorsomorphin, was identified from a screen of 10,000 compounds, and was used initially show to that the therapeutic effect of metformin in cultured hepatocytes was AMPK dependent (121). In this study, compound C was shown to inhibit AMPK with an IC50 of 109nM under assay conditions of 5 μM ATP. But as an ATP competitive inhibitor, compound C loses potency at higher ATP concentrations, and is a relatively poor inhibitor of AMPK at physiological ATP concentrations. In terms of selectivity, compound C was initially reported as “selective” reported as having negligible effects on structurally related kinases including ZAPK, PKA and JAK3 (121). However, several studies have since shown that compound C’s selectivity is poor, inhibiting a number of kinases more potently than AMPK including Src, Lck and DYRK1A (237, 238). In addition, while compound C is able to inhibit AMPK activation by AICAR, it is a poor inhibitor in the context of other cell treatments, such as the mitochondrial uncoupler dinitrophenol (239). Compound C is a type I inhibitor of AMPK, and many of the unwanted qualities of a type I inhibitor persist through its pharmacodynamics, including poor selectivity and weak inhibition in cells where physiological ATP concentrations are able to outcompete compound C at the AMPK active site (183).

Recently, a small molecule that inhibits AMPK in cell free assays was examined. The small molecule Src kinase inhibitor SU6656 was reported to increase fatty acid oxidation, and increase Thr172 phosphorylation on AMPK as well as pACC in skeletal muscle (240). While another kinase inhibitor, sorafenib, was shown to activate AMPK indirectly in cells by disrupting mitochondrial ATP production, SU6656 was shown to paradoxically activate AMPK by binding to the AMPK catalytic site, stimulating phosphorylation by LKB1, and dissociating with AMPK, allowing for catalytic activity (241). While no structural data of SU6656 was reported in this study, it is most likely that it acts as a type I inhibitor, as ATP was able to outcompete SU6656 at the active site.

Type II Type II inhibitors require the kinase to adopt a DFG out conformation, to which they selectivity bind. This conformation involves the Asp reside from the DGF flipping away from the adenine binding pocket, and the Phe residue from DFG moving more than 10Å from its position when the kinase is active (236). It is thought that only a handful of kinases are able to adopt a DFG out conformation, and as such, type II inhibitors may only ever be able to be used for this subset

of kinases (228). It is for this reason that type II inhibitors are intrinsically more selective than type I inhibitors – while type I can bind when the DFG is in or out, type II can only bind when DFG is out.

The first observation that the DFG out conformation could be exploited for kinase inhibitor development came with the solving of the structure of the Abl kinase and imatinib complex (242). Since then, many more type II inhibitors have been developed, and many are clinically available (225).

The only allosteric (type IV) AMPK inhibitor to be reported is MT47-100. MT47-100 is structurally similar to A-769662, and is able to activate β1 AMPK in a CBM and Ser108 dependent mechanism (243). However, MT47-100 inhibits β2 AMPK in a CBM dependent but Ser108 independent mechanism. This means that in cells that contain largely β1 AMPK such as HepG2 cells, MT47-100 causes an overall increase in AMPK signalling. In pancreatic β cells, β2 isoforms constitute ~29% of AMPK expression, but account for a large suppression in glucose stimulated insulin secretion (GSIS) (243). Accordingly, in isolated islets, MT47-100 increased GSIS via inhibition of endogenous β2 AMPK (243).

1.8 Concluding Remarks Our understanding of mammalian energy homeostasis has improved vastly since the discovery of a central metabolic regulating kinase, AMPK, which responds to intracellular adenosine nucleotide ratios and adapts cell metabolism accordingly. The ability of AMPK to lower blood glucose and increase insulin sensitivity has made it an attractive drug target. However, over a decade has passed since the first synthetic direct activator of AMPK was developed, and no drug directly targeting AMPK has advanced through clinical trials. In addition, canonical regulation of AMPK, that is, regulation by adenine nucleotides, is still in dispute, and has serious implications for studying the dynamics of AMPK regulation in a cellular context. Finally, while significant attention has been given to the development of AMPK activators, there remains a lack of effective AMPK inhibitors, despite an emerging role for therapeutic AMPK inhibition.

1.9 Aims and Hypotheses While AMPK regulation by drugs has been intensely investigated for over a decade, current understanding of how small molecules stimulate AMPK pathway activation is limited by an incomplete understanding of adenine nucleotide dynamics. I hypothesise that by directly

measuring intracellular adenine nucleotide concentrations, we will gain additional insight into canonical regulation of AMPK during metabolic stress, and be better equipped to delineate this from novel mechanisms of AMPK activation. Furthermore, while it is known that phosphorylation of AMPK at β1-Ser108 is critical for drug activation of AMPK, endogenous pathways that regulate phosphorylation at this site are unknown. I hypothesise that additional (non-AMPK) kinases are capable of regulating AMPK pathway activation by targeting this site for phosphorylation. Finally, current tools for investigating pharmacological inhibition of AMPK are limited to a handful of poor antagonists. Our understanding of the many roles that AMPK plays in cell signalling and disease progression would benefit from new inhibitors.

To this end, my specific aims are:

• To interrogate fluctuations in cellular adenylate energy charge during metabolic stress and other cell treatments using mass spectrometry. • To investigate the regulation of the AMPK drug-sensitising residue β1-Ser108, and its role in AMPK activation in cells. • To identify and characterise any upstream kinases of AMPK β1-Ser108. • To identify and characterise a novel AMPK inhibitor, SBI-0206965.

CHAPTER 2: Investigating intracellular adenine nucleotide ratios during metabolic stress using mass spectrometry 2.1 INTRODUCTION Canonical regulation of AMPK involves decreases in adenylate energy charge, and AMP and/or ADP binding to the γ-subunit. Mechanisms of adenine nucleotide regulation of AMPK have largely been elucidated, most notably, the discovery of nucleotide binding sites (68, 70), and structural elements that allow transduction of nucleotide binding signals to the α-subunit and catalytic domain (46-48).

There are four CBS domains in AMPK (CBS sites 1-4), which provide three adenine nucleotide binding sites on the γ-subunit - sites 1 and 3 bind AMP, ADP and ATP interchangeably, whereas site 4 appears to be a non-exchangeable AMP binding site, however under some conditions may exchange with ATP after long incubation periods (85). CBS site 2 does not contain an aspartic acid required for interaction with the ribose moiety, and cannot bind nucleotides. Binding of AMP to AMPK promotes activation loop phosphorylation by upstream kinases (92), protects against dephosphorylation by phosphatases (100), directly activates activation loop phosphorylated AMPK (86), can synergistically activate dephosphorylated AMPK (44) and appears to target AMPK to intracellular membranes in a myristoyl-dependent mechanism (63, 64).

Cellular ATP concentrations are relatively high (~1000-fold), compared to AMP, and ATP antagonises activation by AMP by occupying binding sites (100). Thus, AMPK is activated in response to elevated [AMP]/[ATP], rather than absolute increases in [AMP].

Whether or not ADP plays a significant regulatory role for AMPK in the cell is still unclear. Initially, ADP was thought to directly activate AMPK, however this was later attributed to contaminating AMP (244). As ADP is not a direct activator, and exists in the cell in higher (~10-fold) concentrations than AMP, its direct role is inhibitory as it can occupy the AMP binding sites. However as ADP has been shown to promote phosphorylation of pT172 by CaMKK2 (84) and protect against dephosphorylation by PP2C (46), it is still unclear if the net effect of ADP binding to AMPK in cells is agonistic or antagonistic.

Even though the precise effect of ADP on AMPK activity warrants further investigation, it is undisputed that altering adenine nucleotide ratios will affect AMPK activity in cells. For this

reason, it is important to take into account changes in adenine nucleotide ratios when investigating AMPK activity in a cellular environment.

Adenine nucleotide measurements are rarely reported alongside cellular experiments where AMPK activity is monitored (54, 72, 80, 181, 245). In addition, the low abundance of AMP in cells makes it difficult to quantify using high-performance liquid chromatography (HPLC) spectrophotometric assays (absorbance at ~254nm) (86).

In this chapter, I have used liquid chromatography-mass spectrometry (LC-MS) to highlight the importance of adenine nucleotide quantitation in the context of investigating AMPK regulation. I have shown that cell lines are not uniformly susceptible to treatments that are routinely used to induce metabolic stress. In addition, I have shown that small molecules are able to affect adenylate energy charge in a dose-dependent mechanism, and so are likely to indirectly activate AMPK at high concentrations. Lastly, I have shown that the adenylate kinase equilibrium cannot be used to accurately predict increases in AMP in response to most forms of metabolic stress. These findings have important implications in the field of AMPK research by showing that the effect of cell treatments on adenine nucleotides cannot be assumed a priori, but that AMP, ADP and ATP must be measured in order to accurately describe intracellular AMPK regulation.

2.2 MATERIALS AND METHODS Cell culture

All cell lines were cultured in in Dulbecco’s Modified Eagle’s medium (DMEM) containing 10% fetal bovine serum (FBS) and 1% penicillin/streptomycin antibiotics, at 37˚C with 5% CO2. All cells were incubated with fresh DMEM for 3 h prior to incubation with glucose deprived DMEM or compounds as indicated.

Primary hepatocyte isolation

Isolated hepatocytes were kindly prepared by Dr Sandra Galic. Briefly, isolated heaptocytes from wild-type (WT) mice were prepared by the collagenase perfusion method. Livers were perfused with HEPES buffered saline solution by insertion of catheters into the hepatic portal vein, followed by a buffer containing collagenase for 15 minutes. 106 cells were plated into collagen-coated dishes in DMEM supplemented with 10% fetal bovine serum, glutamine and antibiotics. Cells were cultured for 4 hours in high insulin, after which the cells were transferred to low-insulin media and grown overnight.

Primary neuron isolation

Isolated primary neruons were kindly prepared by Dr Sandra Galic. Embryos were collected

from pregnant mice (gestational day 15-16) after they were euthanised by CO2 aphyxiation. The cortical region was aseptically micro-dissected out of the brains of the embryos, free of

meninges and dissociated in Solution 1 (250 ml HBSS, 1.94 ml of 150 mM MgSO4, 0.75 g BSA). The suspended tissues were subjected to trypsin digestion at 37°C for 5 min in Solution 2 (20 ml Solution 1, 80 µl DNAse, 4 mg trypsin) that followed trypsin inactivation by the addition of Solution 4 (16.8 ml Solution 1 and 3.2 ml Solution 3) and centrifuged at 1000 ×g for 5 min at room temperature. The tissue pellet was then subjected to mechanical trituration in Solution 3 (20 ml solution 1, 80 µl DNAse, 200 µl 150 mM MgSO4 and 10.4 mg trypsin inhibitor) and allowed to stand for 0.5 min. The suspension was then transferred to a new sterile 50 ml tube and centrifuged for 5 min at 1000 ×g at room temperature. The cell pellet was then re-suspended in warm (37°C) neurobasal medium (GIBCO, Life Technologies) supplemented with 2.5% B-27 supplement (GIBCO, Life Technologies), 0.25% GlutaMAX-I (GIBCO, Life Technologies), and 1× penicillin-streptomycin (GIBCO, Life technologies). Cells were plated to a density of 5 × 105 cells/well in 24-well plates or 1.5 × 106 cells/well in 6-well plates pre- coated with 0.1 mg/ml sterile poly-D-lysine (Sigma). The cultures were maintained at 37°C in

a humidified incubator containing 5% CO2. In the following day, the medium was replaced with fresh medium containing 2.5% B-27, 0.25% GlutaMAX-I and 1× penicillin-streptomycin. Cultures were grown for seven days (days in vitro 7 or DIV7) before further treatments and half of the medium was changed with fresh medium at day 5 (DIV5).

AMPK α1/2 double knock-out (DKO) MEFs

AMPK α1/2 double knock-out MEFs were generated by another lab by crossing AMPK α1 knock-out mice with AMPK α2 knock-out mice, and immortalising MEFs harvested from 10.5- day postcoitum embryos, as described (246).

Perchloric acid extractions

Treated cells were washed with ice cold PBS and lysed in 200 μl cold 0.5 M perchloric acid. Lysates were collected and clarified at 14,000 rpm for 3 min. 100 μl clarified lysate was collected in a separate tube and neutralised with 25 μl cold KHCO3 and incubated on ice for 5 min. Samples were then centrifuged at 14,000 rpm for 3 min, and supernatant was collected for adenine nucleotide quantitation using liquid chromatography mass spectrometry (LC-MS).

Nucleotide measurements

Adenine nucleotides were measured using a method I developed for another study (243). Briefly, the ABSCIEX 5500 mass spectrometer was operated with the turbo V ion source coupled to a Shimadzu Prominence LC-20AD UFLC pumps, SIL-20ACHT autosampler and CTO-20A column oven. The LC and the MS instrument were controlled and managed with the software Analyst® 1.5.1. Nitrogen was generated using a PEAK nitrogen gas generator (PEAK Scientific). The autosampler was set to 4˚C and the column oven was set to 40˚C. LC conditions were optimised for a 150 mm (length) and 0.5 mm (inner diameter) Hypercarb column (3 µm, Thermo Fisher Scientific Australia). The LC solvent system was (A) 25 mM triethylammonium bicarbonate buffer (TEAB) at pH 7.8 and (B) Acetonitrile with 0.1% trifluoroacetic acid (TFA). Adenine nucleotides were eluted at a flow rate of 500 µL/min in a gradient program consisting of 100% A (5 min) 0 to 25% B (10 min), 50 to 80% B (5 min) and 100% A (5 min). Data was analysed with Multiquant 2.0.2 utilising the area under the LC chromatogram for each corresponding nucleotide peak. Calibration curves were obtained by linear regression of the peak area ratio of each individual nucleotide. The MS conditions and MRM values for AMP, ADP and ATP were optimized by separate infusion of 1 μg/mL solution in 25 mM TEAB at a flow rate of 50 µL/min (see table).

Mass quantifier qualifier CE CXP acquisition time (monoisotopic) Da (Q3) Da (Q3) (volts) (volts) (msec) Da (Q1) AMP 346.06 79 -62 -11 200 97 -24 -11 50 ADP 426.03 79 -62 -13 200 134 -62 -17 50 ATP 505.99 159 -52 -43 200 79 -52 -25 50

All data were acquired in negative ion mode. For all samples the spray voltage was set to -4500 V, ion source gas 1 and 2 were set to 30 and 60 respectively, the source temperature was set to 250, the collision gas was set to high and the curtain gas set at 20. The declustering potential and the entrance potential were fixed at -30 and -10, respectively. The calibration curve was required to have a correlation coefficient (R) of 0.995 or better.

A typical mass chromatogram for AMP, ADP and ATP using the described method is shown below, as well as standard curves showing linearity from 1 pmole to 250 pmoles adenine nucleotide (Figure 2.1).

Adenylate Energy Charge

Adenylate energy charge was calculated using the formula ([ATP] + 0.5 x [ADP])/([AMP] + [ADP] + [ATP])

Statistical Analysis

The data are presented as mean ± SEM of at least three independent experiments. The unpaired, two-tailed Student’s t-test was used for all comparisons. A p-value of <0.05 was considered to be significant.

Figure 2- 1 Mass chromatogram and standard curves of AMP, ADP and ATP (A) Mass chromatogram of Q3 masses of AMP (79 Da in blue, 97 Da in red), ADP (79 Da in green, 134 Da in grey) and ATP (159 Da in magenta, 79 Da in light blue). (B) Nucleotide standards (Sigma) are linear when plotted against the area under the curve.

2.3 RESULTS 2.3.1 Endogenous AMPK activity reduces stress induced changes to adenylate energy charge Because of its role in energy homeostasis, AMPK activation is frequently reported to result in an increase in intracellular [ATP], and subsequent decrease in [AMP]/[ATP] and increase in adenylate energy charge (AEC) (111). Thus, endogenous AMPK activity would buffer changes in intracellular adenine nucleotide ratios during metabolic stress. Indeed, this has been observed in response to muscle contraction (247), ischaemia (248), and metformin (171), whereby cells with LKB1 or AMPK knocked-out have had greater increases in [AMP]/[ATP]. However, in these examples, adenine nucleotide levels were determined using UV absorbance methods, and have not been confirmed using an LC-MS approach.

To confirm the role that AMPK plays in buffering AEC during energy stress using LC-MS, wild-type (WT) MEFs and AMPK α1/2 double knockout (DKO) MEFs were treated with either

1 mM H2O2 or 1 mM phenformin and adenine nucleotides were collected using perchloric acid extraction at indicated time points (Figure 2-2). Both H2O2 and phenformin are known to activate AMPK indirectly by lowering intracellular AEC (249).

WT and AMPK DKO cells that were treated with 1 mM H2O2 showed a rapid decrease in AEC after 10 minutes which was largely recovered after 20-30 minutes (Figure 2-2A). This quick

recovery was most likely due to the endogenous catalase, which metabolises H2O2 into water and oxygen (249). At the 10 minute time point, AMPK DKO MEFs experienced a significantly larger decrease in AEC compared to WT MEFs, (ΔAEC = 0.29 ±0.01 for AMPK DKO MEFs, 0.12 ± 0.01 for WT MEFs, p<0.001), and took significantly longer to recover (Figure 2-2Ai).

Differences between H2O2-treated WT and AMPK DKO MEFs were consistent across AEC, [AMP]/[ATP] and [ADP]/[ATP] (Figures 2-2A i, ii and iii).

Cells that were treated with 1 mM phenformin took approximately 30 minutes to show a reduction in AEC. After 60 minutes, AMPK DKO MEFs had a larger reduction in AEC compared to WT MEFs in response to the 1 mM phenformin (ΔAEC = 0.036 ± 0.002 for AMPK DKO MEFs, 0.023 ± 0.002 for WT MEFs, p<0.05) (Figure 2-2Bi). Differences between phenformin-treated WT and DKO MEFs were consistent across AEC [AMP]/[ATP] and [ADP]/[ATP] (Figures 2-2B i, ii and iii). However, at 30, 40, 50 and 60 minute time points, control AMPK DKO MEFs had significantly higher [AMP]/[ATP], and at 50 and 60 minute time points, had significant lower AEC than control WT MEFs (Figure 2-2B1 and ii). This

44 suggests a level of dysregulation of energy homeostasis under basal conditions in AMPK DKO cells, which may have been exacerbated during the experiment.

Figure 2- 2 Endogenous AMPK activity reduces stress induced changes to adenylate energy charge (A) (i) Adenylate energy charge, (ii) [AMP]/[ATP] and (iii) [ADP]/[ATP] ratios for WT MEFs (black) and AMPK DKO MEFs (red) treated with vehicle control or 1 mM H2O2 for time indicated. (B) (i) Adenylate energy charge, (ii) [AMP]/[ATP] and (iii) [ADP]/[ATP] ratios for WT MEFs (black) and AMPK DKO MEFs (red) treated with vehicle control or 1 mM phenformin for time indicated. Results are shown as mean ± SEM. N=3. * = p < 0.05, ** = p <0.01, *** = p < 0.001 indicating significant difference between WT and AMPK DKO MEFs treated with H2O2 or phenformin. # = p < 0.05, ## = p <0.01 indicating significant difference between WT and AMPK DKO MEFs control samples. Statistical analyses were done using an unpaired, two-tailed Student’s t test.

2.3.2 Measuring changes in adenylate energy charge in response to glucose starvation Metabolism of glucose in normal mammalian cells involves glycolysis, an oxygen-independent breakdown of glucose into pyruvate. Pyruvate is then actively transported into the mitochondrial matrix, where it is converted into acetyl-CoA by pyruvate dehydrogenase, before entering the tricarboxylic acid (TCA) cycle. Both the pyruvate dehydrogenase reaction and TCA cycle reduce nicotinamide adenine dinucleotide (NAD) to produce NADH, and the TCA

cycle also reduces Flavin adenine dinucleotide (FAD) to produce FADH2. Electrons from

NADH and FADH2 are then donated to oxygen via electron carriers in mitochondria, ultimately establishing a proton gradient across the mitochondrial membrane, allowing for ATP synthesis. Cells are able to regenerate NAD+ for further glycolysis in the absence of oxygen via lactic acid fermentation. While only 2 ATP molecules can be formed from one glucose molecule via fermentation, 38 molecules of ATP can be formed via the TCA cycle and oxidative phosphorylation.

Glucose deprivation is a commonly used laboratory technique to activate intracellular AMPK, and has been used to induce AMPK activation in HEK293 cells (140), ovarian carcinoma cell lines (250), COS7 cells (63), and more. However, it was recently reported that mouse embryonic fibroblasts (MEFs) had no change in adenine nucleotide ratios in response to glucose starvation (193). I sought to determine whether other commonly used cell lines were susceptible to changes in glucose concentrations in the culture medium. To test this, I used both primary neuronal cells derived from mouse embryos, and the neuroblast-like cell line SH- SY5Y, as well as other cell lines commonly used in a laboratory setting, and tested the effect of low glucose concentrations in the media on adenylate energy charge (Figure 3). Cells were incubated in glucose deprived media for 4 hours before being harvested, as this time period was sufficient to affect adenine nucleotide ratios in other experiments (193). As previously reported (193), HEK293 cells were sensitive to low glucose ≤2 mM, whereas adenine nucleotide ratios were unchanged in MEFs (figure 3A, B). Similarly, the human hepatocarcinoma cell line HepG2 was resistant to changes in media glucose across all concentrations (figure 3C). The AEC in COS7 cells was significantly reduced at glucose concentrations <2 mM (figure 3D). This is consistent with findings that glucose starvation in COS7 cells appears to activate an AMP-dependent myristoyl switch mechanism (63). SH- SY5Y cell line had no change in AEC from 10 mM to 0.1 mM glucose, but complete glucose starvation (0 mM glucose) caused a significant decrease in AEC (Figure 3E). Interestingly, primary neurons had a much lower basal AEC than all other cell lines, (~0.60 compared to

~0.95). This is likely because the primary neurons were cultured in the absence of astrocytes, which assist in normal metabolic function of neurons (251). At glucose concentrations lower than 2 mM, AEC dropped significantly and at 0 mM glucose was at 0.30 (Figure 3F).

Figure 2- 3 The effect of glucose starvation on the adenylate energy charge of different cells Calculated adenylate energy charge from (A) HEK293 cells, (B) MEFs, (C) HepG2, (D) COS7, (E) SH-SY5Y and (F) primary neurons cultured for 4 hours in DMEM + indicated glucose concentration before harvest. Results are shown as mean ± SEM. N=3. * = p < 0.05, ** = p <0.01, *** = p < 0.001 indicating significant decrease in adenylate energy charge compared to 10 mM glucose. Statistical analyses were done using an unpaired, two-tailed Student’s t test.

2.3.3 High concentrations of small molecules can increase [AMP]/[ATP] Many clinically used drugs have the side effect of mitochondrial toxicity leading to metabolic disorders. Examples include biguanides e.g. metformin, that inhibits complex 1 in electron transport chain (252), or NSAIDs e.g. salicylate, that uncouples electron transport from ATP synthesis (253, 254). Both of these compounds are classified as indirect activators of AMPK, as the increase in [AMP]/[ATP] that they cause is able to increase AMPK activity. Other classes of indirect AMPK activators include polyphenols like resveratrol (255), thiazlidinediones (252), ginsenoside (256, 257) and α-lipoic acid (258). In theory, any compound that causes an increase [AMP]/[ATP] could indirectly activate AMPK.

It has been reported that high doses of the direct allosteric AMPK activator A-769662 reduces cell viability (259). In addition, studies with 991 show an increase in pT172 at high doses (> 50 μM), indicating that mitochondrial ATP production may be inhibited at these concentrations (260). This may suggest that at least some of this drug’s effect on AMPK activation may be caused by increases in [AMP]/[ATP]. To demonstrate that compounds that aren’t classified as indirect activators of AMPK can cause an increase in [AMP]/[ATP] I treated HepG2 cells with a range of doses of compounds (figure 4). Salicylate incubations > 1 mM significantly increased cellular [AMP]/[ATP] compared to untreated cells (Figure 4A). A-769662 did not increase [AMP]/[ATP] at 10, 30 or 100 μM (Figure 3B). In fact, at 100 μM, A-769662 significantly lowered [AMP]/[ATP], likely due to the catabolic effects of AMPK allosteric activation. Both 300 μM and 1000 μM A-769662 increased [AMP]/[ATP] compared to the lower [AMP]/[ATP] observed in response to 100 μM A-769662, but not compared to the untreated cells. PF-06409577, another β1-selective allosteric activator did not change [AMP]/[ATP] ratios at any concentration up to 1 mM (Figure 3C) (81). The anti-allergic drug tranilast inhibits the antigen-induced release of histamine from mast cells, and is thought to act independently of AMPK (261). However, tranilast is reported to affect cell viability and cause cytotoxicity at high doses (>100 μM) (262). I tested whether these concentrations are associated with increased [AMP]/[ATP]. No change in [AMP]/[ATP] was detected at concentrations up to 300 μM, however at 1 mM tranilast caused a significant increase in [AMP]/[ATP] (Figure 3D).

Figure 2- 4 High concentrations of small molecules can increase [AMP]/[ATP] [AMP]/[ATP] in HepG2 cells treated with increasing concentrations of (A) salicylate, (B) A- 769662, (C) PF-06409577 and (D) tranilast for 1 hour before being harvested. Results are shown as mean ± SEM. N=3. * = p < 0.05, ** = p <0.01 indicating significant change in [AMP]/[ATP] compared to untreated samples (0 μM). # = p < 0.05 indicating significant increase in [AMP]/[ATP] compared to 100 μM A-769662. Statistical analyses were done using an unpaired, two-tailed Student’s t test.

2.3.4 The degree to which [AMP]/[ATP] increases in response to metabolic stress is stimulus-dependent and cannot be predicted by the adenylate kinase equilibrium While both AMP and ADP have been shown to play a role in regulating the activity of AMPK (46, 72, 84, 86), there remains contention over the contributions of ADP to overall AMPK activity. This contention arises from three points: 1) the dynamic range of ADP, as dictated by the adenylate kinase equilibrium, is smaller than AMP and therefore it would act as a poorer signalling molecule; 2) ADP has been shown to protect AMPK against pT172 dephosphorylation, but there is conflicting evidence of its ability to stimulate pT172 by upstream kinases, whereas AMP has been consistently shown to do both; 3) ADP is unable to directly stimulate AMPK activity.

The first of these three points arises from the reversible adenylate kinase reaction, which utilises 2 ADP molecules to produce 1 ATP and 1 AMP molecule. If this reaction is at equilibrium, then [AMP]/[ATP] should increase as the square of [ADP]/[ATP] (263). This is shown in the box below.

Due to the difficulty of measuring AMP with HPLC and UV absorbance, surrogate measurements of ADP and ATP have been used to mathematically predict AMP based on adenylate kinase equilibrium (86, 263, 264). However, AMP concentrations are affected by

more than the adenylate kinase reaction, including AMP deamination by AMP deaminase (265), and hydrolysis by 5′-nucleotidase (266) (Figure 2-5). This means that AMP concentrations could be lower than that predicted by the adenylate kinase reaction alone. Indeed, it has been shown that inhibition of 5′-nucleotidase augments AMPK activation in cells (267) and overexpression of AMP deaminase or 5′-nucleotidase suppresses AMPK activation (268).

I aimed to test whether predictions of [AMP]/[ATP] derived from the adenylate kinase equilibrium reflected changes to [AMP]/[ATP] observed by LC-MS. I used a range of metabolic stressors on HepG2 cells, each of which reduce ATP in the cell via a different mechanism (Figure 2-6). I then expressed the data as measured [ADP]/[ATP] (filled squares) and [AMP]/[ATP] (open circles), and [AMP]/[ATP] as would be predicted if the adenylate kinase reaction is at equilibrium ie., [AMP]/[ATP] would vary as the square of [ADP]/[ATP] (red triangles) (263). In response to salicylate, both 10 and 100 mM concentrations significantly increased [AMP]/[ATP] and [ADP]/[ATP] ratios. However, the predicted increase in [AMP]/[ATP] as calculated by the square of [ADP]/[ATP] was significantly higher than measured [AMP]/[ATP] (Figure 2-7A). In fact, changes in [AMP]/[ATP] were statistically indistinguishable from changes in [ADP]/[ATP] at all concentrations of salicylate. Glucose starvation did not statistically increase [AMP]/[ATP] or [ADP]/[ATP] ratios at any time point (Figure 2-7B). 1 mM phenformin increased both [AMP]/[ATP] and [ADP]/[ATP], however, like salicylate, 1 mM phenformin did not increase [AMP]/[ATP] to the degree that would be predicted using the adenylate kinase equilibrium, and the fold increase in both were equal (Figure 2-7C). In the presence 100 µM of the AMP deaminase inhibitor erythro-9-(2-Hydroxy- 3-nonyl)adenine hydrochloride (EHNA) and 1 mM phenformin, the measured [AMP]/[ATP] ratio was significantly higher compared to [ADP]/[ATP] after 50 minutes (Figure 2-7D). This is consistent with Plaideau et al. who observed a larger fold increase in [AMP]/[ATP] compared to [ADP]/[ATP] in electrically stimulated muscle only after treatment with an AMP deaminase inhibitor (269). There is evidence that metformin is able to inhibit AMP deaminase at high concentrations (~10 mM) (270). To test whether phenformin was able to do this, I measured the effect of 5 mM phenformin on HepG2 cells. I measured a statistically higher increase in

[AMP]/[ATP] compared to [ADP]/[ATP] after 50 minutes (Figure 2-7E). However, predicted [AMP]/[ATP] was still significantly higher than measured [AMP]/[ATP] at all time points. Oxidative inactivation has been reported to be an effective way to inhibit AMP deaminase in cells (271). In response to 1 mM H2O2, fold increase in [AMP]/[ATP] was significantly higher

than [ADP]/[ATP], and statistically indistinguishable to changes predicted by the adenylate kinase equilibrium alone (Figure 2-7F). Finally, I tested the proton ionophore dinitrophenol (DNP), which uncouples mitochondrial respiration from ATP synthesis, on HepG2 cells. While fold changes in [AMP]/[ATP] and [ADP]/[ATP] were large (approximately 15-fold) following 100 μM DNP treatment, predicted [AMP]/[ATP] was significantly higher than measured [AMP]/[ATP] at all time points (Figure 2-7G). Interestingly, measured [AMP]/[ATP] was significantly higher than [ADP]/[ATP] at the 10 minute time point, perhaps indicating that very large increases in AMP saturate AMP deaminase, as observed previously (265).

Figure 2- 5 The adenylate kinase reaction is a reversible reaction which utilises 2 ADP molecules to produce 1 ATP and 1 AMP molecule. AMP deaminase metabolises AMP to produce IMP. 5′- nucleotidases hydrolyse AMP and IMP to produce adenosine and inosine respectively. Adenosine deaminase metabolises adenosine into inosine, and adenosine kinase phosphorylates adenosine into AMP. H2O2 has been shown to inhibit the activity of AMP deaminase, and EHNA is an adenosine deaminase inhibitor.

Figure 2- 6 Mechanisms of indirect activation of AMPK Indirect activation of AMPK involves increasing [AMP]/[ATP] and/or [ADP]/[ATP] (i.e. decreasing adenylate energy charge). Glucose deprivation prevents glycolysis and the production of pyruvate for metabolism in the TCA cycle, and subsequent mitochondrial ATP production. H2O2, phenformin, resveratrol, berberine and galegine inhibit electron transport chain complexes. Dinitrophenol and salicylate uncouples substrate oxidation from ATP synthesis in mitochondria through proton leak. Increased consumption of ATP, such during muscle contraction, can also activate AMPK by decreasing adenylate energy charge.

(Figure legend on page 56)

Figure 2- 7 The effect of metabolic stress on [AMP]/[ATP], [ADP]/[ATP] and predictions of [AMP]/[ATP] made using the adenylate kinase equilibrium LC-MS quantitation of [AMP]/[ATP] (open circles) and [ADP]/[ATP] (filled squares), and [AMP]/[ATP] predications made using the adenylate kinase equilibrium (red triangles) in response to (A) salicylate, (B) glucose starvation, (C) 1 mM phenformin, (D) 1 mM phenformin + 100 μM EHNA, (E) 5 mM phenformin, (F) 1 mM H2O2 and (G) 100 μM dinitrophenol in HepG2 cells. Results are shown as mean ± SEM. N=3. * = p < 0.05, ** = p <0.01, *** = p<0.001 indicating significant difference between LC-MS measured [AMP]/[ATP] and LC-MS measured [ADP]/[ATP]. * = p < 0.05, ** = p <0.01, *** = p<0.001 indicating significant difference between measured [AMP]/[ATP] and predicted [AMP]/[ATP]. Statistical analyses were done using an unpaired, two-tailed Student’s t test.

2.4 DISCUSSION Activating ATP conserving pathways in response to decreasing adenylate energy charge (increasing [AMP]/[ATP] and/or [ADP]/[ATP]) was a homeostatic mechanism first proposed in the 60s. The activities of many enzymes are regulated by AMP, ADP and/or ATP, including phosphofructokinase, (272), chloride ion channels (ClCs) (273) and isocitrate dehydrogenase (274). The discovery of AMPK, which is activated by both AMP and ADP, has linked relative adenine nucleotide ratios to the rate of a huge number of anabolic and catabolic pathways that are related directly and indirectly to ATP metabolism. But the way that adenine nucleotide ratios change in response to different cellular environments is not entirely understood, and must be taken into account when investigating AMPK regulation in cells.

Firstly, I have shown here that some cell lines that are used in a laboratory setting to investigate intracellular AMPK signalling are refractory to glucose starvation with respect to adenylate energy charge. HEK293, COS7, SH-SY5Y and primary neuronal cells were all sensitive to changes in glucose in the culture medium, while MEFs and HepG2 cells had no change in adenylate energy charge after four hours in glucose free media. While this does suggest that AMPK will be indirectly activated in HEK293, COS7, SH-SY5Y and primary neuronal cells following glucose starvation, it does not rule out AMPK activation by nucleotide independent mechanisms in these cells or in MEFs and HepG2 cells in glucose-free conditions. For example, AMPK is activated by the absence of fructose-1,6-bisphosphate (FBP), an intermediate in the glycolytic pathway, in response to glucose starvation in MEFs (Zhang et al., 2017). In addition, observations made here are limited to glucose starvation periods of 4 hours. Periods of longer glucose starvation (>24 hours) are common in experimental settings, and the effects on cell metabolism are most likely also cell-specific.

However, while quantitation of adenine nucleotides is fundamental to distinguish canonical activation by AMP and ADP from non-canonical pathways such as FBP sensing, the physiological relevance of glucose starvation in cultured cells is questionable. Physiological ranges of blood glucose range between approximately 3.9 mM and 6.1 mM (Cryer, 2007). Homeostatic mechanisms like β cell glucagon secretion and increased feeding act as defences against blood glucose falling below 3 mM, and levels below 2.7 mM lead to seizures and coma (Cryer, 2007). Therefore it is unlikely that culturing cells in glucose-free media is representative of any realistic physiological environment. It is known that brain glucose levels are much lower than peripheral blood glucose levels (~15-20% of total) (275), which may suggest that increased [AMP] in response to glucose concentrations as low as 0.1 mM is a

physiologically relevant mechanism to activate neuronal AMPK, however additional experiments are required to test this.

Secondly, I have shown that nucleotide quantitation is crucial to characterising direct activators as distinct from indirect activators. There are many compounds that are able to increase [AMP]/[ATP] in cells and therefore activate endogenous AMPK. These compounds are categorised as indirect activators of AMPK, and include resveratrol, thiazlidinediones, ginsenoside, α-lipoic acid, and biguanides e.g. phenformin and metformin, the latter of which remains a first line treatment for type 2 diabetes. Therefore when observing AMPK signalling in cells, to be able to define direct activation of AMPK, measuring adenine nucleotides is critical. I have demonstrated that three molecules, (salicylate, A-769662 and tranilast), were able to cause significant increases in [AMP]/[ATP], and are therefore capable of activating AMPK indirectly. While this is a known feature of salicylate (Krause et al., 2003), A-769662 is a direct AMPK activator (AMP-independent), and tranilast is an anti-inflammatory drug that is thought to act independently of AMPK signalling. These results are proof-of-principle that high concentrations of small molecules are capable of activating AMPK indirectly.

Thirdly, I investigated the degree to which AMP and ADP concentrations change in relation to ATP during stress. The relative contributions of AMP and ADP to net AMPK activation in cells is contentious. AMP is able to increase activation loop phosphorylation by LKB1 and CaMKK2, protect against activation loop dephosphorylation by PP2C, directly increase Thr172 phosphorylated AMPK activity, synergistically activate AMPK with A-769662 and target AMPK to membranes. ADP however has only been shown to increase activation loop phosphorylation and prevent its dephosphorylation. While ADP exists at roughly 10 times the concentration of AMP, the adenylate kinase equilibrium predicts that [AMP]/[ATP] should increase as the square of [ADP]/[ATP] during metabolic stress (263). I have shown that [AMP]/[ATP] increases of this magnitude are only observed during oxidative stress, and under other conditions where mitochondrial function is inhibited, AMP metabolism via other pathways appears to prevent increases in [AMP] that are significantly higher than [ADP], with respect to [ATP]. This has also been demonstrated in MEFs that were treated with the mitochondrial inhibitor berberine, and so is unlikely to be a HepG2 specific effect (Zhang et al., 2017).

This observation suggests that AMP may not be a more sensitive indicator of cell stress, and may point toward a physiological mechanism for more robust activation of AMPK under oxidative stress.

While ADP exists in higher concentrations in the cell, AMP generally has a more potent effect of AMPK activity. A comprehensive analysis of these effects were detailed by Gowens et al. (86). They showed that AMP was 10-fold more potent at protecting rat liver AMPK from dephosphorylation by PP2C than ADP, but calculated (using the adenylate kinase equilibrium) a larger fold increase in [AMP]/[ATP] than [ADP]/[ATP] in response to the mitochondrial uncoupler berberine, thus concluding that [AMP] was the predominant regulator of this mechanism (86). In this thesis, I have shown that the adenylate kinase equilibrium cannot be used to accurately calculate [AMP], and that in response to 1 mM phenformin, salicylate or dinitrophenol, [AMP]/[ATP] does not increase to a larger extent than [ADP]/[ATP], perhaps

due to the activity of AMP deaminase. Meanwhile, 1 mM H2O2, a known inhibitor of AMP deaminase activity, induced an increase in [AMP]/[ATP] identical to predictions made by the adenylate kinase equilibrium. These data strongly suggest a significant role for ADP in AMPK signalling, as well as a role for AMP metabolising pathways in controlling the degree of AMPK activation.

Accordingly, oxidative inactivation of AMP deaminase may represent a physiologically relevant mechanism for allowing larger increases in AMP, and subsequently a larger dynamic range of AMPK activation. During periods of hypoxia, inflammation or during prolonged aerobic respiration, mitochondrial reactive oxygen species (mtROS) is produced (276). During these periods, larger AMPK activation may be required to protect cells from massive ATP depletion, such as during ischaemic injury or endurance exercise (277, 278), or as a feedback mechanism to increase mitochondrial biogenesis in response to exercise (279).

In conclusion, I have used an LC-MS approach to show that the dynamics of adenine nucleotide ratios during metabolic stress are affected by nature of metabolic stress, most likely due to the impact of other AMP metabolising enzymes such as AMP deaminase. I have also shown that cell lines respond differently to glucose starvation. Finally, we have demonstrated that high concentrations of direct activators of AMPK are able to increase AMP/ATP, highlighting the potential for secondary, indirect mechanisms of AMPK activation, which should be taken into consideration when investigating AMPK drugs. Taken together, these data show that

61 quantitation of all three adenosine nucleotides (AMP, ADP and ATP) is crucial for properly understanding AMPK regulation in a cellular environment.

CHAPTER 3: The autophagy initiator ULK1 sensitises AMPK to allosteric drugs 3.1 Introduction Since the development of A-769662 - the first synthetic, allosteric activator of AMPK - residues that are critical for drug activation of AMPK have been identified. A-769662 has been shown to only activate β1 containing AMPK heterotrimers, not β2 (57). It has also been shown that deletion of the carbohydrate binding module (CBM) on the β1 subunit abolishes A-769662 activation, and that this effect can be isolated to a single autophosphorylation site, β1-Ser108 (280). Structural studies have revealed that the drug binding site for most allosteric drugs is formed by the CBM of the β-subunit creating a hydrophobic pocket with the small lobe of the α-subunit kinase domain (PDB: 4CFF, 4CFE, 4QFR, 5KQ5, 5UFU) (54, 80-82). In these structures, the phosphorylated Ser108 residue stabilises the drug binding site by electrostatically interacting by the positively charged residues α2-Lys31 and β1-Asn111. This drug binding site, named the allosteric drug and metabolite (ADaM) site, is the structurally- confirmed binding site for A-769662, 991, PF-739 and PF-06409577, and mutagenic studies suggest that is likely the binding site of many other compounds, including salicylate (81, 167, 281). The presence of a well-defined, regulated AMPK drug site has led to speculation that AMPK drugs may mimic an endogenous metabolite, however none have been discovered.

The significance of phosphorylation of β1-Ser108 for drug activation of AMPK was further emphasised by a study published by our lab in 2014, showing that activation loop phosphorylation could be completely bypassed by A-769662 activation of a β1-Ser108 phosphorylated AMPK heterotrimer (44). In this study, we also showed that AMPK autophosphorylation of β1-Ser108 occurred only through cis-autophosphorylation via an intramolecular mechanism reliant on prior phosphorylation of Thr172 (44). This means that in the absence of alternate signalling, aberrant phosphatase regulation or conditions leading to trans-autophosphorylation, stoichiometries of pSer108 and pThr172 are intrinsically linked, and for the most part equivalent. This has important implications for the development of direct activators of AMPK with clinical potential. Since the development of A-769662 in 2006, a number of pharmaceutical companies have patented direct activators of AMPK that are structurally similar to A-769662, and are likely to bind to the same site on AMPK (180). The β1-selective, ADaM site ligand PF-06409577 is the only direct AMPK activator to have advanced to clinical trials, and showed promise in a preclinical model of diabetic nephropathy,

however was unsuccessful in phase I (81, 282). No other direct AMPK activators have reached clinical trials, highlighting the difficulties achieving potent AMPK activation in vivo while avoiding off-target effects.

It may be possible to increase AMPK specific effects of direct activators by increasing the drug-sensitive pool of AMPK in cells. Autophosphorylation is insufficient to amplify the pool of Ser108 phosphorylated AMPK in cells, as this is restricted to cis-autophosphorylation. However, other kinases may exist in the cell that are able to amplify pSer108 in cells independently of AMPK activation, and therefore increase drug activation of AMPK pathways.

To determine whether any other kinases are capable of phosphorylating AMPK β1-Ser108, I synthesised a peptide covering the β1 Ser108 sequence, and screened a library of kinases using a kinase profiling service (Kinexus). The autophagy initiating kinase ULK1 was identified in this screen, and confirmed as an upstream kinase of AMPK β1-Ser108. Phosphorylation by ULK1 resulted in increased AMPK activation in response to A-769662 and salicylate, and conditions associated with ULK1 activation in cells led to an increase in cellular pS108, and activation of AMPK pathways in the absence of Thr172 phosphorylation. This is reported in the published manuscript.

ARTICLE

DOI: 10.1038/s41467-017-00628-y OPEN The autophagy initiator ULK1 sensitizes AMPK to allosteric drugs

Toby A. Dite1, Naomi X.Y. Ling1, John W. Scott2,3, Ashfaqul Hoque1, Sandra Galic2, Benjamin L. Parker4, Kevin R.W. Ngoei2, Christopher G. Langendorf2, Matthew T. O’Brien2, Mondira Kundu5, Benoit Viollet 6,7,8, Gregory R. Steinberg9, Kei Sakamoto10,11, Bruce E. Kemp 2,3 & Jonathan S. Oakhill1,3

AMP-activated protein kinase (AMPK) is a metabolic stress-sensing enzyme responsible for maintaining cellular energy homeostasis. Activation of AMPK by salicylate and the thienopyridone A-769662 is critically dependent on phosphorylation of Ser108 in the β1 regulatory subunit. Here, we show a possible role for Ser108 phosphorylation in cell cycle regulation and promotion of pro-survival pathways in response to energy stress. We identify the autophagy initiator Unc-51-like kinase 1 (ULK1) as a β1-Ser108 kinase in cells. Cellular β1-Ser108 phosphorylation by ULK1 was dependent on AMPK β-subunit myristoylation, metabolic stress associated with elevated AMP/ATP ratio, and the intrinsic energy sensing capacity of AMPK; features consistent with an AMP-induced myristoyl switch mechanism. We further demonstrate cellular AMPK signaling independent of activation loop Thr172 phosphorylation, providing potential insight into physiological roles for Ser108 phosphorylation. These ﬁndings uncover new mechanisms by which AMPK could potentially maintain cellular energy homeostasis independently of Thr172 phosphorylation.

1 Metabolic Signalling Laboratory, St Vincent’s Institute of Medical Research, University of Melbourne, Melbourne, VIC, Australia. 2 Protein Chemistry & Metabolism, St Vincent’s Institute of Medical Research, University of Melbourne, Melbourne, VIC, Australia. 3 Mary MacKillop Institute for Health Research, Australian Catholic University, Melbourne, VIC, Australia. 4 Charles Perkins Centre, School of Molecular Bioscience, The University of Sydney, Sydney, NSW, Australia. 5 Department of Pathology, St Jude Children’s Research Hospital, Memphis, TN, USA. 6 INSERM, U1016, Institut Cochin, Paris, France. 7 CNRS, UMR8104, Paris, France. 8 Université Paris Descartes, Sorbonne Paris Cité, Paris, France. 9 Divisions of Endocrinology and Metabolism, Department of Medicine, and Department of Biochemistry and Biomedical Sciences, McMaster University, Hamilton, ON, Canada. 10 MRC Protein Phosphorylation and Ubiquitylation Unit, School of Life Sciences, University of Dundee, Scotland, UK. 11Present address: Nestlé Institute of Health Sciences SA, Lausanne, Switzerland. Toby A. Dite and Naomi X.Y. Ling contributed equally to this work. Correspondence and requests for materials should be addressed to J.S.O. (email: [email protected])

NATURE COMMUNICATIONS | 8: 571 | DOI: 10.1038/s41467-017-00628-y | www.nature.com/naturecommunications 1 ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/s41467-017-00628-y

he evolutionarily conserved AMP-activated protein kinase AMPK/drug complexes12, 14. These structures revealed that the (AMPK) is a key regulator of cellular and whole-body phosphate group of pSer108 forms electrostatic interactions with T α energy homeostasis that controls multiple branches of 2-kinase domain residues Thr21, Lys31, and Lys33, thereby metabolism to redress energy imbalances caused by physiological stabilizing the ADaM site and explaining the role of pSer108 in and pathological processes1, 2. AMPK senses increased cellular mediating drug sensitization. The precise mechanism by which AMP/ATP ratio during periods of energy stress (hypoxia, nutri- ADaM site drugs activate AMPK is not fully understood. pSer108 ent deprivation, exercise) and protects the cell from these events is absolutely required for AMPK activation by A-76966211, 13, by switching off energy-consuming anabolic pathways and salicylate15 and MT47-10016, and increases by 40-fold potency of switching on catabolic pathways to restore ATP levels. Multiple the high affinity drug 99112. Using purified enzyme we further physiological processes are regulated by AMPK including autop- demonstrated that activation loop phosphorylation is dispensable hagy, appetite control, mitochondrial biogenesis and cell growth, for AMPK stimulation by A-76966214. The well-defined character and proliferation. Consequently, extensive efforts have been made of the AMPK drug site, and its regulation through reversible to develop AMPK-activating drugs for potential therapeutic use in phosphorylation, has led to speculation that synthetic activators treating metabolic diseases (type 2 diabetes, obesity, cardiovascular (991, A-769662) and salicylate are mimicking an endogenous disease) and also cancer and inflammatory diseases. metabolite(s) that would be capable of sustaining AMPK signal- The AMPK αβγ heterotrimer comprises an α-catalytic subunit ing in the absence of pThr1721, 17. and regulatory β- and γ-subunits. Multiple isoforms of each Thr172 phosphorylation is considered a marker of AMPK subunit exist in mammals (α1/2, β1/2, γ1/2/3) and isoform- activity; identification of upstream kinases, and the mechanisms specific variations in tissue distribution, regulation, and function underpinning pThr172 regulation, have been the subject of have been demonstrated. Both β-isoforms contain a intense investigation over several decades. LKB1 and CaMKK2 carbohydrate-binding module (CBM) and are myristoylated at (Ca2+/calmodulin-dependent protein kinase kinase 2) have been position Gly2, a modification that targets AMPK to intracellular identified as in vivo Thr172 kinases18. Despite some conflicting membranes and is important for temporospatial regulation of evidence, current models of AMPK regulation by adenine AMPK signaling3, 4. γ-subunits possess three allosteric adenylate nucleotides describe a tripartite mechanism in which ATP nucleotide-binding sites that bind ATP, ADP and AMP inter- exchange for AMP and ADP at γ-sites (i) promotes Thr172 changeably, enabling AMPK to sense fluctuations in cellular phosphorylation, (ii) suppresses pThr172 dephosphorylation, and – energy state5 7. (iii) (for AMP) allosterically activates Thr172-phosphorylated In most instances, ligand-induced allosteric regulation of AMPK3, 5, 6. Hierarchical phosphorylation events in the AMPK is governed by distinct phosphorylation events that either α-subunit Ser/Thr rich ST-loop (human α1(472–525)) have also sensitize AMPK to nucleotides/drugs binding at γ-subunit sites been reported to negatively regulate pThr172, either by sup- (phosphorylation of Thr172 (pThr172) in the α-subunit activa- pressing Thr172 phosphorylation (α-Ser487 auto-, Akt- or PKA- – tion loop8 10), or small compounds binding at the ADaM phosphorylation1, 2), or by promoting pThr172 depho- (allosteric drug and metabolism) site (phosphorylation of sphorylation (α-Thr479 phosphorylation by GSK319). Activity of – β-Ser108 (pSer108) in the β-CBM11 13). An exception is syner- the autophagy initiator Unc-51-like kinase (ULK), itself an gistic activation of unphosphorylated AMPK when γ- and ADaM AMPK substrate, is associated with reduced pThr172 via an sites are occupied simultaneously10, 13. The ADaM site, a largely uncharacterized negative feedback loop20. hydrophobic cavity formed between the α-kinase domain small In contrast, regulation and function of β-Ser108 phosphor- lobe and β-subunit CBM, was identified in crystal structures of ylation have been largely unexplored. Ser108 is highly conserved

a b 1/2-dKO iMEF iMEF: WT 1/2-dKO 2.5 Basal A-769662 β1 lenti: –– 2.0 ** WT S108A kD β1 37 1.5 α 75 50 1.0 pACC 200 pACC/ACC 0.5 ACC 200 (fold increase vs. basal)

Tubulin 50 0 β1 lenti: WT S108A

β Basal A-769662 1 lenti: kD pACC 200 WT ACC 200

pACC 200 S108A ACC 200

Fig. 1 A-769662 activation of cellular AMPK signaling is dependent on β1-pSer108. a Reconstitution of basal AMPK signaling in AMPK β1/2 double knockout (β1/2-dKO) iMEFs by lentiviral transduction of AMPK β1 WT or S108A mutant. b Immunoblots for pACC from β1/2-dKO iMEFs-expressing β1WT or S108A mutant, stimulated with 20 μM A-769662 for 90 min. n = 3. Error bars, mean pACC fold change relative to basal ± s.e.m. Statistical analysis was performed using unpaired two-tailed Student’s t-test. **P < 0.01 indicates signiﬁcant increase in pACC compared to basal

2 NATURE COMMUNICATIONS | 8: 571 | DOI: 10.1038/s41467-017-00628-y | www.nature.com/naturecommunications NATURE COMMUNICATIONS | DOI: 10.1038/s41467-017-00628-y ARTICLE

a c Replicate 123 iMEF: 1/2-dKO OSB11_MOUSE (S179) β1 lenti: S108A S108E PAK2_MOUSE (S141, S152) DOCK7_MOUSE (S452) Phenformin: – + – + kD ANS1A_MOUSE (S663) pT172 75 PALM_MOUSE (T141, T145) 50 TB182_MOUSE(S1133) α 75 50 TR150_MOUSE (S238, S243) β1 37 TCOF_MOUSE (S1191) TCOF_MOUSE (S593) Tubulin 50 GLCE_MOUSE (S73) IBTK_MOUSE (S1046) ML12B_MOUSE (T19, S20) b 1/2-dKO iMEF CND1_MOUSE (S1320, S1323) (n=3) BIN1_MOUSE (S296) β β 1 lenti: S108A 1 lenti: S108E NHRF1_MOUSE (S285)

d iMEF: 1/2-dKO 10 Phenformin, 1 h * 8 S108A S108E Cell Lysis Acetone 6 kD Precipitation pPAK2 Reduction 4 50 Alkylation NH PAK2

pPAK2/PAK2 50 NH 2 2 Trypsin digestion 2 2 2 NH NH (arbitrary units) Tubulin 50 2 P P P NH 0 2 P NH NH 2 β NH2 1 lenti: Dimethyl labeling S108A S108E CH CH2O 3 CD O 2 CHD2 R-NH R-N R-NH R-N DHRS9 2 2 C2CD2L NaBH CN CH CHD e BIN2 DKK3 3 3 NaBH3CN 2 BIN1 EGF +28 Da, Light +32 Da, Medium EPHB3 Akt F2 Mix (Medium:Light, 1:1) ARHGEF4 FGF2 APP Run LC-MS/MS for global proteome GH1 TiO phosphopeptide enrichment APC 2 GLCE ANKS1A Run LC-MS/MS for phosphoproteome HMCES AGFG1 HTT miR-19b-3p MBD4 Cyanocobalamin Intensity MMP10 m/z ZDHHC8 NAT2 TNKS1BP1 OSBPL11 Raw data analysis, peptide identification, SUCLG1 protein assembly PAK2 SLC9A3R1 RLBP1 PALM PRPF40B PDK2 PI3K (complex)

Fig. 2 Quantitative global and phosphoproteomic analysis uncovers cellular roles for β1-pSer108. a Representative immunoblots for β1/2-dKO iMEFs-expressing β1 mutants S108A or S108E, stimulated with 2 mM phenformin for 1 h. b Workflow showing the stable isotope dimethyl labeling-based quantitative proteomic and phosphoproteomic approach. c Heatmap showing significantly perturbed cellular phosphoproteins and corresponding phosphosites. Red indicates increased, and green decreased, phosphorylation in β1-S108E compared to β1-S108A-expressing cells. Gray indicates missing phosphopeptide in that replicate. d Immunoblot/densitometry analysis confirming increased PAK2-Ser141 phosphorylation in β1-S108E-expressing cells. n = 3, representative immunoblot is shown. Error bars, mean PAK2-Ser141 phosphorylation (arbitrary units) ± s.e.m. Statistical analysis was performed using unpaired two-tailed Student’s t-test. *P < 0.05 indicates significant increase in PAK2-pSer141 in S108E-expressing cells compared to S108A- expressing cells. e ‘‘Cell cycle, connective tissue development and function, cellular movement’’ is one of the top networks associated with changes in phosphoproteome between S108A and S108E-expressing iMEFs, as identified by Ingenuity Pathway Analysis software in eukaryotes and was identified as an autophosphorylation site in that AMPK β1-Ser108 is a substrate for ULK1 under conditions rat liver AMPK preparations21. Closer examination using kinase associated with elevated AMP. We also provide unambiguous inactive (KI) AMPK expressed in COS-7 cells revealed that demonstration of AMPK signaling independently of Thr172 Ser108 is a cis-autophosphorylation site (dependent on intra- phosphorylation. These findings underpin a ULK-mediated molecular Thr172 phosphorylation) that is dephosphorylated “ligand switch” model of AMPK allosteric control, in which the following removal of the AMPK-activating stimulus. Thus, as for adenylate charge-sensing role of β1-AMPK is replaced by an Thr172, Ser108 is largely unphosphorylated under basal condi- ability to detect perturbations in endogenous metabolite(s) acting tions13. Identification of alternate upstream kinases for Ser108 at the ADaM site. would provide advances in two important areas: firstly, characterization of novel Ser108 kinases implicates their therapeutic modulation as a strategy to increase potency of AMPK-targeting Results drugs; secondly, feed-forward regulation by Ser108 kinases might β1-pSer108 confers cellular AMPK drug sensitivity. The denote the cellular processes under which concentrations of importance of Ser108 phosphorylation in sensitizing AMPK to natural AMPK ligands become elevated. Here, we demonstrate ADaM site ligands has been well-characterized using purified

NATURE COMMUNICATIONS | 8: 571 | DOI: 10.1038/s41467-017-00628-y | www.nature.com/naturecommunications 3 ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/s41467-017-00628-y

AMPK enzyme in vitro, but whether this translates to cellular performed a stable isotope dimethyl labeled-based quantitative AMPK signaling has not been unequivocally demonstrated. To proteomic and phosphoproteomic analysis using β1/2-dKO exclude signaling from endogenous AMPK we generated an iMEFs, transduced with either β1 mutant S108A or S108E immortalized mouse embryonic fibroblast (iMEF) cell line (Fig. 2a). We previously showed that Glu at position 108 acts as derived from mouse embryos harboring genetic deletion of both an effective phosphomimetic, with regard sensitizing AMPK to β1 and β2 isoforms22 (β1/2-dKO). As expected, β1/2-dKO iMEFs activation by A-76966213. Following 1 h phenformin (2 mM) were devoid of detectable AMPK α-orβ-subunit expression and treatment, β1 S108A and S108E transduced iMEF lysates were AMPK signaling as evidenced by lack of phosphorylation of the compared to trace out the changes in proteome and phospho- AMPK substrate ACC-Ser79 (Fig. 1a). Lentiviral-transduction of proteome (Fig. 2b). Fifteen cellular phosphoproteins showed either FLAG-tagged wild-type (WT) or S108A mutant β1inβ1/2- significant changes in phosphorylation between S108A and dKO iMEFs reconstituted expression of AMP-sensitive AMPK S108E-expressing cells in our study (Fig. 2c, Supplementary heterotrimers, and recovered phenformin-sensitive AMPK sig- Table 1) with no detectable differences in AMPK expression or naling in transduced cells (Fig. 1a and Supplementary Fig. 1a, b). phenformin-induced Thr172 phosphorylation (Fig. 2a, Supple- Incubation with the direct AMPK agonist A-769662 (20 μM) led mentary Fig. 2a). For example, we identified increased phos- to a significant increase (1.9-fold) in pACC-Ser79 in β1/2-dKO phorylation of p21-activated kinase 2 (PAK2) on Ser141/Ser152, iMEFs-expressing WT β1, but not the β1 S108A mutant (Fig. 1b). located in the kinase inhibitory domain, in phenformin-treated A-769662-stimulation was mediated exclusively through the iMEFs-expressing β1 S108E, whereas global proteome data ADaM site, since phosphorylation of Thr172 was not increased at showed no changes in PAK2 protein level (Supplementary this dose (Supplementary Fig. 1c). These results confirm a Fig. 2b, c). Immunoblot analysis with a phosphospecific antibody requirement for Ser108 phosphorylation in drug activation of confirmed significant increase in pPAK2-Ser141 in S108E- AMPK in cells. expressing cells, validating our representative phosphoproteome data (Fig. 2d). Collectively, pathway and network analysis using Phosphoproteomic analysis hints at roles for β1-pSer108.To ingenuity pathway analysis (IPA) identified ‘‘Cell cycle, con- investigate the cellular fate of β1-Ser108 phosphorylation, we nective tissue development and function, cellular movement’’ as

a WT AMPK α P β Ser108 γ GST α kinase

β P P α cis- α c 1/2-dKO iMEF LKB1/ autophos. kD γ CaMKK2 P α1 lenti: –WTKI β β α 75 γ γ 50 β1 37 KI AMPK α pACC 200 α(D141)βγ P β Ser108 ACC 200 γ GST KI-α kinase Tubulin 50 β P P α α γ LKB1/ P CaMKK2 β β γ γ

b HEK293T: GST pulldown d 1/2-dKO iMEF: FLAG IP Lenti: KI-α1 C KI-AMPK kD Phenformin: –+ + β1-pS108 37 A-769662: –+ +kD 37 β1 β1-pS108 37

O 2 β 37 2 1 Basal H AICAR A-769662 Ionomycin Phenformin Glucose free Amino-acid free

Fig. 3 β1-Ser108 trans-phosphorylation occurs via an AMPK independent mechanism. a Rationale for employing kinase inactive (KI) AMPK to examine cellular Ser108 phosphorylation. β1-Ser108 phosphorylation (blue) can potentially be performed by an upstream kinase in both WT and KI α1(D141A) AMPK mutant. LKB1/CaMKK2-mediated phosphorylation of α-Thr172 (red) activates WT AMPK (orange) leading to background Ser108 cis-autophosphorylation. This is excluded using KI AMPK, which can be phosphorylated on Thr172 but remains inactive. b Immunoblot for β1-pSer108 in KI-α1β1γ1 puriﬁed from HEK293T cells treated with AMPK-activating agents/conditions: glucose free (glucose-free DMEM + 10% serum, 4 h), AICAR

(2 mM, 1 h), H2O2 (1 mM, 45 min), A-769662 (300 μM, 1 h), phenformin (2 mM, 1 h), ionomycin (2.5 μM, 15 min) and amino-acid free (EBSS medium, 4 h). n = 3, representative immunoblots shown. C: Bacterial expressed, CaMKK2-treated α1β1γ1 standard. c Reconstitution of basal AMPK signaling in AMPK α1/2 double knockout (α1/2-dKO) iMEFs by lentiviral transduction of AMPK α1 WT, but not the KI mutant. d Immunoblot for β1-pSer108 from α1/2-dKO iMEFs-expressing KI-α1, stimulated with 2 mM phenformin for 1 h. n = 3, representative immunoblots shown

4 NATURE COMMUNICATIONS | 8: 571 | DOI: 10.1038/s41467-017-00628-y | www.nature.com/naturecommunications NATURE COMMUNICATIONS | DOI: 10.1038/s41467-017-00628-y ARTICLE

a c ULK1 (min) ULK2 (min) S108tide: KLPLTRsHNNFVARRR C 0102030 C 0 10 20 30 kD β1-pS108 37

β1 37 75 α-pT172 50 α 75 50 % pS108 70 0 21 38 52 70 0 0 0 0

b 600 150 d Basal ULK1 A-769662 35 ULK1 complex Salicylate ) ) –1 –1 ULK2 30 400 100 **** mg mg 25 –1 –1 20 200 50 15 ULK1/2 activity (nmol min (nmol min (+/– drug) **

ULK1 complex activity 10

0 0 AMPK fold activation 5 –6 –5 –4 –3 –2 0 log[S108tide] (M) ULK1 –+ pre-treatment:

K (mean ± s.e.m.) m ULK1: – + kD ULK1 75.8 ± 5.6 μM β-pS108 37 ULK1 complex 123.6 ± 16.3 μM ULK2 1309.0 ± 166.8 μM β1 37

Fig. 4 ULK1 phosphorylates β1-Ser108 in vitro. a Sequence alignment of the ULK consensus motif/favorable substitutions23 with human AMPK β1- and β2-residues 104–112 (Ser108 in lower case). x denotes positions with no demonstrated preference. Preferred/favored consensus β-residues are in bold. Sequence of the synthetic peptide S108tide is shown in red. b Dose curve of S108tide phosphorylation by ULK1 and ULK2 (plotted to left y-axis), and ULK1/ FIP200/Atg13 complex (plotted to right y-axis). n = 3. Error bars, mean activity ± s.e.m. c Immunoblots for β1-pSer108 and α-pThr172 in bacterial-expressed KI-α1β1γ1 phosphorylated with ULK1 (left) or ULK2 (right) for 30 min. n = 3, representative immunoblots shown. C: CaMKK2-treated α1β1γ1 control. d Activity of ULK1-phosphorylated α1(C176S)β1γ1 in the presence of 20 μM A-769662 or 10 mM salicylate. n = 3, representative immunoblots of pSer108 in assayed AMPK preparations are shown. Error bars, mean fold AMPK activation relative to ULK1-untreated ± s.e.m. Statistical analyses were performed using one way ANOVA with post hoc Dunnett’s multiple comparison test. **P < 0.01, ****P < 0.0001 indicate significant increase in AMPK activation compared to ULK1-untreated one of the top networks associated with these perturbed cellular KI-α1 reconstituted AMPK expression; as expected only WT α1 phosphoproteins (Fig. 2e). transduction recovered AMPK signaling (Fig. 3c). However, phenformin induced phosphorylation of Ser108 in α1/2-dKO iMEFs transduced with KI-α1, which was not increased by β trans 1-Ser108 can be phosphorylated in . We previously additional incubation with 100 μM A-769662 (Fig. 3d). Our β demonstrated that AMPK 1-Ser108 is cis-autophosphorylated results demonstrate that, under certain metabolic stress condi- via an intramolecular mechanism reliant on prior phosphoryla- tions, Ser108 is a substrate for a kinase(s) other than AMPK α 13 tion of -Thr172 . This implies that, in the absence of alternate autophosphorylation. signaling, aberrant phosphatase regulation or conditions leading to trans-autophosphorylation, stoichiometries of pSer108 and pThr172 are intrinsically linked and, for the most part, equiva- ULK1 phosphorylation of β1-Ser108 induces drug sensitivity. lent. To interrogate this model we screened AMPK-activating To identify upstream kinases for β1-Ser108 we screened a syn- conditions/agents under which Ser108 becomes a substrate for a thetic peptide corresponding to AMPK β1(102–114) (S108tide) trans-phosphorylation event in cells. To prevent background cis- (Fig. 4a) against a panel of 284 Ser/Thr kinases. 92% of the signaling we expressed a GST-fusion of the KI AMPK mutant α1 kinases screened yielded low activities against S108tide (<10% vs. (D141A)β1γ1 in HEK293T cells. This complex is a suitable top hit). Several kinases from diverse groups demonstrated substrate for Thr172-phosphorylating kinases LKB1 and comparable (>35%) specific activity relative to AMPK α1β1γ1, CaMKK2, but does not undergo Ser108 cis-autophosphorylation e.g., BRSK1/2, NEK2/9, TAOK1 (Supplementary Table 2). (Fig. 3a). Glucose starvation, or incubation with H2O2 or phen- Among positive hits was ULK1, which is in accordance with formin, each induced phosphorylation at Ser108 (Fig. 3b). Incu- identity of the sequence surrounding β1-Ser108 to the consensus bation with ionomycin (activator of CaMKK2 signaling), AICAR motif for substrates of this kinase24, particularly Leu, Thr, and (5-aminoimidazole-4-carboxamide ribonucleotide) or A-769662, Arg directly N-terminal to Ser108 (positions P-3, P-2, and P-1, or amino-acid deprivation, each failed to elicit Ser108 phos- respectively) (Fig. 4a). We performed a more detailed analysis phorylation at the single experimental time point used. We used and phosphorylated the S108tide substrate using purified FLAG- α 23 −2 μ 1/2-dKO iMEFs to exclude the possibility that Ser108 is trans- ULK1, either alone (kcat/Km 1.23 × 10 /s/ M) or complexed to −3 autophosphorylated in phenformin-treated cells. Lentiviral- interaction partners FIP200 and Atg13 (kcat/Km 1.86 × 10 /s/ transduction of α1/2-dKO iMEFs with FLAG-tagged WT- or μM) (Fig. 4b and Supplementary Fig. 3a). ULK2 demonstrated

NATURE COMMUNICATIONS | 8: 571 | DOI: 10.1038/s41467-017-00628-y | www.nature.com/naturecommunications 5 ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/s41467-017-00628-y

a ULK1: – + – – ULK1: –+–– CaMKK2: ––+ +kD CaMKK2: ––++ kD 37 β1-pS108 37 β1-pS108

37 37 β1 β2

75 75 α-pT172 α-pT172 50 50 75 75 αα 50 50

KI WT KI WT α1β1γ1 α1β2γ1

b HEK293T:GST pulldown

H2O2 H2O2

p-β1 – + kD p-β2 –+ kD 37 β1-pS108 37 β2-pS108

37 37 β1 β2

Kl-α1β1γ1 Kl-α1β2γ1

Fig. 5 ULK1 phosphorylation of Ser108 is speciﬁc for the AMPK β1 isoform. a Immunoblots for β1-pSer108, β2-pSer108, and α-pThr172 in bacterial- expressed KI-α1β1γ1(left) or KI-α1β2γ1(right) phosphorylated with ULK1 or CaMKK2 for 30 min. Controls: CaMKK2-treated WT α1β1γ1(left)orα1β2γ1

(right). b Immunoblots for β-pSer108 in KI-α1β1γ1(left) or KI-α1β2γ1(right) purified from HEK293T cells stimulated with 1 mM H2O2 for 45 min. Controls: p-β1 and p-β2, CaMKK2-treated WT α1β1γ1 and α1β2γ1, respectively. In both panels n = 3, representative immunoblots shown reduced efficiency in phosphorylating S108tide (k /K ULK1 phosphorylation of β-Ser108 is specific to AMPK β1. − cat m 2.63 × 10 4/s/μM), compared to ULK1 (Fig. 4b). We examined Ser108 is conserved between mammalian AMPK β-isoforms 1 ULK phosphorylation of the AMPK heterotrimer using purified and 2, however identity of the β2 sequence to the ULK consensus/ KI-α1β1γ1 expressed in bacteria, since we previously found the favored motif is restricted to Leu at P-3 (Fig. 4a). ULK1 was WT complex extracted from this source is autophosphorylated at previously shown to phosphorylate multiple AMPK β2 residues β1-Ser108 with ~ 60% stoichiometry despite lacking pThr17213. in vitro, but not Ser10820. To confirm β-isoform specificity of ULK1 phosphorylated β1-Ser108, but not α1-Thr172, in ULK1 we generated a phosphospecific antibody to β2-pSer108 KI-AMPK, whereas we did not detect phosphorylation of either (Supplementary Fig. 4). Incubation of WT α1β1γ1orα1β2γ1 with residue by ULK2 (Fig. 4c). MS/MS analysis confirmed CaMKK2 resulted in Thr172 phosphorylation and subsequent ULK1-phosphorylation of Ser108 in KI-α1β1γ1 (Supplementary Ser108 autophosphorylation (Fig. 5a). ULK1 treatment of Fig. 3b, c). KI-α1β2γ1 resulted in reduced electrophoretic migration of β2, We found that expression of the mutant α1(C176S)β1γ1in indicative of multiple phosphorylation events. However, we did bacteria produced AMPK that was devoid of pSer108 (Supple- not detect β2-pSer108 by immunoblot following ULK1 treatment mentary Fig. 3d) yet retained full allosteric regulatory mechan- (Fig. 5a, right panel). Furthermore, exposure of HEK293T cells to α β γ isms, such as A-769662/AMP and A-769662/C2 synergistic H2O2 failed to induce Ser108 phosphorylation in KI- 1 2 1 activation (Supplementary Fig. 3e)10, 13, or individual A-769662 (Fig. 5b, right panel). (Supplementary Fig. 3f) and AMP (Supplementary Fig. 3g) activation following phosphorylation by CaMKK2. One explana- tion is that pThr172-independent basal activity, and hence Ser108 β1-Ser108 is a cellular substrate for ULK. We examined whether autophosphorylation, is induced by modification of Cys176 in the ULK phosphorylates β1-Ser108 in HEK293T cells. Initially, we activation loop, occurring as a result of oxidative stress during used the highly selective, ULK1 small molecule inhibitor AMPK expression in bacteria. Redox sensitive mechanisms are SBI-0206965 (termed 6965)24. We found that pre-incubation of known to regulate receptor tyrosine kinases25 and a variety of transfected HEK293T cells with 10 μM 6965, a concentration – Ser/Thr kinases including CaMK2, PKA and PKC26 28. The α1 previously shown to have no effect on AMPK signaling24, sig- β γ fi (C176S) 1 1 construct allowed us to investigate the effect of ULK ni cantly reduced H2O2- (Fig. 6a) and phenformin- (Fig. 6b) phosphorylation on ligand-mediated regulation of AMPK with- induced phosphorylation of Ser108 in KI-AMPK α1β1γ1, relative out the need for prior and extensive phosphatase treatment. We to 6965 untreated cells. Treatment with 6965 caused a significant fi α α β γ found that ULK pre-treatment sensitized puri ed AMPK 1 increase in pThr172 in KI- 1 1 1 in response to both H2O2 and (C176S)β1γ1 to activation by A-769662 or salicylate (Fig. 4d). We phenformin (Fig. 6a, b); this is consistent with a role for ULK1 as detected ULK1-phosphorylation at several other AMPK α1, β1, a negative regulator of Thr172 phosphorylation. and γ1 sites (Supplementary Fig. 3c). These included γ1 residues We investigated regulation of Ser108 phosphorylation in Ser261 and Ser270, located in proximity to nucleotide site 3, iMEFs in which both ULK1 and ULK2 had been genetically which is important for AMP allosteric regulation3, 7. However, deleted (ulk1/2-dKO)29 (Supplementary Fig. 5). Under basal AMP allosteric activation of WT α1β1γ1 was not significantly conditions, pSer108 in endogenous AMPK was significantly affected by pre-treatment with ULK1 (Supplementary Fig. 3g). higher in both WT and ulk1/2-dKO iMEFs compared to

6 NATURE COMMUNICATIONS | 8: 571 | DOI: 10.1038/s41467-017-00628-y | www.nature.com/naturecommunications NATURE COMMUNICATIONS | DOI: 10.1038/s41467-017-00628-y ARTICLE

HEK293T cells, despite no detectable increase in pThr172 immunoprecipitation. KI-AMPK expressed in iMEFs was devoid (Fig. 6c). We bypassed this high basal signal in iMEFs by of basal pSer108 (Fig. 6d). Phenformin-induced phosphorylation examining pSer108 in KI-AMPK, expressed by lentiviral of Ser108 was signiﬁcantly decreased (>90%) in KI-AMPK from transduction of a FLAG-tagged KI-α1(D141A) mutant and ulk1/2-dKO iMEFs, compared to WT iMEFs (Fig. 6d). Combined, isolated with endogenous β- and γ-subunits by FLAG these results conﬁrm β1-Ser108 as a cellular substrate for ULK1.

a HEK293T:GST pulldown b HEK293T:GST pulldown 250 250 pT172 pT172 200 pS108 200 pS108 **** ** 150 150 -treated) 2 O

2 100 100

## ###

(vs. H 50 50

% Phosphorylation #### % Phosphorylation ### ### (vs. phenformin-treated) #### 0 0 Phenformin: – ++ H2O2: –++ 6965: ––+ 6965: ––+kD kD 100 100 α-pT172 α-pT172 75 75 α 100 α 100 75 75

β1-pS108 37 β1-pS108 37

β1 37 β1 37

c Endogenous AMPK pT172 pS108 40 kD 75 30 **** α-pT172 50 **** panα 75 20 50 β1-pS108 37 10 37 Basal phosphorylation β1

(fold increase vs. HEK293T) 0

WT WT

HEK293T HEK293T ulk1/2-dKO ulk1/2-dKO iMEF iMEF

iMEF:FLAG IP d iMEF: WT ulk1/2-dKO 40 pT172 Phenformin: – + – + kD pS108 75 30 α-pT172 50 α 75 20 50 β1-pS108 37 10 * β1 37 Phosphorylation (fold 0 increase +/– phenformin) iMEF: WT

ulk 1/2-dKO

Fig. 6 ULK phosphorylates β1-Ser108 in cells. Statistical analyses were performed using one-way ANOVA with post hoc Dunnett’s multiple comparison test, unless indicated. Immunoblots for β1-pSer108 and α-pThr172 in KI-α1β1γ1 purified from HEK293T cells incubated with a 1mMH2O2 and 10 μM 6965 for 45 min, or b 2 mM phenformin and 10 μM 6965 for 1 h. n = 3, representative immunoblots shown. Error bars, mean % phosphorylation relative to H2O2- or phenformin-treated ± s.e.m. **P < 0.01 indicates significant increase, and ##P < 0.01, ###P < 0.001 and ####P< 0.0001 indicate significant decrease, compared to H2O2- or phenformin-treated. c Immunoblots for β1-pSer108 and α-pThr172 in lysates from HEK293T cells, WT or ulk1/2-dKO iMEFs incubated in 25 mM glucose DMEM + 10% serum. n = 3 individual cultures per cell line, representative immunoblots shown. Error bars, mean fold increase in phosphorylation relative to HEK293T cells ± s.e.m. ****P < 0.0001 indicates significant increase in phosphorylation compared to HEK293T cells. d Immunoblots for β1-pSer108 and α-pThr172 in KI-α1 AMPK purified from WT or ulk1/2-dKO iMEFs stimulated with 2 mM phenformin for 1 h. n = 3, representative immunoblots shown. Error bars, mean fold increase in phosphorylation relative to basal ± s.e.m. Statistical analysis was performed using unpaired two-tailed Student’s t-test. *P < 0.05 indicates significant decrease compared to WT iMEFs

NATURE COMMUNICATIONS | 8: 571 | DOI: 10.1038/s41467-017-00628-y | www.nature.com/naturecommunications 7 ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/s41467-017-00628-y

a b 0.98 **** HEK293T:GST pulldown 0.96 * Phenformin: –+–+ kD 0.94 100 α-pT172 0.92 75 α 100 0.90 75 0.88 β1-pS108 37 0.86 Adenylate energy charge β1 37 0.84 O 2 H 2 γ1: WT D245A Basal INK128 AZD8055 Phenformin

c HEK293T d HEK293T:GST pulldown C AZD8055: – ––+ Phenformin: – + +– kD INK128: ––– + kD β1-pS108 37 ULK1-pS757 150 β1 37 Tubulin 50 lysate 100 α-pT172 75 Glucose: + ––+ 100 β1-pS108 37 α 75 β 37 β1-pS108 37 1

37 pulldown GST β1 β1: myr Non-myr

e 3 *** Basal AMP ULK1: –+++ – + 2 AMP: + –––– + kD β1-pS108 37 1 β 37 pSer108/ β 1 1 (arbitrary units)

C Non-myr myr 0 Non-myr myr Purified α1β1γ1

Fig. 7 An AMP-myristoyl switch triggers ULK1 phosphorylation of β1-Ser108. a Adenine nucleotides extracted from HEK293T cells incubated with phenformin (2 mM, 1 h), H2O2 (1 mM, 45 min), AZD8055 (1 μM, 1 h), or INK128 (1 μM, 1 h) were quantitated by mass spectrometry. Adenylate energy charge was calculated as described in Online Methods. n = 3. Error bars, mean adenylate energy charge ± s.e.m. Statistical analyses were performed using one-way ANOVA with post hoc Dunnett’s multiple comparison test. ****P > 0.001, *P < 0.05 indicate significant decrease in mean adenylate energy charge compared to basal. b Immunoblots for β1-pSer108 and α-pThr172 in KI-α1β1γ1 or KI-α1β1γ1(D245A) purified from HEK293T cells stimulated with 2 mM phenformin for 1 h. n = 3, representative immunoblots shown. c Immunoblots for ULK1-pSer757, and β1-pSer108 and α-pThr172 in KI-α1β1γ1 purified from HEK293T cells incubated with 1 μM mTOR inhibitors AZD8055 or INK128 for 1 h. n = 3, representative immunoblots shown. C: Bacterial expressed, CaMKK2-treated α1β1γ1 standard. d Immunoblots for β1-pSer108 in KI-α1β1γ1 (myr) or KI-α1β1(G2A)γ1 (non-myr) purified from HEK293T cells incubated with 2 mM phenformin for 1 h (upper) or glucose free medium for 4 h (lower). n = 3, representative immunoblots shown. e Immunoblot for β1-pSer108 in bacterial-expressed, non-myristoylated (non-myr) or myristoylated (myr) KI-α1β1γ1 phosphorylated with ULK1 for 30 min in the presence of 100 μM AMP. n = 3, representative immunoblots shown. C: CaMKK2-treated α1β1γ1 standard. Error bars, mean increase in β1-pSer108 relative to ULK1-untreated ± s.e.m. Statistical analysis was performed using unpaired two-tailed Student’s t-test. ***P < 0.001 indicates significant decrease in β1-pSer108 relative to non-myristoylated AMPK

An AMP myristoyl switch triggers β1-Ser108 trans- AMP3 (Fig. 7b). ULK1 can also be activated in response to small phosphorylation.H2O2 and phenformin indirectly activate molecule mTOR inhibitors (AZD8055, INK128) that suppress AMPK through perturbation of adenine nucleotide ratios mTOR-mediated phosphorylation at the ULK1 inhibitory site (increased AMP/ATP ratio, reduced adenylate charge). We found Ser75724. Incubation of HEK293T cells with 1 μM AZD8055 or that H2O2 and phenformin treatments of HEK293T cells that INK128 induced almost complete loss of ULK-pSer757, without induced Ser108 phosphorylation (Fig. 3b) also produced sig- signiﬁcantly affecting adenylate charge or stimulating phosphor- ﬁ α β γ ni cant falls in adenylate energy charge (AEC), with H2O2 incu- ylation of Ser108 in KI- 1 1 1 expressed in these cells (Fig. 7a, c). bation having the greater effect (AMP/ATP ratios: basal 0.0093 ± These data indicate that ULK1 phosphorylation of β1-Ser108 in ± 0.0015; phenformin 0.0177 0.0041; H2O2 0.0418 ± 0.0033) cells requires a reduction in adenylate charge and is dependent on (Fig. 7a). Therefore, we investigated whether elevated AMP was a the AMP-sensing abilities of AMPK. requirement for phosphorylation of Ser108 by ULK1. Phenformin N-terminal myristoylation of β-Gly2 plays important roles in treatment of HEK293T cells failed to induce Ser108 phosphor- AMPK temporospatial regulation, being required for AMP- ylation in KI-α1β1γ1 carrying a mutation in the γ1 nucleotide site stimulation of Thr172 phosphorylation by upstream kinases and 3(γ1-D245A) that renders AMPK insensitive to stimulation by metabolic stress-induced co-localization of AMPK to both

8 NATURE COMMUNICATIONS | 8: 571 | DOI: 10.1038/s41467-017-00628-y | www.nature.com/naturecommunications NATURE COMMUNICATIONS | DOI: 10.1038/s41467-017-00628-y ARTICLE

a 1/2-dKO iMEF 8 7 *** α2-pT172 α2-lenti: WT T172A 6 ACC-pS79 ** Phenformin: –+–+–+–+ 5 * A-769662: ––++ –– + + kD 4 * pACC 200 p/total 3 ACC 200 2 #### α-pT172 75 1 50 ## α 75 (fold change vs. WT control) (fold change vs. WT 0 50 Phenformin: –+–+– + – + Tubulin 37 A-769662: ––++– – + +

α2-lenti: WT T172A

b 1/2-dKO iMEF # # 1.0 β1-pS108 *** α2-lenti: T172A 0.8 ULK1-pS555 Phenformin: –+–+ A-769662: ––++kD 0.6 * * β1-pS108 37

p/total 0.4 β1 37

(arbitrary units) 0.2 ULK1-pS555 150 Tubulin 50 0.0 Phenformin: –+–+ A-769662: ––++

Fig. 8 Cellular AMPK signaling occurs independently of α1-pThr172. Statistical analyses were performed using one-way ANOVA with post hoc Dunnett’s multiple comparison test. a Immunoblots for α2-pThr172 and pACC from α1/2-dKO iMEFs-expressing α2 WT or T172A mutant, stimulated with 2 mM phenformin and/or 100 μM A-769662 for 1 h. n = 3, representative immunoblots shown. Error bars, mean pThr172 and pACC (fold change vs. untreated WT α2-expressing cells) ± s.e.m. ***P < 0.001, **P > 0.01, *P < 0.05 indicate significant increase compared to basal WT α2 cells. ####P > 0.0001, ##P < 0.01 indicate significant increase compared to basal α2(T172A) cells. b Immunoblots for β1-pSer108 and ULK1-pSer555 from α1/2-dKO iMEFs- expressing α2 T172A mutant, stimulated with 2 mM phenformin and 100 μM A-769662 for 1 h. n = 3, representative immunoblots shown. Error bars, mean Ser108 and Ser555 phosphorylation (arbitrary units) ± s.e.m. ***P < 0.001, *P < 0.05 indicate significant increase compared to basal. #P < 0.05 indicates significant increase compared to phenformin treated upstream kinases and protein targets located at intracellular cells treated with phenformin (Fig. 8a). Neither of these effects on membranes3, 30, 31. We examined the requirement for β- pACC were detected in α1/2-dKO iMEFs transduced with empty myristoylation in directing Ser108 phosphorylation during lentivirus (Supplementary Fig. 6), confirming that AMPK was the cellular metabolic stress. We found that KI-α1β1γ1 containing a only kinase phosphorylating ACC-Ser79 under these conditions. myristoylation-deficient β1-G2A mutant (non-myr) was insensi- A similar phosphorylation profile was seen for Ser108 in α2 tive to phenformin- or glucose starvation-induced phosphoryla- (T172A) cells, with phenformin inducing a small increase that tion of Ser108 in HEK293T cells (Fig. 7d). We examined the effect was amplified in the additional presence of A-769662 (Fig. 8b). of β-myristoylation on cell-free ULK1-phosphorylation of Ser108 Phenformin/A-769662 co-incubation was the only condition that using myristoylated and non-myristoylated forms of KI-α1β1γ1. induced a detectable increase in ULK1-Ser555 phosphorylation, In contrast to our observation in cells, β-myristoylation resulted an AMPK substrate that is important for ULK1 activation32 in significant suppression of ULK1-mediated Ser108 phosphor- (Fig. 8b). ylation, which was not relieved by addition of AMP (Fig. 7e). Discussion pThr172 is not absolutely required for AMPK signaling.To In this study, we demonstrate that β1-Ser108 in AMPK, a central assess whether AMPK cellular signaling can be triggered inde- co-ordinator of energy homeostasis, is a phosphorylation target pendently of Thr172 phosphorylation, we expressed FLAG- for ULK1, a major regulator of autophagy initiation. Conse- tagged α2 WT or T172A mutant in α1/2-dKO iMEFs at similar quently, ULK1 sensitizes AMPK to A-769662 and salicylate, the levels (Fig. 8a). The T172A mutant possesses negligible basal and active metabolic break-down product of acetylsalycilic acid AMP-stimulated activities, but importantly can be sensitized to (aspirin), independently of Thr172 phosphorylation (Fig. 4d). ADaM site metabolites/drugs through Ser108 phosphorylation13. Salicylate stimulates fat utilization and reduces plasma fatty acids In α1/2-dKO iMEFs-expressing WT α2, phenformin, but not in vivo15, reduces de novo lipogenesis in human hepatocytes and A-769662, induced robust increases in pThr172 and pACC from MEFs33, 34, and reduces fatty acid and sterol synthesis in mac- a high basal level. In cells expressing α2(T172A), phenformin and rophages35. Daily aspirin prophylaxis is also associated with A-769662, either alone or in combination, failed to induce oncosuppression, an effect mirrored by the indirect AMPK acti- Thr172 phosphorylation as expected. Phenformin, but not vator metformin. Other β1-AMPK-specific direct activators A-769662, induced a small increase in pACC from an undetect- MT63-7836 and the indole acid derivative able basal level. Phenformin/A-769662 co-incubation resulted in PF-0640957737 have shown promise as treatments for either a further 5.8-fold increase in pACC compared to phenformin prostate cancer or diabetic neuropathy, respectively. Although not alone, producing >30% the pACC signal in WT α2-expressing investigated, Ser108 phosphorylation is likely a requirement for

NATURE COMMUNICATIONS | 8: 571 | DOI: 10.1038/s41467-017-00628-y | www.nature.com/naturecommunications 9 ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/s41467-017-00628-y

AMPK activation by these compounds; MT63-78 shares common AMP structural features with A-769662, whereas the crystal structure of PF-06409577 bound to the ADaM site contains the phosphomi- pT172 α metic residue Asp instead of Ser108. Our finding that ULK1 α AMPK β1 phosphorylates Ser108 is significant given the emergence of this β1 γ γ signaling modification as a vital mechanistic constant for AMPK drugs. Discovery of a mechanism that induces AMPK drug sensitization pS108 α independently of autophosphorylation also provides a potential P P P β1 AMPK γ signaling strategy to treat non-small-cell lung and cervical carcinomas, ULK1 associated with genetic loss of LKB138. Accumulated evidence now depicts the ADaM site as an Drugs/ “orphan” allosteric site for an unidentified endogenous AMPK metabolite ligand: (i) β1-Ser108 and contacting α-residues are highly con- β served among eukaryotes; (ii) the β1-ADaM site is transient and Fig. 9 An integrated model for ULK1 regulation of 1-AMPK signaling. The regulated; (iii) both ULK1 phosphorylation of Ser108, and acti- initial ULK stimulus (e.g., Ser555 phosphorylation) is provided by AMPK, vation of AMPK by small molecule drugs, share a common activated itself in response to energy stress and elevated AMP:ATP. Once fi β activated, ULK1 suppresses Thr172 phosphorylation and AMPK sensitivity speci city for the 1-isoform; (iv) synergistic activation of 20 unphosphorylated AMPK, orchestrated across all three AMPK to AMP via negative feedback (red arrows) . ULK1 simultaneously subunits and topographically distant β1-ADaM and γ-sites13, phosphorylates Ser108, sensitizing AMPK to drugs/metabolites acting at appears a highly intrinsic activation mechanism. Additionally, the ADaM site. This would promote AMP-independent AMPK signaling independent studies have described a disconnect between AMPK and maintain ULK1 activity via positive feedback (green arrows). In our cell α signaling (using elevated pACC as an index) and apparent AMPK model, activity of 2(T172A) AMPK is not elevated in response to AMP and activation (no increase in pThr172), most commonly in exam- requires additional stimulation with A-769662 to achieve ULK1 – inations of the role of skeletal muscle AMPK during exercise39 42, phosphorylation but also in response to reactive oxygen species (HeLa cells)43 and berberine (LAMTOR1-KO MEFs)44. We now demonstrate that of β1-AMPK signaling (Fig. 8). An integrated model, involving significant cellular AMPK signaling can be triggered indepen- positive and negative feedback loops, is shown in Fig. 9. AMPK dently of pThr172 (Fig. 8a), although we are unable to determine signaling in these cells is hampered by insensitivity to AMP; whether this arises exclusively by Ser108 phosphorylation, consequently phenformin is unable to stimulate ULK1 activity, AMP/drug synergistic activation of unphosphorylated enzyme, or contributing to the weak pSer108 response (Fig. 8b). Further a combination of both. addition of A-769662 was sufficient to rescue AMPK activity and Applying quantitative phosphoproteomics to investigate a stimulate phosphorylation of ULK1-Ser555 and Ser108, poten- physiological role for Ser108 phosphorylation, we identified sig- tially via positive feedback. cis-autophosphorylation of Ser108 in nificant differences in the phosphorylation profiles of several cell the α2(T172A) mutant is unlikely since activity of the unpho- cycle-associated proteins in response to the phosphomimetic sphorylated complex can only be achieved synergistically; in the mutant β1-S108E (Fig. 2). Our approach was limited in that we AMP/drug-bound state Ser108 is presumably sequestered to the do not know whether potential metabolite(s) acting at the S108E- ADaM site and away from the AMPK active site. The close stabilized ADaM site were elevated in response to phenformin, correlation between pACC, pSer108 and ULK-pSer555 profiles thus our findings are likely an underestimation of the global across all conditions provides support for a positive feedback effects of Ser108 phosphorylation. It should also be emphasized mechanism between ULK and AMPK (Fig. 8a, b). Consistent with that the differences reflect both primary and downstream phos- this, co-incubation of cells expressing only KI-AMPK with phorylation events, so that the phosphosites detected encompass phenformin/A-769662 did not induce a further increase in more than direct AMPK sites. None of these phosphorylated pSer108, compared to phenformin treatment alone (Fig. 3d). Our residues have been identified as direct AMPK substrates, however, model raises the intriguing prospect that, on ULK1-populated both DOCK7_Ser452 and NHRF1_Ser285 are potential candi- membranes (autophagosome, mitochondria), AMPK may be dates based on consensus with the AMPK substrate recognition desensitized to AMP through suppression of α1-Thr172 phos- motif (Supplementary Table 1). β1-S108E expression induced a phorylation, yet simultaneously sensitized to an alternate reg- significant increase in downstream phosphorylation of the PAK2 ulatory ligand through phosphorylation of β1-Ser108. regulatory domain residue Ser141, an event required for full Initial attempts to examine trans-phosphorylation of Ser108 in kinase activity and subsequent induction of the cytostatic and cells were confounded by strong background autopho- anti-apoptotic functions of full-length PAK245, 46. Other down- sphorylation signals in HEK293T endogenous AMPK, and high stream targets are associated with regulation of cell cycle arrest basal levels of pSer108 in immortalized MEFs (Fig. 6c). The cause (Bin1; GO:0071156), p53-mediated cell cycle arrest (TB182; of elevated pSer108 in iMEFs is unknown. Since KI-AMPK was GO:0006977) and cell division (CND1; GO:0051301). Although not phosphorylated on Ser108 under basal conditions (Fig. 3d) we functional roles for the majority of sites detected in our analysis expect loss of a pSer108 phosphatase activity, leading to accu- are unknown, our findings with PAK2 in particular point to a role mulation of Ser108 autophosphorylation in endogenous AMPK, for β1-Ser108 phosphorylation in promoting energy stress- may be a contributing factor. We were able to exploit the intra- induced pro-survival pathways over cell death pathways. This molecular “limitation” of Ser108 cis-autophosphorylation by model is consistent with other autophagy-inducing roles for using KI-AMPK in cells to uncouple background autopho- AMPK, and Ser108-phosphorylation as a direct mechanism to sphorylation from input by alternate signaling pathways. We fi activate AMPK by the autophagy initiator ULK1. Future studies identi ed conditions (glucose starvation, H2O2 and phenformin) will delineate the cellular signaling mechanisms specificto under which Ser108 is phosphorylated independently of AMPK β1-Ser108 phosphorylation, and bridge the gap between AMPK activity (Fig. 3b). These conditions also increase cellular LC3-II or – and the other identified downstream substrates. ULK1-pSer555, both standard markers for autophagic activity47 49. Closer examination of our results from MEFs expressing only Confirmation of ULK1-mediated phosphorylation of Ser108 in α2(T172A) AMPK provides further insight into ULK regulation cells was provided by small molecule (Fig. 6a, b) or

10 NATURE COMMUNICATIONS | 8: 571 | DOI: 10.1038/s41467-017-00628-y | www.nature.com/naturecommunications NATURE COMMUNICATIONS | DOI: 10.1038/s41467-017-00628-y ARTICLE genetic (Fig. 6d) inhibition of ULK activity, each of requirement for myristoyl-group membrane embedment, or an which resulted in almost complete loss of stress-induced Ser108 additional cellular component is required to derepress ULK trans-phosphorylation. phosphorylation through myristoyl group sequestration. These ULK1 phosphorylation of Ser108 is specific to the AMPK results may represent a protective mechanism to ensure Ser108 β1-isoform (Fig. 5); this is likely due to sequence heterogeneity phosphorylation by ULK1 occurs exclusively at membrane proximal to Ser108 in the β2-isoform that diverges from the ULK surfaces. consensus found in the corresponding β1 residues24. Interest- In summary, we have identified an additional layer of com- ingly, β1 residues C-terminal to Ser108 (His109-Phe112) are munication between AMPK and ULK1, two major regulators of considered unfavorable for ULK substrates. However, Ser108 is cellular energy homeostasis. ULK1 phosphorylation of Ser108, a located at the end of a β-hairpin loop structure12 and post-translational modification important for AMPK drug action, C-terminal residues may not participate in substrate recognition may yield strategic opportunity to increase potency of AMPK- by ULK1. β1-Ser108 does not appear to be a favorable substrate targeting therapeutics. Differential phosphorylation of AMPK at for ULK2 (Fig. 4b, c), although we cannot rule out the possibility distinct cellular organelles/membranes raises the intriguing pos- that the two ULK2 preparations used in this study may have sibility of localized AMPK ligand sensitization. Myristoyl- possessed inherently low activity, given the current lack of switching is a consistent driving force for both mechanisms; knowledge regarding cellular regulation and bona fide substrates whether secondary targeting signals are required to partition for this kinase. Overall, our findings are in agreement with distinct AMPK complexes between organelles represents an area autophagy network mapping showing interaction with AMPK β1 of great interest for the future. is restricted to ULK1, whereas ULK2 association is limited to AMPK β241. Further investigation is warranted to determine Methods subsets of ULK1 and ULK2 downstream targets in vivo. Reagents. DNA oligos were from Sigma (Supplementary Table 3). Antibodies for Our results here, and those of others, provide strong evidence pan AMPK α (#2793, clone F6, 1:1000 dilution), FLAG (#2368, 1:1000 dilution), for potentiation of Ser108 phosphorylation by an AMP-myristoyl myc (#2276, clone 9B11, 1:1000 dilution), HA (#2367, clone 6E2, 1:1000 dilution), PAK1/2/3 (#2604, 1:1000 dilution), ULK1 (#4773, clone R600, 1:1000 dilution) and switch mechanism, in which AMP-induced conformational tubulin (#3873, clone DM1A, 1:1000 dilution), and phosphospecific antibodies for changes lead to ejection of the β-subunit myristoyl group from an AMPK α-pThr172 (#2535, clone 40H9, 1:1000 dilution), AMPK β1-pSer108 intramolecular-binding site3. The exposed myristoyl group pro- (#4181, 1:1000 dilution), ACC-pSer79 (#3661, 1:1000 dilution), PAK1/2-pSer144/ motes AMPK targeting to intracellular membranes and, in this Ser141 (#2606, 1:1000 dilution), ULK1-pSer555 (#5869, clone D1H4, 1:1000 dilution), and ULK1-pSer757 (#6888, 1:1000 dilution) were from Cell Signaling case, ULK1 co-localization. Cell treatments that promote pSer108 Technology. AMPK-β1 antibody (#ab58175, 1:1000 dilution), A-769662 independently of autophosphorylation (glucose starvation, H2O2 (#ab120335) and AICAR were from Abcam. IRDye 680RD- or 800CW-labeled and phenformin) (Fig. 3b) are associated with increased AMP/ anti-immunoglobulin G antibodies (1:10,000 dilution) and IRDye 680RD-labeled ATP ratio, either through mitochondrial toxicity or disruption of streptavidin (1:20,000 dilution) were from LI-COR Biosciences. Glutathione 50 Sepharose 4B and Streptavidin Sepharose high performance were from GE Life ATP production . Treatments that did not induce Ser108 Sciences. FLAG synthetic peptide (DYKDDDK) was provided by GL Biochem phosphorylation activate AMPK independently of AMP (A- (Shanghai). Other synthetic peptides were from Purar Chemicals. All peptides were 769662, AICAR, ionomycin, amino-acid deprivation), or activate purified by reversed-phase chromatography and stored as lyophilized powder. ULK independently of changes in adenylate charge (mTOR ULK2 recombinant protein was from Abcam. SBI-0206965, AZD8055 and INK128 inhibitors) (Fig. 7a, c). The AMPK myristoyl-switching effect of were from ApexBio. FuGENE HD transfection reagent was from Promega Cor- poration. All other reagents were from Sigma. ZMP (the metabolized product of AICAR and an AMP mimetic) has not yet been examined. Additionally, removal of either the γ Cell culture. COS7 and HEK293T cell lines were purchased from American Type AMP-sensing ability of AMPK with 1(D245A) mutation Culture Collection. All cell lines were maintained in Dulbecco’s modified Essential (Fig. 7b), or β-subunit myristoylation with β1(G2A) mutation medium (DMEM) containing 10% fetal bovine serum and antibiotics (penicillin, (Fig. 7d), abrogated glucose starvation- and/or phenformin- streptomycin) at 37 °C with 5% CO2. To generate iMEF cell lines, MEFs were induced Ser108 phosphorylation. β1(G2A) mutation was pre- extracted from WT or homozygous AMPK β1β2 null embryos (days 12–14 post-coitum), generated by crossing homozygous β1 and β2 null mice54. WT and viously shown to suppress AMPK partitioning to intracellular β β 3 AMPK 1/2 double knockout ( 1/2-dKO) MEFs were immortalized by Fugene membranes following glucose starvation . ULK1 is also recruited HD-mediated transfection with an SV40 large-T antigen expression construct. to membrane structures during autophagy initiation, notably the AMPK α1/2 double knockout (α1/2-dKO) and ULK1/2 double knockout autophagosome formation sites located near the ER51, and (ulk1/2-dKO) iMEFs were described previously23, 29. mitochondria52. ULK activity is associated with recruitment of other autophagy-associated proteins to the developing phago- Protein expression constructs. All mutants were generated using QuikChange site-directed mutagenesis kits (Stratagene). All constructs were sequence verified. phore, including VPS34 and Beclin-1, both of which are phos- Mammalian cell expression constructs were gifts from Reuben Shaw (pcDNA3 phorylated by AMPK to achieve differential regulation of pro- mouse FLAG-ULK1 (Addgene #27636) and pcDNA3 mouse FLAG-ULK2 and nonautophagy pathways53. Finally, β-myristoylation is (Addgene #27637)), Noboru Mizushima (pME18s-3xHA-human FIP200 (Addgene strongly implicated in AMPK recruitment to LC3-containing #24303)) and Do-Hyung Kim (pRK5 human myc-Atg13 (Addgene #31965)). SV40 puncta in the LKB1-deficient H23 human cancer cell line31. large-T antigen expression construct pBSSVD2005 was a gift from David Ron (Addgene plasmid #21826). Complementary DNA (cDNAs) for human AMPK α1 Collectively, these studies point to a close association of AMPK or α2 were generated with N-terminal FLAG-tag and cloned into pcDNA3 using with ULK1 and the developing phagophore, and an important XhoI/EcoRI (AMPK FLAG-α1) or XhoI/HindIII (AMPK FLAG-α2) restriction role for the exposed AMPK myristoyl group (myr switch ON) in sites. Lentivirus expression constructs LeGO-iG2, second generation viral packa- targeting AMPK to phagophore membranes. A similar mechan- ging vector psPax2 and ecotropic envelope vector pHCMV-EcoEnv were gifts from Carl Walkley (St Vincent’s Institute of Medical Research). cDNAs for human ism has been proposed to control AXIN-mediated AMPK AMPK β1 (WT and mutants), α1 (WT and kinase inactive D141A mutant), and α2 recruitment to the late endosome/lysosome membrane for (WT and T172A mutant) were generated with C-terminal FLAG-tag and cloned Thr172 phosphorylation and AMPK activation by LKB130.An into LeGO-iG2 using BamHI/NotI(β1) or EcoRI/NotI(α1 and α2) restriction sites. unexpected finding was myristoyl-dependent suppression of Ser108 phosphorylation by ULK1 in cell-free assays (Fig. 7e), Protein expression and purification. Heterotrimeric human AMPK (α1β1γ1 and α β γ α indicating that AMPK adopts a conformation in the myristoyl- 1 2 1 expressed as N-terminal His6- fusions; WT or mutants as indicated) was expressed in E. coli strain Rosetta (DE3) and purified using nickel-Sepharose and buried state (myr switch OFF) that is unfavorable for ULK1 size exclusion chromatography as described previously13. Heterotrimeric human phosphorylation of Ser108. Repression of Ser108 phosphorylation AMPK (α1β1γ1 and α1β2γ1 expressed as an N-terminal GST-α fusion; WT or was not alleviated by AMP; either this demonstrates a mutants as indicated) was expressed in HEK293T cells as described previously3.

NATURE COMMUNICATIONS | 8: 571 | DOI: 10.1038/s41467-017-00628-y | www.nature.com/naturecommunications 11 ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/s41467-017-00628-y

Briefly, HEK293T cells at 40% confluency were triply transfected with expression with 1° antibodies (diluted 1:500–1:2000 in PBS-T), prior to 30 min incubation constructs for AMPK α-, β-, and γ-subunits, using transfection reagent FuGENE with anti-rabbit or anti-mouse IgG 2° antibody fluorescently labeled with IR680 or HD according to the manufacturer’s protocols. Cells were harvested 48 h post- IR800. Immunoreactive bands were visualized on an Odyssey membrane imaging transfection in ice cold lysis buffer (50 mM Tris-HCl (pH 7.4), 150 mM NaCl, system (LI-COR Biosciences) and densitometry analyses performed using inte- 50 mM NaF, 1 mM sodium pyrophosphate, 1 mM EDTA, 1 mM EGTA, 1% (v/v) grated software. Uncropped western blot images associated with this study are Triton X-100, protease inhibitors). Lysates were clarified by centrifugation at presented in Supplementary Fig. 7. fl 14,000 rpm for 5 min and ash frozen in liquid N2 until processing. AMPK was fi puri ed on glutathione Sepharose 4B and eluted in 50 mM Tris-HCl, (pH 7.4), Quantitative global and phosphoproteomic analysis. β1/2-dKO iMEFs, trans- 100 mM NaCl, 10% glycerol supplemented with 20 mM glutathione. ULK1/2 and duced with AMPK β1-S108A or β1-S108E, as described above, were incubated with ULK1/FIP200/Atg13 complex was expressed by FuGENE HD transient transfec- 2 mM phenformin for 1 h. Treated lysates were acetone (−20 °C) precipitated and fi tion of HEK293T cells. Lysates were harvested after 48 h, clari ed by centrifugation resuspended in 8 M urea/50 mM triethyl ammonium bicarbonate (TEAB) solution. fi fi and ULK puri ed on anti-FLAG M2 af nity gel. ULK was eluted in 50 mM Protein lysates were reduced with 10 mM tris(2-carboxyethyl)phosphine (TCEP) Tris-HCl (pH 7.4), 100 mM NaCl, 10% glycerol, 0.01% Tween-20 supplemented for 45 min at 37 °C, and alkylated with 55 mM iodoacetamide for 30 min at room with 1 mg/ml FLAG peptide. CaMKK2 isoform 1 was produced in Sf21 insect cells temperature in the dark. Samples were diluted to 1 M urea in 25 mM TEAB and 3 fl as a C-terminal FLAG fusion as described previously . Brie y, Sf21 cells were trypsin (1:60, w/w) digested overnight at 37 °C. Digested tryptic peptides were infected at a multiplicity of infection of 10 and harvested 72 h post-infection. Cell cleaned up using Oasis HLB (hydrophilic lipophilic balance) solid phase extraction fi lysates were prepared as for recombinant AMPK, and protein was puri ed on anti- (SPE) cartridges (Waters) and freeze-dried overnight. Resuspended tryptic digests fi FLAG M2 af nity gel. CaMKK2 was eluted in 50 mM Tris-HCl (pH 7.4), 100 mM were used for stable-isotope dimethyl labeling as described previously55. β1-S108A NaCl, 10% glycerol, 0.01% Tween-20 supplemented with 1 mg/ml FLAG peptide. and β1-S108E derived tryptic peptides were used for light and medium labeling, respectively. Labeled peptides were mixed (1:1) and an aliquot (15 µl) was taken out Lentiviral-mediated AMPK expression and purification. Ecotropic lentivirus and run on liquid chromatography–mass spectrometry (LC-MS)/MS for total was generated by transient transfection of HEK293T cells using calcium phosphate. proteome changes. Remaining mixed labeled peptides were SPE cleaned up and 6 Briefly, 1 day prior to transfection, 2–2.5 × 10 HEK293T cells were seeded per freeze-dried before used for TiO2 phosphopeptides enrichment. Phosphopeptide 56 10 cm culture dish. LeGO-iG2, psPax2 and pHCMV-EcoEnv plasmids (10 μg, enrichment by TiO2 microcolumns was carried out as described previously . 6.3 μg and 3.8 μg per 10 cm culture dish, respectively) were mixed together with Mixed labeled peptides and TiO2 enriched phosphopeptides were analyzed by LC- fi μ fi fi 2 M CaCl2 solution (244 mM nal concentration in 500 l). The DNA/CaCl2 MS/MS using a Q-Exactive plus mass spectrometer (Thermo Scienti c) tted with solution was added drop-wise, with vortexing, into 500 μl of 2xHEPES-buffered nanoflow reversed-phase-high-performance liquid chromatography (HPLC) saline, pH 7.06. After 20 min incubation at RT, the mixture was transferred drop- (Ultimate 3000 RSLC, Dionex). The nano-LC system was equipped with an wise onto HEK293T cell culture and incubated. Within 16 h of transfection, cell Acclaim Pepmap nano-trap column (Dionex—C18, 100 Å, 75 μm × 2 cm) and an were washed with phosphate-buffered saline (PBS) and replaced with 6 ml fresh Acclaim Pepmap RSLC analytical column (Dionex—C18, 100 Å, 75 μm × 50 cm). media. The lentivirus-containing supernatant was harvested after 48 and 72 h post- Typically for each LC-MS/MS experiment, 5 μl of the peptide mix was loaded onto fl μ transfection and stored at −80 °C. the enrichment (trap) column at an isocratic ow of 5 l/min of 3% CH3CN iMEFs were transduced in 6-well plates by spinoculation. Briefly, 1 day prior to containing 0.1% formic acid for 6 min before the enrichment column is switched transduction, 0.5 × 105 cells per well in a 6-well plate were seeded in 2 ml media. in-line with the analytical column. The eluents used for the LC were 0.1% v/v 200 μl lentivirus supernatant in the presence of 8 μg/ml polybrene in 2 ml media formic acid (solvent A) and 100% CH3CN/0.1% formic acid v/v (solvent B). The was spun for 98 min at 25 °C, 1100×g (Heraeus Megafuge 2.0 R). Lentivirus- gradient used was 3% B to 25% B for 23 min, 25% B to 40% B in 2 min, 40% B to containing media was replaced with an equal volume of fresh media after 24 h. 80% B in 2 min and maintained at 85% B for the final 2 min before equilibration Seventy-two hourspost-transduction, iMEFs were incubated with fresh media for 1 for 9 min at 3% B prior to the next analysis. All spectra were acquired in positive h and treated as indicated. Cells were harvested by washing with ice-cold PBS, mode with full scan MS spectra scanning from m/z 375–1400 at 70,000 resolution followed by rapid lysis in situ using 100 μl ice-cold lysis buffer (50 mM Tris.HCl with AGC target of 3e6 with maximum accumulation time of 50 ms. Lockmass of (pH 7.4), 150 mM NaCl, 50 mM NaF, 1 mM sodium pyrophosphate, 1 mM EDTA, 445.120024 was used. The 15 most intense peptide ions with charge states ≥2–5 1 mM EGTA, 1 mM dithiothreitol, 1% (v/v) Triton X-100 and protease inhibitors), were isolated with isolation window of 1.2 m/z and fragmented with normalized and cellular debris was removed by centrifugation. collision energy of 30 at 35,000 resolution with AGC target of 1e5 with maximum FLAG-tagged AMPK α1orα2, and acetyl-CoA carboxylase (ACC), were accumulation time of 120 ms. Underfill threshold was set to 2% for triggering of isolated from iMEF cell lysates using anti-FLAG M2 affinity gel or Streptavidin precursor for MS2. Dynamic exclusion was activated for 30 s. Mass spectrometric Sepharose high performance, respectively. Immobilized ACC was washed raw data were processed and analyzed using Proteome Discoverer 2.1 (Thermo extensively with buffer A and eluted in 2×sodium dodecyl sulfate polyacrylamide Scientific) with Mascot search algorithm against mouse SwissProt database. Perseus 57 gel electrophoresis (SDS-PAGE) loading sample buffer for immunoblotting. 1.5.6.0 was used for further data analysis . IPA software (QIAGEN Redwood City) was used for network and pathway analysis. Kinase activity assays. ULK and AMPK activities were determined by phosphorylation of a synthetic peptide (S108tide: KLPLTRSHNNFVARRR, corre- In-gel digestion of protein bands and mass spectrometry. SDS-PAGE protein sponding to AMPK β1(102–114) with three additional C-terminal Arg residues to bands were excised and simultaneously reduced and alkylated with 10 mM TCEP promote binding to P81 phosphocellulose paper) using 200 μM[γ-32P]ATP and and 40 mM 2-choloracetamide for 1 h at room temperature. Proteins were digested μ with 13 ng/μl of sequencing grade trypsin (Promega) overnight at 37 °C and 5 mM MgCl2 in a 25 l reaction volume at 30 °C. Reactions were terminated after 10 min by spotting 15 μl onto P81 phosphocellulose paper (Whatman) and peptides desalted using in-house made microC18 columns (3 M empore). Peptides washing in 1% phosphoric acid. Radioactivity was quantified by scintillation were resuspended in 0.1% formic acid, 5% acetonitrile and analyzed on a Dionex counting. 3500RS nanoUHPLC coupled to an Orbitrap Fusion mass spectrometer with Tune v2.0.1258 in positive mode. Peptides were separated using an in-house packed 75 μm × 40 cm pulled column (1.9 μm particle size, C18AQ; Dr Maisch, Germany) Kinase screen. Kinase screening (IKPT service) was performed by Kinexus, with a gradient of 2–30% acetonitrile containing 0.1% FA over 60 min at 250 nl/ Canada. S108tide (100 μM) was screened against a panel of 284 selected Ser/Thr min at 55 °C. An MS1 scan was acquired from 350–1550 (120,000 resolution, 5e5 kinases with single replicate. Most assays were performed for 15 min duration, AGC, 100 ms injection time) followed by MS/MS data-dependent acquisition with 33 50 μM[γ- P]ATP, in a 25 μl reaction volume at 30 °C. HCD and detection in the Orbitrap (60,000 resolution, 2e5 AGC, 120 ms injection time, 40 NCE, 2.0 m/z quadrupole isolation width) and, EThcD and detection in 5 Generation of AMPK β2-pSer108 phosphospecific antibody. A phosphorylated the orbitrap (60,000 resolution, 2e AGC, 120 ms injection time, calibrated charge- synthetic peptide (CSTKIPLIKpSHNDFVAILD, corresponding to AMPK β2 dependent ETD reaction times [2 + 121; 3 + 54; 4 + 30; 5 + 20; 6 + 13; 7 + ; 10 ms], – 25 NCE for HCD supplemental activation, 2.0 m/z quadrupole isolation width). All (100 117)) was coupled to keyhole limpet hemocyanin via the peptide N-terminal 58 cysteine residue using the coupling reagent N-succinimidyl-3(-2-pyridyldithio) raw data were analyzed with MaxQuant v1.5.3.25 and searched against the propionate. Rabbits were immunized with 2 mg of peptide conjugate initially in human UniProt database with default settings including phosphorylation of S, T, fi 50% (v/v) Freunds complete adjuvant and in 50% (v/v) Freunds incomplete and Y as a variable modi cation and match between runs enabled. adjuvant for subsequent immunizations. Rabbits were boosted fortnightly with 2 mg of peptide conjugate and bled 7 days after booster injections. The pSer108 Nucleotide measurements. Adenine nucleotides from HEK293T perchlorate antibody was then purified from serum by peptide affinity chromatography. Spe- extracts were measured by LC-MS on an ABISCIEX 5500 QTRAP mass cificity for β2-pSer108 was evaluated by immunoblot against CaMKK2- spectrometer16. AEC was calculated from ratios of [AMP], [ADP], and [ATP] phosphorylated AMPK α1β2γ1, and α1β1γ1. (Equation 1):

Immunoblotting. Samples were electrophoresed by 12% SDS-PAGE and trans- ﬂ ferred to Immobilon FL polyvinylidine- ouride membrane (Millipore). The ½þðÞ: ½ membrane was blocked with 2% nonfat dry milk in PBS +0.1% Tween 20 (PBS-T) ¼ ATP 0 5 ADP ð Þ AEC ½þ½þ½ 1 for 1 h at room temperature. Membranes were incubated either overnight or for 1 h ATP ADP AMP

12 NATURE COMMUNICATIONS | 8: 571 | DOI: 10.1038/s41467-017-00628-y | www.nature.com/naturecommunications NATURE COMMUNICATIONS | DOI: 10.1038/s41467-017-00628-y ARTICLE

Statistical analysis. The data are presented as mean values ± s.e.m. of at least 26. Erickson, J. R. et al. A dynamic pathway for calcium-independent activation of three independent experiments. The unpaired two-tailed Student’s t-test was used CaMKII by methionine oxidation. Cell 133, 462–474 (2008). for all comparisons unless stated. 27. Humphries, K. M., Deal, M. S. & Taylor, S. S. Enhanced dephosphorylation of cAMP-dependent protein kinase by oxidation and thiol modification. J. Biol. – Data availability. The mass spectrometry proteomics data have been deposited to Chem. 280, 2750 2758 (2005). the ProteomeXchange Consortium via the PRIDE59 partner repository with the 28. Cosentino-Gomes, D., Rocco-Machado, N. & Meyer-Fernandes, J. R. Cell data set identifier PXD006365. The authors declare that all other data supporting signaling through protein kinase C oxidation and activation. Int. J. Mol. Sci. 13, the findings of this study are available within the paper and its Supplementary 10697–10721 (2012). Information files, or are available from the authors upon reasonable request. 29. Cheong, H., Lindsten, T., Wu, J., Lu, C. & Thompson, C. B. Ammonia-induced autophagy is independent of ULK1/ULK2 kinases. Proc. Natl Acad. Sci. USA 108, 11121–11126 (2011). Received: 27 September 2016 Accepted: 14 July 2017 30. Zhang, Y.-L. et al. AMP as a low-energy charge signal autonomously initiates assembly of AXIN-AMPK-LKB1 complex for AMPK activation. Cell Metab. 18, 546–555 (2013). 31. Liang, J. et al. Myristoylation confers noncanonical AMPK functions in References autophagy selectivity and mitochondrial surveillance. Nat. Commun. 6, 7926 1. Hardie, D. G. AMPK - sensing energy while talking to other signaling pathways. (2015). Cell Metab. 20, 939–952 (2014). 32. Egan, D. F. et al. Phosphorylation of ULK1 (hATG1) by AMP-activated protein 2. Steinberg, G. R. & Kemp, B. E. AMPK in health and disease. Physiol. Rev. 89, kinase connects energy sensing to mitophagy. Science 331, 456–461 (2011). 1025–1078 (2009). 33. Ford, R. J. et al. Metformin and salicylate synergistically activate liver AMPK, 3. Oakhill, J. S. et al. β-Subunit myristoylation is the gatekeeper for initiating inhibit lipogenesis and improve insulin sensitivity. Biochem. J. 468, 125–132 metabolic stress sensing by AMP-activated protein kinase (AMPK). Proc. Natl (2015). Acad. Sci. USA 107, 19237–19241 (2010). 34. O’Brien, A. J. et al. Salicylate activates AMPK and synergizes with metformin to 4. Houde, V. P. et al. Investigation of LKB1 Ser431 phosphorylation and Cys433 reduce the survival of prostate and lung cancer cells ex vivo through inhibition farnesylation using mouse knockin analysis reveals an unexpected role of of de novo lipogenesis. Biochem. J. 469, 177–187 (2015). prenylation in regulating AMPK activity. Biochem. J. 15,41–56 (2014). 35. Fullerton, M. D. et al. Salicylate improves macrophage cholesterol homeostasis 5. Xiao, B. et al. Structure of mammalian AMPK and its regulation by ADP. via activation of Ampk. J. Lipid Res. 56, 1025–1033 (2015). Nature 472, 230–233 (2011). 36. Zadra, G. et al. A novel direct activator of AMPK inhibits prostate cancer 6. Oakhill, J. S. et al. AMPK is a direct adenylate charge-regulated protein kinase. growth by blocking lipogenesis. EMBO Mol. Med. 6, 519–538 (2014). – Science 332, 1433 1435 (2011). 37. Cameron, K. O. et al. Discovery and preclinical characterization of 6-chloro-5- 7. Chen, L. et al. AMP-activated protein kinase undergoes nucleotide-dependent [4-(1-hydroxycyclobutyl)phenyl]-1H-indole-3-carboxylic acid (PF-06409577), – conformational changes. Nat. Struct. Mol. Biol. 19, 716 718 (2012). a direct activator of adenosine monophosphate-activated protein kinase 8. Hawley, S. A. et al. Characterization of the AMP-activated protein kinase kinase (AMPK), for the potential treatment of diabetic nephropathy. J. Med. Chem. 59, fi from rat liver and identi cation of threonine 172 as the major site at which it 8068–8081 (2016). – phosphorylates AMP-activated protein kinase. J. Biol. Chem. 271, 27879 27887 38. Shackleford, D. B. & Shaw, R. J. The LKB1-AMPK pathway: metabolism and (1996). growth control in tumour suppression. Nat. Rev. Cancer 9, 563–575 (2009). 9. Scott, J. W. et al. CBS domains form energy-sensing modules whose binding of 39. Wojtaszewski, J. F. et al. Dissociation of AMPK activity and ACCβ adenosine ligands is disrupted by disease mutations. J. Clin. Invest. 113, phosphorylation in human muscle during prolonged exercise. Biochem. 274–284 (2004). Biophys. Res. Commun. 298, 309–316 (2002). 10. Langendorf, C. G. et al. Structural basis of allosteric and synergistic activation of 40. Jørgensen, S. B. et al. Knockout of the alpha2 but not alpha1 5′-AMP-activated AMPK by furan-2-phosphonic derivative C2 binding. Nat. Commun. 7, 10912 protein kinase isoform abolishes 5-aminoimidazole-4-carboxamide-1-beta-4- (2016). ribofuranoside but not contraction-induced glucose uptake in skeletal muscle. J. 11. Sanders, M. J. et al. Defining the mechanism of activation of AMP-activated Biol. Chem. 279, 1070–1079 (2004). protein kinase by the small molecule A-769662, a member of the – 41. Dzamko, N. et al. AMPK-independent pathways regulate skeletal muscle fatty thienopyridone family. J. Biol. Chem. 282, 32539 32548 (2007). – 12. Xiao, B. et al. Structural basis of AMPK regulation by small molecule activators. acid oxidation. J. Physiol. 586, 5819 5831 (2008). 42. Schwalm, C. et al. Activation of autophagy in human skeletal muscle is Nat. Commun. 4, 3017 (2013). – 13. Scott, J. W. et al. Small molecule drug A-769662 and AMP synergistically dependent on exercise intensity and AMPK activation. FASEB J. 29, 3515 3526 activate naïve AMPK independent of upstream kinase signaling. Chem. Biol. 21, (2015). – 43. Alexander, A. et al. ATM signals to TSC2 in the cytoplasm to regulate 619 627 (2014). – 14. Calabrese, M. F. et al. Structural basis for AMPK activation: natural and mTORC1 in response to ROS. Proc. Natl Acad. Sci. USA 107, 4153 4158 synthetic ligands regulate kinase activity from opposite poles by different (2010). molecular mechanisms. Structure 22, 1161 –1172 (2014). 44. Zhang, S.-C. et al. The lysosomal v-ATPase-ragulator complex is a common 15. Hawley, S. A. et al. The ancient drug salicylate directly activates AMP-activated activator for AMPK and mTORC1, acting as a switch between catabolism and – protein kinase. Science 336, 918–922 (2012). anabolism. Cell Metab. 20, 526 540 (2014). 16. Scott, J. W. et al. Inhibition of AMP-activated protein kinase at the allosteric 45. Jung, J. H. & Traugh, J. A. Regulation of the interaction of Pak2 with Cdc42 via – drug-binding site promotes islet insulin release. Chem. Biol. 22, 705–711 (2015). autophosphorylation of serine 141. J. Biol. Chem. 280, 40025 40031 (2005). 17. Scott, J. W. et al. Thienopyridone drugs are selective activators of 46. Jakobi, R., Moertl, E. & Koeppel, M. A. p21-activated protein kinase gamma- fi AMP-activated protein kinase beta1-containing complexes. Chem. Biol. 15, PAK suppresses programmed cell death of BALB3T3 broblasts. J. Biol. Chem. – 1220–1230 (2008). 276, 16624 16634 (2001). 18. Oakhill, J. S., Scott, J. W. & Kemp, B. E. AMPK functions as an adenylate 47. Kundu, M. et al. Ulk1 plays a critical role in the autophagic clearance of charge-regulated protein kinase. Trends Endocrinol. Metab. 23, 125–132 (2012). mitochondria and ribosomes during reticulocyte maturation. Blood 112, 19. Suzuki, T. et al. Inhibition of AMPK catabolic action by GSK3. Mol. Cell 50, 1493–1502 (2008). 407–419 (2013). 48. Mitter, S. K. et al. Dysregulated autophagy in the RPE is associated with 20. Löffler, A. S. et al. Ulk1-mediated phosphorylation of AMPK constitutes a increased susceptibility to oxidative stress and AMD. Autophagy 10, 1989–2005 negative regulatory feedback loop. Autophagy 7, 696–706 (2011). (2014). 21. Mitchelhill, K. I. et al. Posttranslational modifications of the 5’-AMP-activated 49. Behrends, C., Sowa, M. E., Gygi, S. P. & Harper, J. W. Network organization of protein kinase β1 subunit. J. Biol. Chem. 272, 24475–24479 (1997). the human autophagy system. Nature 466,68–76 (2010). 22. O’Neill, H. M. et al. AMP-activated protein kinase (AMPK) beta1beta2 muscle 50. Hardie, D. G. AMPK: positive and negative regulation, and its role in whole- null mice reveal an essential role for AMPK in maintaining mitochondrial body energy homeostasis. Curr. Opin. Cell Biol. 33,1–7 (2015). content and glucose uptake during exercise. Proc. Natl Acad. Sci. USA 108, 51. Russell, R. C., Yuan, H. X. & Guan, K. L. Autophagy regulation by nutrient 16092–16097 (2011). signaling. Cell Res. 24,42–57 (2014). 23. Laderoute, K. R. et al. 5-AMP-activated protein kinase (AMPK) is induced by 52. Wu, W. et al. ULK1 translocates to mitochondria and phosphorylates FUNDC1 low-oxygen and glucose deprivation conditions found in solid-tumor to regulate mitophagy. EMBO Rep. 15, 566–575 (2014). microenvironments. Mol. Cell. Biol. 26, 5336–5347 (2016). 53. Kim, J. et al. Differential regulation of distinct Vps34 complexes by AMPK in 24. Egan, D. F. et al. Small molecule inhibition of the autophagy kinase ULK1 and nutrient stress and autophagy. Cell 152, 290–303 (2013). identification of ULK1 substrates. Mol. Cell 59, 285–297 (2015). 54. Davies, M. et al. Novel mechanisms of Na+ retention in obesity: 25. Truong, T. H. & Carroll, K. S. Redox regulation of protein kinases. Crit. Rev. phosphorylation of NKCC2 and regulation of SPAK/OSR1 by AMPK. Am. J. Biochem. Mol. Biol. 48, 332–356 (2013). Physiol. Renal Physiol. 307, F96–F106 (2014).

NATURE COMMUNICATIONS | 8: 571 | DOI: 10.1038/s41467-017-00628-y | www.nature.com/naturecommunications 13 ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/s41467-017-00628-y

55. Boersema, P. J., Raijmakers, R., Lemeer, S., Mohammed, S. & Heck, A. J. Additional information Multiplex peptide stable isotope dimethyl labeling for quantitative proteomics. Supplementary Information accompanies this paper at doi:10.1038/s41467-017-00628-y. Nat. Protoc. 4, 484–494 (2009). 56. Thingholm, T. E., Jorgensen, T. J., Jensen, O. N. & Larsen, M. R. Highly Competing interests: The authors declare no competing financial interests. selective enrichment of phosphorylated peptides using titanium dioxide. Nat. Protoc. 1, 1929–1935 (2006). Reprints and permission information is available online at http://npg.nature.com/ 57. Tyanova, S. et al. The Perseus computational platform for comprehensive reprintsandpermissions/ analysis of (prote)omics data. Nat. Methods 13, 731–740 (2016). 58. Cox, J. & Mann, M. MaxQuant enables high peptide identification rates, Publisher's note: Springer Nature remains neutral with regard to jurisdictional claims in individualized p.p.b.-range mass accuracies and proteome-wide protein published maps and institutional affiliations. quantification. Nat. Biotechnol. 26, 1367–1372 (2008). 59. Vizcaíno, J. A. et al. 2016 update of the PRIDE database and related tools. Nucleic Acids Res. 44, D447–D456 (2016). Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, Acknowledgements adaptation, distribution and reproduction in any medium or format, as long as you give This work was supported by grants from the Australian Research Council (ARC) and the appropriate credit to the original author(s) and the source, provide a link to the Creative National Health and Medical Research Council (NHMRC). G.R.S. is a Canada Research Commons license, and indicate if changes were made. The images or other third party Chair in Metabolism and Obesity and the J. Bruce Duncan Chair in Metabolic Diseases. material in this article are included in the article’s Creative Commons license, unless B.E.K. is an NHMRC Research Fellow. J.S.O. is an ARC Future Fellow. Supported in part indicated otherwise in a credit line to the material. If material is not included in the by the Victorian Government’s Operational Infrastructure Support Program. article’s Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from Author contributions the copyright holder. To view a copy of this license, visit http://creativecommons.org/ T.A.D., N.X.Y.L., and J.S.O. designed and coordinated the study. T.A.D., N.X.Y.L., J.W.S., licenses/by/4.0/. A.H., B.L.P., K.R.W.N., M.T.O., and C.G.L. performed the experiments. S.G., M.K., and B.V. provided immortalized mouse embryonic fibroblasts. J.W.S., G.R.S., K.S., and B.E.K. provided intellectual input. All authors contributed to writing the manuscript. © The Author(s) 2017

14 NATURE COMMUNICATIONS | 8: 571 | DOI: 10.1038/s41467-017-00628-y | www.nature.com/naturecommunications CHAPTER 4: AMPK is selectively inhibited by the type II inhibitor SBI-0206965 4.1 Introduction The idea that active site kinase inhibitors could be useful for the treatment of diseases has been encouraged by the outstanding effectiveness of Imatinib, an Abl kinase inhibitor, in the treatment of chronic myeloid leukaemia. It was thought that high intracellular ATP concentrations would prevent active site inhibitors from being effective, and conservation of the ATP binding pocket between kinases would result in a large number of off-target effects (283). However today there are 28 FDA approved small-molecule kinase inhibitors, and new inhibitors constitute a significant percentage of drug-discovery programs in the pharmaceutical industry (284, 285).

Imatinib is a type II inhibitor, which binds selectively to the DFG-out conformation of Abl kinase. As type I inhibitors are able to bind to either the DFG-in or DFG-out conformation of kinases, type II inhibitors are considered intrinsically more selective than type I (236). In addition type II inhibitors maintain high potency in cells where the concentration of ATP is high (236).

The potential strategy to treat obesity, type 2 diabetes, stroke, cancer and neurodegeneration via targeted pharmacological AMPK inhibition is encumbered by the lack of potent and selective antagonists of AMPK.

I have discovered that the small molecule SBI-0206965, originally reported as a ULK1 inhibitor, is a direct, type IIb AMPK inhibitor that has improved potency and kinase selectivity relative to the existing type I inhibitor, compound C. A crystal structure of the AMPK α2 kinase domain/SBI-0206965 complex shows that SBI-0206965 binds to AMPK at the active site, however biochemical characterisation reveals mixed-competitive kinetics with ATP. SBI- 0206965 inhibited AMPK signalling in a variety of cell types and suppressed proliferation of the prostate cancer cell line PC3. This is reported in the published manuscript.

ARTICLE cro

AMP-activated protein kinase selectively inhibited by the type II inhibitor SBI-0206965 Received for publication, April 18, 2018 Published, Papers in Press, April 25, 2018, DOI 10.1074/jbc.RA118.003547 Toby A. Dite‡, Christopher G. Langendorf§1, Ashfaqul Hoque‡, Sandra Galic§, Richard J. Rebello¶ʈ, Ashley J. Ovens‡, Lisa M. Lindqvist**, Kevin R. W. Ngoei§, Naomi X. Y. Ling‡, Luc Furic¶ʈ‡‡2, Bruce E. Kemp§ §§3, John W. Scott§§§, and Jonathan S. Oakhill‡§§4 From the ‡Metabolic Signalling Laboratory and §Protein Chemistry and Metabolism Unit, St. Vincent’s Institute of Medical Research, University of Melbourne, Fitzroy 3065, Victoria, Australia, the ¶Prostate Cancer Translational Research Laboratory, Peter MacCallum Cancer Centre, Melbourne, Victoria 3000, Australia, the ʈCancer Program, Biomedicine Discovery Institute, and Department of Anatomy and Developmental Biology, Monash University, Clayton 3800, Victoria, Australia, the **Cell Signalling and Cell Death Division, Walter and Eliza Hall Institute of Medical Research, Parkville, Victoria 3052, Australia, the ‡‡Sir Peter MacCallum Department of Oncology, University of Melbourne, Parkville, Victoria 3010, Australia, and the §§Mary MacKillop Institute for Health Research, Australian Catholic University, Melbourne, Victoria 3000, Australia Downloaded from Edited by Henrik G. Dohlman

Inhibition of the metabolic regulator AMP-activated protein whole-body metabolism (1, 2). AMPK is an evolutionarily con- kinase (AMPK) is increasingly being investigated for its thera- served serine/threonine kinase that senses, and is activated by, peutic potential in diseases where AMPK hyperactivity results low adenylate charge (elevated AMP/ATP and ADP/ATP http://www.jbc.org/ in poor prognoses, as in established cancers and neurodegenera- ratios) resulting from energetically demanding processes such tion. However, AMPK-inhibitory tool compounds are largely as muscle contraction or reduced energy supply caused by limited to compound C, which has a poor selectivity profile. hypoxia or nutrient deprivation. AMPK redirects cellular Here we identify the pyrimidine derivative SBI-0206965 as metabolism by down-regulating numerous ATP-consuming a direct AMPK inhibitor. SBI-0206965 inhibits AMPK with anabolic processes (e.g. protein, cholesterol, and fatty acid syn- 40-fold greater potency and markedly lower kinase promiscuity thesis) and up-regulating similarly diverse ATP-producing cat- at UNIVERSITY OF DUNDEE on June 8, 2018 than compound C and inhibits cellular AMPK signaling. Bio- abolic processes (e.g. fat oxidation, glycolysis, and autophagy) to chemical characterization reveals that SBI-0206965 is a mixed- restore energy balance. It does this by direct phosphorylation, type inhibitor. A co-crystal structure of the AMPK kinase either modulating the activities of rate-limiting enzymes in domain/SBI-0206965 complex shows that the drug occupies a multiple metabolic processes or regulating transcriptional pocket that partially overlaps the ATP active site in a type IIb activities of the factors governing their expression. For exam- inhibitor manner. SBI-0206965 has utility as a tool compound ple, AMPK phosphorylation of cytosolic acetyl-CoA carboxyl- for investigating physiological roles for AMPK and provides ase 1 (ACC1) and mitochondrial-associated ACC2, inhibits de fresh impetus to small-molecule AMPK inhibitor therapeutic novo lipogenesis and promotes fat oxidation, respectively. development. AMPK signaling has also been associated with a range of non- metabolic regulatory roles (e.g. circadian rhythm, mitochondrial fission, and appetite control). The ability to maintain energy homeostasis during acute or AMPK is an ␣␤␥ heterotrimeric complex, consisting of a chronic periods of nutrient shortfall is an essential characteris- catalytic ␣ subunit (isoforms ␣1 and ␣2) and regulatory sub- tic of all living organisms. A direct molecular link between units ␤ (isoforms ␤1 and ␤2) and ␥ (isoforms ␥1, ␥2, and ␥3). nutrient supply and energy demand is provided by AMP-acti- The ␣ subunit contains a canonical, bi-lobed kinase domain at vated protein kinase (AMPK),5 a key regulator of cellular and the N terminus, followed by autoinhibitory (AID) and scaffolding domains, and a Ser/Thr-rich loop region (ST loop). The ␤ This work was supported by grants from the Australian Research Council (ARC), subunit contains a mid-molecule carbohydrate-binding mod- Cancer Australia, the National Health and Medical Research Council (NHMRC), ule (CBM) and C-terminal scaffolding domain, whereas the ␥ and the Jack Brockhoff foundation (Grant JBF-4206, 2016). The authors declare ␤ that they have no conflicts of interest with the contents of this article. subunit consists of four cystathionine -synthase domains and This article contains Table S1 and Figs. S1–S4. three adenine nucleotide binding sites (termed ␥-sites 1, 3, and The atomic coordinates and structure factors (code 6BX6) have been deposited in 4) that endow AMPK with its energy-sensing capabilities. the Protein Data Bank (http://wwpdb.org/). 1 An NHMRC Early Career Research Fellow. Supported in part by the Victorian AMPK signaling is dynamically and tightly controlled by Government’s Operational Infrastructure Support Program. numerous activating/inhibiting and localization mechanisms. 2 Supported by the Department of Health and Human Services acting Spatially, AMPK is targeted to various subcellular organelles through the Victorian Cancer Agency (MCRF16007). 3 An NHMRC Research Fellow. 4 To whom correspondence should be addressed: Metabolic Signalling Lab- drate-binding module; CaMKK2, Ca2ϩ/calmodulin-dependent protein oratory, St. Vincent’s Institute of Medical Research, Fitzroy 3065, Victoria, kinase kinase 2; AICAR, 5-aminoimidazole-4-carboxamide ribonucleoside; Australia. Tel.: 61-3-9231-2480; E-mail: [email protected]. PDB, Protein Data Bank; DMEM, Dulbecco’s modified Eagle’s medium; 5 The abbreviations and trivial name used are: AMPK, AMP-activated protein ANOVA, analysis of variance; SBI-0206965, 2-(5-bromo-2-(3,4,5-trime- kinase; ACC, acetyl-CoA carboxylase; AID, autoinhibitory; CBM, carbohy- thoxyphenylamino)pyrimidin-4-yloxy)-N-methylbenzamide.

and compartments in response to leptin and circadian rhythm cytes (25). Compound C is an ATP-competitive inhibitor and (nucleus) (3, 4), glucose starvation (lysosomes, plasma mem- binds to the highly conserved active site of AMPK (26). How- branes) (5, 6), mitochondrial damage (mitochondria), and ever, in vitro screening has shown that compound C is promis- increased autophagic flux (autophagosomes) (7, 8). Activation cuous, inhibiting multiple kinases with similar or greater is triggered primarily by binding of AMP or ADP to exchange- potency than AMPK (27). Numerous studies have since able ␥-sites 1 and 3, stimulating phosphorylation of the kinase described off-target or AMPK-independent cellular effects, activation loop residue Thr-172 (pThr-172) by upstream including inhibition of bone morphogenetic protein type I ϩ kinases LKB1 and Ca2 /calmodulin-dependent protein kinase receptors ALK2, ALK3, and ALK6 (28), hypoxia-induced HIF-1 kinase 2 (CaMKK2) and co-localizing AMPK with substrates on activation (29), preadipocyte proliferation (30), and macro- intracellular membranes (6, 9). pThr-172–independent activity phage chemotaxis (31). Compound C also blocks AICAR cellu- has been demonstrated, although the contribution this makes lar uptake through competition for adenosine transporter to overall AMPK metabolic control is presently unclear (8, 10). binding sites, which largely accounts for its suppressive effects AMPK signaling is negatively regulated by multiple mecha- on AICAR-mediated AMPK activation (13). In light of these nisms, including exchange of AMP/ADP for ATP, AID-medi- undesirable properties, conclusions drawn from using com- ated autoinhibition, pThr-172–dephosphorylating actions of pound C as an AMPK inhibitor are now viewed with extreme phosphatases, phosphorylation of suppressive regulatory sites caution, and recommendations are to avoid its application alto- in the ␣-ST loop, and a variety of nutrients (glucose, amino gether (27). Other characterized direct antagonists include Downloaded from acids, and lipids), hormones (insulin, leptin, resistin), and cyto- MT47-100, a low-potency, ␤2-AMPK allosteric inhibitor that kines (tumor necrosis factor ␣, ciliary neurotrophic factor, and intriguingly also activates ␤1-complexes (32), and SU6656, that interleukin-6) (11). paradoxically stimulates net cellular AMPK signaling by pro- The metabolic dimensions associated with major human dis- moting phosphorylation of Thr-172 by LKB1 (33).

eases, such as type 2 diabetes, cancer, and inflammatory dis- Here, we report the discovery of the small-molecule SBI- http://www.jbc.org/ orders, have encouraged efforts to develop small-molecule 0206965 as a direct, type IIb AMPK inhibitor that demon- AMPK activators. Patented examples now number in the hun- strated improved potency and kinase selectivity relative to dreds. One of the first pharmacological AMPK activators dis- compound C. A crystal structure of the AMPK ␣2 kinase covered was 5-aminoimidazole-4-carboxamide ribonucleoside domain/SBI-0206965 complex revealed a binding pocket that (AICAR) (12). AICAR is taken up by cells via the adenosine partially overlapped the ATP site; however, SBI-0206965 at UNIVERSITY OF DUNDEE on June 8, 2018 transport system (13) and converted to the monophosphorylat- displayed mixed-competitive kinetics. SBI-0206965 inhibited ed derivative ZMP, which functions as an AMP-mimetic. Other AMPK signaling in a variety of cell types. Our study also high- AMPK agonists can be broadly classified as direct activators lights limitations pertaining to SBI-0206965 when used in con- (those that bind to drug sites located either between the kinase junction with AICAR. domain and ␤-CBM (e.g. A-769662, salicylate, 991, and PF-937 ␥ (14–17)) or within the subunit (e.g. C2 (18, 19)) or as indirect Results activators (those that commonly induce energy imbalance SBI-0206965: A potent, small-molecule AMPK inhibitor through mitochondrial toxicity, including metformin, xenobi- otics, and other natural products). Using an active site competitive screen, Egan et al. (34) Pharmacological AMPK inhibition provides potential strat- recently reported the 2-aminopyrimidine derivative SBI- egies to treat obesity (appetite suppression), type 2 diabetes 0206965 (Fig. 1A) was a potent and highly selective inhibitor of (enhanced insulin secretion), and stroke (neuroprotection) the autophagy initiator ULK1. ␣1- and ␣2-AMPK complexes (11), and AMPK hyperactivity has also been linked to pathogen- were also reported as hits in this screen. Subsequent studies in esis of neurodegeneration (20). The role of AMPK in cancer is MEFs showed that SBI-0206965 (50 ␮M) produced a disconnect complex; whereas initial studies demonstrated a tumor-sup- between AMPK-pThr-172 (elevated) and pACC (unchanged), pressive role, AMPK signaling also contributes to the metabolic indicative of direct inhibition of AMPK activity (34). Using cell- adaptations associated with tumor growth (e.g. increased glyco- free phosphorylation of the AMPK and ULK1 synthetic peptide lytic flux (the Warburg effect) and maintenance of ATP and substrate S108tide as a measure of kinase activity (8), we found NADPH) and promotes anchorage-independent proliferation that SBI-0206965 (1 ␮M) was a more effective inhibitor of (21–23). AMPK also promotes autophagic processes, via phos- AMPK ␣1␤1␥1 (80% inhibition) compared with ULK1 (63% phorylation of ULK1, to maintain homeostasis in the neoplastic inhibition) at 200 ␮M ATP (Fig. 1B). Conversely, compound C cell (24). Thus, AMPK is considered pro-tumorigenic under (Fig. 1A) was more effective as an inhibitor of ULK1 (36% inhi- certain circumstances, underpinning the attraction of AMPK bition) than AMPK (19% inhibition) at 200 ␮M ATP (Fig. 1B). inhibition as a strategy for cancer treatments. Under our assay conditions, SBI-0206965 inhibited both ␣ ␤ ␥ ␣ ␤ ␥ Current availability of small-molecule AMPK inhibitors, 1 1 1 and 2 1 1 with nanomolar IC50 (Fig. 1C), whereas either for clinical application or as research tools to delineate compound C inhibited both ULK1 and ␣1␤1␥1 with micromo-

AMPK’s physiological roles, is extremely limited. By far the lar IC50 (Fig. 1D). Using purified enzymes, we were unable to most widely applied AMPK inhibitor, the pyrazolopyrimidine detect an effect of SBI-0206965 on the rate of LKB1-mediated derivative compound C (dorsomorphin), was originally selected phosphorylation of Thr-172 (Fig. 1E). from a high-throughput screen and used to confirm AMPK-de- To assess kinase selectivity in terms of activity inhibition, we pendent effects of AICAR and metformin in cultured hepato- profiled SBI-0206965 (0.25 ␮M) against a diverse panel of 50

J. Biol. Chem. (2018) 293(23) 8874–8885 8875 High-potency AMPK inhibitor Downloaded from http://www.jbc.org/ at UNIVERSITY OF DUNDEE on June 8, 2018

8876 J. Biol. Chem. (2018) 293(23) 8874–8885 High-potency AMPK inhibitor

elevation in [ATP] (from 20 to 2000 ␮M), whereas compound C displayed conventional ATP-competitive inhibitor kinetics, Ϯ Ϯ ␮ with IC50 increasing 8-fold from 1.91 0.04 to 15.36 0.06 M across a 10-fold elevation in [ATP] (from 20 to 200 ␮M)(Fig. 3A). Whereas ATP (2000 ␮M) and compound C (20 ␮M) blocked binding of Thr-172–phosphorylated, AMPK ␣2 kinase domain to ATP-agarose, SBI-0206965 (Յ20 ␮M) was ineffective (Fig. 3B). However, SBI-0206965 (1 ␮M) inhibited the activity of phosphorylated ␣2 kinase domain to a similar extent as the AMPK heterotrimer at 200 ␮M ATP (Fig. 3C), confirming that the kinase domain alone encapsulates the SBI-0206965 binding site. We next tested the effect of varying ATP concentration on the activity of phosphorylated ␣2 kinase domain across a range of [SBI-0206965]. SBI-0206965 displayed kinetic properties

indicative of mixed-type inhibitor modality, as the Ks for ATP increased and Vmax decreased with increasing [SBI-0206965] (Fig. 3D and Fig. S2, A and B). A secondary plot displaying the Downloaded from

relationship between Km/Vmax (slopes on the Lineweaver–Burk plot (Fig. 3D)) and [SBI-0206965] generated a Ki of 261.4 nM ␣ (Fig. S2C). This is in contrast to the apparent Ki (also com- Ј monly termed Ki ) of 892.3 nM (Fig. S2D), which indicates that SBI-0206965 has a lower affinity for AMPK when ATP is http://www.jbc.org/ bound. Equally, ATP has a lower affinity for AMPK when SBI- ϭ ␮ ␣ ϭ 0206965 is bound (Ks(ATP) 82.1 M (Fig. 3D) versus Ks(ATP) ␮ ␣ ϭ ␣ 280.5 M (derived from the equation, Ki/Ki Ks/Ks)). In the context of AMPK regulation, these kinetic properties of SBI-

0206965 are consistent with mixed-type inhibition. at UNIVERSITY OF DUNDEE on June 8, 2018 To confirm the binding mode of SBI-0206965, we solved a 2.9 Å resolution crystal structure of the inhibitor complexed to ␣2 Figure 2. Kinase selectivity profile for SBI-0206965 (0.25 ␮M) and compound C (2.5 ␮M). Profiling (n ϭ 2 per kinase, with or without inhibitor) was kinase domain (residues 6–278), in which the activation loop performed using the MRC-PPU Express Screen service. MKK1 (activated 180% Thr-172 was mutated to the phosphomimetic Asp (Table S1). A by SBI-0206965) is excluded from the left-hand panel for clarity. similar crystallization construct was used previously to visual- kinases (Fig. S1A). Compound C (2.5 ␮M) was screened in par- ize the compound C–binding pocket (26) (PDB entry 3AQV) allel for direct comparison (Fig. S1B). AMPK in this panel was and adopted a structure essentially identical to that of the Thr- assayed at 20 ␮M ATP, hence the need to profile inhibitors 172–phosphorylated WT kinase domain in the heterotrimeric complex (16, 19). SBI-0206965 was found to occupy a pocket below the IC50 concentrations determined in our assays at 200 ␮M ATP (Fig. 1, C and D). Of the kinases in the panel, AMPK located between the kinase N- and C-lobes and hinge region, was inhibited to the greatest extent by both SBI-0206965 (22 Ϯ which partially overlaps the compound C–binding site (Fig. 4A 2% residual activity) and compound C (4 Ϯ 1%); however, SBI- and Fig. S3A). Specifically, the trimethoxyphenyl and pyrimidi- 0206965 displayed a preferable selectivity profile, inhibiting nyl rings of SBI-0206965 lie approximately within the same activity of only five other kinases by Ͼ50% (compared with 23 plane as, and overlap substantially with, the phenyl and pyra- kinases with compound C) and seven other kinases by Ͼ30% zolo[1,5-a]pyrimidinyl rings, respectively, of the compound C (compared with 31 kinases with compound C) (Fig. 2 and Fig. core. The positioning of the electronegative bromine atom of S1). SBI-0206965 inhibited LKB1 and CaMKK2 activities by 4 SBI-0206965, probably contributing to increased potency rela- and 17%, respectively (Fig. 2). tive to compound C, was confirmed by anomalous scattering at 13.6 keV (Fig. 4B) that revealed a large single peak for bromine SBI-0206965 is a mixed-type AMPK inhibitor (Table S1). The bromine moiety was found to occupy a large We investigated the inhibitory mechanism of SBI-0206965 cavity bordered by residues Val-30, Lys-45, Ile-77, Met-93, Ala- by performing dose-response assays at different ATP concen- 156, and Asn-162. ␣ ␤ ␥ trations using 1 1 1 AMPK. SBI-0206965 IC50 increased We could not unambiguously place the N-methylbenzamide Ͻ4-fold from 0.16 Ϯ 0.06 to 0.59 Ϯ 0.09 ␮M with 100-fold group of SBI-0206965 within the electron density; therefore, we

Figure 1. Biochemical characterization of SBI-0206965. A, structures of SBI-0206965 and compound C. B, inhibition of AMPK ␣1␤1␥1- and ULK1-phosphorylation of S108tide by SBI-0206965 and compound C. Error bars, mean percentage kinase activity versus untreated Ϯ S.E. (error bars)(n ϭ 3). **, p Ͻ 0.01; ***, p Ͻ 0.001; ****, p Ͻ 0.0001 by one-way ANOVA with post hoc Dunnett’s multiple-comparison test versus untreated kinase. ####, p Ͻ 0.0001 by one-way ANOVA with post hoc Dunnett’s multiple-comparison test versus compound C–treated AMPK. Shown are dose-response curves for SBI-0206965 inhibition of ULK1, ␣ ␤ ␥ ␣ ␤ ␥ ␣ ␤ ␥ ␮ AMPK 1 1 1, and AMPK 2 1 1(C) and compound C inhibition of ULK1 and AMPK 1 1 1(D). For C and D, assays were performed at fixed 200 M ATP. IC50 values (␮M) Ϯ S.E. were calculated from triplicate experiments. E, Thr-172 phosphorylation assay. Bacterially expressed AMPK ␣2␤1␥1 was incubated with LKB1 in the presence of SBI-0206965, and pThr-172 was measured by immunoblotting. Error bars, mean pThr-172/␣ (arbitrary units) Ϯ S.E. (n ϭ 3).

J. Biol. Chem. (2018) 293(23) 8874–8885 8877 High-potency AMPK inhibitor Downloaded from http://www.jbc.org/ at UNIVERSITY OF DUNDEE on June 8, 2018

Figure 3. SBI-0206965 is an AMPK type II inhibitor. A, inhibition of AMPK ␣1␤1␥1 phosphorylation of S108tide by SBI-0206965 and compound C at varying ATP concentrations (20, 200, and 2000 ␮M for SBI-0206965; 20 and 200 ␮M for compound C). Error bars, mean percentage kinase activity versus untreated Ϯ S.E. ϭ (error bars)(n 3). The table shows calculated IC50 values for each combination of inhibitor/ATP concentration and -fold increase in IC50 with increasing ATP concentration. B, ATP-agarose competitive binding assay. Thr-172–phosphorylated AMPK ␣2 kinase domain was incubated with ATP-agarose in the presence of SBI-0206965, ATP, or compound C (CC), as indicated. Immobilized AMPK was measured by immunoblot for ␣-subunit. Representative immunoblots of two individual experiments are shown. C, inhibition of ␣2 kinase domain phosphorylation of S108tide by SBI-0206965 (1 ␮M). Error bars, mean percentage kinase activity versus untreated Ϯ S.E. (n ϭ 3). ***, p Ͻ 0.001 by unpaired two-tailed Student’s t test versus untreated kinase domain. D, Lineweaver–Burk (double reciprocal) plot showing inhibition of ␣2 kinase domain at four fixed SBI-0206965 concentrations (0–400 nM as indicated). Assays were performed in the ␮ presence of varying ATP concentrations (25–200 M). The table shows apparent Km and Vmax values across SBI-0206965 dose range, calculated from data presented in Fig. S2A.

modeled two distinct conformations with equal occupancy keeper residue Met-93), and C-lobe (Ile-77, Gly-99, Glu-100, (Fig. 4C). In one position (conformation A), the aromatic ring is Leu-146, and Ala-156) residues. rotated almost 80° with respect to the compound plane, and the Comparisons with either apo- (PDB entry 2YZA), compound N-methyl group extends into the space occupied by the cata- C-complexed (26), or activated (PDB entry 4ZHX) (19) kinase lytic loop, Mg2ϩ-chelating residue Asn-144, and the ATP domain structures reveal the ␣2/SBI-0206965 complex con- ␣-phosphate in structures of active kinases (35). In the alternate tains many of the hallmarks of an unproductive kinase. In both position (conformation B), the benzamide ring is rotated 180° compound C- and SBI-0206965-bound structures, the C-␣-he- relative to conformation A, and the N-methyl group is directed lix adopts a “swung-out” position, ensuring that the glutamate– toward P-loop residues Leu-22 and Gly-23 (Fig. 4C). In both lysine salt bridge, required for efficient phosphoryl transfer, is conformations, SBI-0206965 makes two electrostatic contacts unformed (Fig. S3A). These residues (Lys-45 and Glu-64) are with the main chain of Val-96 in the kinase hinge region; oth- instead interdigitated by the activation loop residue Ser-161, erwise, drug binding is mainly stabilized by hydrophobic con- hydrogen bonding with Glu-64. Glu-64 forms a further hydro- tacts with hinge (Tyr-95), N-lobe (Leu-22, Ala-43, and the gate- gen bond with the backbone of Leu-160. Compared with the

8878 J. Biol. Chem. (2018) 293(23) 8874–8885 High-potency AMPK inhibitor Downloaded from http://www.jbc.org/ at UNIVERSITY OF DUNDEE on June 8, 2018

Figure 4. A, crystal structure of ␣2 kinase domain (residues 6–278) (T174D) complexed to SBI-0206965 in conformation A. Pink, N-lobe; blue, C-lobe. B, anomalous difference map at 3.0 Å of SBI-0206965 bromine contoured to 5.0 ␴ collected at the bromine edge (13.6 keV). The anomalous bromine peak (white mesh) strongly supports the placement of SBI-0206965 within the electron density (blue mesh). C, left, overlapping SBI-0206965 (green, conformation A) and compound C (cyan) binding sites. Surface represents SBI-0206965–bound ␣2. Gatekeeper residue ␣2-Met-93 and catalytic loop residue Asn-144 are shown (green, SBI-0206965 complex; cyan, compound C complex). Right, SBI-0206965 in conformation B. D, comparison of HRD motifs in the active ␣2 kinase conformation (PDB 4ZHX) (left) and ␣2 kinase domain complexed to SBI-0206965 (right).

compound C-bound structure, the DFG motif (Asp157-Phe- conformation; Asp-157 is sterically hindered from adopting the Gly159) in the activation loop of our SBI-0206965-bound struc- “DFG-in” conformation by the compound N-methylbenzamide ture is displaced away from the ATP active site by a maximum and is oriented toward the N-lobe (Fig. S3B), with Phe-158 of 2.9 Å (Phe-158 ␣C), presumably as a consequence of the vacating the regulatory R-spine to displace His-137 in the nonplanar structure of SBI-0206965 relative to compound C. kinase HRD motif (Fig. 4D). A consequence of this displace- However, in common with the ␣2/compound C structure, this ment is rearrangement of the HRD backbone and positioning of motif adopts a unique conformation that is neither “DFG-in,” in the Arg-138 side chain to prevent coordination with the phos- which Asp-157 is appropriately positioned to chelate Mg2ϩ phate group of pThr-172. ions required for coordination of ATP phosphate groups, nor classical “DFG-out,” in which the motif flips by 180° such that SBI-0206965 inhibits cellular AMPK signaling the Phe side chain now occludes the ATP pocket and Asp-157 is We explored the efficacy of SBI-0206965 to inhibit AMPK no longer able to coordinate Mg-ATP for catalysis (Fig. S3B). signaling in a range of cell lines. In HEK293 cells, glucose star- Instead, the inhibited ␣2 DFG motif adopts an intermediate vation triggered a 1.9-fold increase in pThr-172, indicative of

J. Biol. Chem. (2018) 293(23) 8874–8885 8879 High-potency AMPK inhibitor Downloaded from http://www.jbc.org/ at UNIVERSITY OF DUNDEE on June 8, 2018 Figure 5. SBI-0206965 suppresses AMPK signaling in diverse cultured cell lines. Shown are representative immunoblots for pACC and pThr-172 from prepared lysates. HEK293 cells were incubated with glucose-free medium (A) and SBI-0206965 for4horAICAR and SBI-0206965 for1h(B), as indicated. Error bars, mean -fold change in phosphorylation Ϯ S.E. (error bars)(n ϭ 3). *, p Ͻ 0.05; **, p Ͻ 0.01; ***, p Ͻ 0.001 by one-way ANOVA with post hoc Dunnett’s multiple-comparison test versus glucose-starved (A) or AICAR-treated (B) cells. C, SH-SY5Y neuronal cells were co-incubated with CaMKK2 activator ionomycin and SBI-0206965 for 1 h, as indicated. Error bars, mean -fold change in phosphorylation Ϯ S.E. (n ϭ 3). *, p Ͻ 0.05 by one-way ANOVA with post hoc Dunnett’s multiple-comparison test versus ionomycin-treated cells.

energy stress, and this was accompanied by a 3.8-fold increase MEFs (evidenced as ACC phosphorylation) was unchanged fol- in pACC (Fig. 5A). The addition of SBI-0206965 to glucose-free lowing incubation with 50 ␮M SBI-0206965, despite increased medium significantly suppressed pACC at concentrations of Thr-172 phosphorylation, which may have arisen from fluctu- Ն5 ␮M, probably due to direct AMPK inhibition because SBI- ations in adenine nucleotide ratios at high concentrations (Fig. 0206965 did not induce reductions in pThr-172, reflecting sup- S4B). SBI-0206965 has subsequently been shown to suppress pressed LKB1 activity (Fig. 5A). SBI-0206965 concentrations of non-small-cell lung cancer cell growth (36). We now show that Ն ␮ 5 M also blocked AMPK signaling in HEK293 cells treated SBI-0206965 can be repurposed as a potent and kinase-selec- with AICAR (Fig. 5B); however, this was probably due to inhi- tive inhibitor of the metabolic coordinator AMPK, providing a bition of AICAR cellular uptake, reminiscent of compound C, useful alternative to the almost singularly used, yet promiscu- Ն ␮ because SBI-0206965 5 M also resulted in significantly ous, compound C. Because AMPK is a major positive regulator reduced intracellular ZMP accumulation (Fig. S4A). SBI- of autophagy through direct phosphorylation of ULK1 (37) and 0206965 Յ 30 ␮M did not significantly affect ratios of adenine other autophagy-associated proteins (e.g. VSP34 and Beclin-1), nucleotides in HEK293 cells, indicative of mitochondrial toxic- simultaneous inhibition of both AMPK and ULK1 signaling ity; however, we note a trend toward reduced adenylate energy supports the application of SBI-0206965 as a highly effective charge at 30 ␮M (Fig. S4B). 5 ␮M SBI-0206965 also significantly suppressor of prosurvival autophagic responses in tumor cells. We suppressed pACC in the neuronal cell line SH-SY5Y, in which CaMKK2-mediated AMPK signaling was triggered by the addi- previously used SBI-0206965 as an agent to confirm ULK1-medi- ␤ tion of the calcium ionophore ionomycin (Fig. 5C). ated phosphorylation of the drug-sensitizing AMPK 1-subunit residue Ser-108, until recently regarded as an autophosphoryla- Discussion tion site (8). Our demonstration now that SBI-0206965 also inhib- SBI-0206965 was originally identified as an ATP-competitive its AMPK does not detract from our previous conclusions, because inhibitor of the autophagy initiator kinase ULK1, with the abil- Ser-108 autophosphorylation is a cis event (10), and in our study, ity to suppress cellular ULK signaling and ULK1-mediated sur- kinase-inactive AMPK enzyme was used as the cellular substrate. vival of lung cancer and glioblastoma cells when coupled with Our biochemical characterization of SBI-0206965 as an nutrient stress (34). In this initial report, AMPK signaling in AMPK antagonist highlights several limitations of use that

8880 J. Biol. Chem. (2018) 293(23) 8874–8885 High-potency AMPK inhibitor

demand consideration. Both Egan et al. (34) and our study here from of mixed inhibition) or uncompetitive kinetic properties indicate a highly selective kinase profile for SBI-0206965; how- (47). Active site binding by mixed inhibitors has also been ever, it must be noted that our combined, activity-based analy- inferred by direct methods (51, 52) or from close structural ses cover only 57 kinases besides AMPK ␣1 and ␣2, represent- relationship to the substrate (53, 54). Type II kinase inhibitors ing just 11% of the human kinome. In light of our data such as SBI-0206965 probably stabilize the “DFG-out” confor- implicating AMPK as being able to accommodate mixed-type mation, an arrangement less favorable for ATP binding (55). inhibitors, we consider that activity, rather than ATP-compet- Thus, the binding sites occupied by SBI-0206965 or ATP rep- itive binding, is a more definitive metric of inhibitor selectivity. resent distinct conformational arrangements, each reducing Besides ULK1 and -2, JAK3 and Src are flagged in both analyses the binding affinity of the competing ligand. as possible kinase targets for SBI-0206965, and we cannot SBI-0206965 increased Km(ATP), indicating a preference for exclude the possibility that others exist. We strongly advise binding to the free enzyme rather than the enzyme-substrate against using SBI-0206965 in combination with AICAR when complex. This is consistent with the location of the inhibitor- investigating cellular AMPK signaling. We also note that cellu- binding site resolved in our structure and our observation that lar incubations at high concentrations may induce Thr-172 SBI-0206965 cannot displace AMPK from ATP-agarose (Fig. phosphorylation through fluctuations in AMP/ATP and ADP/ 3C). This indicates that SBI-0206965 can only bind to an ATP ratios, rather than via stimulation of LKB1-mediated phos- actively cycling AMPK, as the binding site must first become phorylation as with SU6656, which has the potential to con- available following release of the ADP product from the active Downloaded from found interpretation of results. site. Furthermore, for the inhibitor-bound AMPK complexes, Structural characteristics used to classify small-molecule [ATP] in the activity assay may be below Ks ATP, precluding an kinase inhibitors label SBI-0206965 as a type IIb AMPK inhib- increase of IC at the same rate as increase of [ATP], as dic- itor (38, 39). Foremost, the SBI-0206965–binding pocket over- 50 tated by the Cheng–Prusoff equation (56). This probably also laps the ATP site in the ␣2 kinase domain, with the DFG motif http://www.jbc.org/ explains why SBI-0206965 is very effective at inhibiting AMPK positioned in a nonclassical “out” conformation. The Asp-157 in cells; type II inhibitors typically display high cellular potency, side chain is not “in” (i.e. unable to coordinate phosphate-sta- ϩ whereas millimolar [ATP] often prevents type I inhibitors (e.g. bilizing Mg2 ions), and Asn-144/Phe-158 (7.7 Å) and Glu-64/ compound C) from maintaining the potency observed in cell- Phe-158 (10.1 Å) C␣ atomic distances are consistent with a free activity assays (43). “DFG-out” classification (40). Additionally, the SBI-0206965 In summary, we show here that SBI-0206965 displays prefer- at UNIVERSITY OF DUNDEE on June 8, 2018 molecule does not extend into the back cleft of the ␣2 kinase able characteristics, relative to compound C, as an AMPK domain, and the R-spine is distorted upon binding. Type IIb inhibitors bind to sites that incompletely overlap the ATP inhibitor in vitro. Further studies are required to reveal its effi- active site and are usually regarded as ATP-competitive. How- cacy in vivo; however, biochemical and structural data provided ever, members of this inhibitor class have also been described as in this study suggest that SBI-0206965 is a promising lead for “ATP noncompetitive” (41, 42) and “indirectly ATP-competi- the development of a new class of AMPK inhibitors with ther- tive” (43), perhaps because comprehensive enzyme kinetic pro- apeutic potential. filing is rarely reported alongside structural studies. We have confirmed that SBI-0206965 partially shares its Experimental procedures binding site on AMPK kinase domain with ATP but under our Reagents assay conditions is not ATP-competitive, instead displaying a Antibodies for pan-AMPK ␣ (catalog no. 2793, 1:1,000 dilu- mixed-type inhibition profile (Fig. 3D). Mixed-type inhibition, tion), and phosphospecific antibodies for ␣-pThr-172 (catalog by definition, is a conceptual mixture of competitive (increased no. 2535, 1:1,000 dilution) and pACC (catalog no. 3661, 1:1,000 Km substrate) and uncompetitive inhibition (reduced Vmax), in dilution) were from Cell Signaling Technology. Streptavidin- which binding of the inhibitor affects substrate binding, and IRDye 680RD (1:20,000 dilution) and anti-rabbit IgG IRDye 680 vice versa (44). From a structural perspective, mixed-type inhi- and anti-mouse IgG IRDye 800 fluorescent-labeled secondary bition would appear incompatible with overlapping inhibitor/ antibodies (1:10,000 dilution) were from LI-COR Biosciences. substrate binding sites. However, numerous studies have GSH-Sepharose 4B (catalog no. 17075601) and streptavidin- reported mixed-type inhibition by agents that have either been Sepharose were from GE Life Sciences. Recombinant LKB1/ demonstrated to bind at the enzyme’s active site or most likely MO25/STRAD was from Sigma (catalog no. SRP0246). SBI- do (i.e. transition state inhibitor analogues) (45–50). Specifi- 0206965 was from ApexBio. ATP-agarose was from Innova cally, a type II, active site binding inhibitor of insulin-like Biosciences. cOmplete protease inhibitor mixture was from growth factor 1 receptor tyrosine kinase activity has been Roche Applied Science. DNA oligonucleotides and all other reported to display mixed-type kinetics with respect to ATP reagents were from Sigma. (45), whereas the tyrphostin inhibitor AG1296 displayed either competitive or mixed-type characteristics, depending on the activation state of its target kinase in platelet-derived growth Kinase inhibitor profiling factor receptor (46). Additionally, crystal structures of adeny- Inhibitor profiling (Express screen) was performed by the late cyclase complexed to P-site inhibitors (adenine nucleo- International Centre for Kinase Profiling, Medical Research sides/adenine nucleoside 3Ј-phosphates) show active site bind- Council Protein Phosphorylation and Ubiquitylation Unit ing; these inhibitors display either noncompetitive (a special (MRC-PPU), University of Dundee, UK.

J. Biol. Chem. (2018) 293(23) 8874–8885 8881 High-potency AMPK inhibitor

Expression constructs Buster (https://www.globalphasing.com/buster/)6 (61), respectively. Data for the bromine anomalous map was collected on Assays—pET28b plasmid for expression of N-terminal His6 tag ␣2 kinase domain (residues 6–279) fusion protein was a the MX2 beamline at the Australian Synchrotron near the bro- generous gift from Dr. Dean Littler (Monash University). pET mine edge at 13.6 KeV, calculated on the X-ray anomalous DUET constructs for Escherichia coli expression of AMPK het- scattering website (http://skuld.bmsc.washington.edu/scatter/ 6 erotrimer ␣2␤1␥1 have been described (10, 19). AS_form.html). An anomalous map was generated by Phe- Crystallization—pET21b plasmid for expression of ␣2 kinase nix.refine (62). SBI-0206965 molecular coordinates and restraints domain (residues 6–278) incorporating T172D mutation, with were generated using the PRODRG web server (63). Structural validation was performed using Molprobity (64). Omit maps a PreScission protease cleavable N-terminal His6 tag, was syn- thesized by General Biosystems. Constructs were sequence- were generated using Buster, and figures were created using verified, and expressed constructs were mass-verified by TOF PyMOL. MS. S108tide activity assay Protein production AMPK and ULK1 activities were determined by phosphory- ␤ ␮ Heterotrimeric human AMPK GST-␣1␤1␥1 and GST-␣2␤1␥1, lation of the AMPK 1 Ser-108 peptide (S108tide) using 100 M ␥ 32 ␮ S108tide, 5 mM MgCl2, and [ - P]ATP in a 25- l reaction vol-

and human FLAG-ULK1 were produced in mammalian COS7 Downloaded from cells as described (8). AMPK and ULK1 were purified using ume at 30 °C. Reactions were terminated after 10 min by spot- ␮ GSH-Sepharose and anti-FLAG M2 affinity gel, respectively. ting 15 l onto P81 phosphocellulose paper (Whatman) and FLAG-CaMKK2 was produced in Sf21 insect cells as described washing in 1% phosphoric acid. Radioactivity was quantified by (6)andpurifiedusinganti-FLAGM2affinitygel.Thr-172–phos- scintillation counting. phorylated AMPK His-␣2(6–279) kinase domain was pro- http://www.jbc.org/ duced in bacteria as described (10). Briefly, the pET28b-␣2(6– LKB1 phosphorylation assays 279) construct was transformed into E. coli strain Rosetta (DE3) 200 ng of unphosphorylated His-␣2␤1␥1 was incubated with and expressed by isopropyl 1-thio-␤-D-galactopyranoside 20 ng of LKB1/STRAD␣/MO25 in the presence of buffer A ␮ induction. Cells were ruptured using a precooled Emulsi- supplemented with 2 mM MgCl2, 200 M ATP, and compounds Flex-C5 homogenizer (Avestin) and protein immobilized on (final 1% DMSO) or 1% DMSO vehicle, as indicated, at 32 °C. nickel-Sepharose before phosphorylation with CaMKK2. Pro- Reactions were terminated by the addition of SDS sample at UNIVERSITY OF DUNDEE on June 8, 2018 tein was eluted using 400 mM imidazole and buffer-exchanged buffer and immunoblotted for pThr-172. with buffer A (50 mM Tris⅐HCl, pH 7.4, 150 mM NaCl, 10% glycerol, 0.02% Tween 20) using a PD10 desalt column. A sim- Immunoblotting ilar procedure was used to produce unphosphorylated AMPK Samples were separated by SDS-PAGE on a 12% gel (7% for ␣ ␤ ␥ 2 1 1, except the CaMKK2 incubation step was omitted. ACC), previously enriched using streptavidin-Sepharose (10), and transferred to Immobilon-FL polyvinylidene difluoride Protein crystallization membrane (EMD Millipore). Membrane was blocked in PBS ϩ His-␣2(6–278) (T174D) was generated as described above 0.1% Tween 20 (PBST) with 2% nonfat milk for 30 min at 22 °C for His-␣2(6–279), except the CaMKK2 incubation step was and then incubated for either 1 h or overnight with primary omitted. His tag was removed with GST-tagged PreScission antibodies (dilutions in PBST). After washes with PBST, mem- protease treatment (overnight, 4 °C) after the buffer exchange branes were incubated with anti-rabbit or anti-mouse IgG sec- step. PreScission protease was removed using GSH-Sepharose ondary antibodies, fluorescently labeled with IR680 or IR800 and ␣2(6–279) (T174D) repurified by SEC. Concentrated dye, for 1 h. Immunoreactive bands were visualized on an ␣2(6–279) (T174D) (4 mg/ml) was incubated with 0.5 mM SBI- Odyssey௡ IR imaging system with densitometry analyses deter- 0206965 on ice for 30 min and centrifuged at 10,000 rpm for 3 mined using ImageStudioLite software (LI-COR Biosciences). min before setting crystallization experiments. Protein was mixed equally 1:1 at room temperature with a reservoir solution ATP-agarose immobilization

containing 9–14% ethanol, 5 mM MgCl2,10mM tris(2-carboxy- 20 ␮g of Thr-172–phosphorylated His-␣2(6–279) kinase ethyl)phosphine, and 3–7 mM MnCl2. Diffraction quality crys- domain was incubated with ATP/compounds for 30 min at tals were obtained through streak seeding with a cat’s whisker. room temperature, before the addition of 10 ␮l of ATP-agarose Crystals appeared after 2–3 days and reached full size after 1–2 suspended in 200 ␮l of buffer B (50 mM HEPES, pH 7.4, 100 mM weeks. Crystals were then incubated in a cryoprotectant-con- NaCl, 10% glycerol, 5 mM MgCl2). Beads were incubated on a taining reservoir solution with an addition of 25% ethylene gly- rotating wheel at 4 °C for 2 h, before washing three times with col. Data were collected on both MX1 and MX2 beamlines at 500 ␮l of buffer A supplemented with 0.1% Tween 20. Beads the Australian Synchrotron (Melbourne, Australia). Data were were resuspended in 15 ␮l of SDS sample buffer, boiled, and processed and integrated using XDS (57) and scaled using AIM- immunoblotted for AMPK ␣. LESS from the CCP4 suite (58). The structure was solved by molecular replacement using Phaser from the CCP4 suite (59) and 3AQV as the search model. Iterative rounds of model 6 Please note that the JBC is not responsible for the long-term archiving and building and refinement were performed using Coot (60) and maintenance of this site or any other third party hosted site.

8882 J. Biol. Chem. (2018) 293(23) 8874–8885 High-potency AMPK inhibitor

Cell culture and incubations References HEK293 and SH-SY5Y mammalian cells were cultured in 1. Steinberg, G. R., and Kemp, B. E. (2009) AMPK in health and disease. Dulbecco’s modified Eagle’s medium (DMEM) containing 10% Physiol. Rev. 89, 1025–1078 CrossRef Medline 2. Hardie, D. G., Schaffer, B. E., and Brunet, A. (2016) AMPK: an energy- fetal bovine serum and 1% penicillin/streptomycin antibiotics, sensing pathway with multiple inputs and outputs. Trends Cell Biol. 26, at 37 °C with 5% CO2. Cells were incubated with fresh DMEM 190–201 CrossRef Medline for 3 h, before simultaneous incubation with glucose-free 3. Suzuki, A., Okamoto, S., Lee, S., Saito, K., Shiuchi, T., and Minokoshi, Y. DMEM and or/compounds as indicated. Glucose-starved cells (2007) Leptin stimulates fatty acid oxidation and peroxisome proliferator- were washed 4 h post-treatment with ice-cold PBS and har- activated receptor ␣ gene expression in mouse C2C12 myoblasts by changing the subcellular localization of the ␣2 form of AMP-activated vested by rapid lysis using ice-cold lysis buffer (50 mM Tris⅐HCl, protein kinase. Mol. Cell. Biol. 27, 4317–4327 CrossRef Medline pH 7.4, 150 mM NaCl, 50 mM NaF, 1 mM sodium pyrophos- 4. Lamia, K. A., Sachdeva, U. M., DiTacchio, L., Williams, E. C., Alvarez, J. G., phate, 1 mM EDTA, 1 mM EGTA, 1 mM DTT, 1% Triton X-100, Egan, D. F., Vasquez, D. S., Juguilon, H., Panda, S., Shaw, R. J., Thompson, and cOmplete protease inhibitor mixture). AICAR/ionomycin- C. B., and Evans, R. M. (2009) AMPK regulates the circadian clock by treated cells were harvested similarly, except this was per- cryptochrome phosphorylation and degradation. Science 326, 437–440 formed 1 h post-treatment. Lysates were clarified by centrifu- CrossRef Medline 5. Zhang, C. S., Jiang, B., Li, M., Zhu, M., Peng, Y., Zhang, Y. L., Wu, Y. Q., Li, gation (14,000 rpm; 3 min; 4 °C) and flash-frozen in liquid N2 until processed. T. Y., Liang, Y., Lu, Z., Lian, G., Liu, Q., Guo, H., Yin, Z., Ye, Z., Han, J., Wu, J. W., Yin, H., Lin, S. Y., and Lin, S. C. (2014) The lysosomal v-ATPase- Downloaded from Ragulator complex is a common activator for AMPK and mTORC1, act- Nucleotide measurements ing as a switch between catabolism and anabolism. Cell Metab. 20, Adenine nucleotides and adenylate energy charge were 526–540 CrossRef Medline measured by LC-MS from HEK293 perchlorate extracts, as 6. Oakhill, J. S., Chen, Z. P., Scott, J. W., Steel, R., Castelli, L. A., Ling, N., Macaulay, S. L., and Kemp, B. E. (2010) ␤-Subunit myristoylation is the

described previously (32). http://www.jbc.org/ gatekeeper for initiating metabolic stress sensing by AMP-activated protein kinase (AMPK). Proc. Natl. Acad. Sci. U.S.A. 107, 19237–19241 ZMP measurements CrossRef Medline All ZMP measurements were acquired using LC-MS from 7. Liang, J., Xu, Z. X., Ding, Z., Lu, Y., Yu, Q., Werle, K. D., Zhou, G., Park, Y. Y., Peng, G., Gambello, M. J., and Mills, G. B. (2015) Myristoylation perchlorate extracts, using a method similar to that described confers noncanonical AMPK functions in autophagy selectivity and mito- previously for nucleotides. Briefly, LC conditions were opti- chondrial surveillance. Nat. Commun. 6, 7926 CrossRef Medline at UNIVERSITY OF DUNDEE on June 8, 2018 mized for a 150-mm (length) and 0.5-mm (inner diameter) 8. Dite, T. A., Ling, N. X. Y., Scott, J. W., Hoque, A., Galic, S., Parker, B. L., Hypercarb column (3 ␮m; Thermo Fisher Scientific Australia Ngoei, K. R. W., Langendorf, C. G., O’Brien, M. T., Kundu, M., Viollet, B., Pty. Ltd.). The LC solvent system was (A) 25 mM triethylammo- Steinberg, G. R., Sakamoto, K., Kemp, B. E., and Oakhill, J. S. (2017) The nium bicarbonate buffer at pH 7.8 and (B) acetonitrile with autophagy initiator ULK1 sensitizes AMPK to allosteric drugs. Nat. Com- 0.1% TFA. ZMP was eluted at a flow rate of 500 ␮l/min in a mun. 8, 571 CrossRef Medline 9. Ross, F. A., Jensen, T. E., and Hardie, D. G. (2016) Differential regulation by gradient program consisting of 100% A (5 min), 0–25% B (10 AMP and ADP of AMPK complexes containing different ␥ subunit iso- min), 50–80% B (5 min), and 100% B (5 min). Data were ana- forms. Biochem. J. 473, 189–199 CrossRef Medline lyzed with Multiquant 2.0.2, utilizing the area under the LC 10. Scott, J. W., Ling, N., Issa, S. M., Dite, T. A., O’Brien, M. T., Chen, Z. P., chromatogram for the corresponding ZMP peak. Calibration Galic, S., Langendorf, C. G., Steinberg, G. R., Kemp, B. E., and Oakhill, J. S. curves were obtained by linear regression of the peak area ratio (2014) Small molecule drug A-769662 and AMP synergistically activate of a ZMP standard. The MS conditions and MRM values for naive AMPK independent of upstream kinase signaling. Chem. Biol. 21, 619–627 CrossRef Medline ZMP were optimized by separate infusion of 1 ␮g/ml solution 11. Viollet, B., Horman, S., Leclerc, J., Lantier, L., Foretz, M., Billaud, M., Giri, in 25 mM triethylammonium bicarbonate buffer at a flow rate of S., and Andreelli, F. (2010) AMPK inhibition in health and disease. Crit. 50 ␮l/min. Rev. Biochem. Mol. Biol. 45, 276–295 CrossRef Medline 12. Corton, J. M., Gillespie, J. G., Hawley, S. A., and Hardie, D. G. (1995) Statistical analysis 5-aminoimidazole-4-carboxamide ribonucleoside: a specific method for Ϯ activating AMP-activated protein kinase in intact cells? Eur. J. Biochem. The data are presented as mean values S.E. of at least three 229, 558–565 CrossRef Medline independent experiments unless stated. Statistical analyses 13. Fryer, L. G., Parbu-Patel, A., and Carling, D. (2002) Protein kinase inhibi- were performed using GraphPad Prism version 7.0c. Ordinary tors block the stimulation of the AMP-activated protein kinase by one-way analysis of variance (ANOVA) with Dunnett’s multi- 5-amino-4-imidazolecarboxamide riboside. FEBS Lett. 531, 189–192 ple-comparison test was used for all comparisons unless stated. CrossRef Medline 14. Cool, B., Zinker, B., Chiou, W., Kifle, L., Cao, N., Perham, M., Dickinson, R., Adler, A., Gagne, G., Iyengar, R., Zhao, G., Marsh, K., Kym, P., Jung, P., Author contributions—T. A. D., C. G. L., A. H., S. G., R. J. R., A. J. O., Camp, H. S., and Frevert, E. (2006) Identification and characterization of a L. M. L., K. R. W. N., and N. X. Y. L. data curation; T. A. D., A. H., small molecule AMPK activator that treats key components of type 2 S. G., R. J. R., A. J. O., L. M. L., K. R. N., N. X. Y. L., B. E. K., J. W. S., diabetes and the metabolic syndrome. Cell Metab. 3, 403–416 CrossRef and J. S. O. formal analysis; T. A. D., C. G. L., and A. H. methodol- Medline ogy; T. A. D., C. G. L., J. W. S., and J. S. O. writing-original draft; 15. Hawley, S. A., Fullerton, M. D., Ross, F. A., Schertzer, J. D., Chevtzoff, C., L. F., J. W. S., and J. S. O. supervision; L. F., B. E. K., and J. S. O. fund- Walker, K. J., Peggie, M. W., Zibrova, D., Green, K. A., Mustard, K. J., ing acquisition; B. E. K., J. W. S., and J. S. O. conceptualization; Kemp, B. E., Sakamoto, K., Steinberg, G. R., and Hardie, D. G. (2012) The J. S. O. project administration; J. S. O. writing-review and editing. ancient drug salicylate directly activates AMP-activated protein kinase. Science 336, 918–922 CrossRef Medline

J. Biol. Chem. (2018) 293(23) 8874–8885 8883 High-potency AMPK inhibitor

16. Xiao, B., Sanders, M. J., Carmena, D., Bright, N. J., Haire, L. F., Underwood, N. C., Steinberg, G. R., Kemp, B. E., and Oakhill, J. S. (2015) Inhibition of E., Patel, B. R., Heath, R. B., Walker, P. A., Hallen, S., Giordanetto, F., AMP-activated protein kinase at the allosteric drug-binding site promotes Martin, S. R., Carling, D., and Gamblin, S. J. (2013) Structural basis of islet insulin release (2015) Chem. Biol. 22, 705–711 CrossRef Medline AMPK regulation by small molecule activators. Nat. Commun. 4, 3017 33. Ross, F. A., Hawley, S. A., Auciello, F. R., Gowans, G. J., Atrih, A., Lamont, Medline D. J., and Hardie, D. G. (2017) Mechanisms of paradoxical activation of 17. Cokorinos, E. C., Delmore, J., Reyes, A. R., Albuquerque, B., Kjøbsted, R., AMPK by the kinase inhibitors SU6656 and Sorafenib. Cell Chem. Biol. 24, Jørgensen, N. O., Tran, J. L., Jatkar, A., Cialdea, K., Esquejo, R. M., Meissen, 813–824.e4 CrossRef Medline J., Calabrese, M. F., Cordes, J., Moccia, R., Tess, D., et al. (2017) Activation 34. Egan, D. F., Chun, M. G., Vamos, M., Zou, H., Rong, J., Miller, C. J., Lou, of skeletal muscle AMPK promotes glucose disposal and glucose lowering H. J., Raveendra-Panickar, D., Yang, C. C., Sheffler, D. J., Teriete, P., Asara, in non-human primates and mice. Cell Metab. 25, 1147–1159.e10 J. M., Turk, B. E., Cosford, N. D., and Shaw, R. J. (2015) Small molecule CrossRef Medline inhibition of the autophagy kinase ULK1 and identification of ULK1 sub- 18. Hunter, R. W., Foretz, M., Bultot, L., Fullerton, M. D., Deak, M., Ross, F. A., strates. Mol. Cell. 59, 285–297 CrossRef Medline Hawley, S. A., Shpiro, N., Viollet, B., Barron, D., Kemp, B. E., Steinberg, 35. Zheng, J., Trafny, E. A., Knighton, D. R., Xuong, N. H., Taylor, S. S., Ten G. R., Hardie, D. G., and Sakamoto, K. (2014) Mechanism of action of Eyck, L. F., and Sowadski, J. M. (1993) A refined crystal structure of the compound-13: an alpha1-selective small molecule activator of AMPK. catalytic subunit of cAMP-dependent protein kinase complexed with Chem. Biol. 21, 866–879 CrossRef Medline MnATP and a peptide inhibitor. Acta Crystallogr. D Biol. Crystallogr. 49, 19. Langendorf, C. G., Ngoei, K. R., Scott, J. W., Ling, N. X., Issa, S. M., Gor- 362–365 CrossRef Medline man, M. A., Parker, M. W., Sakamoto, K., Oakhill, J. S., and Kemp, B. E. 36. Tang, F., Hu, P., Yang, Z., Xue, C., Gong, J., Sun, S., Shi, L., Zhang, S., Li, Z., (2016) Structural basis of allosteric and synergistic activation of AMPK by Yang, C., Zhang, J., and Xie, C. (2017) SBI0206965, a novel inhibitor of Downloaded from furan-2-phosphonic derivative C2 binding. Nat. Commun. 7, 10912 Ulk1, suppresses non-small cell lung cancer cell growth by modulating CrossRef Medline both autophagy and apoptosis pathways. Oncol. Rep. 37, 3449–3458 20. Liu, Y. J., and Chern, Y. (2015) AMPK-mediated regulation of neuronal CrossRef Medline metabolism and function in brain diseases. J. Neurogenet. 29, 50–58 37. Egan, D. F., Shackelford, D. B., Mihaylova, M. M., Gelino, S., Kohnz, R. A., CrossRef Medline Mair, W., Vasquez, D. S., Joshi, A., Gwinn, D. M., Taylor, R., Asara, J. M., 21. Hardie, D. G., and Alessi, D. R. (2013) LKB1 and AMPK and the cancer- Fitzpatrick, J., Dillin, A., Viollet, B., Kundu, M., Hansen, M., and Shaw, R. J. http://www.jbc.org/ metabolism link: ten years after. BMC Biol. 11, 36 CrossRef Medline (2011) Phosphorylation of ULK1 (hATG1) by AMP-activated protein ki- 22. Popovics, P., Frigo, D. E., Schally, A. V., and Rick, F. G. (2015) Targeting nase connects energy sensing to mitophagy. Science 331, 456–461 the 5Ј-AMP-activated protein kinase and related metabolic pathways for CrossRef Medline the treatment of prostate cancer. Expert Opin. Ther. Targets 19, 617–632 38. Dar, A. C., and Shokat, K. M. (2011) The evolution of protein kinase CrossRef Medline inhibitors from antagonists to agonists of cellular signaling, Annu. Rev. 23. Jeon, S. M., and Hay, N. (2015) The double-edged sword of AMPK signal- Biochem. 80, 769–795 CrossRef Medline ing in cancer and its therapeutic implications. Arch. Pharm. Res. 38, at UNIVERSITY OF DUNDEE on June 8, 2018 39. Roskoski, R., Jr. (2016) Classification of small molecule protein kinase 346–357 CrossRef Medline inhibitors based upon the structures of their drug-enzyme complexes. 24. Galluzzi, L., Pietrocola, F., Bravo-San Pedro, J. M., Amaravadi, R. K., Bae- Pharmacol. Res. 103, 26–48CrossRef Medline hrecke, E. H., Cecconi, F., Codogno, P., Debnath, J., Gewirtz, D. A., 40. Vijayan, R. S., He, P., Modi, V., Duong-Ly, K. C., Ma, H., Peterson, J. R., Karantza, V., Kimmelman, A., Kumar, S., Levine, B., Maiuri, M. C., Martin, Dunbrack, R. L., Jr., and Levy, R. M. (2015) Conformational analysis of the S. J., et al. (2015) Autophagy in malignant transformation and cancer DFG-out kinase motif and biochemical profiling of structurally validated progression. EMBO J. 34, 856–880 CrossRef Medline Type II inhibitors. J. Med. Chem. 58, 466–479 CrossRef Medline 25. Zhou, G., Myers, R., Li, Y., Chen, Y., Shen, X., Fenyk-Melody, J., Wu, M., 41. Blanc, J., Geney, R., and Menet, C. (2013) Type II kinase inhibitors: an Ventre, J., Doebber, T., Fujii, N., Musi, N., Hirshman, M. F., Goodyear, opportunity in cancer for rational design. Anticancer Agents Med. Chem. L. J., and Moller, D. E. (2001) Role of AMP-activated protein kinase in mechanism of metformin action. J. Clin. Invest. 108, 1167–1174 CrossRef 13, 731–747 CrossRef Medline Medline 42. Garuti, L., Roberti, M., and Bottegoni, G. (2010) Non-ATP competitive 26. Handa, N., Takagi, T., Saijo, S., Kishishita, S., Takaya, D., Toyama, M., protein kinase inhibitors. Curr. Med. Chem. 17, 2804–2821 CrossRef Terada, T., Shirouzu, M., Suzuki, A., Lee, S., Yamauchi, T., Okada-Iwabu, Medline M., Iwabu, M., Kadowaki, T., Minokoshi, Y., and Yokoyama, S. (2011) 43. Liu, Y., and Gray, N. S. (2006) Rational design of inhibitors that bind to Structural basis for compound C inhibition of the human AMP-activated inactive kinase conformations. Nat. Chem. Biol. 2, 358–364 CrossRef protein kinase ␣2 subunit kinase domain. Acta Crystallogr. D Biol. Crys- Medline tallogr. 67, 480–487 CrossRef Medline 44. Cleland, W. W. (1963) The kinetics of enzyme-catalyzed reactions with 27. Bain, J., Plater, L., Elliott, M., Shpiro, N., Hastie, C. J., McLauchlan, H., two or more substrates or products. II. Inhibition: nomenclature and the- Klevernic, I., Arthur, J. S., Alessi, D. R., and Cohen, P. (2007) The selectivity ory. Biochim. Biophys. Acta 67, 173–187 CrossRef Medline of protein kinase inhibitors: a further update. Biochem. J. 408, 297–315 45. Lesuisse, D., Mauger, J., Nemecek, C., Maignan, S., Boiziau, J., Harlow, G., CrossRef Medline Hittinger, A., Ruf, S., Strobel, H., Nair, A., Ritter, K., Malleron, J. L., Dagal- 28. Yu, P. B., Hong, C. C., Sachidanandan, C., Babitt, J. L., Deng, D. Y., Hoyng, lier, A., El-Ahmad, Y., Guilloteau, J. P., et al. (2011) Discovery of the first S. A., Lin, H. Y., Bloch, K. D., and Peterson, R. T. (2008) Dorsomorphin non-ATP competitive IGF-1R kinase inhibitors: Advantages in compari- inhibits BMP signals required for embryogenesis and iron metabolism. son with competitive inhibitors. Bioorg. Med. Chem. Lett. 21, 2224–2228 Nat. Chem. Biol. 4, 33–41 CrossRef Medline CrossRef Medline 29. Emerling, B. M., Viollet, B., Tormos, K. V., and Chandel, N. S. (2007) 46. Kovalenko, M., Ro¨nnstrand, L., Heldin, C. H., Loubtchenkov, M., Gazit, Compound C inhibits hypoxic activation of HIF-1 independent of AMPK. A., Levitzki, A., and Bo¨hmer, F. D. (1997) Phosphorylation site-specific FEBS Lett. 581, 5727–5731 CrossRef Medline inhibition of platelet-derived growth factor ␤-receptor autophosphoryla- 30. Nam, M., Lee, W. H., Bae, E. J., and Kim, S. G. (2008) Compound C inhibits tion by the receptor blocking tyrphostin AG1296. Biochemistry 36, clonal expansion of preadipocytes by increasing p21 level irrespectively of 6260–6269 CrossRef Medline AMPK inhibition. Arch. Biochem. Biophys. 479, 74–81CrossRef Medline 47. Dessauer, C. W., Tesmer, J. J., Sprang, S. R., and Gilman, A. G. (1999) The 31. Lee, Y., Park, B. H., and Bae, E. J. (2016) Compound C inhibits macrophage interactions of adenylate cyclases with P-site inhibitors. Trends Pharma- chemotaxis through an AMPK-independent mechanism. Biochem. Bio- col. Sci. 20, 205–210 CrossRef Medline phys. Res. Commun. 469, 515–520 CrossRef Medline 48. Westmark, P. R., Kelly, J. P., and Smith, B. D. (1993) Photoregulation of 32. Scott, J. W., Galic, S., Graham, K. L., Foitzik, R., Ling, N. X., Dite, T. A., Issa, enzyme activity: photochromic, transition-state-analogue inhibitors of S. M., Langendorf, C. G., Weng, Q. P., Thomas, H. E., Kay, T. W., Birnberg, cysteine and serine proteases. J. Am. Chem. Soc. 115, 3416–3419 CrossRef

8884 J. Biol. Chem. (2018) 293(23) 8874–8885 High-potency AMPK inhibitor

49. Marmor, S., Petersen, C. P., Reck, F., Yang, W., Gao, N., and Fisher, S. L. 57. Kabsch, W. (2010) XDS. Acta Crystallogr. D Biol. Crystallogr. 66, 125–132 (2001) Biochemical characterization of a phosphinate inhibitor of Esche- CrossRef Medline richia coli MurC. Biochemistry 40, 12207–12214 CrossRef Medline 58. Evans, P. R., and Murshudov, G. N. (2013) How good are my data and what 50. Thomson, J. M., Prati, F., Bethel, C. R., and Bonomo, R. A. (2007) Use of is the resolution? Acta Crystallogr. D Biol. Crystallogr. 69, 1204–1214 novel boronic acid transition state inhibitors to probe substrate affinity in CrossRef Medline SHV-type extended-spectrum ␤-lactamases. Antimicrob. Agents Che- 59. McCoy, A. J., Grosse-Kunstleve, R. W., Adams, P. D., Winn, M. D., Sto- mother. 51, 1577–1579 CrossRef Medline roni, L. C., and Read, R. J. (2007) Phaser crystallographic software. J. Appl. 51. Davidson, W., Frego, L., Peet, G. W., Kroe, R. R., Labadia, M. E., Lukas, Crystallogr. 40, 658–674 CrossRef Medline S. M., Snow, R. J., Jakes, S., Grygon, C. A., Pargellis, C., and Werneburg, 60. Emsley, P., Lohkamp, B., Scott, W. G., and Cowtan, K. (2010) Features and B. G. (2004) Discovery and characterization of a substrate selective p38R development of Coot. Acta Crystallogr. D Biol. Crystallogr. 66, 486–501 CrossRef Medline inhibitor. Biochemistry 43, 11658–11671 CrossRef Medline 61. Bricogne, G., Blanc, E., Brandl, M., Flensburg, C., Keller, P., Paciorek, P., 52. Ghani, U., and Ullah, N. (2010) New potent inhibitors of tyrosinase: novel Roversi, P., Sharff, A., Smart, O., Vonrhein, C., and Womack, T. (2011) clues to binding of 1,3,4-thiadiazole-2(3H)-thiones, 1,3,4-oxadiazole- BUSTER version 2.10.0, Global Phasing Ltd., Cambridge, UK 2(3H)-thiones, 4-amino-1,2,4-triazole-5(4H)-thiones, and substituted hy- 62. Afonine, P. V., Grosse-Kunstleve, R. W., Echols, N., Headd, J. J., Moriarty, drazides to the dicopper active site. Bioorg. Med. Chem. 18, 4042–4048 N. W., Mustyakimov, M., Terwilliger, T. C., Urzhumtsev, A., Zwart, P. H., CrossRef Medline and Adams, P. D. (2012) Towards automated crystallographic structure 53. Bdolah, A., and Feingold, D. S. (1968) Uridine diphosphate D-glucose de- refinement with phenix.refine. Acta Crystallogr. D Biol. Crystallogr. 68, hydrogenase of Aerobacter aerogenes. J. Bacteriol. 96, 1144–1149 Medline 352–367 CrossRef Medline 54. Patel, M. S. (1972) The effect of phenylpyruvate on pyruvate metabolism 63. Schüttelkopf, A. W., and van Aalten, D. M. (2004) PRODRG: a tool for Downloaded from in rat brain. Biochem. J. 128, 677–684 CrossRef Medline high-throughput crystallography of protein-ligand complexes. Acta Crys- 55. Knight, Z. A., and Shokat, K. M. (2005) Features of selective kinase inhib- tallogr. D Biol. Crystallogr. 60, 1355–1363 CrossRef Medline itors. Chem Biol. 12, 621–637 CrossRef Medline 64. Chen, V. B., Arendall, W. B., 3rd, Headd, J. J., Keedy, D. A., Immormino, 56. Cheng, Y., and Prusoff, W. H. (1973) Relationship between the inhibition R. M., Kapral, G. J., Murray, L. W., Richardson, J. S., and Richardson, D. C. constant (K1) and the concentration of inhibitor which causes 50 per cent (2010) MolProbity: all-atom structure validation for macromolecular inhibition (I50) of an enzymatic reaction. Biochem. Pharmacol. 22, crystallography. Acta Crystallogr. D Biol. Crystallogr. 66, 12–21 CrossRef http://www.jbc.org/ 3099–3108 CrossRef Medline Medline at UNIVERSITY OF DUNDEE on June 8, 2018

J. Biol. Chem. (2018) 293(23) 8874–8885 8885 AMP-activated protein kinase selectively inhibited by the type II inhibitor SBI-0206965 Toby A. Dite, Christopher G. Langendorf, Ashfaqul Hoque, Sandra Galic, Richard J. Rebello, Ashley J. Ovens, Lisa M. Lindqvist, Kevin R. W. Ngoei, Naomi X. Y. Ling, Luc Furic, Bruce E. Kemp, John W. Scott and Jonathan S. Oakhill J. Biol. Chem. 2018, 293:8874-8885. doi: 10.1074/jbc.RA118.003547 originally published online April 25, 2018

Access the most updated version of this article at doi: 10.1074/jbc.RA118.003547

Downloaded from Alerts: • When this article is cited • When a correction for this article is posted

Click here to choose from all of JBC's e-mail alerts http://www.jbc.org/

This article cites 63 references, 10 of which can be accessed free at http://www.jbc.org/content/293/23/8874.full.html#ref-list-1

at UNIVERSITY OF DUNDEE on June 8, 2018 High potency AMPK inhibitor

Supporting Information

AMP-activated protein kinase selectively inhibited by the type II inhibitor SBI-0206965

Authors: Toby A. Dite1, Christopher G. Langendorf2, Ashfaqul Hoque1, Sandra Galic2,

Richard J. Rebello3,4, Ashley J. Ovens1, Lisa M. Lindqvist5, Kevin R. W. Ngoei2, Naomi X. Y.

Ling1, Luc Furic3,4,6, Bruce E. Kemp2,7, John W. Scott2,7 and Jonathan S. Oakhill1,7*

1Metabolic Signalling Laboratory, St Vincent’s Institute of Medical Research, University of

Melbourne, Victoria, Australia.

2Protein Chemistry & Metabolism, St Vincent’s Institute of Medical Research, University of

Melbourne, Victoria, Australia.

3Prostate Cancer Translational Research Laboratory, Peter MacCallum Cancer Centre,

Victoria, Australia.

4Cancer Program, Biomedicine Discovery Institute and Department of Anatomy and

Developmental Biology, Monash University, Victoria, Australia.

5Cell Signalling and Cell Death Division, The Walter and Eliza Hall Institute of Medical

Research, Victoria, Australia.

6Sir Peter MacCallum Department of Oncology, University of Melbourne, Victoria, Australia

7Mary MacKillop Institute for Health Research, Australian Catholic University,

Victoria, Australia.

*email: [email protected].

1 High potency AMPK inhibitor

SBI-0206965 250 nM 140 120 100 80 60 40 20 % activity vs untreated 0 Src Lck TTK CK2 BTK SYK PKA TrkA PIM1 S6K1 TAK1 JAK3 PLK1 EF2K JNK1 PAK4 LKB1 TBK1 CK1 δ RSK1 PKD1 HER4 PDK1 NEK6 PRK2 PKBa PKCa MST2 MLK3 CHK2 SGK1 MSK1 MKK1 HIPK2 RIPK2 IRAK4 IGF-1R SRPK1 GSK3b MARK3 CAMK1 EPH-A2 VEG-FR ROCK 2 DYRK1A Aurora B CAMKKb SmMLCK p38a MAPK AMPK (hum) Kinase

B Compound C 2.50 µM 140 120 100 80 60 40 20 % activity vs untreated 0 Src Lck TTK CK2 SYK BTK PKA TrkA PIM1 S6K1 TAK1 PLK1 EF2K JNK1 JAK3 PAK4 LKB1 CK1 δ TBK1 PDK1 PKCa NEK6 MST2 PRK2 PKBa RSK1 HER4 PKD1 MLK3 CHK2 SGK1 MSK1 MKK1 RIPK2 HIPK2 IRAK4 IGF-1R SRPK1 GSK3b MARK3 CAMK1 EPH-A2 VEG-FR ROCK 2 DYRK1A Aurora B CAMKKb SmMLCK p38a MAPK AMPK (hum) Kinase

Supporting Figure 1. Kinase inhibitor profiles for (A) SBI-0206965 (250 nM) and (B) compound C (2.5 µM). Inhibitor profiling (Express screen) was performed by the International

Centre for Kinase Profiling, Medical Research Council Protein Phosphorylation and

Ubiquitylation Unit (MRC-PPU), University of Dundee, UK.

2 High potency AMPK inhibitor

A B 120 [SBI-0206965] Ks Kp

) 100 0 nM

1 E + S ES E + P -

ctivity 100 nM a 80 .mg

1 200 nM ⍺K - Ki i 60 400 nM ⍺Ks 40 EI + S EIS no reaction

(nmol.min 20 AMPKspecific 0 0 100 200 300 400 [ATP] (μM)

C D slope 1/Vapp 2.5 0.015

2 R = 0.9571 R2 = 0.9735 2.0 0.010 1.5

1.0 0.005 Ki = 261.4 nM ⍺Ki = 892.3 nM 0.5

-400 -200 0 200 400 [SBI-0206965] (nM) [SBI-0206965] (nM)

Supporting Figure 2. SBI-0206965 is a mixed-competitive AMPK inhibitor (A) ATP dose curves for activity of phosphorylated a2 kinase domain in the presence of fixed SBI-0206965 as indicated. Data was used to generate Lineweaver-Burk plot in Fig. 3D. (B) The effect of a mixed inhibitor (I) on kinetics of product (P) formation by an enzyme (E)/substrate (S) complex. Secondary plots showing [SBI-0206965] vs. (C) Km/V (slope of Lineweaver-Burk plots in Fig 3D) and (D) reciprocal of apparent maximal velocity (1/Vapp). Data points and x-

axis intercepts for (C) (-Ki) and (D) (-aKi) were calculated by GraphPad Prism version 7.0c.

3 High potency AMPK inhibitor

Supplemental Figure 3

K45

S161 E64 L160

DFG DFG DFG

Supporting Figure 3. Structural features of SBI-0206965-bound a2 kinase domain.

Comparisons are based on structural alignment of hinge regions. (A) SBI-0206965-induced rearrangements of P-loop (magenta), C-ahelix (red) and activation loop (blue), relative to apo- a2(T172D) kinase domain (yellow, PDB 2YZA). SBI-0206965 in green. H-bonds indicated by dotted lines. K45 and E64, forming an intact salt bridge in the active a2 kinase conformation

(PDB 4ZHX), in cyan. Residues in orange are from SBI-0206965-bound complex. The K45

CE and NZ atoms in this complex could not be resolved. (B) Comparison of DFG motifs from

(left, DFG-in, yellow) epidermal growth factor receptor tyrosine kinase domain complexed to erlotinib (PDB 1M17), (middle, classical DFG-out, red) c-Abl kinase domain complexed to imatinib (PDB 1IEP) and (right, non-classical DFG-out, orange) a2 kinase domain complexed to SBI-0206965. ATP active site circled in pink.

4 High potency AMPK inhibitor

A! B! 14! 0.96!

12! 0.95! cells) !

5 10! 0.94! 8! *! 0.93! 6! 0.92! 4!

2! ****! 0.91! ZMP ( pmoles /1x10 ZMP

****! Adenylate energy charge ! ****! ****! 0! 0.90! AICAR (1 mM)! -! +! +! +! +! +! +! +! SBI-0206965 (μM)! 0! 1! 5! 10! 30! SBI-0206965 (μM)! -! 0! 1! 5! 10! 30! -! -! -! -! -! -! -! -! 10! 20! compound C (μM)!

Supporting Figure 4. Compatibility of SBI-0206965 to investigate cellular AMPK signalling.

(A) ZMP accumulation in HEK293 cells in response to co-incubation of AICAR with SBI-

0206965 or compound C. (B) Adenylate energy charge of HEK293 cells, treated for 1 h with

SBI-0206965 as indicated.

5 High potency AMPK inhibitor

Supporting Table 1. Data collection and refinement statistics.

AMPK a2 kinase domain Anomalous Br

Data collection

Resolution range (Å) 35.93 - 2.9 (3.004 - 2.9)* 37.97 (3.136 – 3.027)

Wavelength (Å) 0.95374 0.91165

Space group I121 I121 Unit cell - a, b, c (Å) 38.43 54.78 143.19 38.45 54.82 143.42 - a, b, g (˚) 90 96.25 90 90 96.31 90 Total reflections 46866 (7755) 40351 (6527)

Unique reflections 6688 (1072) 5889 (571)

Multiplicity 7.0 (7.2) 6.9 (6.9)

Completeness (%) 99.93 (100.00) 99.59 (98.28)

Mean I/sigma(I) 12.70 (2.2) 20.2 (3.6)

Wilson B-factor 98.94 106.74

R-merge 0.068 (0.793) 0.051 (0.534)

R-pim 0.028 (0.315) 0.021 (0.218)

CC1/2 0.999 (0.909) 0.999 (0.922)

Anomalous completeness - 99.2 (96.9)

Anomalous multiplicity - 3.6 (3.5)

Refinement

R-work 0.2520 -

R-free 0.2800 -

Number of atoms 1879 -

macromolecules 1817 -

ligands 62 -

water 0 -

Protein residues 243 -

RMS(bonds) 0.011 - RMS(angles) 1.60 -

Ramachandran favored (%) 96 - Ramachandran outliers (%) 0 - Average B-factor 111.10 - macromolecules 110.50 - ligands 126.30 -

6 High potency AMPK inhibitor

solvent n/a - *Statistics for the highest-resolution shell are shown in parentheses.

7 CHAPTER 5: General Discussion This project aimed to investigate the mechanisms of activation and inhibition of the metabolic regulator AMP activated protein kinase (AMPK). AMPK is canonically activated by metabolic stress that leads to a decrease in the intracellular adenylate energy charge (AEC). The central role that AMPK plays in responding to changes in AEC and adjusting cellular metabolism has encouraged the development of direct activating drugs of AMPK, and hundreds have been patented to date. The potential of therapeutic AMPK inhibition has also emerged in circumstances where hyperactivity of AMPK leads to disease progression, such as in stroke, neurodegeneration and tumour cell survival.

However, our understanding of how adenine nucleotide ratios change in response to metabolic stress and other cell treatments is incomplete. Furthermore, the ADaM site, which is tightly regulated by phosphorylation of the β1-Ser108 residue, has been confirmed as the drug binding site of many compounds that stimulate AMPK activity, however, endogenous pathways that regulate β1-Ser108 phosphorylation are unknown. Finally, research tools to investigate AMPK inhibition are severely limited.

In my PhD I employed a novel LC-MS method to examine how adenylate energy charge, and individual AMP- and ADP-to-ATP ratios change during metabolic stress, and to demonstrate the importance of adenine nucleotide quantitation to contextualise AMPK signalling in cells. I also identified and confirmed ULK1 as an upstream kinase of the drug sensitising residue AMPK β1-Ser108. Finally, I identified and characterised a highly selective, type II inhibitor of AMPK, SBI-0206965.

AMPK links the anabolic and catabolic output of the cell to adenylate energy charge, and altered [AMP]/[ATP] and/or [ADP]/[ATP] is universally considered to alter AMPK pathway signalling in cells (33, 111, 286). Despite this, monitoring AMP, ADP and ATP ratios is not routine when investigating AMPK regulation in cells (72, 181, 245). To accurately interrogate mechanisms of AMPK activation or inhibition by drugs or novel endogenous ligands, it is crucial to simultaneously quantitate intracellular AMP, ADP and ATP so as to distinguish novel and canonical regulation of AMPK.

Off-target effects of many compounds can result in an increase in [AMP]/[ATP] in cells, thereby indirectly activating endogenous AMPK pathways. Examples of these include compounds that inhibit mitochondrial oxidative phosphorylation such as resveratrol, thiazlidinediones, ginsenoside and α-lipoic acid, or mitochondrial uncouplers, such as

101

dinitrophenol. In theory, any compound that causes an increase in [AMP]/[ATP] can activate AMPK indirectly, and, in the absence of nucleotide quantitation, could be mistaken as a direct activator of AMPK. In my PhD studies I demonstrated that three molecules, (salicylate, A- 769662 and tranilast), were able to cause significant increases in [AMP]/[ATP], and are therefore capable to activating AMPK indirectly. While this is a known feature of salicylate (254), A-769662 is a direct AMPK activator (AMP independent), and tranilast is an antiinflammatory drug that is thought to act independently of AMPK signalling (287). These results are precedent that compounds are unpredictably capable of activating AMPK indirectly in cells at high concentrations, and should be taken into account in all intracellular AMPK studies.

In my PhD I also examined how AMP and ADP change with respect to ATP during metabolic stress. AMP binding to the γ-subunit has the ability to stimulate AMPK activity through multiple mechanisms, whereas ADP binding only indirectly increases AMPK activity by increasing Thr172 phosphorylation. While ADP exists at roughly 10 times the concentration of AMP, the adenylate kinase equilibrium predicts that [AMP]/[ATP] should increase as the square of [ADP]/[ATP] during metabolic stress (263). In fact, adenylate kinase-mediated increases in AMP has been proposed to be the main mechanism through which metabolic stress is relayed to AMPK (288-291). In my PhD I have shown that fold increases in [AMP]/[ATP] are statistically indistinguishable from fold increases in [ADP]/[ATP] under most circumstances of metabolic stress. This has also been demonstrated in MEFs that were treated with the mitochondrial inhibitor berberine, and so is unlikely to be a cell-specific effect (193). However, in conditions associated with inhibition of AMP metabolising pathways, larger increases of [AMP]/[ATP] were observed.

AMP metabolising pathways have previously been proposed to prevent large decreases in adenylate energy charge during cell stress by removing AMP and reducing the total pool of adenine nucleotides in the cell (265). I have shown that only during oxidative stress (which is known to inhibit AMP deaminase) (271), do [AMP]/[ATP] ratios increase as the square of [ADP]/[ATP]. In line with this, pharmacological AMP deaminase inhibition has been shown to increase [AMP]/[ATP], and the degree of AMPK activation in contracting muscle (269). Work presented in my PhD suggests that oxidative inactivation of AMP deaminase may represent a physiological mechanism for increased AMPK by allowing increases in [AMP]/[ATP] during stress. This may occur during times of severe or prolonged cellular stress, such as hypoxia or endurance exercise, which results in mtROS production.

102

Next I investigated regulation by phosphorylation of the drug-sensitising residue, AMPK β1- Ser108. Work conducted prior to my PhD identified phosphorylation of Ser108 on the β1- subunit of AMPK as critical for activation by compounds that bind to the allosteric drug and metabolite (ADaM) site (44). Our group determined that ADaM site ligands are able to activate AMPK that is not phosphorylated on its activation loop, provided that β1-Ser108 is phosphorylated, or AMP is bound to the γ-subunit. In addition, AMPK autophosphorylation of β1-Ser108 occurs via cis-autophosphorylation, limiting the pool of drug-sensitive AMPK to that which has been phosphorylated at Thr172 (44).

Understanding how endogenous pathways regulate β1-Ser108 phosphorylation may assist in the development of more efficient strategies to pharmacologically activate AMPK pathways. Over a decade has passed since the discovery of the first synthetic small molecule activator of AMPK, and yet no drug has advanced through clinical trials. The discovery that AMPK could be activated synergistically by ADaM site and AMP/C2 has offered new opportunities for combinatorial therapies to enhance AMPK activation in cells, without the need for activation loop phosphorylation. Similarly, if the pool of Ser108 phosphorylated AMPK in the cell could be increased by exploiting endogenous pathways, the effectiveness of existing AMPK activators may be enhanced.

My PhD studies identified that β1-Ser108 phosphorylation is required for AMPK pathway activation by ADaM site drugs in cells. Using recombinant, kinase-inactive AMPK that could not be autophosphorylated, I demonstrated that under conditions of metabolic stress that increased [AMP]/[ATP], β1-Ser108 could be trans-phosphorylated by a non-AMPK kinase. In addition, my PhD studies unambiguously demonstrated the first example of activation loop- independent AMPK signalling in cells.

Using a synthetic peptide covering the β1-Ser108 sequence, and western-blot and mass spectrometric analysis, I identified AMPK β1-Ser108 as a phosphorylation target for ULK1, a major regulator of autophagy initiation. Consequently, ULK1 sensitises AMPK to A-769662 and salicylate independently of Thr172 phosphorylation. Using pharmacological inhibition and genetic deletion of ULK1, I was able to confirm that ULK1 is a physiologically relevant upstream kinase of AMPK β1-Ser108 in cells.

This discovery has important implications given that phosphorylation of β1-Ser108 is now recognised as vital to drug activation of β1-AMPK, and may be useful for exploiting AMPK pathways in circumstances where Thr172 dependent activation of AMPK is limited, such as

103

non-small-cell lung and cervical carcinomas, which are associated with genetic loss of LKB1, and where AMPK activity may be therapeutic (292). Furthermore, identification of ULK1 as an upstream kinase of Ser108 may point toward a functional role for ADaM site ligand- regulated AMPK activity. While salicylate is currently the only naturally occurring compound known to bind at the ADaM site, the presence of a post-translationally-regulated drug binding site has led to speculation that it is an “orphan” allosteric site for an unidentified endogenous AMPK ligand. Many endogenous metabolites are elevated during autophagy, which are mainly associated with amino acid, energy, carbohydrate and lipid metabolism (293). Phosphorylation of AMPK by ULK1 may act as a priming event for AMPK pathway activation during autophagy, while certain metabolites are elevated. Future studies may be able to identify which, if any, of these metabolites are able to regulate AMPK activity, however this was beyond the scope of my PhD.

The physiological role for Ser108 phosphorylation was examined using a quantitative phosphoproteomic approach. Significant differences in the phosphorylation profiles of several cell-cycle-associated proteins were identified in response to the phosphomimetic mutant β1- S108E. In particular, phosphorylation of the cell-cycle regulator PAK2 at Ser141 was significantly increased by β1-S108E expression. This is a known functional site of PAK2, required for full kinase activity and can lead to cell cycle arrest and the prevention of apoptosis (294, 295). Along with other downstream targets associated with cell-cycle arrest, these data point to a role for β1-Ser108 phosphorylation in promoting cell survival during cell stress, via AMPK-ULK1 cross-talk.

Finally, my PhD studies have identified and characterised a novel, type II inhibitor of AMPK, SBI-0206965.

While many small molecule activators of AMPK have been developed, existing tools for pharmacological inhibition of AMPK are limited to a handful of non-selective and/or weak antagonists. This is despite accumulating evidence that AMPK activity may actually play a negative role in the progression of some diseases (183). By far the most widely applied AMPK inhibitor, the pyrazolopyrimidine derivative compound C (also called dorsomorphin), was originally selected from a high-throughput screen used to confirm AMPK dependent effects of AICAR and metformin in cultured hepatocytes (121). However, the off-target effects and lack of potency in cells where [ATP] is high has severely restricted compound C’s usefulness as a research tool.

104

In the course of investigating ULK1 phosphorylation of AMPK in cells, I discovered that the pyrimidine derivative SBI-0206965 is a potent inhibitor of AMPK activity. Inhibitor selectivity profiling was used to show that SBI-0206965 is preferably selective for AMPK inhibition, relative to compound C. SBI-0206965 was able to inhibit AMPK in a cell-free activity assays in the nanomolar range for both the α1 and α2 AMPK heterotrimers. SBI-0206965 was also able to inhibit the α2 kinase domain (6-278), and enzyme kinetic profiling revealed that SBI- 0206965 behaves as a mixed, ATP non-competitive inhibitor of AMPK under assay conditions tested in my PhD.

A crystal structure of the α2 kinase domain/inhibitor complex was solved, revealing that SBI- 0206965 binds to the active site of AMPK, partially overlapping with the ATP binding site. The N-methylbenzamide group of SBI-0206965, which diverges from the active site occupancy of compound C, sterically clashes with the Asp of the AMPK DFG motif. This limits SBI-0206965 binding to AMPK in the DFG-out conformation. Additionally, binding of SBI-0206965 to the AMPK kinase domain does not extend to the back of the catalytic cleft.

Taken together, these features classify SBI-0206965 as a type IIb inhibitor of AMPK (225). Type II inhibitors are intrinsically more selective than type I inhibitors such as staurosporine and compound C, and are able to maintain high potency in cells. While compound C is a poor inhibitor of AMPK activity in cells (183, 239), SBI-0206965 was able to inhibit AMPK activity in cells in response to increased Ca2+ in the neuronal cell line SH-SY5Y, and glucose starvation in HEK293 cells. SBI-0206965 reduced AMPK activity in response to AICAR in HEK293s, however this was also due to reduced ZMP accumulation. Finally, in the prostate cancer cell line PC3, SBI-0206965 was able to prevent cell proliferation, consistent with the role that AMPK plays prostate cancer, providing one example of how AMPK inhibition may be exploited therapeutically.

Future in vivo studies using SBI-0206965 will reveal its usefulness as an AMPK inhibitor in disease models. While increased AMPK activity has been proposed to contribute to the severity of some neurodegenerative diseases (219, 220), cancers(212-214) and stroke (203), loss of AMPK activity has been shown to exacerbate many other diseases. For example, AMPK activity promotes cell death in T cell acute lymphoblastic leukemia and suppresses tumour growth in vivo, suggesting that untargeted AMPK inhibition would be detrimental for patients (296, 297). Advanced targeted therapies, such as drug nanocarriers, may be an essential consideration for any AMPK inhibitor to reach the clinic (298).

105

In conclusion my study has shown aspects of AMPK activation and inhibition that are of critical importance in relation to exploiting AMPK as a drug target. I have demonstrated that only by employing adenine nucleotide quantitation while investigating AMPK regulation in cells is it possible to determine when AMPK is being activated indirectly through decreases in adenylate energy charge, or through other mechanisms such as ADaM site ligand binding. Additionally, I have shown that the autophagy initiating kinase ULK1 is able to phosphorylate AMPK β1- Ser108 in cells, and that phosphorylation at this site is critical for drug activation of AMPK in cells. My project also identified, and biochemically and structurally characterised, a selective and potent AMPK inhibitor that may assist in the development of a new class of AMPK inhibitors with therapeutic potential.

References

1. D. E. Atkinson, Biological feedback control at the molecular level. Science (New York, N.Y.) 150, 851-857 (1965). 2. D. E. Atkinson, G. M. Walton, Adenosine triphosphate conservation in metabolic regulation. Rat liver citrate cleavage enzyme. J biol chem 242, 3239-3241 (1967). 3. M. Carlson, B. C. Osmond, D. Botstein, Mutants of yeast defective in sucrose utilization. Genetics 98, 25-40 (1981). 4. J. L. Celenza, M. Carlson, Cloning and genetic mapping of SNF1, a gene required for expression of glucose-repressible genes in Saccharomyces cerevisiae. Mol Cell Biol 4, 49-53 (1984). 5. J. L. Celenza, M. Carlson, A yeast gene that is essential for release from glucose repression encodes a protein kinase. Science (New York, N.Y.) 233, 1175-1180 (1986). 6. K. Hedbacker, M. Carlson, SNF1/AMPK pathways in yeast. Front Biosci 13, 2408-2420 (2008). 7. M. Johnston, J. S. Flick, T. Pexton, Multiple mechanisms provide rapid and stringent glucose repression of GAL gene expression in Saccharomyces cerevisiae. Mol Cell Biol 14, 3834-3841 (1994). 8. L. G. Vallier, M. Carlson, Synergistic release from glucose repression by mig1 and ssn mutations in Saccharomyces cerevisiae. Genetics 137, 49-54 (1994). 9. S. Pessina et al., Snf1/AMPK promotes S-phase entrance by controlling CLB5 transcription in budding yeast. Cell Cycle 9, 2189-2200 (2010). 10. S. Busnelli et al., Snf1/AMPK promotes SBF and MBF-dependent transcription in budding yeast. Biochimica et biophysica acta 1833, 3254-3264 (2013).

106

11. F. Tripodi, R. Fraschini, M. Zocchi, V. Reghellin, P. Coccetti, Snf1/AMPK is involved in the mitotic spindle alignment in Saccharomyces cerevisiae. Sci Rep 8, 5853 (2018). 12. J. L. Celenza, F. J. Eng, M. Carlson, Molecular analysis of the SNF4 gene of Saccharomyces cerevisiae: evidence for physical association of the SNF4 protein with the SNF1 protein kinase. Mol Cell Biol 9, 5045-5054 (1989). 13. F. C. Smith, S. P. Davies, W. A. Wilson, D. Carling, D. G. Hardie, The SNF1 kinase complex from Saccharomyces cerevisiae phosphorylates the transcriptional repressor protein Mig1p in vitro at four sites within or near regulatory domain 1. FEBS Lett 453, 219-223 (1999). 14. J. Sheen, Metabolic repression of transcription in higher plants. Plant Cell 2, 1027-1038 (1990). 15. A. von Schaewen, M. Stitt, R. Schmidt, U. Sonnewald, L. Willmitzer, Expression of a yeast- derived invertase in the cell wall of tobacco and Arabidopsis plants leads to accumulation of carbohydrate and inhibition of photosynthesis and strongly influences growth and phenotype of transgenic tobacco plants. EMBO J 9, 3033-3044 (1990). 16. C. D. Dickinson, T. Altabella, M. J. Chrispeels, Slow-growth phenotype of transgenic tomato expressing apoplastic invertase. Plant Physiol 95, 420-425 (1991). 17. I. A. Graham, K. J. Denby, C. J. Leaver, Carbon Catabolite Repression Regulates Glyoxylate Cycle Gene Expression in Cucumber. Plant Cell 6, 761-772 (1994). 18. S. M. Yu, Y. H. Kuo, G. Sheu, Y. J. Sheu, L. F. Liu, Metabolic derepression of alpha-amylase gene expression in suspension-cultured cells of rice. J biol chem 266, 21131-21137 (1991). 19. A. Alderson et al., Complementation of snf1, a mutation affecting global regulation of carbon metabolism in yeast, by a plant protein kinase cDNA. PNAS 88, 8602-8605 (1991). 20. N. G. Halford, D. G. Hardie, SNF1-related protein kinases: global regulators of carbon metabolism in plants? Plant Mol Biol 37, 735-748 (1998). 21. S. Dale et al., Bacterial expression of the catalytic domain of 3-hydroxy-3-methylglutaryl-CoA reductase (isoform HMGR1) from Arabidopsis thaliana, and its inactivation by phosphorylation at Ser577 by Brassica oleracea 3-hydroxy-3-methylglutaryl-CoA reductase kinase. Eur J Biochem 233, 506-513 (1995). 22. C. Sugden, P. G. Donaghy, N. G. Halford, D. G. Hardie, Two SNF1-related protein kinases from spinach leaf phosphorylate and inactivate 3-hydroxy-3-methylglutaryl-coenzyme A reductase, nitrate reductase, and sucrose phosphate synthase in vitro. Plant Physiol 120, 257-274 (1999). 23. M. Bachmann et al., Identification of Ser-543 as the major regulatory phosphorylation site in spinach leaf nitrate reductase. Plant Cell 8, 505-517 (1996). 24. S. Laurie, R. S. McKibbin, N. G. Halford, Antisense SNF1-related (SnRK1) protein kinase gene represses transient activity of an alpha-amylase (alpha-Amy2) gene promoter in cultured wheat embryos. J Exp Bot 54, 739-747 (2003). 25. N. G. Halford, M. J. Paul, Carbon metabolite sensing and signalling. Plant Biotechnol J 1, 381- 398 (2003). 26. F. V. Mayer et al., ADP regulates SNF1, the Saccharomyces cerevisiae homolog of AMP- activated protein kinase. Cell metabolism 14, 707-714 (2011). 27. C. A. Carlson, K. H. Kim, Regulation of hepatic acetyl coenzyme A carboxylase by phosphorylation and dephosphorylation. Arch Biochem Biophys 164, 478-489 (1974). 28. Z. H. Beg, D. W. Allmann, D. M. Gibson, Modulation of 3-hydroxy-3-methylglutaryl coenzyme A reductase activity with cAMP and wth protein fractions of rat liver cytosol. Biochem Biophys Res Commun 54, 1362-1369 (1973). 29. L. A. Yeh, K. H. Lee, K. H. Kim, Regulation of rat liver acetyl-CoA carboxylase. Regulation of phosphorylation and inactivation of acetyl-CoA carboxylase by the adenylate energy charge. J biol chem 255, 2308-2314 (1980).

107

30. D. Carling, V. A. Zammit, D. G. Hardie, A common bicyclic protein kinase cascade inactivates the regulatory enzymes of fatty acid and cholesterol biosynthesis. FEBS Lett 223, 217-222 (1987). 31. M. R. Munday, D. G. Campbell, D. Carling, D. G. Hardie, Identification by amino acid sequencing of three major regulatory phosphorylation sites on rat acetyl-CoA carboxylase. Eur J Biochem 175, 331-338 (1988). 32. D. Carling, P. R. Clarke, V. A. Zammit, D. G. Hardie, Purification and characterization of the AMP-activated protein kinase. Copurification of acetyl-CoA carboxylase kinase and 3- hydroxy-3-methylglutaryl-CoA reductase kinase activities. Eur J Biochem 186, 129-136 (1989). 33. G. R. Steinberg, B. E. Kemp, AMPK in Health and Disease. Physiol Rev 89, 1025-1078 (2009). 34. A. Woods et al., Characterization of AMP-activated protein kinase beta and gamma subunits. Assembly of the heterotrimeric complex in vitro. J biol chem 271, 10282-10290 (1996). 35. D. Stapleton et al., Mammalian AMP-activated protein kinase subfamily. J biol chem 271, 611-614 (1996). 36. F. A. Ross, C. MacKintosh, D. G. Hardie, AMP-activated protein kinase: a cellular energy sensor that comes in 12 flavours. FEBS J 283, 2987-3001 (2016). 37. M. Foretz et al., Maintenance of red blood cell integrity by AMP-activated protein kinase alpha1 catalytic subunit. FEBS Lett 584, 3667-3671 (2010). 38. H. Zhu et al., PRKAA1/AMPKalpha1 is required for autophagy-dependent mitochondrial clearance during erythrocyte maturation. Autophagy 10, 1522-1534 (2014). 39. B. Viollet et al., The AMP-activated protein kinase alpha2 catalytic subunit controls whole- body insulin sensitivity. J Clin Invest 111, 91-98 (2003). 40. B. Viollet et al., AMPK: Lessons from transgenic and knockout animals. Front Biosci (Landmark Ed) 14, 19-44 (2009). 41. B. Nolen, S. Taylor, G. Ghosh, Regulation of protein kinases; controlling activity through activation segment conformation. Mol Cell 15, 661-675 (2004). 42. A. P. Kornev, S. S. Taylor, Defining the conserved internal architecture of a protein kinase. Biochimica et biophysica acta 1804, 440-444 (2010). 43. S. A. Hawley et al., Characterization of the AMP-activated protein kinase kinase from rat liver and identification of threonine 172 as the major site at which it phosphorylates AMP- activated protein kinase. J biol chem 271, 27879-27887 (1996). 44. J. W. Scott et al., Small molecule drug A-769662 and AMP synergistically activate naive AMPK independent of upstream kinase signaling. Chemistry & biology 21, 619-627 (2014). 45. B. E. Crute, K. Seefeld, J. Gamble, B. E. Kemp, L. A. Witters, Functional domains of the alpha1 catalytic subunit of the AMP-activated protein kinase. J biol chem 273, 35347-35354 (1998). 46. B. Xiao et al., Structure of mammalian AMPK and its regulation by ADP. Nature 472, 230-233 (2011). 47. L. Chen et al., Conserved regulatory elements in AMPK. Nature 498, E8-10 (2013). 48. F. J. Xin, J. Wang, R. Q. Zhao, Z. X. Wang, J. W. Wu, Coordinated regulation of AMPK activity by multiple elements in the alpha-subunit. Cell research 23, 1237-1240 (2013). 49. T. J. Iseli et al., AMP-activated protein kinase beta subunit tethers alpha and gamma subunits via its C-terminal sequence (186-270). J biol chem 280, 13395-13400 (2005). 50. T. Suzuki et al., Inhibition of AMPK Catabolic Action by GSK3. Mol Cell 50, 407-419 (2013). 51. S. A. Hawley et al., Phosphorylation by Akt within the ST loop of AMPK-alpha1 down- regulates its activation in tumour cells. Biochem J 459, 275-287 (2014). 52. B. Xiao et al., Structural basis for AMP binding to mammalian AMP-activated protein kinase. Nature 449, 496-500 (2007). 53. C. Thornton, M. A. Snowden, D. Carling, Identification of a novel AMP-activated protein kinase beta subunit isoform that is highly expressed in skeletal muscle. J biol chem 273, 12443-12450 (1998).

108

54. E. C. Cokorinos et al., Activation of Skeletal Muscle AMPK Promotes Glucose Disposal and Glucose Lowering in Non-human Primates and Mice. Cell metabolism 25, 1147-1159 e1110 (2017). 55. J. Wu et al., Chemoproteomic analysis of intertissue and interspecies isoform diversity of AMP-activated protein kinase (AMPK). J biol chem 288, 35904-35912 (2013). 56. X. Stephenne et al., Metformin activates AMP-activated protein kinase in primary human hepatocytes by decreasing cellular energy status. Diabetologia 54, 3101-3110 (2011). 57. J. W. Scott et al., Thienopyridone drugs are selective activators of AMP-activated protein kinase beta1-containing complexes. Chemistry & biology 15, 1220-1230 (2008). 58. E. R. Hudson et al., A novel domain in AMP-activated protein kinase causes glycogen storage bodies similar to those seen in hereditary cardiac arrhythmias. Current biology : CB 13, 861- 866 (2003). 59. G. Polekhina et al., AMPK beta subunit targets metabolic stress sensing to glycogen. Current biology : CB 13, 867-871 (2003). 60. A. Koay, K. A. Rimmer, H. D. Mertens, P. R. Gooley, D. Stapleton, Oligosaccharide recognition and binding to the carbohydrate binding module of AMP-activated protein kinase. FEBS Lett 581, 5055-5059 (2007). 61. A. Ray, N. Jatana, L. Thukral, Lipidated proteins: Spotlight on protein-membrane binding interfaces. Prog Biophys Mol Biol, 128, 74-84 (2017). 62. K. I. Mitchelhill et al., Posttranslational modifications of the 5'-AMP-activated protein kinase beta1 subunit. J biol chem 272, 24475-24479 (1997). 63. J. S. Oakhill et al., beta-Subunit myristoylation is the gatekeeper for initiating metabolic stress sensing by AMP-activated protein kinase (AMPK). PNAS 107, 19237-19241 (2010). 64. N. Ali et al., beta-subunit myristoylation functions as an energy sensor by modulating the dynamics of AMP-activated Protein Kinase. Sci Rep 6, 39417 (2016). 65. T. Rubio, S. Vernia, P. Sanz, Sumoylation of AMPKbeta2 subunit enhances AMP-activated protein kinase activity. Mol Biol Cell 24, 1801-1811, S1801-1804 (2013). 66. Y. Oligschlaeger et al., The recruitment of AMP-activated protein kinase to glycogen is regulated by autophosphorylation. J biol chem 290, 11715-11728 (2015). 67. A. S. Loffler et al., Ulk1-mediated phosphorylation of AMPK constitutes a negative regulatory feedback loop. Autophagy 7, 696-706 (2011). 68. P. C. Cheung, I. P. Salt, S. P. Davies, D. G. Hardie, D. Carling, Characterization of AMP- activated protein kinase gamma-subunit isoforms and their role in AMP binding. Biochem J 346 Pt 3, 659-669 (2000). 69. M. Mahlapuu et al., Expression profiling of the gamma-subunit isoforms of AMP-activated protein kinase suggests a major role for gamma3 in white skeletal muscle. Am J Physiol Endocrinol Metab 286, E194-200 (2004). 70. J. W. Scott et al., CBS domains form energy-sensing modules whose binding of adenosine ligands is disrupted by disease mutations. J Clin Invest 113, 274-284 (2004). 71. J. Ereno-Orbea, I. Oyenarte, L. A. Martinez-Cruz, CBS domains: Ligand binding sites and conformational variability. Arch Biochem Biophys 540, 70-81 (2013). 72. F. A. Ross, T. E. Jensen, D. G. Hardie, Differential regulation by AMP and ADP of AMPK complexes containing different gamma subunit isoforms. Biochem J 473, 189-199 (2016). 73. P. Mertins et al., Proteogenomics connects somatic mutations to signalling in breast cancer. Nature 534, 55-62 (2016). 74. K. Sharma et al., Ultradeep human phosphoproteome reveals a distinct regulatory nature of Tyr and Ser/Thr-based signaling. Cell Rep 8, 1583-1594 (2014). 75. Y. Bian et al., An enzyme assisted RP-RPLC approach for in-depth analysis of human liver phosphoproteome. J Proteomics 96, 253-262 (2014). 76. R. Townley, L. Shapiro, Crystal structures of the adenylate sensor from fission yeast AMP- activated protein kinase. Science (New York, N.Y.) 315, 1726-1729 (2007).

109

77. X. Jin, R. Townley, L. Shapiro, Structural insight into AMPK regulation: ADP comes into play. Structure (London, England : 1993) 15, 1285-1295 (2007). 78. G. A. Amodeo, M. J. Rudolph, L. Tong, Crystal structure of the heterotrimer core of Saccharomyces cerevisiae AMPK homologue SNF1. Nature 449, 492-495 (2007). 79. D. R. Littler et al., A conserved mechanism of autoinhibition for the AMPK kinase domain: ATP-binding site and catalytic loop refolding as a means of regulation. Acta Crystallogr Sect F Struct Biol Cryst Commun 66, 143-151 (2010). 80. B. Xiao et al., Structural basis of AMPK regulation by small molecule activators. Nature communications 4, 3017 (2013). 81. K. O. Cameron et al., Discovery and Preclinical Characterization of 6-Chloro-5-[4-(1- hydroxycyclobutyl)phenyl]-1H-indole-3-carboxylic Acid (PF-06409577), a Direct Activator of Adenosine Monophosphate-activated Protein Kinase (AMPK), for the Potential Treatment of Diabetic Nephropathy. Journal of medicinal chemistry 59, 8068-8081 (2016). 82. M. F. Calabrese et al., Structural basis for AMPK activation: natural and synthetic ligands regulate kinase activity from opposite poles by different molecular mechanisms. Structure (London, England : 1993) 22, 1161-1172 (2014). 83. C. G. Langendorf et al., Structural basis of allosteric and synergistic activation of AMPK by furan-2-phosphonic derivative C2 binding. Nature communications 7, 10912 (2016). 84. J. S. Oakhill et al., AMPK is a direct adenylate charge-regulated protein kinase. Science (New York, N.Y.) 332, 1433-1435 (2011). 85. L. Chen et al., AMP-activated protein kinase undergoes nucleotide-dependent conformational changes. Nat Struct Mol Biol 19, 716-718 (2012). 86. G. J. Gowans, S. A. Hawley, F. A. Ross, D. G. Hardie, AMP is a true physiological regulator of AMP-activated protein kinase by both allosteric activation and enhancing net phosphorylation. Cell metabolism 18, 556-566 (2013). 87. S. P. Hong, F. C. Leiper, A. Woods, D. Carling, M. Carlson, Activation of yeast Snf1 and mammalian AMP-activated protein kinase by upstream kinases. Proceedings of the National Academy of Sciences of the United States of America 100, 8839-8843 (2003). 88. S. A. Hawley et al., Complexes between the LKB1 tumor suppressor, STRAD alpha/beta and MO25 alpha/beta are upstream kinases in the AMP-activated protein kinase cascade. J Biol 2, 28 (2003). 89. D. Fan, C. Ma, H. Zhang, The molecular mechanisms that underlie the tumor suppressor function of LKB1. Acta Biochim Biophys Sin (Shanghai) 41, 97-107 (2009). 90. J. M. Evans, L. A. Donnelly, A. M. Emslie-Smith, D. R. Alessi, A. D. Morris, Metformin and reduced risk of cancer in diabetic patients. BMJ (Clinical research ed.) 330, 1304-1305 (2005). 91. D. B. Shackelford et al., LKB1 inactivation dictates therapeutic response of non-small cell lung cancer to the metabolism drug phenformin. Cancer Cell 23, 143-158 (2013). 92. S. A. Hawley et al., 5'-AMP activates the AMP-activated protein kinase cascade, and Ca2+/calmodulin activates the calmodulin-dependent protein kinase I cascade, via three independent mechanisms. J biol chem 270, 27186-27191 (1995). 93. R. L. Hurley et al., The Ca2+/calmodulin-dependent protein kinase kinases are AMP-activated protein kinase kinases. The Journal of biological chemistry 280, 29060-29066 (2005). 94. S. A. Hawley et al., Calmodulin-dependent protein kinase kinase-beta is an alternative upstream kinase for AMP-activated protein kinase. Cell metabolism 2, 9-19 (2005). 95. A. Woods et al., Ca2+/calmodulin-dependent protein kinase kinase-beta acts upstream of AMP-activated protein kinase in mammalian cells. Cell metabolism 2, 21-33 (2005). 96. M. Momcilovic, S. P. Hong, M. Carlson, Mammalian TAK1 activates Snf1 protein kinase in yeast and phosphorylates AMP-activated protein kinase in vitro. J biol chem 281, 25336- 25343 (2006). 97. M. Xie et al., A pivotal role for endogenous TGF-beta-activated kinase-1 in the LKB1/AMP- activated protein kinase energy-sensor pathway. PNAS 103, 17378-17383 (2006).

110

98. A. C. Ferretti et al., AMPK and PKA interaction in the regulation of survival of liver cancer cells subjected to glucose starvation. Oncotarget 7, 17815-17828 (2016). 99. Y. Dagon et al., p70S6 kinase phosphorylates AMPK on serine 491 to mediate leptin's effect on food intake. Cell metabolism 16, 104-112 (2012). 100. S. P. Davies, N. R. Helps, P. T. Cohen, D. G. Hardie, 5'-AMP inhibits dephosphorylation, as well as promoting phosphorylation, of the AMP-activated protein kinase. Studies using bacterially expressed human protein phosphatase-2C alpha and native bovine protein phosphatase-2AC. FEBS Lett 377, 421-425 (1995). 101. N. Kudo et al., Characterization of 5'AMP-activated protein kinase activity in the heart and its role in inhibiting acetyl-CoA carboxylase during reperfusion following ischemia. Biochimica et biophysica acta 1301, 67-75 (1996). 102. B. H. Goodpaster, L. M. Sparks, Metabolic Flexibility in Health and Disease. Cell metabolism 25, 1027-1036 (2017). 103. H. Sano et al., Insulin-stimulated phosphorylation of a Rab GTPase-activating protein regulates GLUT4 translocation. J biol chem 278, 14599-14602 (2003). 104. H. M. O'Neill, AMPK and Exercise: Glucose Uptake and Insulin Sensitivity. Diabetes Metab J 37, 1-21 (2013). 105. C. Pehmoller et al., Genetic disruption of AMPK signaling abolishes both contraction- and insulin-stimulated TBC1D1 phosphorylation and 14-3-3 binding in mouse skeletal muscle. Am J Physiol Endocrinol Metab 297, E665-675 (2009). 106. D. An et al., TBC1D1 regulates insulin- and contraction-induced glucose transport in mouse skeletal muscle. Diabetes 59, 1358-1365 (2010). 107. M. D. Bruss, E. B. Arias, G. E. Lienhard, G. D. Cartee, Increased phosphorylation of Akt substrate of 160 kDa (AS160) in rat skeletal muscle in response to insulin or contractile activity. Diabetes 54, 41-50 (2005). 108. H. M. O'Neill et al., AMP-activated protein kinase (AMPK) β1β2 muscle null mice reveal an essential role for AMPK in maintaining mitochondrial content and glucose uptake during exercise. PNAS 108, 16092-16097 (2011). 109. D. Carling, D. G. Hardie, The substrate and sequence specificity of the AMP-activated protein kinase. Phosphorylation of glycogen synthase and phosphorylase kinase. Biochimica et biophysica acta 1012, 81-86 (1989). 110. W. G. Aschenbach et al., Effect of AICAR treatment on glycogen metabolism in skeletal muscle. Diabetes 51, 567-573 (2002). 111. D. G. Hardie, F. A. Ross, S. A. Hawley, AMPK: a nutrient and energy sensor that maintains energy homeostasis. Nat rev. Mol cell biol 13, 251-262 (2012). 112. A. S. Marsin et al., Phosphorylation and activation of heart PFK-2 by AMPK has a role in the stimulation of glycolysis during ischaemia. Current biology : CB 10, 1247-1255 (2000). 113. E. Domenech et al., AMPK and PFKFB3 mediate glycolysis and survival in response to mitophagy during mitotic arrest. Nature cell biology 17, 1304-1316 (2015). 114. L. Abu-Elheiga, W. Oh, P. Kordari, S. J. Wakil, Acetyl-CoA carboxylase 2 mutant mice are protected against obesity and diabetes induced by high-fat/high-carbohydrate diets. PNAS 100, 10207-10212 (2003). 115. D. G. Hardie, D. A. Pan, Regulation of fatty acid synthesis and oxidation by the AMP-activated protein kinase. Biochem Soc Trans 30, 1064-1070 (2002). 116. I. M. Graham, A. L. Catapano, N. D. Wong, Current guidelines on prevention with a focus on dyslipidemias. Cardiovasc Diagn Ther 7, S4-S10 (2017). 117. M. D. Fullerton et al., Salicylate improves macrophage cholesterol homeostasis via activation of Ampk. J Lipid Res 56, 1025-1033 (2015). 118. D. M. Muoio, K. Seefeld, L. A. Witters, R. A. Coleman, AMP-activated kinase reciprocally regulates triacylglycerol synthesis and fatty acid oxidation in liver and muscle: evidence that

111

sn-glycerol-3-phosphate acyltransferase is a novel target. Biochem J 338 ( Pt 3), 783-791 (1999). 119. H. Park et al., Coordinate regulation of malonyl-CoA decarboxylase, sn-glycerol-3-phosphate acyltransferase, and acetyl-CoA carboxylase by AMP-activated protein kinase in rat tissues in response to exercise. The Journal of biological chemistry 277, 32571-32577 (2002). 120. M. J. Watt et al., Reduced plasma FFA availability increases net triacylglycerol degradation, but not GPAT or HSL activity, in human skeletal muscle. Am J Physiol Endocrinol Metab 287, E120-127 (2004). 121. G. Zhou et al., Role of AMP-activated protein kinase in mechanism of metformin action. J Clin Invest 108, 1167-1174 (2001). 122. Y. Li et al., AMPK phosphorylates and inhibits SREBP activity to attenuate hepatic steatosis and atherosclerosis in diet-induced insulin-resistant mice. Cell metabolism 13, 376-388 (2011). 123. P. Zarogoulidis et al., mTOR pathway: A current, up-to-date mini-review (Review). Oncol Lett 8, 2367-2370 (2014). 124. K. Inoki, Y. Li, T. Zhu, J. Wu, K. L. Guan, TSC2 is phosphorylated and inhibited by Akt and suppresses mTOR signalling. Nat cell biol 4, 648-657 (2002). 125. E. Vander Haar, S. I. Lee, S. Bandhakavi, T. J. Griffin, D. H. Kim, Insulin signalling to mTOR mediated by the Akt/PKB substrate PRAS40. Nature cell biology 9, 316-323 (2007). 126. K. Inoki, T. Zhu, K. L. Guan, TSC2 mediates cellular energy response to control cell growth and survival. Cell 115, 577-590 (2003). 127. D. M. Gwinn et al., AMPK phosphorylation of raptor mediates a metabolic checkpoint. Mol Cell 30, 214-226 (2008). 128. J. Kaur, J. Debnath, Autophagy at the crossroads of catabolism and anabolism. Nat rev. Mol cell biol 16, 461-472 (2015). 129. J. Huang, D. J. Klionsky, Autophagy and human disease. Cell Cycle 6, 1837-1849 (2007). 130. J. L. Guan et al., Autophagy in stem cells. Autophagy 9, 830-849 (2013). 131. P. Strzyz, Autophagy: Nuclear autophagy in tumour suppression. Nat rev. Mol cell biol 16, 700-701 (2015). 132. G. Marino, M. Niso-Santano, E. H. Baehrecke, G. Kroemer, Self-consumption: the interplay of autophagy and apoptosis. Nat rev. Mol cell biol 15, 81-94 (2014). 133. D. J. Klionsky et al., A unified nomenclature for yeast autophagy-related genes. Dev Cell 5, 539-545 (2003). 134. S. Wesselborg, B. Stork, Autophagy signal transduction by ATG proteins: from hierarchies to networks. Cell Mol Life Sci 72, 4721-4757 (2015). 135. D. J. Klionsky et al., Guidelines for the use and interpretation of assays for monitoring autophagy (3rd edition). Autophagy 12, 1-222 (2016). 136. R. C. Russell, H. X. Yuan, K. L. Guan, Autophagy regulation by nutrient signaling. Cell research 24, 42-57 (2014). 137. G. E. Mortimore, C. M. Schworer, Induction of autophagy by amino-acid deprivation in perfused rat liver. Nature 270, 174-176 (1977). 138. E. F. Blommaart, J. J. Luiken, P. J. Blommaart, G. M. van Woerkom, A. J. Meijer, Phosphorylation of ribosomal protein S6 is inhibitory for autophagy in isolated rat hepatocytes. J biol chem 270, 2320-2326 (1995). 139. A. Iwamaru et al., Silencing mammalian target of rapamycin signaling by small interfering RNA enhances rapamycin-induced autophagy in malignant glioma cells. Oncogene 26, 1840- 1851 (2007). 140. J. Kim, M. Kundu, B. Viollet, K. L. Guan, AMPK and mTOR regulate autophagy through direct phosphorylation of Ulk1. Nat cell biol 13, 132-141 (2011). 141. D. F. Egan et al., Phosphorylation of ULK1 (hATG1) by AMP-activated protein kinase connects energy sensing to mitophagy. Science (New York, N.Y.) 331, 456-461 (2011).

112

142. P. B. Martinez de Morentin, A. Urisarri, M. L. Couce, M. Lopez, Molecular mechanisms of appetite and obesity: a role for brain AMPK. Clinical science (London, England : 1979) 130, 1697-1709 (2016). 143. Y. Minokoshi et al., AMP-kinase regulates food intake by responding to hormonal and nutrient signals in the hypothalamus. Nature 428, 569-574 (2004). 144. J. A. Villena et al., Induced adiposity and adipocyte hypertrophy in mice lacking the AMP- activated protein kinase-alpha2 subunit. Diabetes 53, 2242-2249 (2004). 145. M. Claret et al., AMPK is essential for energy homeostasis regulation and glucose sensing by POMC and AgRP neurons. J Clin Invest 117, 2325-2336 (2007). 146. D. G. Nicholls, E. Rial, A novel regulatory mechanism for the brown-fat uncoupling protein? Nat Struct Mol Biol 23, 364-365 (2016). 147. A. Herwig, A. W. Ross, K. N. Nilaweera, P. J. Morgan, P. Barrett, Hypothalamic thyroid hormone in energy balance regulation. Obes Facts 1, 71-79 (2008). 148. A. D. van Dam, S. Kooijman, M. Schilperoort, P. C. Rensen, M. R. Boon, Regulation of brown fat by AMP-activated protein kinase. Trends Mol Med 21, 571-579 (2015). 149. K. Eckel-Mahan, P. Sassone-Corsi, Metabolism and the circadian clock converge. Physiol Rev 93, 107-135 (2013). 150. K. A. Stokkan, S. Yamazaki, H. Tei, Y. Sakaki, M. Menaker, Entrainment of the circadian clock in the liver by feeding. Science (New York, N.Y.) 291, 490-493 (2001). 151. K. A. Lamia et al., AMPK regulates the circadian clock by cryptochrome phosphorylation and degradation. Science (New York, N.Y.) 326, 437-440 (2009). 152. B. Viollet et al., Cellular and molecular mechanisms of metformin: an overview. Clinical science (London, England : 1979) 122, 253-270 (2012). 153. K. Cusi, A. Consoli, R. A. DeFronzo, Metabolic effects of metformin on glucose and lactate metabolism in noninsulin-dependent diabetes mellitus. J Clin Endocrinol Metab 81, 4059- 4067 (1996). 154. R. S. Hundal et al., Mechanism by which metformin reduces glucose production in type 2 diabetes. Diabetes 49, 2063-2069 (2000). 155. A. Natali, E. Ferrannini, Effects of metformin and thiazolidinediones on suppression of hepatic glucose production and stimulation of glucose uptake in type 2 diabetes: a systematic review. Diabetologia 49, 434-441 (2006). 156. C. Wilcock, C. J. Bailey, Accumulation of metformin by tissues of the normal and diabetic mouse. Xenobiotica 24, 49-57 (1994). 157. E. Perland, R. Fredriksson, Classification Systems of Secondary Active Transporters. Trends Pharmacol Sci 38, 305-315 (2017). 158. M. Y. El-Mir et al., Dimethylbiguanide inhibits cell respiration via an indirect effect targeted on the respiratory chain complex I. J biol chem 275, 223-228 (2000). 159. M. R. Owen, E. Doran, A. P. Halestrap, Evidence that metformin exerts its anti-diabetic effects through inhibition of complex 1 of the mitochondrial respiratory chain. Biochem J 348 Pt 3, 607-614 (2000). 160. M. Zang et al., AMP-activated protein kinase is required for the lipid-lowering effect of metformin in insulin-resistant human HepG2 cells. J biol chem 279, 47898-47905 (2004). 161. M. D. Fullerton et al., Single phosphorylation sites in Acc1 and Acc2 regulate lipid homeostasis and the insulin-sensitizing effects of metformin. Nat med 19, 1649-1654 (2013). 162. H. Z. Lin et al., Metformin reverses fatty liver disease in obese, leptin-deficient mice. Nat med 6, 998-1003 (2000). 163. B. Cool et al., Identification and characterization of a small molecule AMPK activator that treats key components of type 2 diabetes and the metabolic syndrome. Cell metab 3, 403- 416 (2006). 164. O. Goransson et al., Mechanism of action of A-769662, a valuable tool for activation of AMP- activated protein kinase. J biol chem 282, 32549-32560 (2007).

113

165. M. J. Sanders, P. O. Grondin, B. D. Hegarty, M. A. Snowden, D. Carling, Investigating the mechanism for AMP activation of the AMP-activated protein kinase cascade. Biochem J 403, 139-148 (2007). 166. G. A. Higgs, J. A. Salmon, B. Henderson, J. R. Vane, Pharmacokinetics of aspirin and salicylate in relation to inhibition of arachidonate cyclooxygenase and antiinflammatory activity. PNAS 84, 1417-1420 (1987). 167. S. A. Hawley et al., The Ancient Drug Salicylate Directly Activates AMP-Activated Protein Kinase. Science (New York, N.Y.) 336, 918-922 (2012). 168. C. G. Langendorf, B. E. Kemp, Choreography of AMPK activation. Cell research 25, 5-6 (2015). 169. A. J. O'Brien et al., Salicylate activates AMPK and synergizes with metformin to reduce the survival of prostate and lung cancer cells ex vivo through inhibition of de novo lipogenesis. Biochem j 469, 177-187 (2015). 170. R. J. Ford et al., Metformin and salicylate synergistically activate liver AMPK, inhibit lipogenesis and improve insulin sensitivity. Biochem J, 468, 125-132 (2015). 171. M. Foretz et al., Metformin inhibits hepatic gluconeogenesis in mice independently of the LKB1/AMPK pathway via a decrease in hepatic energy state. J Clin Invest 120, 2355-2369 (2010). 172. S. Ducommun et al., Enhanced activation of cellular AMPK by dual-small molecule treatment: AICAR and A769662. Am J Physiol Endocrinol Metab 306, E688-696 (2014). 173. J. E. Sullivan, F. Carey, D. Carling, R. K. Beri, Characterisation of 5'-AMP-activated protein kinase in human liver using specific peptide substrates and the effects of 5'-AMP analogues on enzyme activity. Biochem Biophys Res Commun 200, 1551-1556 (1994). 174. J. M. Corton, J. G. Gillespie, S. A. Hawley, D. G. Hardie, 5-aminoimidazole-4-carboxamide ribonucleoside. A specific method for activating AMP-activated protein kinase in intact cells? Eur J Biochem 229, 558-565 (1995). 175. K. Fabianowska-Majewska, J. A. Duley, H. A. Simmonds, Effects of novel anti-viral adenosine analogues on the activity of S-adenosylhomocysteine hydrolase from human liver. Biochem Pharmacol 48, 897-903 (1994). 176. M. F. Vincent, M. D. Erion, H. E. Gruber, G. Van den Berghe, Hypoglycaemic effect of AICAriboside in mice. Diabetologia 39, 1148-1155 (1996). 177. L. J. Goodyear, The exercise pill--too good to be true? N Engl J Med 359, 1842-1844 (2008). 178. J. E. Gomez-Galeno et al., A Potent and Selective AMPK Activator That Inhibits de Novo Lipogenesis. ACS med chem letters 1, 478-482 (2010). 179. R. W. Hunter et al., Mechanism of action of compound-13: an alpha1-selective small molecule activator of AMPK. Chemistry & biology 21, 866-879 (2014). 180. F. Giordanetto, D. Karis, Direct AMP-activated protein kinase activators: a review of evidence from the patent literature. Expert Opin Ther Pat 22, 1467-1477 (2012). 181. R. Willows, N. Navaratnam, A. Lima, J. Read, D. Carling, Effect of different gamma-subunit isoforms on the regulation of AMPK. Biochem j 474, 1741-1754 (2017). 182. D. G. Hardie, AMPK: a target for drugs and natural products with effects on both diabetes and cancer. Diabetes 62, 2164-2172 (2013). 183. B. Viollet et al., AMPK inhibition in health and disease. Crit Rev Biochem Mol Biol 45, 276-295 (2010). 184. M. Kelly et al., AMPK activity is diminished in tissues of IL-6 knockout mice: the effect of exercise. Biochem Biophys Res Commun 320, 449-454 (2004). 185. Y. Wu, P. Song, J. Xu, M. Zhang, M. H. Zou, Activation of protein phosphatase 2A by palmitate inhibits AMP-activated protein kinase. J biol chem 282, 9777-9788 (2007). 186. M. Y. Wang, R. H. Unger, Role of PP2C in cardiac lipid accumulation in obese rodents and its prevention by troglitazone. Am J Physiol Endocrinol Metab 288, E216-221 (2005). 187. G. da Silva Xavier et al., Role for AMP-activated protein kinase in glucose-stimulated insulin secretion and preproinsulin gene expression. Biochem J 371, 761-774 (2003).

114

188. C. E. Gleason, D. Lu, L. A. Witters, C. B. Newgard, M. J. Birnbaum, The role of AMPK and mTOR in nutrient sensing in pancreatic beta-cells. J biol chem 282, 10341-10351 (2007). 189. I. P. Salt, G. Johnson, S. J. Ashcroft, D. G. Hardie, AMP-activated protein kinase is activated by low glucose in cell lines derived from pancreatic beta cells, and may regulate insulin release. Biochem J 335 ( Pt 3), 533-539 (1998). 190. Y. Ido, D. Carling, N. Ruderman, Hyperglycemia-induced apoptosis in human umbilical vein endothelial cells: inhibition by the AMP-activated protein kinase activation. Diabetes 51, 159-167 (2002). 191. G. da Silva Xavier et al., Role of AMP-activated protein kinase in the regulation by glucose of islet beta cell gene expression. PNAS 97, 4023-4028 (2000). 192. S. I. Itani et al., Glucose autoregulates its uptake in skeletal muscle: involvement of AMP- activated protein kinase. Diabetes 52, 1635-1640 (2003). 193. C. S. Zhang et al., Fructose-1,6-bisphosphate and aldolase mediate glucose sensing by AMPK. Nature, (2017). 194. W. Derave et al., Dissociation of AMP-activated protein kinase activation and glucose transport in contracting slow-twitch muscle. Diabetes 49, 1281-1287 (2000). 195. J. F. Wojtaszewski, S. B. Jorgensen, Y. Hellsten, D. G. Hardie, E. A. Richter, Glycogen- dependent effects of 5-aminoimidazole-4-carboxamide (AICA)-riboside on AMP-activated protein kinase and glycogen synthase activities in rat skeletal muscle. Diabetes 51, 284-292 (2002). 196. G. K. McConell et al., Differential attenuation of AMPK activation during acute exercise following exercise training or AICAR treatment. J Appl Physiol (1985) 105, 1422-1427 (2008). 197. J. N. Nielsen et al., Role of 5'AMP-activated protein kinase in glycogen synthase activity and glucose utilization: insights from patients with McArdle's disease. J Physiol 541, 979-989 (2002). 198. A. McBride, S. Ghilagaber, A. Nikolaev, D. G. Hardie, The glycogen-binding domain on the AMPK beta subunit allows the kinase to act as a glycogen sensor. Cell metab 9, 23-34 (2009). 199. I. Leclerc, G. A. Rutter, AMP-activated protein kinase: a new beta-cell glucose sensor?: Regulation by amino acids and calcium ions. Diabetes 53 Suppl 3, S67-74 (2004). 200. C. Culmsee, J. Monnig, B. E. Kemp, M. P. Mattson, AMP-activated protein kinase is highly expressed in neurons in the developing rat brain and promotes neuronal survival following glucose deprivation. J Mol Neurosci 17, 45-58 (2001). 201. N. Kuramoto et al., Phospho-dependent functional modulation of GABA(B) receptors by the metabolic sensor AMP-dependent protein kinase. Neuron 53, 233-247 (2007). 202. L. D. McCullough et al., Pharmacological inhibition of AMP-activated protein kinase provides neuroprotection in stroke. J biol chem 280, 20493-20502 (2005). 203. J. Li, Z. Zeng, B. Viollet, G. V. Ronnett, L. D. McCullough, Neuroprotective effects of adenosine monophosphate-activated protein kinase inhibition and gene deletion in stroke. Stroke 38, 2992-2999 (2007). 204. H. Motoshima, B. J. Goldstein, M. Igata, E. Araki, AMPK and cell proliferation--AMPK as a therapeutic target for atherosclerosis and cancer. J Physiol 574, 63-71 (2006). 205. D. G. Hardie, D. R. Alessi, LKB1 and AMPK and the cancer-metabolism link - ten years after. BMC Biol 11, 36 (2013). 206. S. L. Bowker, S. R. Majumdar, P. Veugelers, J. A. Johnson, Increased cancer-related mortality for patients with type 2 diabetes who use sulfonylureas or insulin. Diabetes Care 29, 254-258 (2006). 207. M. Buzzai et al., Systemic treatment with the antidiabetic drug metformin selectively impairs p53-deficient tumor cell growth. Cancer Res 67, 6745-6752 (2007). 208. R. M. Sutherland, Cell and environment interactions in tumor microregions: the multicell spheroid model. Science (New York, N.Y.) 240, 177-184 (1988).

115

209. G. Helmlinger, F. Yuan, M. Dellian, R. K. Jain, Interstitial pH and pO2 gradients in solid tumors in vivo: high-resolution measurements reveal a lack of correlation. Nat med 3, 177-182 (1997). 210. J. M. Brown, A. J. Giaccia, The unique physiology of solid tumors: opportunities (and problems) for cancer therapy. Cancer Res 58, 1408-1416 (1998). 211. J. M. Brown, Tumor microenvironment and the response to anticancer therapy. Cancer Biol Ther 1, 453-458 (2002). 212. K. Kato et al., Critical roles of AMP-activated protein kinase in constitutive tolerance of cancer cells to nutrient deprivation and tumor formation. Oncogene 21, 6082-6090 (2002). 213. Z. T. Schafer et al., Antioxidant and oncogene rescue of metabolic defects caused by loss of matrix attachment. Nature 461, 109-113 (2009). 214. J. T. Isaacs, W. B. Isaacs, Androgen receptor outwits prostate cancer drugs. Nat med 10, 26- 27 (2004). 215. C. D. Chen et al., Molecular determinants of resistance to antiandrogen therapy. Nat med 10, 33-39 (2004). 216. D. E. Frigo et al., CaM kinase kinase beta-mediated activation of the growth regulatory kinase AMPK is required for androgen-dependent migration of prostate cancer cells. Cancer Res 71, 528-537 (2011). 217. P. Popovics, D. E. Frigo, A. V. Schally, F. G. Rick, Targeting the 5'-AMP-activated protein kinase and related metabolic pathways for the treatment of prostate cancer. Expert Opin Ther Targets 19, 617-632 (2015). 218. A. Nakano et al., AMPK controls the speed of microtubule polymerization and directional cell migration through CLIP-170 phosphorylation. Nat cell biol 12, 583-590 (2010). 219. M. Domise, V. Vingtdeux, AMPK in Neurodegenerative Diseases. EXS 107, 153-177 (2016). 220. J. S. Won, Y. B. Im, J. Kim, A. K. Singh, I. Singh, Involvement of AMP-activated-protein-kinase (AMPK) in neuronal amyloidogenesis. Biochem Biophys Res Commun 399, 487-491 (2010). 221. T. Ma et al., Inhibition of AMP-activated protein kinase signaling alleviates impairments in hippocampal synaptic plasticity induced by amyloid beta. J Neurosci 34, 12230-12238 (2014). 222. G. Mairet-Coello et al., The CAMKK2-AMPK kinase pathway mediates the synaptotoxic effects of Abeta oligomers through Tau phosphorylation. Neuron 78, 94-108 (2013). 223. K. C. Duong-Ly, J. R. Peterson, The Human Kinome and Kinase Inhibition as a therapeutic strategy. Curr Protoc Pharmacol 0 2, Unit2 9 (2013). 224. D. G. Hardie, Regulation of AMP-activated protein kinase by natural and synthetic activators. Acta Pharm Sin B 6, 1-19 (2016). 225. R. Roskoski, Jr., Classification of small molecule protein kinase inhibitors based upon the structures of their drug-enzyme complexes. Pharmacol Res 103, 26-48 (2016). 226. A. C. Dar, K. M. Shokat, The evolution of protein kinase inhibitors from antagonists to agonists of cellular signaling. Annu Rev Biochem 80, 769-795 (2011). 227. S. R. Hubbard, L. Wei, L. Ellis, W. A. Hendrickson, Crystal structure of the tyrosine kinase domain of the human insulin receptor. Nature 372, 746-754 (1994). 228. Z. Zhao et al., Exploration of type II binding mode: A privileged approach for kinase inhibitor focused drug discovery? ACS Chem Biol 9, 1230-1241 (2014). 229. L. Gavrin, E. Saiah, Approaches to discover non-ATP site kinase inhibitors. MedChemComm 4, 41-51 (2013). 230. V. Lamba, I. Ghosh, New directions in targeting protein kinases: focusing upon true allosteric and bivalent inhibitors. Curr Pharm Des 18, 2936-2945 (2012). 231. J. J. Liao, Molecular recognition of protein kinase binding pockets for design of potent and selective kinase inhibitors. J med chem 50, 409-424 (2007). 232. S. Omura et al., A new alkaloid AM-2282 OF Streptomyces origin. Taxonomy, fermentation, isolation and preliminary characterization. J Antibiot (Tokyo) 30, 275-282 (1977).

116

233. T. Tamaoki et al., Staurosporine, a potent inhibitor of phospholipid/Ca++dependent protein kinase. Biochem Biophys Res Commun 135, 397-402 (1986). 234. A. M. Lawrie et al., Protein kinase inhibition by staurosporine revealed in details of the molecular interaction with CDK2. Nat Struct Biol 4, 796-801 (1997). 235. M. W. Karaman et al., A quantitative analysis of kinase inhibitor selectivity. Nat Biotechnol 26, 127-132 (2008). 236. Y. Liu, N. S. Gray, Rational design of inhibitors that bind to inactive kinase conformations. Nat Chem Biol 2, 358-364 (2006). 237. J. Bain et al., The selectivity of protein kinase inhibitors: a further update. Biochem j 408, 297-315 (2007). 238. J. Vogt, R. Traynor, G. P. Sapkota, The specificities of small molecule inhibitors of the TGFss and BMP pathways. Cell Signal 23, 1831-1842 (2011). 239. L. G. Fryer, A. Parbu-Patel, D. Carling, Protein kinase inhibitors block the stimulation of the AMP-activated protein kinase by 5-amino-4-imidazolecarboxamide riboside. FEBS Lett 531, 189-192 (2002). 240. E. Yamada et al., Fyn phosphorylates AMPK to inhibit AMPK activity and AMP-dependent activation of autophagy. Oncotarget 7, 74612-74629 (2016). 241. F. A. Ross et al., Mechanisms of Paradoxical Activation of AMPK by the Kinase Inhibitors SU6656 and Sorafenib. Cell Chem Biol 24, 813-824 e814 (2017). 242. T. Schindler et al., Structural mechanism for STI-571 inhibition of abelson tyrosine kinase. Science (New York, N.Y.) 289, 1938-1942 (2000). 243. J. W. Scott et al., Inhibition of AMP-Activated Protein Kinase at the Allosteric Drug-Binding Site Promotes Islet Insulin Release. Chemistry & biology 22, 705-711 (2015). 244. D. G. Hardie, D. Carling, A. T.R. Sim, The AMP-activated protein kinase: a multisubstrate regulator of lipid metabolism. Trends in Biochem Sci 14, 20-23 (1989). 245. C. S. Zhang et al., Metformin Activates AMPK through the Lysosomal Pathway. Cell metabolism 24, 521-522 (2016). 246. K. R. Laderoute et al., 5'-AMP-activated protein kinase (AMPK) is induced by low-oxygen and glucose deprivation conditions found in solid-tumor microenvironments. Mol Cell Biol 26, 5336-5347 (2006). 247. K. Sakamoto et al., Deficiency of LKB1 in skeletal muscle prevents AMPK activation and glucose uptake during contraction. EMBO J 24, 1810-1820 (2005). 248. K. Sakamoto et al., Deficiency of LKB1 in heart prevents ischemia-mediated activation of AMPKalpha2 but not AMPKalpha1. Am J Physiol Endocrinol Metab 290, E780-788 (2006). 249. F. R. Auciello, F. A. Ross, N. Ikematsu, D. G. Hardie, Oxidative stress activates AMPK in cultured cells primarily by increasing cellular AMP and/or ADP. FEBS Lett 588, 3361-3366 (2014). 250. A. Priebe et al., Glucose deprivation activates AMPK and induces cell death through modulation of Akt in ovarian cancer cells. Gynecol Oncol 122, 389-395 (2011). 251. A. M. Brown, B. R. Ransom, Astrocyte glycogen and brain energy metabolism. Glia 55, 1263- 1271 (2007). 252. B. Brunmair et al., Thiazolidinediones, like metformin, inhibit respiratory complex I: a common mechanism contributing to their antidiabetic actions? Diabetes 53, 1052-1059 (2004). 253. A. C. Nulton-Persson, L. I. Szweda, H. A. Sadek, Inhibition of cardiac mitochondrial respiration by salicylic acid and acetylsalicylate. J Cardiovasc Pharmacol 44, 591-595 (2004). 254. M. M. Krause et al., Nonsteroidal antiinflammatory drugs and a selective cyclooxygenase 2 inhibitor uncouple mitochondria in intact cells. Arthritis Rheum 48, 1438-1444 (2003). 255. S. A. Hawley et al., Use of cells expressing gamma subunit variants to identify diverse mechanisms of AMPK activation. Cell metab 11, 554-565 (2010).

117

256. K. J. Jeong, G. W. Kim, S. H. Chung, AMP-activated protein kinase: An emerging target for ginseng. J Ginseng Res 38, 83-88 (2014). 257. L. Shen et al., Ginsenoside Rb1 reduces fatty liver by activating AMP-activated protein kinase in obese rats. J Lipid Res 54, 1430-1438 (2013). 258. Q. W. Shen, M. J. Zhu, J. Tong, J. Ren, M. Du, Ca2+/calmodulin-dependent protein kinase kinase is involved in AMP-activated protein kinase activation by alpha-lipoic acid in C2C12 myotubes. Am J Physiol Cell Physiol 293, C1395-1403 (2007). 259. D. Moreno, E. Knecht, B. Viollet, P. Sanz, A769662, a novel activator of AMP-activated protein kinase, inhibits non-proteolytic components of the 26S proteasome by an AMPK- independent mechanism. FEBS Lett 582, 2650-2654 (2008). 260. Y. C. Lai et al., A small-molecule benzimidazole derivative that potently activates AMPK to increase glucose transport in skeletal muscle: comparison with effects of contraction and other AMPK activators. The Biochem j 460, 363-375 (2014). 261. H. Azuma, K. Banno, T. Yoshimura, Pharmacological properties of N-(3',4'- dimethoxycinnamoyl) anthranilic acid (N-5'), a new anti-atopic agent. Br J Pharmacol 58, 483-488 (1976). 262. S. Darakhshan, A. Ghanbari, Tranilast enhances the anti-tumor effects of tamoxifen on human breast cancer cells in vitro. J Biomed Sci 20, 76 (2013). 263. D. G. Hardie, S. A. Hawley, AMP-activated protein kinase: the energy charge hypothesis revisited. Bioessays 23, 1112-1119 (2001). 264. R. A. Howlett et al., Regulation of skeletal muscle glycogen phosphorylase and PDH at varying exercise power outputs. Am J Phys 275, R418-425 (1998). 265. A. G. Chapman, A. L. Miller, D. E. Atkinson, Role of the adenylate deaminase reaction in regulation of adenine nucleotide metabolism in Ehrlich ascites tumor cells. Cancer Res 36, 1144-1150 (1976). 266. V. Bianchi, J. Spychala, Mammalian 5'-nucleotidases. J biol chem 278, 46195-46198 (2003). 267. S. S. Kulkarni et al., Suppression of 5'-nucleotidase enzymes promotes AMP-activated protein kinase (AMPK) phosphorylation and metabolism in human and mouse skeletal muscle. J biol chem 286, 34567-34574 (2011). 268. C. Plaideau et al., Overexpression of AMP-metabolizing enzymes controls adenine nucleotide levels and AMPK activation in HEK293T cells. FASEB J 26, 2685-2694 (2012). 269. C. Plaideau et al., Effects of pharmacological AMP deaminase inhibition and Ampd1 deletion on nucleotide levels and AMPK activation in contracting skeletal muscle. Chemistry & biology 21, 1497-1510 (2014). 270. J. Ouyang, R. A. Parakhia, R. S. Ochs, Metformin activates AMP kinase through inhibition of AMP deaminase. J biol chem 286, 1-11 (2011). 271. D. R. Janero, C. Yarwood, Oxidative modulation and inactivation of rabbit cardiac adenylate deaminase. Biochem j 306 ( Pt 2), 421-427 (1995). 272. J. V. Passonneau, O. H. Lowry, Phosphofructokinase and the Pasteur effect. Biochem Biophys Res Commun 7, 10-15 (1962). 273. G. Zifarelli, M. Pusch, Intracellular regulation of human ClC-5 by adenine nucleotides. EMBO Rep 10, 1111-1116 (2009). 274. P. F. Cohen, R. F. Colman, Diphosphopyridine nucleotide dependent isocitrate dehydrogenase from pig heart. Charactgerization of the active substrate and modes of regulation. Biochem 11, 1501-1508 (1972). 275. A. A. Dunn-Meynell et al., Relationship among brain and blood glucose levels and spontaneous and glucoprivic feeding. J Neurosci 29, 7015-7022 (2009). 276. S. S. Sabharwal, P. T. Schumacker, Mitochondrial ROS in cancer: initiators, amplifiers or an Achilles' heel? Nat Rev Cancer 14, 709-721 (2014). 277. C. I. Rousset et al., A dual role for AMP-activated protein kinase (AMPK) during neonatal hypoxic-ischaemic brain injury in mice. J neurochem 133, 242-252 (2015).

118

278. P. Steinbacher, P. Eckl, Impact of Oxidative Stress on Exercising Skeletal Muscle. Biomolecules 5, 356-377 (2015). 279. A. J. Trewin, B. J. Berry, A. P. Wojtovich, Exercise and Mitochondrial Dynamics: Keeping in Shape with ROS and AMPK. Antioxidants (Basel) 7, (2018). 280. M. J. Sanders et al., Defining the mechanism of activation of AMP-activated protein kinase by the small molecule A-769662, a member of the thienopyridone family. J biol chem 282, 32539-32548 (2007). 281. G. Zadra et al., A novel direct activator of AMPK inhibits prostate cancer growth by blocking lipogenesis. EMBO Mol Med 6, 519-538 (2014). 282. U. S. N. L. o. Medicine. (2015), vol. 2018. NCT02286882 283. B. J. Druker, Molecularly targeted therapy: have the floodgates opened? Oncologist 9, 357- 360 (2004). 284. P. Wu, T. E. Nielsen, M. H. Clausen, Small-molecule kinase inhibitors: an analysis of FDA- approved drugs. Drug Discov Today 21, 5-10 (2016). 285. Z. A. Knight, K. M. Shokat, Features of selective kinase inhibitors. Chemistry & biology 12, 621-637 (2005). 286. R. Ke, Q. Xu, C. Li, L. Luo, D. Huang, Mechanisms of AMPK in the maintenance of ATP balance during energy metabolism. Cell Biol Int, 42, 384-392 (2017). 287. S. Darakhshan, A. B. Pour, Tranilast: a review of its therapeutic applications. Pharmacol Res 91, 15-28 (2015). 288. D. G. Hardie, The AMP-activated protein kinase pathway--new players upstream and downstream. Journal of cell science 117, 5479-5487 (2004). 289. D. G. Hardie, AMP-activated protein kinase: a cellular energy sensor with a key role in metabolic disorders and in cancer. Biochem Soc Trans 39, 1-13 (2011). 290. S. C. Lin, D. G. Hardie, AMPK: Sensing Glucose as well as Cellular Energy Status. Cell metab, 27, 299-313 (2017). 291. D. G. Hardie, Keeping the home fires burning: AMP-activated protein kinase. J R Soc Interface 15, (2018). 292. D. B. Shackelford, R. J. Shaw, The LKB1-AMPK pathway: metabolism and growth control in tumour suppression. Nat Rev Cancer 9, 563-575 (2009). 293. S. Shen et al., Metabolomic Analysis of Mouse Embryonic Fibroblast Cells in Response to Autophagy Induced by Acute Starvation. Sci Rep 6, 34075 (2016). 294. J. H. Jung, J. A. Traugh, Regulation of the interaction of Pak2 with Cdc42 via autophosphorylation of serine 141. J biol chem 280, 40025-40031 (2005). 295. R. Jakobi, E. Moertl, M. A. Koeppel, p21-activated protein kinase gamma-PAK suppresses programmed cell death of BALB3T3 fibroblasts. J biol chem 276, 16624-16634 (2001). 296. B. Faubert et al., AMPK is a negative regulator of the Warburg effect and suppresses tumor growth in vivo. Cell metab 17, 113-124 (2013). 297. R. J. Kishton et al., AMPK Is Essential to Balance Glycolysis and Mitochondrial Metabolism to Control T-ALL Cell Stress and Survival. Cell metab 23, 649-662 (2016). 298. E. Perez-Herrero, A. Fernandez-Medarde, Advanced targeted therapies in cancer: Drug nanocarriers, the future of chemotherapy. Eur J Pharm Biopharm 93, 52-79 (2015). 299. I. K. Vila et al., A UBE2O-AMPKalpha2 Axis that Promotes Tumor Initiation and Progression Offers Opportunities for Therapy. Cancer Cell 31, 208-224 (2017).

119

Minerva Access is the Institutional Repository of The University of Melbourne

Author/s: Dite, Toby

Title: Mechanisms of activation and inhibition of the metabolic regulator AMP-activated protein kinase (AMPK)

Date: 2018

Persistent Link: http://hdl.handle.net/11343/214389

File Description: Mechanisms of activation and inhibition of the metabolic regulator AMP-activated protein kinase (AMPK)

Terms and Conditions: Terms and Conditions: Copyright in works deposited in Minerva Access is retained by the copyright owner. The work may not be altered without permission from the copyright owner. Readers may only download, print and save electronic copies of whole works for their own personal non-commercial use. Any use that exceeds these limits requires permission from the copyright owner. Attribution is essential when quoting or paraphrasing from these works.