i Structural and Functional Characterisation of the Nutrient Sensing Kinase GCN2 Alison J. Inglis This dissertation is submitted to the University of Cambridge for the degree of Doctor of Philosophy Downing College March 2018 ii Abstract Title: Structural and Functional Characterisation of the Nutrient Sensing Kinase GCN2 Author: Alison J. Inglis A cell’s ability to recognise and respond to changes in its environment is crucial to its survival. The availability of nutrients is a fundamental part of the environment, and so cells must be able to identify when they are plentiful and when they are scarce, and adapt accordingly. The kinase GCN2 is a key protein within the eukaryotic proteome and is activated in response to a drop in the intracellular concentration of amino acids. Upon activation, GCN2 phosphorylates the translation initiation factor eIF2. Phosphorylation of this factor initiates the Integrated Stress Response, which causes a global inhibition of protein translation whilst upregulating the expression of stress response genes. Under conditions of amino acid starvation, the activation of GCN2 is an important step in mediating the cell’s response, allowing it to adapt to the new conditions. GCN2 has been implicated in a wide range of cellular processes in both health and disease, including the development of neurological disorders and cancer. Despite the important role played by GCN2, the molecular mechanisms that control its regulation and activation remain unclear. Genetic experiments on the yeast homologue have provided some insights, but the highly complex and interconnected nature of the regulatory pathways mean that it is difficult to disentangle the precise mechanistic details in a cellular context. Furthermore, there is evidence for significant differences between the yeast and mammalian proteins, meaning that it is not possible to simply equate the two. For these reasons, the aim of this project was to investigate the structural and functional characteristics of the human GCN2 kinase, and to examine the molecular mechanisms that enable it to act as a cellular sensor of nutritional stress. This thesis initially describes the development of a system to reconstitute GCN2 activation using purified components. This allowed the effects of different regulators to be tested, and identified the ribosome as a potent activator of the kinase. A direct interaction between GCN2 and mammalian ribosomes was identified and characterised by truncation analysis. To identify the site on the ribosome to which GCN2 binds, an extensive analysis of the Abstract iii interaction was performed using Hydrogen-Deuterium exchange-mass spectrometry (HDX- MS). This demonstrated that GCN2 binds to the uL10 protein of the ribosomal P stalk, directly adjacent to the A site. These data taken together allow the proposal of a model concerning how GCN2 senses amino acid deprivation in a cellular context through the surveillance of translational processivity. Further structural insights into this process have been gained using a combination of X-ray crystallography and electron microscopy. These analyses resulted in a high-resolution crystal structure of the pseudokinase domain of GCN2, representing a first glimpse into the structure of human GCN2, and low-resolution insights into the full-length kinase from negative stain electron microscopy. Whilst GCN2 is activated by nutritional stress, mammalian cells have evolved a panoply of responses to environmental stress. Hsp90 is a chaperone that is required for the stability and maintenance of approximately 60 % of the human kinome, yet the mechanism by which it recognises client kinases is still unclear. The final chapter of this thesis describes the use of biochemical methods in combination with HDX-MS to characterise the interactions between Hsp90’s co-chaperone Cdc37 and client kinases. These analyses enabled the identification of a correlation between protein stability and dependence on Hsp90/Cdc37, and revealed that Cdc37 binding causes a dramatic conformational remodelling of the N-lobe of the kinase. iv Preface This dissertation is the result of work carried out at the MRC Laboratory of Molecular Biology between October 2014 and March 2018. This dissertation is the result of my own work and includes nothing that is the outcome of work done in collaboration, except as specified in the text. It is not substantially the same as any that I have submitted, or is being concurrently submitted, for a degree, diploma or any other qualification at the University of Cambridge or any other University or similar institution. I further state that no substantial part of my dissertation has already been submitted, or is being concurrently submitted, for any such degree, diploma or other qualification at the University of Cambridge or any other similar institution. This dissertation does not exceed 60,000 words. Alison J. Inglis March 2018 v Acknowledgements Throughout my PhD, I have been endlessly grateful to all the people who have donated their time and knowledge to help me achieve my aims. I could not have done this alone. First and foremost, I would like to thank Roger Williams, who took me in as an undergraduate summer student and has helped me every step of the way. He has provided tremendous academic support, and without fail has stepped in whenever needed. I would also like to thank Olga, without whom I’m not sure I would have ever got anything done. Her boundless knowledge and conscientiousness have kept me on track, and I am extraordinarily grateful. I would like to thank everyone who has been a member of the Williams lab throughout my PhD for all their help with experiments, presentations and eating up extra cake. Especial thanks to Glenn for helping me with many HDX experiments on the mass spectrometer, and for being the recipient of most of my inane questions over the years. I would also like to thank the many people within the LMB who have helped me at every stage of my work. I am especially grateful to Susan Shao, who prompted me to consider a new perspective on my project, and who has been an ever generous and helpful collaborator. I’d also like to extend my thanks to Stephen McLaughlin and Chris Johnson for all their help in the Biophysics Facility, and to Christos Savva, Shaoxia Chen and Giuseppe Cannone for training me in cryo-electron microscopy. I am very grateful to my supervisors Manu Hegde and Tom Blundell, who have been extremely supportive throughout my PhD. I’d also like to thank my collaborator Tom Bunney at U.C.L. for letting me get involved in his work on chaperones, and for his helpful comments on this thesis. I owe a debt of gratitude to John Burke, who took me under his wing during my time as a summer student. John showed kindness, patience and humour whilst I naïvely miscalculated buffer recipes, dropped samples on the floor and forgot how to clean an ion exchange column. He showed me that research could be fun and infinitely sociable (in stark contrast to my undergraduate practicals) and is probably the main reason I was able to pursue a PhD. I would like to thank my friends and flatmates, who have been constantly supportive throughout my PhD. They have cheered me up when it was all going horribly wrong and Acknowledgements vi pinned my Western blot to the fridge when it worked. I don’t know what I would have done without the stair tea, debriefs, kayaking trips and family Sundays. Lastly, I would like to thank my family. They have continued to believe that I was capable when I wasn’t so sure, and their unwavering support has kept me going. Chris has been unerringly enthusiastic and has helped me maintain my sense of humour throughout, even after the microscope crashed after eight hours of alignments. I’d like to thank my brothers for all their encouragement, and for providing an escape to London whenever required. Finally, none of this would have been possible without my parents. They have constantly inspired me, propped me up during the tough bits and celebrated with me during the successes. It is to them that I dedicate this thesis, in thanks for all they have done. vii Abbreviations and Nomenclatures 4EBP = eIF4E binding protein ABC = ATP-binding cassette ADA = 2-[(2-Amino-2-oxoethyl)-(carboxymethyl)amino]acetic acid AMPPNP = Adenylyl imidodiphosphate ATF4 = Activating Transcription Factor 4 ATP = Adenosine triphosphate BiP = Binding Immunoglobulin Protein BSA = Bovine Serum Albumin CAPS = 3-(Cyclohexylamino)-1-propanesulfonic acid Cdc37 = Cell division cycle protein 37 Cdk4 = Cyclin-dependent kinase 4 cDNA = Complementary DNA CHAPS = 3-[(3-Cholamidopropyl)dimethylammonio]-1-propanesulfonate CHX = Cycloheximide CTD = C-terminal domain CTP = Cytidine triphosphate DAG = Diacylglycerol DDB = Didemnin B DDM = Dodecyl-β-maltoside DeAc tRNA = Deacylated tRNA DEAD = Aspartate-Glutamate-Alanine-Aspartate DEP = Dishevelled, EGL-10, Pleckstrin DEPTOR = DEP domain-containing mTOR-interacting protein DNA = Deoxyribonucleic acid DSF = Differential scanning fluorimetry dsRBD = Double stranded RNA-binding domain dsRNA = Double stranded RNA DTB = Desthiobiotin DTT = Dithiothreitol EDC = 1-Ethyl-3-(3-dimethylaminopropyl)-carbodiimide Abbreviations and Nomenclatures viii EDTA = Ethylenediaminetetraacetic acid eEF1A = Eukaryotic elongation factor 1A eEF2 = Eukaryotic elongation factor 2 eEF3 = Eukaryotic elongation factor 3 EGFR = Epidermal Growth Factor Receptor EGTA = Ethylene glycol-bis(β-aminoethyl ether)-N,N,N’,N’-tetraacetic acid eIF1 = Eukaryotic initiation factor 1 eIF1A = Eukaryotic initiation factor 1A eIF2 = Eukaryotic initiation factor 2 eIF2B = Eukaryotic initiation factor 2B eIF3 = Eukaryotic initiation factor 3 eIF4A = Eukaryotic initiation factor 4A eIF4E = Eukaryotic initiation factor 4E eIF4F = Eukaryotic initiation factor 4F eIF4G = Eukaryotic initiation factor 4G eIF5 = Eukaryotic initiation factor 5 EM = Electron microscopy ER = Endoplasmic reticulum eRF1 = Eukaryotic release factor 1 ESRF = European Synchrotron Radiation Facility FEG = Field emission gun FGFR = Fibroblast Growth Factor Receptor FRS2 = FGFR substrate 2 FSC = Fourier correlation shell GAAC = General Amino Acid Control GAB1 = GRB2-Associated Binding Protein 1 GCN1 = General Control Nonderepressible 1 GCN2 = General Control Nonderepressible 2 GCN20 = General Control Nonderepressible 20 GCN4 = General Control Nonderepressible 4 GDP = Guanosine diphosphate GEF = Guanidine nucleotide exchange factor GRB2 = Growth Factor Receptor-Bound 2 Abbreviations and Nomenclatures ix GST = Glutathione-S-transferase GTP = Guanosine triphosphate GTPBP2 = GTP binding protein 2 HBS = HEPES-buffered saline HDX-MS = Hydrogen-Deuterium exchange-mass spectrometry HEAT = Huntingtin, eEF3, protein phosphatase 2A, mTOR HEPES = 4-(2-Hydroxyethyl)-1-piperazineethanesulfonic acid HisRS-like = Histidyl tRNA synthetase-like HIV = Human Immunodeficiency Virus HRI = Heme-Regulated Inhibitor Hsp90 = Heat shock protein 90 IF2 = Initiation factor 2 (Archaeal) Ig = Immunoglobulin IMPACT = Imprinted Gene with Ancient Domain Protein IMS = Ion mobility spectrometry IP3 = Inositol-1,4,5-trisphosphate IPTG = Isopropyl β-D-1-thiogalactopyranoside IRE = Inositol-Requiring Enzyme ISR = Integrated Stress Response KD = Kinase domain LB = Lysogeny broth LDS = Lithium dodecyl sulphate LMNG = Lauryl maltose neopentyl glycol m2 = Motif 2 m7G = 7-Methylguanylate MAPK = Mitogen-Activated Protein Kinase Met-tRNAiMet = Initiator methionine tRNA MHC = Major Histocompatibility Complex MLST8 = Mammalian Lethal with Sec13 protein 8 mRNA = Messenger RNA mTOR = Mammalian Target of Rapamycin mTORC1 = mTOR Complex 1 mTORC2 = mTOR Complex 2 Abbreviations and Nomenclatures x NHS = N-Hydroxysuccinimide NMR = Nuclear magnetic resonance ORF = Open reading frame PABP = Polyadenylate Binding Protein PBS = Phosphate-buffered saline PCR = Polymerase chain reaction PDB = Protein data bank PEG = Polyethylene glycol PERK = Protein Kinase R-like Endoplasmic Reticulum Kinase PI3K = Phosphatidylinositol-3-kinase PIC = Preinitiation complex PIP2 = Phosphatidylinositol-4,5-bisphosphate PIP3 = Phosphatidylinositol-3,4,5-trisphosphate PKC = Protein Kinase C PKR = Protein Kinase R PLC = Phospholipase C PRAS40 = Proline-rich AKT substrate 1 PVDF = Polyvinylidene difluoride PVOD = Pulmonary Veno-Occlusive Disease RAPTOR = Regulatory-Associated Protein of mTOR RHEB = Ras-Homologue Enriched in Brain RICTOR = Rapamycin-insensitive companion of mTOR (RICTOR) RING = Really Interesting New Gene RNA = Ribonucleic acid RRL = Rabbit reticulocyte lysate rRNA = Ribosomal RNA RWD = RING finger proteins, WD repeat-containing proteins, yeast DEAD-like helicases SAXS = Small-Angle X-ray Scattering SDS = Sodium dodecyl sulphate SDS-PAGE = SDS-Polyacrylamide gel electrophoresis SEC-MALS = Size-exclusion chromatography-multi angle light scattering SH2 = Src Homology 2 SLBP = Stem-Loop Binding Protein Abbreviations and Nomenclatures xi SLFN14 = Schlafen Family Member 14 SOS = Son of Sevenless SPR = Surface plasmon resonance TCEP = Tris(2-carboxyethyl)phosphine TEM = Transmission electron microscope TEV = Tobacco Etch Virus Tm = Melting temperature Tris = Tris(hydroxymethyl)aminomethane tRNA = Transfer RNA TSC = Tuberous Sclerosis Complex TYE = Tryptone yeast extract UBC9 = E2 Ubiquitin Conjugating Enzyme uORF = Upstream open reading frame UPLC = Ultra-performance liquid chromatography UTP = Uridine triphosphate UTR = Untranslated region UV = Ultraviolet v-ATPase = vacuolar ATPase VHP = Villin Headpiece Domain WD = Tryptophan-Aspartate Yih1 = Yeast IMPACT Homologue 1 YKD = Pseudokinase Domain All amino acids are abbreviated with standard three or one letter codes. xii Table of Contents Preface ii Acknowledgements v Abstract Error! Bookmark not defined. Abbreviations and Nomenclatures vii Table of Contents xii List of Figures and Tables xvi Figures xvi Tables xix Chapter One – Introduction 1 Protein Translation 1 1.1. The Integrated Stress Response 5 1.2. The eIF2α Kinases 7 1.3. Amino Acid Sensing 10 1.4. Structure and Function of the kinase GCN2 14 1.5. Regulators of GCN2 22 1.6. 1.6.1. Deacylated tRNA 22 1.6.2. GCN1 24 1.6.3. GCN20 26 1.6.4. IMPACT 27 1.6.5. The Ribosome 29 1.6.6. Translational Elongation Factors 32 A Model for the Activation of GCN2 33 1.7. Aims of this Thesis 35 Chapter Two – Reconstitution of GCN2 kinase activation with purified factors 37 2.1. Introduction 37 2.2. Materials and Methods 38 2.2.1. Protein Expression and Purification 38 2.2.2. Buffer Optimisation by Differential Scanning Fluorimetry 41 2.2.3. Size-Exclusion Chromatography-Multi Angle Light Scattering 41 2.2.4. Surface Plasmon Resonance 42 Table of Contents xiii 2.2.5. Protein-Protein Interaction Pull-Downs 42 2.2.6. Ribosome Purification 42 2.2.7. Radiolabelled Autophosphorylation Assay 43 2.2.8. Western Blot Autophosphorylation Assay 44 2.2.9. Phosphorylation Analysis by Mass Spectrometry 44 2.2.10. eIF2α Phosphorylation Assay 45 2.2.11. Ribosomal Co-migration Assays 46 2.2.12. Ribosomal Pull-Down Assays 47 2.2.13. Gel Filtration of Rabbit Reticulocyte Lysate 47 2.2.14. Mass Spectrometry Analysis of GCN2-interacting proteins 47 2.2.15. Treatment of Rabbit Reticulocyte Lysate with Micrococcal Nuclease 48 2.3. Results and Discussion 49 2.3.1. Expression and Purification of GCN2 49 2.3.2. Biophysical Characterisation of Human GCN2 54 2.3.2.1. Buffer Optimisation by Differential Scanning Fluorimetry 54 2.3.2.2. SEC-MALS 57 2.3.2.3. Surface Plasmon Resonance 58 2.3.3. Testing Potential Interacting Partners of GCN2 60 2.3.3.1. Purification of GCN1 and GCN20 60 2.3.3.2. Identification of protein-protein interactions 62 2.3.4. Functional Analysis of Purified GCN2 64 2.3.4.1. Autophosphorylation Assays 64 2.3.4.2. The effect of GCN1 and GCN20 65 2.3.4.3. Identifying the phosphorylation sites 67 2.3.4.4. Testing the importance of specific phosphorylations 70 2.3.4.5. Purification of eIF2α 71 2.3.4.6. eIF2α Phosphorylation Assay 74 2.3.4.7. Identification of an interaction between GCN2 and Ribosomes 76 2.3.4.8. Domain Mapping of GCN2 84 2.3.4.9. Purification of a GCN2 Truncation Library 84 2.3.4.10. SEC-MALS Analysis of the Construct Library 86 2.3.4.11. Autophosphorylation Activity of the Construct Library 88 2.3.4.12. eIF2α Phosphorylation by the Construct Library 89 2.3.4.13. Ribosome Binding by the Construct Library 91 2.3.4.14. Investigating specificity of activation 93 2.4. Conclusions 96 Chapter Three – Characterisation of the interaction between GCN2 and the ribosome using HDX-MS 101 Table of Contents xiv 3.1. Introduction 101 3.2 Materials and Methods 109 3.2.1. Initial GCN2-Ribosome Interaction Analysis by HDX-MS 109 3.2.2. Optimised GCN2-Ribosome Interaction Analysis by HDX-MS 110 3.3. Results and Discussion 111 3.3.1. Optimisation of HDX-MS to allow the study of ribosomal interactors 111 3.4. Conclusions 119 Chapter Four – Structural Insights into GCN2 120 4.1. Introduction 120 4.2. Materials and Methods 121 4.2.1. Crystallisation Trials 121 4.2.2. Hydrogen-Deuterium Exchange-Mass Spectrometry 121 4.2.3. Limited Proteolysis 121 4.2.4. Nanobody Production 122 4.2.5. Nanobody Interaction Analysis by Pull-Downs 123 4.2.6. Purification of GCN2-Nanobody Complexes 124 4.2.7. Crystallisation of the Pseudokinase domain 124 4.2.8. Negative Stain Electron Microscopy 125 4.2.9. Cryo-Electron Microscopy Sample Preparation 125 4.2.10. GraFix 126 4.2.11. Assembly of GCN2-Ribosome Complexes 127 4.2.12. Cryo-Electron Microscopy Data Collection 133 4.2.13. Cryo-Electron Microscopy Data Processing 133 4.3. Results and Discussion 134 4.3.1. Crystallography of full-length GCN2 134 4.3.1.1. Intrinsic Disorder Analysis by HDX-MS 134 4.3.1.2. Limited Proteolysis 135 4.3.1.3. Nanobodies 136 4.3.2. Crystallography of the Pseudokinase domain 143 4.3.3. Electron Microscopy of GCN2 148 4.3.4. Electron Microscopy of the GCN2-Ribosome Complex 154 4.4. Conclusions 162 Chapter Five – Investigation of the conformational remodelling of kinases by the Hsp90 co- chaperone Cdc37 using HDX-MS 167 5.1. Introduction 167 5.2. Materials and Methods 173 Table of Contents xv 5.2.1. HDX-MS 173 5.3. Results and Discussion 175 5.3.1. Reconstitution of a FGFR3-Cdc37-Hsp90 complex 175 5.3.2. Characterisation of different FGFR3 mutations 176 5.3.3. Investigation of the effects of Cdc37 binding to FGFR3 181 5.3.4. Investigation of the effects of FGFR3 binding to Cdc37 186 5.4. Conclusions 190 Conclusions and Future Directions 194 References 199 Supplementary Materials 221 xvi List of Figures and Tables Figures Figure 1.1. The canonical translational machinery 2 Figure 1.2. The mechanism of ATF4 upregulation by eIF2α phosphorylation 6 Figure 1.3. The four eIF2α kinases 7 Figure 1.4. The domain structures of the eIF2α kinases 8 Figure 1.5. The activation mechanism for the kinase PKR 9 Figure 1.6. The ‘line-up’ activation mechanism for the kinase PERK 10 Figure 1.7. The activation of mTORC1 by amino acids 12 Figure 1.8. The model for GCN2 activation by deacylated tRNA 14 Figure 1.9. The domain structure of human GCN2 16 Figure 1.10. The structure of the RWD domain from mouse GCN2 17 Figure 1.11. The structure of the kinase domain from yeast Gcn2 18 Figure 1.12. The structure of the CTD from mouse and yeast GCN2 20 Figure 1.13. A model for the reorganisation of the domains of GCN2 upon activation 22 Figure 1.14. The domain structure of human GCN1 25 Figure 1.15. The domain structure of Yih1 27 Figure 1.16. The structure of the archaeal P stalk complex 31 Figure 1.17. Schematic of the effect of regulators on GCN2 activation 34 Figure 2.1. Purification of His6-tagged GCN2 50 Figure 2.2. Purification of GST-tagged GCN2 52 Figure 2.3. Purification of Strep-tagged GCN2 53 Figure 2.4. Comparison of GCN2 stability at different pH values by DSF 54 Figure 2.5. Identification of the optimum pH for GCN2 stability by DSF 55 Figure 2.6. Comparison of GCN2 aggregation at different pH values by DSF 56 Figure 2.7. The effects of detergents on GCN2 stability by DSF 57 Figure 2.8. SEC-MALS of GCN2 58 Figure 2.9. SPR of the GCN2-tRNA interaction 58 Figure 2.10. Purification of Strep-tagged GCN1 61 Figure 2.11. Purification of Strep-tagged GCN20 62 List of Figures and Tables xvii Figure 2.12. Pull-down analysis of the interactions between GCN2, GCN1 and GCN20 63 Figure 2.13. The effects of tRNA and ribosomes on GCN2 autophosphorylation 64 Figure 2.14. The effects of GCN1 and GCN20 on GCN2 autophosphorylation 66 Figure 2.15. The effects of tRNA and ribosomes on GCN2 T899 autophosphorylation 67 Figure 2.16. GCN2 T899 autophosphorylation is dependent on ATP 68 Figure 2.17. The effects of removing phosphorylation sites on GCN2 autophosphorylation 71 Figure 2.18. Purification of recombinant eIF2α 73 Figure 2.19. Phosphorylation of eIF2α by GCN2 74 Figure 2.20. The effects of tRNA and ribosomes on eIF2α phosphorylation by GCN2 75 Figure 2.21. The effect of cytosol on the migration of GCN2 through a sucrose gradient 77 Figure 2.22. Pull-down analysis of the interaction between GCN2 and the ribosome 78 Figure 2.23. Analysis of the GCN2-ribosome interaction under different conditions 83 Figure 2.24. The GCN2 truncation library 86 Figure 2.25. The final purity of the GCN2 truncation library 86 Figure 2.26. Autophosphorylation activity of the GCN2 truncation library 88 Figure 2.27. Phosphorylation of eIF2α by the GCN2 truncation library 90 Figure 2.28. Ribosome binding by the GCN2 truncation library 92 Figure 2.29. The effects of antibiotics on GCN2 autophosphorylation 94 Figure 2.30. A summary of the information from GCN2 domain mapping studies 98 Figure 3.1. The chemical structure of a polypeptide 102 Figure 3.2. Schematic of an HDX-MS experiment 106 Figure 3.3. Deuterium uptake of the uL10 peptide identified by initial HDX-MS 112 Figure 3.4. Mapping of the identified peptide on to the structure of the ribosome 113 Figure 3.5. Deuterium uptake of a typical peptide across two HDX-MS datasets 114 Figure 3.6. Deuterium uptake of three uL10 peptides identified by optimised HDX-MS 116 Figure 3.7. The GCN2 binding site on the structure of the ribosome 117 Figure 4.1. Global deuteration profile for full-length GCN2 134 Figure 4.2. Limited proteolysis of GCN2 136 Figure 4.3. The structure of a typical nanobody 137 Figure 4.4. Schematic showing the architecture of different types of antibodies 137 List of Figures and Tables xviii Figure 4.5. Cross-linking of GCN2 and GCN1 138 Figure 4.6. Purification of the library of nanobodies 140 Figure 4.7. Pull-down analysis of GCN2-nanobody interactions 140 Figure 4.8. Purification of GCN2-nanobody complexes 142 Figure 4.9. Crystallisation of the GCN2 pseudokinase domain 145 Figure 4.10. Crystal structure of the human GCN2 pseudokinase domain 146 Figure 4.11. Comparison between the human pseudokinase domain and the yeast kinase domain 148 Figure 4.12. Negative stain EM images of GCN2 149 Figure 4.13. 2D classes from negative stain EM images of GCN2 149 Figure 4.14. Negative stain 3D reconstruction of GCN2 150 Figure 4.15. Cryo-EM image of GCN2 151 Figure 4.16. GraFix cross-linking of GCN2 152 Figure 4.17. Capture of GCN2-ribosome complexes by FLAG tag purification 154 Figure 4.18. Capture of GCN2-ribosome complexes by Strep tag purification 155 Figure 4.19. Initial electron density maps for the putative GCN2-ribosome complex 156 Figure 4.20. Electron density map for the GCN2-ribosome complex after cross-linking 159 Figure 5.1. Crystal structure of the core Hsp90-Cdc37 complex 169 Figure 5.2. The signalling pathways downstream of FGFRs 171 Figure 5.3. SEC analysis of the Hsp90-Cdc37-FGFR3 KD interaction 175 Figure 5.4. Pull-down analysis of the Cdc37-FGFR3 KD interaction 176 Figure 5.5. Quantification of pull-downs between Cdc37 and FGFR3 KD mutants 177 Figure 5.6. Quantification of the incorporation of FGFR3 KD mutants into a ternary complex with Hsp90 and Cdc37 177 Figure 5.7. Stability of FGFR3 KD mutants in comparison to wild-type FGFRs 178 Figure 5.8. Stability of a panel of FGFR3 KD mutants 179 Figure 5.9. Global deuteration profile for FGFR3 KD 180 Figure 5.10. The effects of Cdc37 binding on the deuteration of FGFR3 KD (wild-type) 182 Figure 5.11. The effects of Cdc37 binding on the deuteration of FGFR3 KD mutants 183 Figure 5.12. HDX-MS changes mapped on to the structure of the FGFR3 KD 184 Figure 5.13. The effects of Cdc37 binding on the deuteration of B-Raf KD 185 Figure 5.14. HDX-MS changes mapped on to the structure of the B-Raf KD 185 List of Figures and Tables xix Figure 5.15. The effects of FGFR3 KD E466K binding on the deuteration of Cdc37 187 Figure 5.16. HDX-MS changes mapped on to a model of Cdc37 188 Figure 5.17. SAXS analysis of the Cdc37-FGFR3 KD complex 189 Figure 5.18. Schematic of a model describing how Hsp90/Cdc37 could distinguish between client and non-client kinases 192 Tables Table 1.1. The physiological roles of GCN2 15 Table 2.1. Phosphorylation sites of GCN2 in the presence and absence of activators 69 Table 2.2. GCN2 binding partners as identified by mass spectrometry after incubation in cytosol 80 Table 2.3. Details of the GCN2 truncation library 85 Table 2.4. SEC-MALS results for the GCN2 truncation library 87 Table 2.5. Summary of the results obtained for the GCN2 truncation library 98 Table 4.1. The nanobody library produced 139 Table 4.2. Pull-down analysis of GCN2-nanobody interactions 141 Table 4.3. A summary of the crystallisation trials that were set up 144 Table 4.4. Statistics for the X-ray dataset and model for the pseudokinase domain 147 Table 4.5. A summary of the EM conditions that were tested 153 Table 4.6. Statistics for the EM dataset of the cross-linked GCN2-ribosome complex 160 Table 4.7. A summary of the EM datasets collected on GCN2-ribosome samples 161 1 Chapter One – Introduction A cell’s ability to sense and respond to changes in the environment is crucial to the maintenance of homeostasis. In order to survive, cells must be able to dynamically adapt to changing conditions. This enables optimal use of available resources, and requires integration of information from a wide range of inputs to allow the cell to respond appropriately. The mechanisms by which all these signals are collated and integrated is the subject of much interest, as differences in how cells interpret signals underlie many diseases. Control within the cell can be implemented in many ways and at many different points. One major target of cellular regulation is protein translation, as altering a cell’s protein expression profile represents a powerful way by which to fundamentally alter the homeostasis of the cell. Protein Translation 1.1. Protein translation is the process in which the ribosome decodes information encoded in messenger RNA (mRNA) transcripts and uses the information to synthesise a nascent polypeptide chain. Translation (Figure 1.1) typically begins with the recognition of the initiator methionine tRNA (Met-tRNAiMet) by the guanosine triphosphate (GTP)-bound eukaryotic initiation factor 2 (eIF2) to form the ternary complex. eIF2 is a heterotrimeric complex made up of α, β and γ subunits. High resolution structures of eIF2α and eIF2γ have been solved (Dhaliwal and Hoffman, 2003; Ito et al., 2004; Schmitt et al., 2002), but currently there is no structure of the eukaryotic heterotrimeric complex in the absence of the ribosome. However, there is a crystal structure of the archaeal complex (IF2) from Sulfolobus solfataricus (Yatime et al., 2007) that shows the complex has an extended form with IF2γ acting as the complex core and binding to IF2α and IF2β. This correlates with biochemical data from eukaryotes indicating that eIF2γ interacts with Met-tRNAiMet and GTP, whilst eIF2α and eIF2β stabilise the complex (Naveau et al., 2013; Nika et al., 2001). Furthermore, Llácer and colleagues recently solved a cryo-electron microscopy structure of the eukaryotic preinitiation complex, including the entire ternary complex along with the small ribosomal subunit and other initiation factors (Llácer et al., 2015). This represents the first view of eukaryotic eIF2β, and generally agrees well with the archaeal structures. Moreover, this structure allows analysis of additional interactions between the ternary complex and other factors within the translation initiation machinery. Introduction 2 Figure 1.1. The canonical translation machinery. The small ribosomal subunit (40S) is primed via the binding of the initiation factors eIF1, eIF1A, eIF3 and eIF5, allowing the binding of the ternary complex (consisting of the initiator tRNA (Met-tRNAiMet), GTP and eIF2). This leads to the formation of the 43S preinitiation complex. This 43S ribosomal complex is then able to recruit the mRNA transcript via the initiation factor eIF4F (consisting of eIF4G, eIF4A and eIF4E) bound to the m7G cap. As shown, eIF4F is also associated with the PABP bound to the poly(A) tail of the mRNA, leading to circularisation of the mRNA. For simplicity, this circularisation is omitted in subsequent steps. Once the 48S preinitiation complex is assembled, it is in an open state. The complex then scans the mRNA until it encounters an AUG start codon, at which point the initiator tRNA is able to become fully accommodated in the ribosomal P site, forming the closed 48S preinitiation complex. This conformational rearrangement causes the dissociation of the associated initiation factors including GDP-bound eIF2. This then allows the large ribosomal subunit (60S) to bind, forming the 80S ribosome, which is fully translationally competent and can accommodate an aminoacylated tRNA molecule within the A site to begin elongation. The GDP-bound eIF2 is then recycled to a GTP-bound state by the cognate guanine exchange factor eIF2B. Introduction 3 Once the ternary complex has formed between eIF2, Met-tRNAiMet and GTP, it must be loaded on to the small ribosomal subunit. For this to happen correctly, the small subunit must first be ‘primed’ through the binding of several initiation factors including eIF1, eIF1A and eIF3. It is also thought that eIF5, a GTPase activating protein for eIF2, associates with the preinitiation complex alongside the ternary complex, forming the 43S preinitiation complex (43S PIC). The next stage of translation initiation is the recruitment of the mRNA transcript. This occurs through the recognition of the 7-methylguanylate (m7G) cap present at the 5’ end of all eukaryotic mRNA messages (Shatkin, 1976). The cap is generally recognised by the initiation factors eIF4E, eIF4G and eIF4A, which together form the complex eIF4F. eIF4E is an essential factor that is conserved in all eukaryotes, which makes direct contact with the m7G cap. eIF4A is an RNA helicase that allows the complex to unwind inhibitory secondary structures. Finally, eIF4G functions as a molecular scaffold, and through a series of protein- protein interactions enables the binding of the complex to the 43S assembly, ultimately forming the 48S preinitiation complex (48S PIC). eIF4G also forms interactions with the polyadenylate binding protein (PABP), enabling the circularisation of the mRNA message (Imataka et al., 1998; Tarun and Sachs, 1996). Initially, the 48S PIC exists in an open conformation, with the Met-tRNAiMet unable to become fully accommodated in the ribosomal P site. In this state, the complex is able to scan along the mRNA transcript in a 5’ to 3’ direction, until a start codon (AUG) enters the P site and is bound by the Met-tRNAiMet. During scanning, the GTP within the ternary complex is hydrolysed, but the detached phosphate moiety is retained until recognition of the start codon. Upon this recognition, a conformational rearrangement is triggered and the complex transitions into a closed conformation, with the Met-tRNAiMet fully engaged in the P site (Llácer et al., 2015). This structural rearrangement triggers the release of the initiation factors eIF1, eIF2 and eIF5, along with the dissociated phosphate. eIF1A is then able to promote the recruitment and binding of the 60S ribosomal subunit before itself dissociating. This leaves the 80S ribosome, which is translationally competent and able to accept an aminoacylated tRNA into the A site, to begin an elongation cycle. Once the guanosine diphosphate (GDP)-bound eIF2 has dissociated from the initiation complex, it must be recycled to a GTP-bound state to begin another cycle of translation initiation. This is enabled by the guanidine nucleotide exchange factor (GEF) eIF2B. eIF2B Introduction 4 consists of 5 different subunits: α, β, γ, δ and ε. These subunits are thought to form two distinct subcomplexes which bind independently to eIF2 (Pavitt et al., 1998): eIF2Bγ and eIF2Bε form a catalytically competent subcomplex whilst the remaining subunits form a regulatory subcomplex, which has been postulated to play a key role in the regulation of the GEF (Kuhle et al., 2015). Whilst this is considered to be the canonical model for translation initiation and occurs for the majority of eukaryotic mRNAs, there is significant evidence for alternative pathways to achieve a translationally active ribosome. All these pathways contain the same general steps: 1. Assembly of the 43S PIC 2. Recruitment of the mRNA transcript 3. Identification of the start codon 4. Binding of the 60S subunit However, the proteins and molecular mechanisms that orchestrate each step can be varied. This inbuilt redundancy provides a level of flexibility to the translation process that allows the cell to adapt to a variety of different conditions by adjusting the protein expression profile in a sensitive and selective manner. Examples of non-canonical mechanisms include cap recognition by the initiation factor eIF3, allowing the system to bypass the need for eIF4F (Lee et al., 2015; 2016). This renders the translation of certain transcripts insensitive to the modulation of eIF4F, for example by the release of the inhibitory eIF4E-binding protein (4EBP) as a result of cell proliferation-promoting pathways. Some mRNAs are able to organise their translation to avoid the scanning step altogether. An example of this has been described for the histone H4 mRNA (Martin et al., 2011). Histone mRNAs are structurally distinct from typical eukaryotic mRNAs, as they contain a very short 5’ untranslated region (UTR), lack introns and a poly(A) tail, and instead have a conserved stem-loop structure at their 3’ end that interacts with a stem-loop binding protein (SLBP). As the 5’UTR is too short to initiate translation in the canonical manner, it appears they are able to recruit eIF4E and position the ribosome over the start codon via two internal RNA elements within the open reading frame (ORF). This streamlining of the process could be an advantage during the S phase of the cell cycle, when the cell is replicating DNA very quickly, and therefore needs to synthesise a large quantity of histone proteins as quickly as possible. Introduction 5 The Integrated Stress Response 1.2. Given the crucial nature of the initiation step of translation, it is unsurprising that it is subject to exquisite regulation. Many different signalling pathways converge to control different steps within this process, allowing the cell to integrate signals from a wide variety of sources. One of these pathways is the Integrated Stress Response (ISR). The ISR can be activated by a diverse range of signals, but in each case the activation of the pathway results in the phosphorylation of the initiation factor eIF2α at a conserved serine residue. This phosphorylation causes a significant increase in the affinity of eIF2 for the GEF eIF2B, causing its mode of binding to shift to a non-productive interaction. This means that the phosphorylation of eIF2α causes eIF2 to be converted from a substrate of the GEF eIF2B to a competitive inhibitor, ultimately resulting in a significant decrease in the concentration of GTP-bound eIF2 available to bind to Met-tRNAiMet and therefore less eIF2 that is able to participate in translation initiation. Given that there are significantly fewer ternary complexes available under conditions where the ISR is active and eIF2α has been phosphorylated, it is unsurprising that the rate of translation for the majority of mRNA transcripts falls dramatically. However, this is not universal, as some transcripts are able to escape this translational block. The best studied example of this is the mRNA encoding the transcription factor Activating Transcription Factor 4 (ATF4; a homologue of General Control Nonderepressible 4 (Gcn4) in yeast), which is specifically upregulated upon induction of the ISR. The mechanisms behind this phenomenon have been primarily studied in yeast (Hinnebusch, 2005), but appear to be generally conserved across eukaryotes. The basis of this effect is the presence of two (in human ATF4, four in yeast GCN4) upstream open reading frames (uORFs) in the 5’ UTR of the mRNA transcript. Under all conditions, ribosomes will first initiate translation at the first uORF (uORF1). Normally, once a translating ribosome reaches a stop codon it is released from the mRNA, and cannot reinitiate translation downstream. However, after translating uORF1, some posttermination 40S subunits are able to remain attached to the mRNA and resume scanning (Hinnebusch, 2014). These subunits will then gradually reacquire ternary complexes. Once a ternary complex has bound, the 40S subunit is able to productively scan the mRNA, and reinitiate translation at any downstream open reading frames. Under non- starvation conditions (when the ISR is not active), there is typically plenty of GTP-bound eIF2. This means that a ternary complex is likely to rebind to the small ribosomal subunit in Introduction 6 time to reinitiate translation at the second uORF (uORF2). Translation of uORF2 has an inhibitory effect upon translation of the ATF4-encoding ORF as the ribosomes tend to dissociate after translation of uORF2 (Figure 1.2). Figure 1.2. How a decrease in the concentration of ternary complexes can lead to the skipping of inhibitory upstream ORFs (uORFs) to increase protein translation of the transcription factor ATF4. ATF4 mRNA is depicted as a straight line with uORFs shown as red arrows. The ATF4 ORF is shown as a white box. Translation of uORF1 allows some post-termination 40S ribosomal subunits to remain bound to the mRNA and resume scanning. Under normal conditions, the concentration of ternary complexes is high, meaning that the ribosome is likely to quickly rebind a ternary complex and therefore can reinitiate translation at an AUG codon soon after translation of uORF1, thus enabling translation of uORF2. This then inhibits translation of the ATF4 ORF. Under conditions when the ISR is active, the low concentration of ternary complexes in the cell means that the ribosomes are unlikely to rebind a ternary complex as quickly and therefore will not be competent to reinitiate translation immediately after translation of uORF1. This means that it is more likely that the ribosomes will skip uORF2 and therefore be able to translate the ATF4 ORF. When the ISR is active, the decrease in availability of GTP-bound eIF2 causes a decrease in the concentration of ternary complexes. This means that it is more likely the ribosome scanning the mRNA after translating uORF1 will proceed past the start codon of uORF2 Introduction 7 before reassociating with a ternary complex. The ribosome will thus not translate uORF2 and will therefore not be inhibited from initiating translation at the start codon for the open reading frame for ATF4. Therefore, the induction of the ISR leads to the upregulation of ATF4 expression (Figure 1.2) (Vattem and Wek, 2004). The eIF2α Kinases 1.3. Given the importance of the phosphorylation of eIF2α, it is unsurprising that the kinase(s) that perform this phosphorylation are subject to stringent regulation. In yeast, eIF2α is phosphorylated by a single kinase, called General Control Nonderepressible 2 (Gcn2). Gcn2 is conserved across all eukaryotes, and is activated in response to amino acid starvation (Berlanga et al., 1999; Hinnebusch, 1985). In mammals, there are three additional eIF2α kinases, all of which phosphorylate eIF2α in response to specific stressors (Figure 1.3). [Following convention, the yeast homologue is hereafter denoted Gcn2 whilst the mammalian homologue is GCN2.] Figure 1.3. The specific signals that activate the four eIF2α kinases present in mammals. The Heme-Regulated Inhibitor (HRI) kinase is regulated by the availability of iron. Expressed primarily in erythroid cells, HRI is activated when the intracellular concentration of iron drops significantly (Han et al., 2001). This mechanism prevents the production of globin proteins when there is insufficient intracellular heme to bind to it, thereby preventing aggregation of excess globin and hyperchromic anaemia. Protein Kinase R (PKR) is another eIF2α kinase, which is activated by the presence of double stranded RNA (dsRNA) (Rojas et al., 2010). This is thought to constitute a viral defence mechanism, allowing cells to shut down their protein production machinery to avoid the production of viral proteins. The third additional kinase is called PKR-like Endoplasmic Reticulum Kinase (PERK) (Shi et al., 1998). PERK is activated in response to an accumulation of misfolded proteins within the Introduction 8 endoplasmic reticulum (ER) (Harding et al., 1999), and is most highly expressed in the pancreas. All four eIF2α kinases contain a conserved catalytic domain, but each contains additional regulatory domains enabling each kinase to recognise specific signals (Figure 1.4). This raises an interesting question concerning whether each eIF2α kinase has a common mechanism for their activation, despite the very different input signals. Figure 1.4. The domain structure for each of the four eIF2α kinases present in mammalian cells. Each has a conserved catalytic domain (highlighted in red) as well as additional domains with regulatory roles. The domains that are thought to be key for each protein to detect its cognate stressor are highlighted in blue. The HisRS-like domain of GCN2 shows sequence homology to the histidyl tRNA synthetase, and the IRE domain of PERK is related to the inositol-requiring enzyme (IRE). Perhaps best understood is the mechanism of PKR activation by dsRNA (Figure 1.5). PKR contains two dsRNA-binding domains (dsRBDs) N-terminal to its kinase domain (Figure 1.4), which promote dimerisation of the kinase in the presence of dsRNA. Under normal conditions, the kinase exists in a latent monomeric state, with the dsRBDs autoinhibiting the kinase (Wu and Kaufman, 1997). Dimerisation is key to the activation of the kinase, as shown by the replacement of the two dsRNA-binding domains with a heterologous dimerisation motif leading to the reconstitution of recombinantly expressed PKR activity in Saccharomyces cerevisiae (Ung et al., 2001), whilst the kinase domain alone is unable to phosphorylate eIF2α to any great extent. It is thought that binding of dsRNA to the dsRBDs relieves the autoinhibition, and promotes dimerisation. This permits autophosphorylation of specific threonine residues in the kinase domain, which is required for full activity (Romano et al., 1998). This autophosphorylation appears to occur in trans between dimers, as structural characterisation of the kinase domain in complex with eIF2α shows a back-to-back arrangement which is seemingly incompatible with intra-dimer autophosphorylation (Dar et al., 2005; Dey et al., 2005a). Furthermore, kinetic and biochemical evidence contradict a cis Introduction 9 reaction, as it can be shown that a kinase dead mutant can become autophosphorylated by a wild type protein (Ortega et al., 1996; Thomis and Samuel, 1995). Once autophosphorylation has occurred, eIF2α can bind to the complex and thus become phosphorylated. Figure 1.5. The activation mechanism for the kinase PKR. Initially PKR is in a monomeric state and the kinase domain (KD) is autoinhibited by the two N-terminal dsRBDs. Upon the detection of dsRNA, the dsRBDs are able to bind to the dsRNA. This relieves the autoinhibition and promotes dimerisation of the kinase. The kinase is now able to autophosphorylate, which in turn promotes substrate binding and phosphorylation. The activation pathway for PERK has also been the subject of much study, and many aspects of PKR activation appear to be conserved despite key differences. As shown in Figure 1.4, PERK contains a domain with sequence homology to the Inositol-Requiring Enzyme (IRE), which acts as a sensor of unfolded proteins within the ER lumen. This domain is critical for PERK to become activated in times of ER stress. In the absence of ER stress, PERK is present in the ER membrane as an inactive dimer (Figure 1.6). The luminal domain associates with the ER chaperone Binding Immunoglobulin Protein (BiP) (Bertolotti et al., 2000), and it has been hypothesised that a build up of unfolded proteins within the ER causes BiP to dissociate from the luminal domain of PERK, which could be the PERK activation signal. However, another possibility is that exposed peptides from unfolded proteins are able to bind to PERK directly, via a major histocompatibility complex (MHC)-like groove on the surface of the luminal domain, and this causes BiP to dissociate from the complex (Gardner and Walter, 2011). This finding has led to the ‘line-up model’ in which a thread of unfolded protein is able to bind to multiple PERK homodimers and effectively lines them up in large multimers in the ER membrane (Figure 1.6). It is clear that some form of oligomerisation takes place upon recognition of the unfolded protein, in a step similar to the dimerisation of PKR (Carrara et al., 2015). As for PKR, oligomerisation of PERK leads to trans autophosphorylation (Cui et al., 2011; Ma et al., 2001; Marciniak et al., 2006), which allows eIF2α recruitment and phosphorylation. Introduction 10 Figure 1.6. The ‘line-up’ activation mechanism for the eIF2α kinase PERK. Under non-stressed conditions, PERK exists as a dimer in the ER membrane, with the kinase domains in the cytoplasm and the IRE unfolded protein stress-sensing domains in the lumen of the ER. The IRE domains are associated with the ER chaperone BiP. Upon the induction of ER stress, the unfolded proteins can bind to the IRE domains via their MHC-like groove. At this stage BiP dissociates from the complex, and PERK oligomerises in the ER membrane. This multimerisation drives autophosphorylation of the kinase, which then allows the recruitment and phosphorylation of eIF2α. The similarities between the mechanism of activation for PKR and PERK imply that oligomerisation and autophosphorylation could be characteristic principles of activation for the eIF2α kinases, and could therefore be applicable to both HRI and GCN2 as well. Amino Acid Sensing 1.4. Amino acids are a crucial cellular substrate, as they constitute the building blocks for the vast majority of cellular machinery. There are twenty naturally occurring amino acids in humans, nine of which are essential. This means that the cell is unable to make them de novo, and so they must be included in the diet. These include histidine, isoleucine, leucine, lysine, methionine, phenylalanine, threonine, tryptophan and valine. There are a further seven amino acids that can only be synthesised under certain conditions, so in some cases they must too be gained from the diet (for example in premature infants). These are arginine, cysteine, glycine, glutamine, proline, serine and tyrosine. The cell must be able to recognise when there is an abundance of amino acids, and when there is a shortage. If there is a shortage, the cell may need to increase amino acid biosynthesis, if possible, or produce a signal to communicate that the current diet of the organism is insufficient. This ability to sense the presence and absence of amino acids is generally coordinated by two pathways. The first is the GCN2-dependent branch of the ISR, which recognises a depletion of amino acids, and the second is the nutrient-sensing pathway driven by the mammalian target of rapamycin (mTOR). The protein kinase mTOR forms the catalytic subunit of two Introduction 11 distinct cellular complexes, known as mTOR complex 1 (mTORC1) and mTOR complex 2 (mTORC2). mTORC1 consists of mTOR, the regulatory-associated protein of mTOR (RAPTOR) (Hara et al., 2002; Kim et al., 2002) and mammalian lethal with Sec13 protein 8 (MLST8) (Kim et al., 2003), as well as other more recently identified associated factors such as proline-rich AKT substrate 1 (PRAS40) (Sancak et al., 2007) and DEP domain-containing mTOR-interacting protein (DEPTOR) (Peterson et al., 2009). mTORC2, on the other hand, consists of mTOR, rapamycin-insensitive companion of mTOR (RICTOR) and mLST8. Each complex fulfils a slightly different role: mTORC1 is generally associated with cell growth whilst mTORC2 is involved in the promotion of cell proliferation. Upstream of mTORC1 is a trimeric signalling complex known as the Tuburous Sclerosis complex (TSC). The TSC is able to act as a regulator of mTOR through the GTPase activity of one component of the complex, TSC2 (Inoki et al., 2003a). TSC2 is able to regulate the nucleotide status of the small G protein Ras-homologue enriched in brain (Rheb), which is located at the lysosomal membrane (Inoki et al., 2003a; Sancak et al., 2010; Yamagata et al., 1994). When loaded with GTP, Rheb is a potent activator of mTORC1 (Inoki et al., 2003a), meaning that the TSC acts as a negative regulator of mTORC1 activity. The TSC acts as a signalling hub, and is able to integrate information from a wide range of extracellular and intracellular sources, such as mitogen signalling (Inoki et al., 2002; Manning et al., 2002; Potter et al., 2002), cellular energy levels (Inoki et al., 2003b) and oxygen availability (Brugarolas et al., 2004). These inputs thus feed into the mTORC1 pathway, and contribute to its activation status. Once activated, mTORC1 is able to phosphorylate a diverse range of cellular substrates, all of which are linked to the control of cell growth. Such substrates include the S6 kinases, which are involved in the promotion of ribosome biogenesis, and the eIF4E-binding protein (4EBP), phosphorylation of which prevents its inhibitory interaction with eIF4E and so promotes global protein translation. Alongside the signals fed through to mTORC1 via the TSC, there are also additional levels of control that affect the activity of mTORC1 in different ways. Amino acids are a key example of this. It has been known for many years that an abundance of amino acids positively regulates protein translation (Preedy and Garlick, 1986), and it has been subsequently shown that this is partly under the control of mTORC1. Furthermore, a combination of both mitogen signalling and an abundance of amino acids is necessary for the in vivo phosphorylation of canonical mTORC1 substrates (Hara et al., 1998; Wang et al., 1998). The maintenance of Introduction 12 mTORC1-dependent sensitivity to cellular amino acid levels even in a TSC-/- double knock out mouse implied the existence of a separate, TSC-independent pathway for mTORC1 activation (Smith et al., 2005). This alternative pathway for mTORC1 activation has been the subject of great interest, and now much of it has been elucidated (Figure 1.7). Figure 1.7. The activation of mTORC1 by amino acids. When there are amino acids present within the lumen of the lysosome, this is recognised by the vacuolar ATPase (v-ATPase). This recognition event drives a remodelling of the interface between v-ATPase and Ragulator, a GEF. This remodelling stimulates the activity of Ragulator, which then acts to alter the nucleotide status of the small Rag G proteins. Once the correct nucleotide loading status has been achieved, mTORC1 is recruited through binding to the Rags. This places mTORC1 in the vicinity of its activator Rheb, causing mTORC1 to become activated. When there is an abundance of amino acids, they accumulate in the lysosome (Zoncu et al., 2011). This is sensed by the vacuolar ATPase (v-ATPase) within the lysosomal membrane and causes a remodelling of its interactions with another protein in the lysosomal membrane called Ragulator (Zoncu et al., 2011). Ragulator is a GEF, and the amino acid-dependent Introduction 13 interactions with v-ATPase stimulate its GEF activity towards two small G proteins: Rag A/B and Rag C/D (Bar-Peled et al., 2012), which are tethered to the lysosomal membrane through interactions with Ragulator (Sancak et al., 2010). Alteration of the nucleotide status of the Rags (Figure 1.7) drives the relocalisation of mTORC1 from the cytoplasm to the lysosomal membrane, where it can bind to the Rag proteins (Sancak et al., 2008). Once localised at the lysosomal membrane, mTORC1 can be activated by Rheb. When the cellular concentration of amino acids is low, this mechanism whereby mTOR becomes activated will no longer occur. However, there also exists an active pathway for decreasing the rate of translation and initiating stress responses in times of amino acid deprivation. This pathway, effectively reciprocal to that coordinated by mTOR, is the integrated stress response, orchestrated by the eIF2α kinase GCN2. This response was first discovered in Saccharomyces cerevisiae and in yeast is known as General Amino Acid Control (GAAC) (Delforge et al., 1975; Wolfner et al., 1975). It is typically characterised by the phosphorylation of eIF2α and the induction of Gcn4 (the homologue of mammalian ATF4) expression, as well as an increase in expression of the downstream targets of Gcn4. Interestingly, GAAC can also be induced by growing yeast on media with amino acid imbalances designed to elicit starvation responses through feedback signaling to enzymes that are shared between multiple amino acid biosynthetic pathways (Niederberger et al., 1981). The stable induction of GAAC is dependent on the kinase Gcn2 (Tzamarias et al., 1989). The targets of Gcn4 have been the subject of much interest, and it has been shown that these include genes involved in amino acid biosynthetic pathways, amino acid transporters and regulators of autophagy (Jia et al., 2000; Natarajan et al., 2001), all of which can be easily reconciled with the role of the GAAC to overcome amino acid starvation. However, profiling experiments have been done that show up to 10 % of the yeast genome is upregulated upon induction of GAAC, indicating that the scale of response is much wider and more complex than originally presumed (Natarajan et al., 2001). How GCN2 is able to recognise the cellular state of amino acid deprivation has been a question in the field for many years. The mechanism is thought to be reliant on deacylated tRNA molecules. It is clear that upon a decrease in the intracellular concentration of amino acids, this will cause a reciprocal increase in the concentration of deacylated tRNA molecules. Introduction 14 Binding of the deacylated tRNA molecule to GCN2 is thought to stimulate a conformational rearrangement of GCN2, leading to the adjustment of multiple inter-domain interactions and ultimately resulting in the allosteric activation and autophosphorylation of the kinase domain (Lageix et al., 2014; 2015; Romano et al., 1998) (Figure 1.8). Figure 1.8. Schematic showing the reciprocal relationship between the intracellular concentration of amino acids and the concentration of deacylated tRNA molecules (deAc tRNA). The tRNA can then bind to GCN2, leading to a conformational rearrangement and the autophosphorylation and activation of GCN2. Structure and Function of the kinase GCN2 1.5. GCN2 is an important protein within the eukaryotic proteome, and has been implicated in many biological processes in both health and disease. Many of these putative roles are in the realm of nutrient sensing, but mammalian systems seem to require the proper function of GCN2 in diverse roles in very complex physiological processes, such as neuronal plasticity and memory formation, initiation of an immune response and behavioural adaptation (Table 1.1) (Chesnokova et al., 2017; Costa-Mattioli et al., 2005; Ravindran et al., 2014; Van de Velde et al., 2016; Vasudevan et al., 2017). The kinase has also been shown to play a role in the development of many diseases such as pulmonary veno-occlusive disease (PVOD), Alzheimer’s and cancer (Eyries et al., 2014; Ma et al., 2013; Ye et al., 2010), meaning that the full molecular characterisation of the kinase could have far-ranging implications for therapeutics. Mice lacking GCN2 are viable (Zhang et al., 2002); however, under conditions of amino acid deprivation they show excessive protein production in the liver (Anthony et al., 2004). The action of GCN2 has been suggested to be important for rapid behavioural changes in terms of dietary choices (Hao et al., 2005; Maurin et al., 2005) but this data was not reproducible (Leib and Knight, 2015). Leib and Knight further showed that knocking out GCN2 did not affect long term responses to amino acid deficient diets, and demonstrated no link between GCN2 and dietary responses. Introduction 15 Table 1.1. A table to show the many roles in which GCN2 has been implicated, aside from overcoming amino acid starvation. Function Reference Regulation of fatty-acid homeostasis (Guo and Cavener, 2007) Regulation of lipid storage (Xu et al., 2013) Regulation of metabolism during aging (Maurin et al., 2012) Modulation of insulin sensitivity during amino acid deprivation (Xiao et al., 2011) Induction of anti-microbial immunity (Vasudevan et al., 2017) Loss-of-function substitutions cause familial pulmonary veno-occlusive disease (Eyries et al., 2014) Modulation of protein expression for neuronal plasticity (Chesnokova et al., 2017) Enhancement of antigen presentation after vaccination against yellow fever (Ravindran et al., 2014) Formation of memories (Costa-Mattioli et al., 2005) Limitation of human immunodeficiency virus (HIV) replication (Jaspart et al., 2017) Sensing of nucleolar stress (Nakamura and Kimura, 2017) Trafficking of cytotoxic T cells (Van de Velde et al., 2016) Induction of neuronal dysfunction during Alzheimer’s disease (Ma et al., 2013) Promotion of tumourigenesis (Ye et al., 2010) Induction of autophagy (B'chir et al., 2013) Protection of glomerular endothelial cells from high glucose (Eleftheriadis et al., 2016) Contribution to the resolution of autoimmune neuroinflammation (Keil et al., 2016) Control of intestinal inflammation (Ravindran et al., 2016) In humans, GCN2 is a protein containing 1649 amino acids, with a molecular weight of approximately 190 kDa. It consists of five domains: an N-terminal RWD domain (present in RING finger proteins, WD repeat-containing proteins and the yeast DEAD-like helicases), a pseudokinase domain, a catalytically active kinase domain, a HisRS-like domain (with sequence homology to the histidyl tRNA synthetase enzyme) and finally a C-terminal domain (CTD) (Figure 1.9). Most of the molecular information gained concerning GCN2 has been discovered in yeast; however, the high level of sequence conservation between yeast and human GCN2 indicates these data may be extrapolated to the human construct. A combination of biochemical and genetic studies indicates that GCN2 forms dimers and possibly tetramers under certain conditions, and much work has been done to map the regions that are necessary and sufficient for dimerisation (Diallinas and Thireos, 1994; Qiu et al., 2001; 1998). These studies have indicated that several regions of the protein contribute to dimerisation, including the kinase domain, part of the HisRS-like domain and the CTD (Figure 1.9). Further studies have Introduction 16 indicated that the CTD appears to be most important for the in vivo dimerisation of the protein, and that dimerisation is crucial for kinase activity (Narasimhan et al., 2004). Figure 1.9. The basic domain structure of human GCN2. Residue numbers are for the human construct, and in some cases have been extrapolated from yeast data via sequence alignment using ClustalOmega (Sievers et al., 2011). The amino acids indicated in black above the domain have been implicated in some aspect of functionality. Putative binding determinants have been indicated for dimerisation (green), GCN1 (orange), tRNA (purple), ribosomes (grey) and eEF1A (red). The RWD domain is at the N-terminus of the protein. The structure of this domain in solution has been solved using nuclear magnetic resonance (NMR) spectroscopy and shows an α + β sandwich fold consisting of a four-stranded β sheet alongside three adjacent α helices (Figure 1.10) (Nameki et al., 2004). The structure also shows the location of the invariant YPXXXP motif that is characteristic of RWD domains. Interestingly, this structure shows considerable homology to the E2 ubiquitin conjugating enzyme UBC9 (Tong et al., 1997), and they have since been grouped together into a homologous superfamily clade known as Ubiquitin-conjugating enzyme/RWD-like (IPR016135) (Hunter et al., 2012; Velankar et al., 2005). However, there is no known functional link, and the RWD domain is thought to be involved in mediating protein-protein interactions. Introduction 17 Figure 1.10. The structure of the N-terminal RWD domain from mouse GCN2 (PDB code 1UKX) (Nameki et al., 2004). The structure is coloured from the N-terminus (blue) to the C-terminus (red) and the termini are indicated. Adjacent to the RWD domain, there is a highly charged region consisting of a large number of lysines, arginines, aspartates and glutamates. Following this is a domain with sequence similarity to archetypal kinase domains, however it lacks the key residues that are necessary for catalytic activity, therefore it is categorised as a pseudokinase domain (Hanks et al., 1988). This pseudokinase domain is necessary for yeast Gcn2’s kinase activity both in vitro and in vivo (Wek et al., 1990; Zhu et al., 1996), and genetic evidence has indicated that the pseudokinase domain interacts directly with the kinase domain. Constitutively activating mutations within the pseudokinase domain result in an enhanced interaction between the pseudokinase and the kinase domains, according to in vitro coimmunoprecipitation experiments (Lageix et al., 2014). It has therefore been proposed that this interaction is enhanced upon amino acid starvation, leading to the allosteric activation of the kinase domain. The structure of the yeast Gcn2 kinase domain (Figure 1.11) has been solved in both a constitutively activated and an inactive state (Padyana et al., 2005). These structures along with mutational analysis have shown the kinase domain to be in an intrinsically inactive state, even in the absence of other regulatory domains and inhibitory interactions. Introduction 18 Figure 1.11. The structure of the yeast Gcn2 kinase domain (PDB code 1ZYD) (Padyana et al., 2005). A: The structure of wild-type Gcn2 in complex with ATP and Mg2+. A monomer is shown, coloured from the N-terminus (blue) to the C-terminus (red), with the termini indicated. An ATP molecule along with two Mg2+ ions is accommodated in the catalytic cleft between the N- and the C- lobes of the kinase domain. B: The dimeric arrangement of the kinase domain within the crystal structure. One monomer is shown in pink and the other one in blue. In the inactive state (PDB code 1ZXE: D835N inactivating mutation) the kinase domain contains a very rigid hinge region between the N- and the C- lobes, leading to a closed formation of the lobes and thus obstructing substrate access to the catalytic cleft. Furthermore, this structure exhibits distortion of the key residues for catalysis, for example those involved in ATP binding. Lastly, there appears to be an extended loop acting as a lid over the catalytic cleft, further preventing substrate access. Comparison of this structure with a constitutively activated construct (PDB code 1ZY4: R794G hyperactivating mutation) shows a significant rearrangement of key amino acids, altering the hydrogen-bonding network and thus driving conformational change within the domain. These changes include a reduction in rigidity of the hinge region, allowing an increase in flexibility between the two lobes and so opening up the catalytic cleft. Moreover, the key catalytic residues can now assume their optimal position for catalysis, and the extension that prevented substrate access is rearranged to enable substrates to enter the active site. Once ATP can enter, the kinase can autophosphorylate on two threonine residues on the activation loop of the kinase (threonine 882 and 887 in yeast Gcn2; 899 and 904 in human GCN2), which stabilises the protein in the active conformation. There is no structural information concerning how GCN2 binds to the full-length eIF2α substrate, but the high level of conservation between the kinase domains of the four eIF2α kinases means that the structure of the PKR kinase domain bound to eIF2α can be used to Introduction 19 provide insights into how GCN2 recognises and phosphorylates its substrate (Dar et al., 2005; Dey et al., 2005a). This structure, along with a number of insights from mutational analyses, has allowed the postulation of a mechanism for how eIF2α becomes phosphorylated by GCN2. In the isolated eIF2α, serine 51 (the site of phosphorylation in yeast eIF2α) is buried in a hydrophobic pocket, making it inaccessible to other cellular kinases. Upon binding to the kinase, a conformational rearrangement occurs, leading to the exposure and eventual projection of the peptide containing serine 51 into the active site of the kinase. The binding of eIF2α to the kinase is crucially dependent on eIF2α residues far away from serine 51 (Dey et al., 2005b), explaining the observation that a peptide containing serine 51 and surrounding residues is phosphorylated with a much lower efficiency than the full-length substrate (Mellor and Proud, 1991). Mutational analyses within the kinase domain have provided information about the roles of several key amino acids (highlighted in Figure 1.9). Serine 567 (S577 in yeast – see the sequence alignment in Supplemental Figure 1) is a phosphorylation site just N-terminal to the N-lobe of the kinase domain. It can be phosphorylated in vivo, and when phosphorylated it reduces the affinity of Gcn2 for tRNA (Garcia-Barrio et al., 2002). Its phosphorylation is regulated by the mTOR pathway, providing an aspect of cross-talk between the two reciprocal pathways. Upon nutrient depletion, mTOR will become inactivated, which leads to the dephosphorylation of TAP42, a phosphoprotein that is able to stably associate with phosphatases (Di Como and Arndt, 1996; Jiang and Broach, 1999; Zheng and Jiang, 2005). Dephosphorylation reduces the affinity of this interaction, leading to the release and activation of the phosphatases (Düvel et al., 2003; Yan et al., 2006). These phosphatases are able to dephosphorylate the serine 567 position, thus partially stimulating Gcn2. Interestingly, mutation of the serine for an alanine in yeast leads to constitutively increased Gcn2 autophosphorylation and eIF2α phosphorylation (Garcia-Barrio et al., 2002). The kinase domain contains a highly conserved ATP-binding motif (human GCN2 residues Y616 – I621 in the N-lobe. Substitution of a lysine residue in this motif for an arginine in the yeast protein (K628R) completely abolishes Gcn2 activity, indicating that it is essential for catalysis (Wek et al., 1989). Mutation of an aspartate residue to an asparagine (D848N) at the beginning of the C-lobe also results in a catalytically incompetent protein (Padyana et al., 2005). Introduction 20 The histidyl tRNA synthetase-like domain of GCN2 is generally considered to be the major contributor to the recognition of an intracellular decrease in amino acid levels (Wek et al., 1989). This domain is essential for Gcn2 activity within the cell (Wek et al., 1995), and is able to bind to deacylated tRNA (Dong et al., 2000; Wek et al., 1995). Sequence analysis of this region shows the conservation of the ‘motif 2’ (m2), which in an authentic tRNA synthetase interacts with the acceptor stem of a tRNA molecule during the aminoacylation reaction (Ruff et al., 1991). Mutation of this motif (yeast residues Y1119L; R1120L) completely abrogates Gcn2’s ability to bind to tRNA, as well as abolishing any catalytic activity towards eIF2α (Dong et al., 2000; Wek et al., 1995), further validating the hypothesis that the binding of deacylated tRNA to the HisRS-like domain of GCN2 is the activating signal. Interestingly, the HisRS-like domain does not seem to be the sole determinant for tRNA binding: the C-terminal domain is also thought to be necessary (Dong et al., 2000). The structure of the C-terminal domains of both yeast and mouse GCN2 were solved recently (Figure 1.12), and show a reasonably high level of structural homology despite relatively low levels of sequence identity (He et al., 2014). Figure 1.12. Crystal structures of the C-terminal domain from mouse (on the left) and yeast (on the right) GCN2 (PDB codes 4OTN and 4OTM respectively) (He et al., 2014). The structures are coloured from the N-terminus (blue) to the C-terminus (red). The structures are both dimeric, and show an unusual interlocking arrangement of the two monomers. For the most part, they are also functionally similar. Both constructs form constitutive dimers, and both are able to stably associate with double and single-stranded nucleic acids. Despite this, they seem to differ in their ability to associate with ribosomes, as the yeast construct co-migrates with ribosomes through a sucrose gradient whilst the mouse construct does not under the same conditions. This ability to bind ribosomes is dependent on Introduction 21 the presence of three lysine residues within the CTD, which are not conserved in the murine version of the protein (He et al., 2014). Multiple lines of evidence point to the existence of a network of intra- and intermolecular interactions within the dimeric full-length protein, and unpicking these complex interactions has allowed a model for the conformational rearrangement that underlies GCN2 activation to be put forward (Figure 1.13) (Lageix et al., 2015). Under normal conditions, GCN2 is present as an inactive dimer with interactions occurring between the C-terminal domains, the HisRS- like domains and the kinase domains of each protomer. The kinase domains are likely arranged in the mode of dimerisation observed in the crystal structure (Figure 1.11B), interacting via their N-terminal lobes, and the HisRS-like domains are also likely to dimerise as seen in the crystal structure of the genuine histidyl tRNA synthetase protein (Merritt et al., 2010). The kinase domains are engaged in inhibitory interactions with the C-terminal domains, and this arrangement is stabilised by the two HisRS-like domains. In this scenario, the pseudokinase domains are unable to form a stable interaction with the kinase domains, but they may interact with the C-terminal domains. Upon amino acid starvation, uncharged tRNA builds up and is therefore able to bind to the two m2 motifs in the HisRS-like domains. This causes a conformational remodelling of the HisRS-like domains, including the region that was previously bound to the C-terminal domains, thus triggering dissociation of the CTDs. Consequently, the affinity between the CTDs and the kinase domains is reduced, enabling the pseudokinase domains to outcompete the CTDs and bind to the kinase domains. This is a stimulatory interaction, and leads to a structural rearrangement of the kinase domains and ultimately their activation. Introduction 22 Figure 1.13. A model for the reorganisation of the domains of GCN2 upon activation through the build up of deacylated tRNA (modelled on (Lageix et al., 2014; 2015)). In the inactive state, the kinase domain (KD, green) is engaged in inhibitory interactions with the C-terminal domain (CTD, yellow) and the HisRS-like domain (blue). Upon amino acid starvation, a build up of deacylated tRNA permits tRNA to bind to GCN2 via the HisRS-like domain and the CTD. This causes a conformational remodelling of the HisRS-like domain, leading to dissociation of the CTD. This in turn reduces the affinity of the CTD for the KD, which allows the KD to become engaged in stimulatory interactions with the pseudokinase domain (ΥKD). This ultimately results in the activation of the kinase domain through structural changes allowing substrate access to the active site and ultimately catalysis. Regulators of GCN2 1.6. 1.6.1. Deacylated tRNA Deacylated tRNA is generally considered to be the major regulator of GCN2. Under conditions of amino acid starvation, a decrease in amino acid concentration will negatively affect the efficiency of the tRNA aminoacylation reaction. Without sufficient levels of the cognate amino acid, the tRNA synthetase enzyme will not be able to aminoacylate the tRNA, and so the levels of deacylated tRNA will increase. Large-scale microarrays have demonstrated that the activity level of Gcn2 is directly correlated with the concentration of deacylated tRNA within the cell (Zaborske et al., 2009), providing in vivo validation of this model. It has also been shown that a decrease in only one amino acid is sufficient to downregulate the aminoacylation of other species of tRNA, indicating that there must be additional feedback mechanisms that regulate these reactions. Introduction 23 Both gel shift and Northwestern assays have demonstrated that yeast Gcn2 is able to bind to deacylated tRNA molecules through a domain with sequence homology to the histidyl tRNA synthetase (HisRS-like domain, see Figure 1.9) (Dong et al., 2000; Wek et al., 1995), implicating tRNA as a direct effector of GCN2 activation. The HisRS-like domain is essential for eIF2α phosphorylation by Gcn2 (Zhu et al., 1996), lending further weight to this hypothesis. As described previously, the HisRS-like domain contains a conserved m2 motif, which in authentic tRNA synthetases binds to the acceptor stem of the deacylated tRNA (Ruff et al., 1991). Mutation of this motif completely abrogates Gcn2’s ability to bind to tRNA as well as all catalytic activity (Dong et al., 2000; Wek et al., 1989; 1995). Furthermore, it has also been demonstrated that GCN2 is able to distinguish between deacylated and aminoacylated tRNA species, yet binds to deacylated tRNAs of different types with nearly equal affinities (Dong et al., 2000). Whilst the HisRS-like domain appears to be the major determinant for tRNA binding, the C- terminal domain of GCN2 also appears to contribute. Specifically, three lysine residues (yeast Gcn2 K1552, K1553 and K1556) present on an amphipathic helix of the C-terminal domain are essential for deacylated tRNA binding (as well as ribosome binding, which is discussed later) (Dong et al., 2000; Zhu and Wek, 1998). The dimerisation of the HisRS-like domain is also thought to be essential for tRNA binding (Qiu et al., 2001); however, it is interesting to note that the presence of tRNA does not drive dimerisation, indicating that the change in oligomeric state is not as important for activation as for other eIF2α kinases such as PKR and PERK. GCN2 is also activated in conditions other than amino acid starvation, which has led to the hypothesis that other signals may also activate the kinase, such as glucose starvation, Ultraviolet (UV) radiation and proteasome inhibition (Deng et al., 2002; Jiang and Wek, 2005; Yang et al., 2000). However, in each of these cases there is some evidence for the build up of deacylated tRNA, implying this might still act as the direct GCN2 effector. This is supported by the finding that in every case, the activation of the kinase is dependent on the m2 motif in the HisRS-like domain. Under conditions of glucose starvation, for example, the lack of glucose could prompt a metabolic switch, causing the cell to begin to utilise amino acids as the main carbon source (Yang et al., 2000). This would effectively cause a secondary amino acid starvation state, and so lead to the activation of GCN2. In the presence of UV radiation, it has been suggested that cells will rapidly synthesise nitric oxide via the enzyme Introduction 24 nitric oxide synthase, in order to protect the cell from DNA damage. One of the substrates of this synthetic pathway is arginine, and so a large increase in the rate of this pathway could deplete intracellular arginine levels sufficiently to activate GCN2. Lastly, under conditions in which the proteasome is inhibited, GCN2 is thought to be activated in response to the resulting reduction in the concentration of free amino acids within the cell, as the activation can be inhibited by the addition of free amino acids (Suraweera et al., 2012). 1.6.2. GCN1 General Control Nonderepressible 1 (GCN1 in humans, Gcn1 in yeast) was identified in the same genetic screen as Gcn2 (Hinnebusch, 1985). This screen demonstrated the importance of Gcn1 to the cellular amino acid starvation response, as a yeast strain with Gcn1 deleted did not show GCN2 activation under conditions of nutrient starvation. It has subsequently been shown that Gcn1 is not essential for Gcn2 activity per se, but is rather required for the activation of Gcn2 in response to amino acid starvation. Human GCN1 consists of 2671 amino acids, and has a molecular weight of approximately 290 kDa (Figure 1.14). The middle portion of the protein consists of predicted HEAT repeats (a helical repeating fold first identified in the proteins Huntingtin, eukaryotic elongation factor 3 (eEF3), protein phosphatase 2A and mTOR). Some sequence homology to eEF3 appears to extend past the end of the HEAT repeats, but the significance of this is not clear. eEF3 is an unusual translational elongation factor that is essential only in fungi (Sandbaken et al., 1990; Skogerson and Wakatama, 1976), and it has been postulated to have a role in the transfer of deacylated tRNA out from the ribosomal E site (Andersen et al., 2006; Triana-Alonso et al., 1995). Introduction 25 Figure 1.14. The domain structure of human GCN1. Areas that have been implicated in certain functions are indicated. The middle of GCN1 shows sequence homology to the translation factor eEF3, most of which forms HEAT repeats. Most of the protein sequence seems to contribute to ribosome binding in some way, but two major motifs (M7 and M1) have been identified (residues 793 – 834 and 1508 – 1515 respectively; indicated in grey). The binding determinant for GCN20 (shown in blue) has been narrowed down to the eEF3-like HEAT repeats, and is dependent on the conserved glycine 1494. GCN2 binds to the C-terminus of the protein (green), and is dependent on the conserved arginine 2312. Residue numbers are for the human construct, and in some cases have been extrapolated from yeast data via sequence alignment using ClustalOmega (Sievers et al., 2011). Gcn1 associates with Gcn2 in vivo in yeast (Garcia-Barrio et al., 2000), and this has been corroborated through in vitro pull-downs (Kubota et al., 2000). These studies have demonstrated that Gcn1 appears to bind to the N-terminal RWD domain of Gcn2. The binding regions of Gcn1 are much less clear, but the major determinant seems to consist of the C- terminal end of the protein, from approximately residue 2050 (Figure 1.14). In addition, regions at the N-terminus of the protein also seem to play a role in the interaction, as truncation of this region reduces the efficiency of immunoprecipitation experiments (Sattlegger and Hinnebusch, 2000). Overexpression of the N-terminal RWD domain of Gcn2 leads to impaired cell growth under conditions of amino acid starvation (Garcia-Barrio et al., 2000; Kubota et al., 2000; Sattlegger and Hinnebusch, 2000), and this correlates with a reduction in eIF2α phosphorylation. The interaction is absolutely dependent on the yeast Gcn1 residue arginine 2259: when this is mutated to an alanine, no interaction between Gcn1 and Gcn2 can be observed, either in vivo or in vitro. Furthermore, Gcn2 activation is completely abolished (Sattlegger and Hinnebusch, 2000). Together this evidence points towards a model in which a direct interaction between Gcn2 and Gcn1 is essential for Gcn2’s ability to recognise a state of amino acid deprivation. Gcn1 is also able to interact with ribosomes (Sattlegger and Hinnebusch, 2000). Attempts to narrow down the binding determinant have been relatively challenging, and it seems as though much of the protein is involved in the interaction. However, two conserved, non- contiguous regions have been identified, called M7 and M1, which seem to be particularly important (Figure 1.14) (Sattlegger and Hinnebusch, 2005). The M7 region is made up of a series of basic residues towards the N-terminus of the protein, whilst M1 consists of the Introduction 26 sequence ExxWRTKR and is located in the central, eEF3-like portion of the protein. Mutation of either of these motifs is sufficient to decrease the association of Gcn1 with translating ribosomes, and mutating both has a cumulative effect. Not only is the interaction between Gcn1 and ribosomes significantly reduced, but the cellular levels of phosphorylated eIF2α are similarly reduced, implying that Gcn1’s ability to bind ribosomes is key to its role as an activator of Gcn2 (Sattlegger and Hinnebusch, 2005). The overexpression of Gcn1 confers an increased sensitivity to paramomycin, an antibiotic that binds in the ribosomal A site (Ogle et al., 2003), whilst knocking out the protein confers resistance (Sattlegger and Hinnebusch, 2000). This sensitivity to paramomycin correlates with Gcn1’s ability to interact with ribosomes, as measured for the different constructs, implying that the presence of Gcn1 at the ribosome is able to render the ribosome more susceptible to paramomycin binding. The mechanism for this increase in sensitivity is still unclear, but it does indicate that Gcn1 may be binding to somewhere near the A site of the ribosome. 1.6.3. GCN20 General Control Nonderepressible 20 (GCN20 in humans, Gcn20 in yeast) is another protein first identified in the genetic screen to identify components of the GAAC pathway (Hinnebusch, 1985). GCN20 is a member of the ATP-binding cassette (ABC) family of proteins, and contains two ABC cassettes towards its C-terminus. This protein has also been shown to have sequence homology to the translation factor eEF3. Yeast Gcn20 forms a stable dimer with Gcn1 (Vazquez de Aldana et al., 1995), with the N-terminus of Gcn20 binding to the central portion of Gcn1 (Figure 1.14). This interaction is dependent on glycine 1494 (Marton et al., 1997). Gcn20 also has the intrinsic ability to associate with ribosomes (Marton et al., 1997), but in the presence of ATP the Gcn1-Gcn20 complex has a higher affinity for ribosomes than either component alone (Marton et al., 1997). This is not the case in the absence of ATP: in this case Gcn20 appears to act as a negative regulator of ribosome binding. A mammalian homologue of Gcn20 has not yet been identified, despite much investigation. The most likely candidate is an ABC-family protein known as ABC50, which contains some sequence homology to yeast Gcn20, and interestingly also associates with eIF2. However, ABC50 does not functionally rescue yeast cells without Gcn20 (Tyzack et al., 2000), and so its role is still unclear. Introduction 27 1.6.4. IMPACT The imprinted gene with ancient domain protein (IMPACT) was originally discovered through a mouse screen for imprinted genes (Hagiwara et al., 1997). It is highly conserved amongst most eukaryotes, and is present in yeast where it is denoted Yeast IMPACT Homologue 1 (Yih1). Yih1 has been identified as a negative regulator of Gcn2, as overexpression reduces both Gcn2 autophosphorylation and eIF2α phosphorylation. The results of a yeast two-hybrid screen indicated that Yih1 was able to interact with Gcn1, and sequence analysis demonstrated that Yih1 contains a region of sequence homology to the RWD domain of Gcn2 (Figure 1.15) (Kubota et al., 2000; Sattlegger et al., 2011). It was therefore proposed that Yih1 could interfere with Gcn2 activation through competing with Gcn2 for the RWD domain-binding site on Gcn1. This hypothesis is supported by the observation that the overexpression of the RWD domain of Yih1 is sufficient to inhibit Gcn2 activity and impair cell growth under conditions of amino acid starvation (Kubota et al., 2000). Furthermore, this phenotype can be restored to wild-type via the overexpression of Gcn2 (Sattlegger et al., 2004). It also seems likely that the interaction is very similar between Gcn2 and Gcn1, and Yih1 and Gcn1 as an in vitro binding assay showed that the Yih1-Gcn1 interaction is dependent on the Gcn1 arginine 2259 residue in a similar way to the Gcn2-Gcn1 interaction. Figure 1.15. The domain structure of Yih1. The protein consists of an N-terminal RWD domain followed by a C-terminal ancient domain, connected by a highly charged linker. Areas that have been implicated in binding to ribosomes, Gcn1 and actin are indicated in grey, orange and pink respectively. The evidence suggests that this role for Yih1 is conserved for IMPACT in mammals. Co- immunoprecipitation studies in mouse brain extracts have demonstrated that IMPACT is able to co-precipitate with GCN1 and vice versa, and that this interaction is again dependent on the conserved arginine (R2312) in GCN1’s C-terminus (Pereira et al., 2005). Moreover, when overexpressed in mouse embryonic fibroblasts, IMPACT results in a decreased level of interaction between GCN1 and GCN2 (Cambiaghi et al., 2014), and conversely knockdown of IMPACT in undifferentiated neuronal-like N2a cells results in increased activation of GCN2 during amino acid starvation (Roffé et al., 2013). Lastly, overexpression of IMPACT Introduction 28 is able to functionally complement for Yih1 in yeast cells (Cambiaghi et al., 2014; Pereira et al., 2005). Whilst IMPACT appears to have a key role in the regulation of GCN2’s kinase activity, knock down of either Yih1 or IMPACT in yeast cells or undifferentiated neuronal-like cells respectively does not lead to a global increase in the level of eIF2α phosphorylation, indicating that its role may be important only under certain conditions (Sattlegger et al., 2004). Possible insights into this have come from the finding that Yih1 is actively regulated by actin. Free G actin co-immunoprecipitates with Yih1 in a 1:1 stoichiometric complex, and this interaction is conserved for IMPACT (Sattlegger et al., 2004; Waller et al., 2012). Delineation of the binding regions indicate that the binding sites for Gcn1 and monomeric actin appear to overlap, implying that Yih1 cannot associate with both at the same time (Figure 1.15) (Sattlegger et al., 2011). This suggests that actin could sequester the cellular Yih1, and thus prevent it from inhibiting Gcn2. This interaction appears to have a role in vivo, as the genetic reduction in the amount of actin in a cell leads to impairment of the cell’s ability to recover from amino acid deprivation, and this can be reverted to a near wild-type phenotype by the deletion of Yih1 (Sattlegger et al., 2004). The implications of this finding are still under debate, but it is possible that this represents a level of spatial control of GCN2’s activity, dictated by the level of actin polymerisation. For example, the actin scaffold provides a platform for the translational machinery in areas of the cell that require maximal protein translational efficiency (such as the growing bud of yeast). In these areas, all the actin present is likely to be polymerised, and so would not be available to bind to Yih1. In this case, Yih1 is free to bind to Gcn1, and thus prevent it from stimulating Gcn2’s kinase activity, so allowing translation to proceed at a maximal rate. In areas that do not require such maximal growth, there is more unpolymerised actin present, and so it will be able to sequester Yih1 and thus permit Gcn1 to activate Gcn2 under non-optimal translational conditions. Like many of the proteins associated with the GCN2-dependent branch of the ISR, Yih1/IMPACT is also able to associate with translating ribosomes in a manner independent of GCN1 (Figure 1.15) (Roffé et al., 2013; Waller et al., 2012). This interaction seems to be relatively transient, as only a fraction of the total cellular Yih1/IMPACT is found associated with polysomes, but could have important implications for further regulatory control. Whilst the expression levels of Yih1 do not seem to change significantly throughout the yeast cell cycle, the expression of IMPACT seems to be highly regulated between different tissue Introduction 29 types. IMPACT is highly expressed in neuronal cells, particularly within the suprachiasmatic nucleus of the hypothalamus, which typically displays very low levels of basal eIF2α phosphorylation (Pereira et al., 2005), indicating that IMPACT may have a role in dampening the activation of GCN2 in these areas. Given the association of GCN2 with a variety of complex biological processes in the neurons, such as the control of neuronal synaptic plasticity, the formation of long-term memories and behavioural adaptation (Chesnokova et al., 2017; Costa-Mattioli et al., 2005; Maurin et al., 2005), this could represent a potential mechanism for the differential regulation of GCN2 in neurons. Furthermore, more research in this area could provide insights into the importance of this pathway in neuronal development, which is currently still very abstruse. 1.6.5. The Ribosome The role of the ribosome in the regulation of GCN2 is still relatively unclear, but progress is being made in understanding the link between GCN2 and the translational machinery. Co- migration assays have indicated that yeast Gcn2 is able to loosely associate with translating ribosomes, and that this is mediated by the C-terminal domain of Gcn2 (He et al., 2014; Ramirez et al., 1991). The CTD of Gcn2 shows sequence similarity to the core of the double- stranded RNA binding domain (dsRBD), including three lysine residues within an amphipathic helix that are conserved across the dsRBD family. Substitution of these residues in Gcn2 with three alanine residues significantly impairs Gcn2’s ability to associate with ribosomes in vivo, implying that the RNA binding activity of this domain is at least partially responsible for the interaction. Interestingly, these three lysines have also been implicated in the ability of Gcn2 to bind to deacylated tRNA. The mechanism for this is unclear, but seems to hint at the possibility of a mechanistic link between tRNA and ribosome binding. A scenario in which deacylated tRNA is able to compete with the ribosome for binding to GCN2 has been hypothesised; however, most of the current data indicates that amino acid starvation has no significant effect on the level of Gcn2-ribosome-containing complexes. Nevertheless, the predicted transience of the interaction could mean that changes are occurring, but they are not sufficiently robust to be detected in large-scale, steady state assays (Garcia-Barrio et al., 2000; Marton et al., 1997; Ramirez et al., 1991). The interaction with ribosomes is considered to be essential for Gcn2’s ability to recognise and respond to amino acid starvation, as deletion of the CTD or mutation of the three lysine Introduction 30 residues is sufficient to prevent activation of the kinase (Ramirez et al., 1991). There is also evidence to indicate that other regions of Gcn2 have an effect upon the ribosomal interaction, as deletion of other portions of the protein cause a decrease in the proportion of Gcn2 bound to ribosomes (Ramirez et al., 1991). The ability of many other components of the regulatory pathway (such as GCN1, GCN20 and IMPACT) to associate with the ribosome could also play a role in its recruitment. Theoretically, the ribosome could be acting as a simple platform for the assembly of GCN2- containing complexes, with no real catalytic effect. However, recent evidence indicates that a specific region of the ribosome has a direct activating effect upon GCN2 (Jiménez-Díaz et al., 2013). Ribosomes consist of a huge number of proteins and RNA molecules, some of which are core, highly structured parts of the translational machinery, and others that are more dynamically associated and have been suggested to have roles in translational regulation. An example is the highly flexible P stalk of the eukaryotic ribosome, which is a pentameric complex consisting of the protein uL10 (previously known as P0 (Ban et al., 2014)) as well as two P1/P2 heterodimers (Lee et al., 2010; 2011). The P1 and P2 proteins exist in a dynamic equilibrium between free and ribosome-bound, and their localisation is thought to be a potential mechanism for changes in translation under different conditions. uL10 contains two helices to which each copy of the P1/P2 heterodimer is able to bind via their N-terminal dimerisation domains. Connected to each copy of the N-terminal domains of P1 and P2 are long, unstructured C-terminal tails, which are thought to extend outwards from the ribosome (Lee et al., 2013). This P stalk has been shown to be necessary for the recruitment of translation factors during the elongation stage of protein synthesis (Bargis-Surgey et al., 1999; Mochizuki et al., 2012; Naganuma et al., 2010), and also plays a role in the stimulation of GTPase activity as part of the translational progression (Abo et al., 2004), which posits the idea that these flexible tails are acting as fishing lines that can ‘catch’ elongation factors in the near vicinity and recruit them to the translational apparatus. A crystal structure of the heptameric complex from the archaea Pyrococcus horikoshii has been solved (Figure 1.16) (Naganuma et al., 2010). These proteins are highly conserved between archaea and eukaryotes, but there are key differences. Whilst the stoichiometry of this complex in eukaryotes is uL10 – (P1-P2) – (P1-P2), the archaeal stalk complex is a heptamer consisting of three P1 homodimers binding to the three spine helices of uL10 (uL10 – (P1-P1) –(P1-P1) Introduction 31 – (P1-P1)). Despite this, the archaeal stalk is able to recruit eukaryotic elongation factors at a similar rate to the eukaryotic stalk, indicating that they are fundamentally similar (Nomura et al., 2006). These proteins that make up the eukaryotic pentameric complex have recently been implicated in the activation of yeast Gcn2 (Jiménez-Díaz et al., 2013). When not bound to ribosomes, P1 and P2 were demonstrated to act as potent activators of Gcn2 in vitro, and their deletion in yeast inhibits Gcn2 activation under certain conditions. Puzzlingly, these conditions include glucose starvation or osmotic stress, but not amino acid starvation. More research on this topic is required in order to clarify the mechanism of activation. Figure 1.16. The structure of the archaeal P stalk complex (PDB code 3A1Y) (Naganuma et al., 2010). uL10 (purple) is part of the large ribosomal subunit, and extends outwards from the ribosome. Three P1 homodimers bind to a helical region of uL10 via their N-terminal domains, and their unstructured C- terminal tails extend away from the uL10 spine. The C-terminal tails for uL10 and P1 were not included in the crystal structure, and so have been represented in cartoon form. Each C-terminal tail is predicted to culminate in two helices, and these have been represented by cartoon ovals. The conservation of these mechanisms is a subject under some debate. Co-migration assays indicate that mammalian GCN2 has a much lower ability to associate with ribosomes in comparison to yeast Gcn2 (He et al., 2014); however, it seems likely that some homologous mechanism exists. Indeed, some experiments in mammalian systems have provided evidence that the mechanism of GCN2 activation in mammalian cells may be subject to additional regulation based on the ribosomal state. Recently, evidence has emerged that mammalian GCN2 can be activated in a manner independent of tRNA (Ishimura et al., 2016; 2014). The Introduction 32 authors developed a mouse model with a genetic background deficient in both a neuronal tRNA and the ribosome recycling factor GTP-binding protein 2 (GTPBP2). This combination led to a high incidence of stalled elongation complexes in the neurons of the mice, allowing the authors to examine the effects of widespread ribosomal stalling. They observed a significant rise in GCN2 activity compared to wild-type cells, resulting in an increased level of eIF2α phosphorylation. These data indicate the possibility that the signal for activation is translational stalling, and implicate GCN2 as a direct or indirect sensor of ribosomal processivity. It seems possible that this is the mechanism by which deacylated tRNA activates GCN2, as a large influx of deacylated tRNA into a cell would very likely lead to widespread ribosomal stalling. For GCN2 to fulfil a role as a sensor of ribosome processivity, it must require a level of specificity, i.e. be able to determine the difference between a stalled and an elongating ribosome. As of yet, there is no answer to this question. 1.6.6. Translational Elongation Factors Two translation factors have been identified as having a direct impact upon Gcn2 activity: eEF3 and eEF1A. eEF3 has a role in the transfer of the deacylated tRNA from the ribosomal E site after its cognate amino acid has been incorporated into the nascent polypeptide chain, whilst eEF1A is involved in the delivery of an aminoacylated tRNA to the A site. eEF1A is an essential translation factor in all eukaryotes; however, a mammalian homologue to eEF3 has not yet been found, and it has been suggested that mammalian ribosomes do not require an additional factor to catalyse the release of deacylated tRNA. Overexpression of eEF3 in yeast leads to a significant reduction in the level of eIF2α phosphorylation, indicating that it has some effect on Gcn2 (Visweswaraiah et al., 2012). Data from genetic studies in yeast have indicated that this is due to eEF3 binding to the ribosome and thereby preventing the formation of a functional Gcn1-ribosome complex (Visweswaraiah et al., 2012). This would mean that Gcn1 is no longer able to stimulate the kinase activity of Gcn2, and thus would result in the indirect inhibition of Gcn2 activation. It is not the case that eEF3 prevents a Gcn1-ribosome complex altogether, as overexpression of the elongation factor does not seem to affect the overall amount of the complex, but it is possible that the presence of eEF3 alters the interaction in such a way that the Gcn1-ribosome interaction is no longer a productive one. Introduction 33 In vitro binding studies have demonstrated a direct interaction between eEF1A and Gcn2, mediated by the C-terminal domain of Gcn2, whilst activity assays have shown eEF1A to be a direct inhibitor of eIF2α phosphorylation (but not Gcn2 autophosphorylation) (Visweswaraiah et al., 2011). This suggests that eEF1A could bind to Gcn2 in such a way that prevents substrate recruitment, without affecting the kinase’s ability to undergo the conformational rearrangements that are thought to be essential for autophosphorylation. In vivo studies have shown that under conditions of amino acid starvation the interaction between eEF1A and Gcn2 is significantly reduced (Visweswaraiah et al., 2011), which correlates with data showing that the addition of deacylated tRNA to the in vitro components leads to a decrease in eEF1A-Gcn2 complex formation. Given that Gcn2’s CTD is implicated in both interactions, this raises the possibility of a competition mechanism between deacylated tRNA and eEF1A. A Model for the Activation of GCN2 1.7. From the data discussed above, a model has emerged in the field concerning the mechanism for activation of GCN2 (Figure 1.17). GCN1, GCN20 and the dimeric GCN2 kinase form a heterotrimeric complex, which can associate with a translating ribosome. When amino acid levels become depleted, and therefore there are no cognate aminoacylated tRNAs available, it has been suggested that the corresponding deacylated tRNA will enter the A site (Murchie and Leader, 1978), as occurs in bacteria (Agirrezabala et al., 2013; Brown et al., 2016). However, some argue that this is unlikely, due to the requirement of the translation factor eEF1A for the recruitment of tRNA to the ribosome, and the low affinity of eEF1A for a deacylated tRNA (Gromadski et al., 2007). If deacylated tRNA were to be present in the A site via an unknown mechanism however, the model suggests the tRNA will then be somehow transferred from the ribosomal A site to the bipartite binding site on GCN2. The mechanisms behind this transfer are very unclear; however, it has been suggested that the process could be catalysed by the eEF3-like activity of GCN1. If the tRNA is not present in the A site, then the way in which it is recruited to GCN2 is unknown. This basic mechanism is subject to substantial additional levels of regulation, as described previously. GCN2 activation is inhibited in regions of the cell in which maximal rates of translation are required, via the release of the protein Yih1/IMPACT from the actin cytoskeleton. Activation of the pathway is also linked to the translational cycle via inhibitory Introduction 34 interactions with the elongation factors eEF3 and eEF1A. When elongation is underway, the presence of eEF3 at the ribosome E site inhibits the formation of a productive GCN1- ribosomal complex, thus inhibiting the ability of GCN1 to activate GCN2. Furthermore, the regular recruitment of eEF1A permits complex formation between GCN2 and eEF1A, thus inhibiting eIF2α substrate recruitment. Figure 1.17. Schematic depiction of a model for the effects of different regulators upon GCN2 activation. GCN1, GCN2 and GCN20 form a trimeric assembly near the A site of the ribosome. In the event of a build up of deacylated tRNA, the deacylated tRNA will enter the A site. The eEF3-like activity of GCN1 then acts to somehow transfer the tRNA from the ribosomal A site to the bipartite binding site on GCN2 (1). This activates GCN2 via a conformational rearrangement and autophosphorylation, and allows it to phosphorylate eIF2α (2). IMPACT can inhibit this process by binding to GCN1 (3) and preventing the formation of a GCN1-GCN2 complex. However, IMPACT is negatively regulated by the presence of free actin (4), meaning that in areas in which actin is mainly polymerised, IMPACT will be free to inhibit GCN2 activation. eEF1A, the translation factor responsible for the recruitment of aminoacylated tRNAs during elongation, can inhibit the process by binding to GCN2 and inhibiting substrate access to its active site (5). Finally, eEF3 can also downregulate GCN2 activation by inhibiting the formation of a productive GCN1-ribosomal complex (6). 35 Aims of this Thesis A cell’s ability to monitor its environment and adapt its metabolism accordingly is fundamental to its survival. The protein kinase GCN2 is able to recognise when the cell becomes depleted of amino acids, and initiates a program of stress response to help the cell recover. Despite its importance, the manner by which GCN2 is able to orchestrate this response and the roles of the various auxiliary factors are not well understood. The mechanism of activation of Gcn2 in yeast has been studied extensively, and this has permitted the proposition of a model for how the kinase is able to recognise and respond to amino acid starvation. However, rigorous testing of this model has proved difficult, as the bulk of the work originates from genetic studies performed in yeast. Given the functional proximity of this pathway to the translational machinery, it is difficult to study the regulatory influences of specific factors using a whole cell model, as it is almost impossible to study the effects of adjusting one factor in isolation. Furthermore, some data suggest significant differences between the yeast and the human proteins, undermining the ability to extrapolate from one to the other. For this reason, a system in which the activation of the kinase could be reconstituted in vitro would be very valuable to the field. Such a system would allow the effects of different regulatory inputs to be tested, and potentially allow the characterisation of the molecular details of the process by which GCN2 is activated upon the recognition of amino acid starvation. The role of GCN2 in the development of diseases such as neurological degeneration disorders and cancer means that a comprehensive understanding of the human kinase could be extremely valuable to inform the development of potential therapeutic candidates. The aim of this thesis is to shed light on the workings of this kinase and the intricate layers of regulation that it is subject to. Chapter Two describes the development and optimisation of a system in which the activity of human GCN2 is reconstituted. This allows the characterisation of the effects of various regulators, ultimately enabling the identification of the ribosome as a potent activator of GCN2. A direct interaction between GCN2 and the ribosome was then demonstrated, and initially characterised through biochemical methods and truncation analysis. To gain more information about this interaction, Chapter Three describes an extensive Hydrogen-Deuterium exchange-mass spectrometry (HDX-MS) study that resulted in the identification of the binding site for GCN2 on the ribosome. Together, these data allow Aims of this Thesis 36 the proposition of a model describing how human GCN2 recognises the depletion of amino acids through the ability to monitor translational processivity. Efforts to gain high-resolution structural insights into GCN2 alone or as part of a functional complex with the ribosome are described in Chapter Four. Whilst challenging, these studies resulted in the high-resolution structure of the pseudokinase domain of GCN2 and a low- resolution envelope for the full-length GCN2 from negative stain electron microscopy. These data represent the first structural insights into the human GCN2 kinase. Finally, Chapter Five describes a different and often underappreciated aspect of kinase regulation, introduced in more detail in Chapter Five. The chaperone Hsp90 is known to have an important role in the stability and maintenance of approximately 60 % of the human kinome. Despite this, the mechanisms behind the interaction between the chaperone and kinases are not well understood. Furthermore, the distinction between ‘client’ and ‘non-client’ kinases is not clear, with closely related kinases showing very different Hsp90-dependencies. In this chapter, the interaction between the Hsp90 co-chaperone Cdc37 and several kinases is characterised using a combination of biochemistry and HDX-MS. These studies identified an important link between kinase stability and Hsp90-dependence, and provide key insights into the molecular effects of the interaction. 37 Chapter Two – Reconstitution of GCN2 kinase activation with purified factors 2.1. Introduction By far the majority of the work so far looking at the mechanisms underlying GCN2 activation has been performed by genetic studies on yeast, and then extrapolated to the human protein. However, there is evidence for significant differences between the yeast and the mammalian proteins, and so it is necessary to test these hypotheses on the mammalian protein to check these assumptions. Furthermore, genetic studies make it very challenging to unpick precise molecular mechanisms, as altering a single component of the system in isolation is nearly impossible. For this reason, the development of a reconstituted system of GCN2 activation would be very desirable, as it would permit the modulation of each factor in turn to try to decipher the exact molecular mechanisms and regulation behind GCN2 activation. This requires the ability to purify each potential component of the system in a functional state. Furthermore, these components must be able to be assembled, and activation of the kinase quantitated. Once a stable and tractable assay has been established, the importance of each component could be analysed by manipulating the constituent factors. This would allow the characterisation of the activation mechanism for GCN2 in a detailed and comprehensive manner, as well as providing an opportunity to test the hypotheses that are currently in the literature concerning facets of the activation pathway. In this chapter, the development of such an in vitro system is described. Furthermore, the system is then utilised to begin to characterise this activation pathway in a thorough and detailed manner. The importance of various putative components of the system is probed, allowing the proposition of a new model of activation that is consistent with both in vivo data from the literature as well as the new molecular insights gained from the reconstituted system. Reconstitution of GCN2 Kinase Activation 38 2.2. Materials and Methods 2.2.1. Protein Expression and Purification DNA encoding human GCN2 (NCBI reference number: NP_001013725.2), GCN1 (NCBI reference number: NP_006827.1) and GCN20 (NCBI reference number: NP_001020262.1) constructs were inserted into individual pAceBac1 vectors with an N-terminal tag (either His6, twin StrepII-His6, or glutathione-S-transferase (GST)) followed by a tobacco etch virus (TEV) protease site. Plasmids encoding each gene were integrated into EMBacY baculoviral DNA via Tn7 transposition. Baculoviruses were produced by transfection of Spodoptera frugiperda (Sf9) cells at a density of 0.5x106 cells/mL with 2 – 4 µg of DNA using FuGENE HD Transfection Reagent (Promega E2311). Cells were grown at 27 ˚C for 11 days before the virus was harvested by centrifugation. GCN2, GCN1 and GCN20 were expressed in Sf9 cells by infecting cells at a density of 1.5x106 cells/mL with 15 mL of virus per 500 mL of cells. Cells were grown at 27 ˚C for 55 hours before being harvested, washed with ice-cold phosphate-buffered saline (PBS) and frozen in liquid nitrogen. The Sf9 cell pellets (from 1.5 – 2 L cells) were thawed and lysed in 100 mL lysis buffer E [20 mM tris(hydroxymethyl)aminomethane-HCl (Tris-HCl) pH 8.0 (room temperature), 150 mM NaCl, 5 % v/v glycerol, 2 mM β-mercaptoethanol, one cOmplete EDTA-free protease inhibitor tablet (Roche 04693132001)] per pellet. Cells were lysed via a probe sonicator (Sonics Vibra-Cell VCX750) for 3 minutes (10 s on, 10 s off, 65 % power), then 6 U/mL Universal Nuclease (Pierce 88702) was added. The lysate was then subjected to centrifugation at 140,000g for 45 minutes at 4 ˚C in a Ti45 rotor (Beckman Coulter 339160). The resulting supernatant was passed through a 0.45 µM syringe filter (Millipore Sigma SE2M230I04), before being subjected to affinity purification on the basis of the N-terminal tag. His-tagged proteins The filtered cell lysate was loaded on to a 5 mL HisTrap HP column (GE Healthcare Life Sciences 17-5248-02), equilibrated in Ni-A buffer [20 mM Tris-HCl pH 8.0 (room temperature), 100 mM NaCl, 5 % v/v glycerol, 10 mM imidazole pH 8.0, 2 mM β- mercaptoethanol] at a flow rate of 3 mL/min. The column was then washed with up to 200 mL Ni-A buffer, and the proteins were eluted with a gradient of 0 – 50 % Ni-B buffer [20 mM Reconstitution of GCN2 Kinase Activation 39 Tris-HCl pH 8.0 (room temperature), 100 mM NaCl, 5 % v/v glycerol, 200 mM imidazole pH 8.0, 2 mM β-mercaptoethanol]. The peak fractions were pooled and diluted 1:1 with Q0 buffer [20 mM Tris-HCl pH 8.0 (room temperature), 5 % v/v glycerol, 2 mM β-mercaptoethanol] to reduce the NaCl concentration to 50 mM. Strep-tagged proteins The filtered lysate was loaded on to a 5 mL StrepTrap HP column (GE Healthcare Life Sciences 28-9075-47), equilibrated in Strep-A buffer [20 mM Tris-HCl pH 8.0 (room temperature), 150 mM NaCl, 5 % v/v glycerol, 2 mM β-mercaptoethanol], at a flow rate of 2 mL/min. The column was washed with 50 mL Strep-A buffer, and then the protein was eluted with 6 mM desthiobiotin in Strep-B buffer [20 mM Tris-HCl pH 8.0 (room temperature), 150 mM NaCl, 5 % v/v glycerol, 2 mM β-mercaptoethanol, 6 mM desthiobiotin (IBA 2-1000- 005)]. The peak fractions were pooled and diluted 1:2 with Q0 buffer to reduce the NaCl concentration to 50 mM. GST-tagged proteins The filtered lysate was incubated with 1 mL Glutathione Sepharose 4B resin (GE Healthcare Life Sciences 17-0756-05), equilibrated in GST buffer [20 mM Tris-HCl pH 8.0 (room temperature), 100 mM NaCl, 5 % v/v glycerol, 5 mM dithiothreitol (DTT)] for 90 minutes. The sample was then transferred to a BioRAD gravity flow column, and the beads were washed with approximately 100 mL GST buffer. The protein was then eluted from the beads in 12 mL GST-E buffer [100 mm Tris-HCl pH 8.0 (room temperature), 100 mM NaCl, 15 mM glutathione] and then diluted 1:1 with Q0 buffer to reduce the NaCl concentration to 50 mM. After affinity purifications, each protein was then further purified through ion exchange chromatography. The protein was loaded on to a 5 mL HiTrap Q HP column (GE Healthcare Life Sciences 17-1153-01), equilibrated in Q-A buffer [20 mM Tris-HCl pH 8.0 (room temperature), 50 mM NaCl, 5 % v/v glycerol, 2 mM β-mercaptoethanol], at a flow rate of 3 mL/min. The column was washed with up to 50 mL Q-A buffer, and then the protein was eluted with a gradient of 0 – 100 % Q-B buffer [20 mM Tris-HCl pH 8.0 (room temperature), 1 M NaCl, 5 % v/v glycerol, 2 mM β-mercaptoethanol]. The peak fractions were pooled and concentrated to 1 mL using a 15 mL centrifugal filter (Amicon UFC901024 – UFC910024, depending on protein size) before injection on to a HiLoad 16/60 Superdex 200 column (GE Reconstitution of GCN2 Kinase Activation 40 Healthcare Life Sciences 17-1069-01), equilibrated with GF buffer [20 mM 4-(2- hydroxyethyl)-1-piperazineethanesulfonic acid (HEPES) pH 7.5, 150 mM NaCl, 2 mM tris(2- carboxyethyl)phosphine (TCEP)], at a flow rate of 1 mL/min. The peak fractions were concentrated using a 15 mL centrifugal filter (Amicon UFC901024 – UFC910024, depending on protein size) to 2 - 30 mg/mL before being aliquoted and frozen in liquid nitrogen. Untagged versions of the proteins were purified in the same way, except with the addition of a TEV cleavage step, in between affinity purification and ion exchange. TEV protease (His6- TEV-S219V at 0.5 mg/mL was added to the eluent from the StrepTrap column. The proteins were incubated at 4 ˚C overnight, and then subsequently diluted and loaded on to a HiTrap Q HP column as above. DNA encoding human eIF2α (NCBI reference number: NP_004085.1) was inserted into the vector pOPTH with an N-terminal His6 tag followed by a TEV protease site. The plasmid was transformed into chemically competent BL21 Star (DE3) cells, and cells were grown overnight before being transferred to a 50 mL starter culture in 2xTY media containing 0.1 mg/mL Ampicillin. The starter culture was incubated at 37 ˚C for 90 minutes, then 12 mL of starter culture was added to 4 x 1 L 2xTY media containing Ampicillin. Cultures were incubated at 37 ˚C until the optical density reached 1, and then protein expression was induced by the addition of 0.3 mM isopropyl β-D-1-thiogalactopyranoside (IPTG). Cells were grown for a further 3 hours at 37 ˚C before being harvested, washed with ice-cold phosphate- buffered saline and frozen in liquid nitrogen. The bacterial cell pellets were lysed in 100 mL lysis buffer B [20 mM Tris-HCl pH 8.0 (room temperature), 100 mM NaCl, 5 % v/v glycerol, 2 mM β-mercaptoethanol, 0.5 mg/mL Lysozyme (Sigma L6876), 5 U/mL Universal Nuclease (Pierce 88702), one cOmplete EDTA- free protease inhibitor tablet (Roche 04693132001)] per 2 L cells. The cells were lysed using a probe sonicator (Sonics Vibra-Cell VCX750) for 5 minutes (10 s on, 10 s off, 65 % power), followed by centrifugation at 140,000g for 45 minutes at 4 ˚C in a Ti45 rotor (Beckman Coulter 339160). The supernatant was filtered using a 0.45 µM syringe filter (Millipore Sigma SE2M230I04) and then was loaded on to a 5 mL HisTrap HP column (GE Healthcare Life Sciences 17-5248-02), equilibrated in Ni-A buffer (described above), at a flow rate of 3 mL/min. The column was then washed with 200 mL of Ni-A buffer, followed by the elution of the protein via a gradient of 0 – 100 % Ni-B buffer. The peak fractions were then pooled and diluted 1:1 with Q0 buffer as above, before being loaded on to a 5 mL HiTrap Q HP Reconstitution of GCN2 Kinase Activation 41 column equilibrated in Q-A buffer as above. The column was then washed with 50 mL Q-A buffer before the protein was eluted via a 0 – 100 % gradient of Q-B buffer. Peak fractions were pooled and concentrated down to 1 mL using a 15 mL centrifugal filter (Amicon UFC901024). The sample was then injected on to a HiLoad 16/60 Superdex 75 column (GE Healthcare Life Sciences 17-1068-01), equilibrated with GF buffer at a flow rate of 1 mL/min. The peak fractions were concentrated using a 15 mL centrifugal filter (Amicon UFC901024) to approximately 8 mg/mL before being aliquoted and frozen in liquid nitrogen. 2.2.2. Buffer Optimisation by Differential Scanning Fluorimetry Differential scanning fluorescence changes were measured using a Prometheus NT.48 instrument (NanoTemper Technologies) over a temperature range of 15 – 90 ˚C, with a laser excitation of 50 %. The protein was diluted to 0.4 mg/mL in buffer containing 20 mM Buffer X, 150 mM NaCl, 2 mM TCEP. For the pH screen, the buffers used were from Hampton Research pH screen (Hampton Research HR2-241). The samples were incubated on ice for 90 minutes before thermal denaturation. Each sample was then run in duplicate. Detergent screening was performed in the same manner as described above. The sample was in buffer containing 20 mM HEPES pH 7.5, 150 mM NaCl and 2 mM TCEP ± detergent at different concentrations [3-[(3-Cholamidopropyl)dimethylammonio]-1-propanesulfonate (CHAPS) at 0.05 %, 0.1 % or 0.2 % (v/v); Tween20 at 0.0025 %, 0.005 % or 0.0075 % (v/v); n-Dodecyl-β-Maltoside (DDM) at 0.03 %, 0.04 % or 0.05 % (v/v); Lauryl Maltose Neopentyl Glycol (LMNG) at 0.03 %, 0.04 % or 0.05 % (v/v)]. 2.2.3. Size-Exclusion Chromatography-Multi Angle Light Scattering The SEC-MALS data were collected on a Heleos II 18-angle light scattering instrument (Wyatt) in line with an Optilab rEX online refractive index detector (Wyatt). The instrument was calibrated using Bovine Serum Albumin (BSA) (ThermoFisher Scientific 23209) as a standard. Purified proteins were diluted to 1 mg/mL and 100 µL volumes were injected on to a HiLoad 10/300 S-200 column, with a flow rate of 0.5 mL/min. The buffer was GF buffer (described above). The protein concentration was measured using the refractive index (RI), on the basis of 1 mg/mL protein having a ΔRI of 0.186 mL/g, and the observed scattering intensities were used to calculate the absolute molecular mass using the ASTRA software package (Wyatt). Reconstitution of GCN2 Kinase Activation 42 2.2.4. Surface Plasmon Resonance Surface Plasmon Resonance (SPR) data were collected using a BIAcore T200 instrument (GE Healthcare Life Sciences). GCN2 (at a concentration of 50 ng/µL) was directly immobilised on to a CM5 chip through amine coupling after activation of the chip with 1-ethyl-3-(3- dimethylaminopropyl)-carbodiimide (EDC) and N-hydroxysuccinimide (NHS). The protein was passed over the chip at 5 µL/min in a buffer consisting of 10 mM sodium acetate pH 5.0. The analyte was deacylated tRNA (purified from porcine pancreatic cells by the Hegde group) at a range of concentrations prepared by a 1:1 serial dilution in RNC buffer [50 mM HEPES pH 7.5, 100 mM KOAc, 5 mM MgAc2, 1 mM DTT]. The highest concentration was 40 µM. The tRNA was injected for 120 s in HEPES-buffered saline (HBS) buffer [20 mM HEPES pH 7.5, 150 mM NaCl] followed by a 600 s dissociation step. After each measurement, any remaining tRNA was dissociated with 2 M NaCl for 120 s. In parallel, a reference channel was blankly immobilised and injected with tRNA under the same conditions to check for non- specific binding. The data were analysed in Prism, and the buffer response from the reference channel was subtracted from the data. The KD was calculated using both kinetic and equilibrium methods. 2.2.5. Protein-Protein Interaction Pull-Downs Pull-downs were carried out by incubating purified Strep-tagged proteins at 1.5 µM with 20 µL StrepTactin Sepharose High Performance resin (GE Healthcare Life Sciences 28-9355-99) in GF buffer [20 mM HEPES pH 7.5, 150 mM NaCl, 2 mM TCEP] for 20 minutes. The beads were then washed with 3 x 1 mL GF buffer and then untagged proteins to be tested for binding were added to a final concentration of 3 µM. Samples were then taken for analysis by Sodium Dodecyl Sulphate-Polyacrylamide Gel Electrophoresis (SDS-PAGE) (totals). The reactions were incubated for 20 minutes before the beads were washed with 5 x 1 mL GF buffer. Finally sample buffer was added to elute the bound proteins. The samples were analysed by SDS-PAGE and visualised with Coomassie staining. 2.2.6. Ribosome Purification 5 mL of hemin-treated and nucleased rabbit reticulocyte lysate (RRL) (Green Hectares) (as described in (Sharma et al., 2010) was thawed gently and immediately placed on ice. The RRL was then transferred to two 3 mL polycarbonate thickwall tubes (Beckman Coulter Reconstitution of GCN2 Kinase Activation 43 349622) and centrifuged at 539,994g for 40 minutes at 4 ˚C in a TLA100.3 rotor (Beckman Coulter 349490). The supernatant was subsequently discarded. The pellet was washed with ribosomal wash buffer (RWB) [20 mM HEPES pH 7.5, 100 mM KOAc, 1.5 mM MgAc2, 0.1 mM ethylenediaminetetraacetic acid (EDTA) pH 7.4), 1 mM DTT] and then resuspended in RNC buffer [50 mM HEPES pH 7.5, 100 mM KOAc, 5 mM MgAc2 1 mM DTT] using a wide-bore pipette tip. The sample was then transferred to a 5 mL glass dounce and homogenised. Once entirely resuspended, the concentrations of KOAc and MgAc2 were increased to final concentrations of 750 mM and 15 mM respectively. The sample was then layered over a 1 mL sucrose cushion [20 mM HEPES pH 7.5, 750 mM KOAc, 15 mM MgAc2, 0.1 mM EDTA pH 7.4, 1 mM DTT, 1 M sucrose] and centrifuged at 539,994g for 1 hour at 4 ˚C in a TLA100.3 rotor (Beckman Coulter 349490). The supernatant was then discarded and the pellet washed with RNC buffer. The pellet was then resuspended in RNC buffer as before and homogenised in a 1 mL glass dounce, before being aliquoted and frozen in liquid nitrogen. 2.2.7. Radiolabelled Autophosphorylation Assay The reactions were performed in assay buffer [50 mM HEPES pH 7.5, 100 mM KOAc, 5 mM MgAc2, 1 mM DTT, 6.25 mM β-glycerophosphate] with protein components at a final concentration of 25 nM, ± 5 µM tRNA (deacylated tRNA was purified from porcine pancreatic cells by the Hegde lab) ± 50 nM ribosomes unless otherwise indicated. The reactions were started upon the addition of an ATP mix [final concentrations were 0.125 mM MgATP pH 8.0, 75 µCi/mL γ-33P-ATP, 18.75 mM MgCl2] and the transfer of the reactions from ice to 32 ˚C. The final volume of the reactions was 40 µL. Time course samples were taken at 0, 5 and 10 minutes. Reactions were quenched by the 1:1 addition of 2X Lithium Dodceyl Sulphate (LDS) sample buffer (NuPAGE NP0008). All the samples were boiled for 5 minutes before being run on a 10 % Bis-Tris Protein Gel (NuPAGE NP0303). The gels were then dried and exposed to film (Kodak 870 1302) for 24 – 48 hours. To test the effects of different antibiotics, 50 nM purified ribosomes were mixed with either 50 µg/mL cycloheximide, 200 µM emetine, 10 µM anisomycin or 50 µM didemnin B, and incubated at 32 ˚C for 5 minutes. The reactions were then removed to ice, and the autophosphorylation reactions performed as described above. Reconstitution of GCN2 Kinase Activation 44 2.2.8. Western Blot Autophosphorylation Assay These reactions were assembled exactly as described for the radiolabelled autophosphorylation assay, but were started with the addition of a non-radioactive ATP mix [final concentrations were 0.5 mM MgATP pH 8.0, 18.75 mM MgCl2] and the transfer to 32 ˚C. Samples were taken at 0, 5 and 10 minutes and quenched by the 1:1 addition of 2X LDS sample buffer. All the samples were boiled for 5 minutes before being run on a 10 % Bis-Tris Protein Gel (NuPAGE NP0303). The proteins were then transferred to a polyvinylidene difluoride (PVDF) membrane (Immobilon IPVH00010) and the membrane was blocked for 1 hour in 5 % BSA (Fisher Scientific BP1605) in TBST buffer [100 mM Tris-HCl pH 7.5 (room temperature), 150 mM NaCl, 0.1 % v/v Tween 20], and washed three times in TBST. The membrane was then incubated with the primary antibody [Abcam ab75836 (1:1000) for phospho-T899 GCN2; Abcam ab76949 (1:1000) for Strep-tagged GCN2] for 1 hour, then washed three times in TBST before being incubated with the secondary antibody [Cell Signalling #7074 (1:5000) for both] for 1 hour and subsequently washed three times in TBST. (All antibodies were diluted at the indicated dilutions in 5 % BSA in TBST.) Finally, the membranes were incubated with 2 mL SuperSignal West Pico PLUS chemiluminescent substrate (ThermoFisher Scientific 34577) and then exposed using a ChemiDoc Touch Imaging System (Bio-Rad). 2.2.9. Phosphorylation Analysis by Mass Spectrometry For each sample, a 400 µL phosphorylation reaction was assembled in assay buffer [50 mM HEPES pH 7.5, 100 mM KOAc, 5 mM MgAc2, 1 mM DTT, 6.25 mM β-glycerophosphate], containing 25 nM GCN2, ± 5 µM tRNA ± 50 nM ribosomes. The reactions were started with the addition of an ATP mix [final concentrations were 0.5 mM MgATP pH 8.0, 18.75 mM MgCl2], and incubated at 32 ˚C for 10 minutes. The reactions were then removed to ice. To purify GCN2 from the reactions, the same volume of denaturation buffer [2 % v/v sodium dodecyl sulphate (SDS) in 0.1 M Tris-HCl pH 8.0 (room temperature)] was added to each reaction and the reactions were heated at 95 ˚C for 3 minutes. The SDS was then diluted by the addition of 10 mL RNC-Wash buffer [50 mM HEPES pH 7.5, 100 mM KOAc, 5 mM MgAc2, 1 mM DTT, 0.1 % v/v Triton X-100] to each reaction. 100 µL StrepTactin Sepharose High Performance resin (GE Healthcare Life Sciences 28-9355-99) was added to each sample, and the reactions were mixed at 4 ˚C for 1 hour. The beads were subsequently washed with 3 x 1 mL RNC-Wash buffer, and the reactions were transferred to a fresh tube. The Reconstitution of GCN2 Kinase Activation 45 proteins were then eluted in RNC-Elute buffer [50 mM HEPES pH 7.5, 100 mM KOAc, 5 mM MgAc2, 1 mM DTT, 25 mM desthiobiotin (IBA 2-1000-005)]. The beads were transferred into a spin column (Pierce 69705) and the elutions isolated from the beads by a brief centrifugation. The elutions were then prepared for mass spectrometric analysis by digestion in solution with 6 ng/µL trypsin (Promega) for 8 hours at 37 ˚C, and the peptides extracted in 2 % v/v formic acid and 2 % v/v acetonitrile. The peptides were analysed via nano-scale capillary LC-MS/MS on an Ultimate U3000 HPLC (ThermoScientific). The petides were trapped on a C18 Acclaim PepMap100 5 µm, 100 µm by 20 mm nanoViper (ThermoScientific) and then separated on a C18 Acclaim PepMap100 3 µm, 75 µm by 250 mm nanoViper (ThermoScientific), before being eluted with a gradient of 2 – 80 % v/v acetonitrile over 60 minutes. The peptides were then passed onto a Q-Exactive Plus Orbitrap mass spectrometer (ThermoScientific), and data-dependent analysis was performed. The m/z range was from 300 – 2000. The data were then searched against the UniProt KB protein database using the search programme Mascot (Matrix Science) (Perkins et al., 1999). The search parameters consisted of a precursor tolerance of 10 ppm and a fragment ion mass tolerance of 0.8 Da. The data were analysed using the programme Scaffold (Proteome Software). 2.2.10. eIF2α Phosphorylation Assay Due to problems with eIF2α associating with the reaction tubes, all eIF2α phosphorylation reactions were performed in tubes that had been previously coated in BSA according to the following protocol. 1 mL of a 100 mg/mL solution of BSA (Fisher Scientific BP1605) was dispensed into a 1.5 mL eppendorf tube and the tube was left for 12 hours. The BSA was then aspirated, and the tube washed twice with 1 mL phosphate-buffered saline (PBS). All liquid was then removed and the tube left to dry thoroughly. The reactions were prepared under the same conditions as described for the Western blot autophosphorylation assay, except with the addition of 1 mg/mL BSA (Fisher Scientific BP1605) and 250 nM eIF2α to each reaction. The reactions were again started upon the addition of an ATP mix [final concentrations 0.5 mM MgATP pH 8.0, 18.75 mM MgCl2] and the transfer from ice to 32 ˚C. Reactions were quenched and run on a gel as described above. The proteins were transferred to a nitrocellulose membrane using the iBlot Gel Transfer Device (Invitrogen) in combination with iBlot Transfer Stacks (Invitrogen IB301002). The Reconstitution of GCN2 Kinase Activation 46 membrane was then blocked for 1 hour in 5 % BSA in TBST buffer [100 mM Tris-HCl pH 7.5 (room temperature), 150 mM NaCl, 0.1 % v/v Tween 20], and washed three times in TBST. The membrane was then incubated with the primary antibody [Cell Signalling #9721 (1:1000) for phospho-eIF2α; Santa Cruz Biotech sc-133132 (1:1000) for total eIF2α] for 1 hour, then washed three times in TBST before being incubated with the secondary antibody [Cell Signalling #7074 (1:5000) for phospho-eIF2α; Pierce 31430 (1:10,000 dilution) for total eIF2α] for 1 hour and subsequently washed three times in TBST. Finally, the membranes were incubated with 2 mL SuperSignal West Pico PLUS chemiluminescent substrate (ThermoFisher Scientific 34577) and then exposed using a ChemiDoc Touch Imaging System (Bio-Rad). The blots were quantified using ImageJ. Band intensities were measured, and then the background values subtracted. The values were then normalised to the amount of eIF2α phosphorylation by the full-length protein in the presence of ribosomes. Values plotted are the means ± the standard deviations of three independent experiments. 2.2.11. Ribosomal Co-migration Assays 50 µL reactions containing 100 nM Strep-tagged proteins and 35 µL RRL (Green Hectares) were assembled in RNC buffer [50 mM HEPES pH 7.5, 100 mM KOAc, 5 mM MgAc2, 1 mM DTT] and then incubated at 32 ˚C for 15 minutes. 200 µL gradients containing 10 – 50 % sucrose in RNC buffer were assembled in five 40 µL steps in polycarbonate centrifuge tubes (Beckman Coulter 343775) and left to sit on ice for 30 minutes. 20 µL of the reaction was loaded on top of the gradients, and then the gradients were subjected to centrifugation at 259,000g in a TLS-55 rotor (Beckman Coulter 346936) for 30 minutes at 4 ˚C, with an acceleration of 9 and a deceleration of 0. Eleven 20 µL fractions were manually collected from the top of the gradient and mixed with 20 µL 2X LDS sample buffer (NuPAGE NP0008). Samples were run on a 10 % Bis-Tris Protein Gel (NuPAGE NP0303), and then were Western blotted as described above. The primary antibody was Anti-StrepII (abcam 76949, 1:1000 dilution) and the secondary antibody was Anti-Rabbit (Cell Signalling #7074, 1:5000 dilution). Reconstitution of GCN2 Kinase Activation 47 2.2.12. Ribosomal Pull-Down Assays 500 µL reactions including 100 nM Strep-tagged proteins and 100 nM ribosomes were assembled in RNC buffer, and then incubated at 32 ˚C for 15 minutes. 20 µL of StrepTactin Sepharose High Performance resin (GE Healthcare Life Sciences 28-9355-99), equilibrated in RNC-Wash buffer [50 mM HEPES pH 7.5, 100 mM KOAc, 5 mM MgAc2, 1 mM DTT, 0.1 % Triton X-100] was added to each reaction and the reactions were rotated at 4 ˚C for 90 minutes. The beads were then sedimented through centrifugation at 500g for 1 minute at 4 ˚C, before the supernatant was removed and the beads were resuspended in 1 mL RNC-Wash buffer. This washing step was repeated seven times in total and the reactions transferred to fresh tubes. All solution was then aspirated and the proteins eluted from the beads through the addition of 18 µL 2X LDS sample buffer (NuPAGE NP0008). Samples were then run on a 4 – 12 % Bis-Tris Protein Gel (NuPAGE NP0329) and stained using InstantBlue Protein Stain (Expedeon 1SB1L). For quantification, three ribosomal protein bands were chosen, and their intensity measured in each lane using Image Lab 2 (BioRad) and normalised to the pull-down efficiency of the full-length protein. Values plotted are the means ± the standard deviations of these three bands, across three independent experiments (giving nine values in total). 2.2.13. Gel Filtration of Rabbit Reticulocyte Lysate A PD-10 column (GE Healthcare 17-0851-01) was equilibrated in RNC buffer [50 mM HEPES pH 7.5, 100 mM KOAc, 5 mM MgAc2, 1 mM DTT], and 2.5 mL crude RRL (Green Hectares) was applied to the top of the column. Once the RRL had fully passed into the column, 3.5 mL RNC buffer was added and the elution of the RRL was collected. 2.2.14. Mass Spectrometry Analysis of GCN2-interacting proteins Reactions containing 100 nM full-length GCN2, 100 nM kinase domain (residues 585-1024), or no protein, and 350 µL gel-filtered RRL (Green Hectares) were incubated at 32 ˚C for 15 minutes (the final volume of the reactions were 500 µL). 20 µL of StrepTactin Sepharose High Performance resin, equilibrated in RNC-Wash buffer [50 mM HEPES pH 7.5, 100 mM KOAc, 5 mM MgAc2, 1 mM DTT, 0.1 % Triton X-100] was added to each reaction and the reactions were rotated at 4 ˚C for 90 minutes. The beads were then sedimented through centrifugation at 500g for 1 minute at 4 ˚C, before the supernatant was removed and the beads were resuspended in 1 mL RNC-Wash buffer. This washing step was repeated seven times in Reconstitution of GCN2 Kinase Activation 48 total and the reactions transferred to fresh tubes. All solution was then aspirated and the proteins eluted from the beads through the addition of 18 µL 2X LDS sample buffer (NuPAGE NP0008). Samples were then run on a 4 – 12 % Bis-Tris Protein Gel (NuPAGE NP0329) and stained using InstantBlue Protein Stain (Expedeon 1SB1L). The lanes of the gel were then excised and sectioned into slices of approximately 1 mm. The sections were then prepared for mass spectrometric analysis via the Janus liquid handling system (PerkinElmer). Briefly, the gel pieces were destained with 50 % v/v acetonitrile and 50 mM ammonium bicarbonate, then reduced with 10 mM DTT and alkylated with 55 mM iodoacetamide. The proteins were then digested with 6 ng/µL trypsin (Promega) for 8 hours at 37 ˚C, and the peptides extracted in 2 % v/v formic acid and 2 % v/v acetonitrile. The peptides were analysed via nano-scale capillary LC-MS/MS on an Ultimate U3000 HPLC (ThermoScientific). The petides were trapped on a C18 Acclaim PepMap100 5 µm, 100 µm by 20 mm nanoViper (ThermoScientific) and then separated on a C18 Acclaim PepMap100 3 µm, 75 µm by 250 mm nanoViper (ThermoScientific), before being eluted with a gradient of 2 – 80 % v/v acetonitrile over 60 minutes. The peptides were then passed onto a Q-Exactive Plus Orbitrap mass spectrometer (ThermoScientific), and data-dependent analysis was performed. The m/z range was from 300 – 2000. The data were then searched against the UniProt KB protein database using the search programme Mascot (Matrix Science) (Perkins et al., 1999). The search parameters consisted of a precursor tolerance of 10 ppm and a fragment ion mass tolerance of 0.8 Da. The data were analysed using the programme Scaffold (Proteome Software). 2.2.15. Treatment of Rabbit Reticulocyte Lysate with Micrococcal Nuclease Gel-filtered rabbit reticulocyte lysate was thawed and a final concentration of 1 mM CaCl2 was added. Micrococcal Nuclease (Roche 10107921001), dissolved in 50 mM HEPES pH 7.5, was then added to a final concentration of 150 U/mL. The lysate was incubated at 25 ˚C for 12 minutes, with occasional mixing. The reaction was then removed to ice and 2 mM ethylene glycol-bis(β-aminoethyl ether)-N,N,N’,N’-tetraacetic acid (EGTA) pH 8.0 was then added. Reconstitution of GCN2 Kinase Activation 49 2.3. Results and Discussion 2.3.1. Expression and Purification of GCN2 Initially, DNA encoding human GCN2 was cloned into a pAceBac1 vector with an N- terminal hexahistidine (His6) tag followed by a cleavage site for Tobacco Etch Virus (TEV) protease. The plasmid was integrated into EMBacY baculoviral DNA and then transfected into Spodoptera frugiperda (Sf9) cells for protein expression. The cells were grown for 55 hours before being harvested and lysed via probe sonication. After sonication, the cell lysate was subjected to ultracentrifugation at 140,000g for 45 minutes to eliminate cell debris from the sample. GCN2 was then purified from the cell lysate following the purification protocol described in Materials and Methods. Briefly, the cell lysate was passed over a 5 mL nickel affinity column to capture GCN2 via the N-terminal His6 tag. The column was then washed extensively before the bound protein was eluted with 100 mM imidazole. The His6 tag was then cleaved with TEV protease before the sample was subjected to ion exchange chromatography. GCN2 was eluted from the ion exchange column using a gradient of NaCl, and the fractions were analysed for the presence of GCN2 via denaturing gel electrophoresis (SDS-PAGE). The GCN2-containing fractions were then passed over a clean nickel affinity column for a second time to remove the His6-tagged TEV protease before being concentrated and injected on to a gel filtration column. The purity of the protein was ultimately assessed by SDS-PAGE (Figure 2.1). In general, the gel filtration profiles after each purification showed a single, monodisperse peak, indicating a pure and homogenous sample at the end of the purification protocol. Furthermore, the final gel tended to show very pure material with minimal contaminants. Reconstitution of GCN2 Kinase Activation 50 Figure 2.1. Results of a typical purification for the His6-tagged human GCN2 (from 4 L of Sf9 cells). The gel filtration profile is shown above, with the fractions pooled indicated with a bracket. A gel showing the purity of the fractions (labelled above the gel) is shown below. As shown in Figure 2.1, purification via the His6 tag gave pure and homogenous material. However, the yield was generally low (typically 0.4 – 0.6 mg protein per litre of cell culture). indicating that the His6 tag may not be the best strategy. Moreover, even in the absence of the His6 tag (i.e. after TEV cleavage), there was significant binding between GCN2 and the nickel affinity resin. Whilst this isn’t surprising given the naturally occurring histidine Reconstitution of GCN2 Kinase Activation 51 residues in full-length GCN2, it may present a problem for analysis of protein-protein interactions via affinity capture. For these reasons, other purification tags were tested to further optimise the purification protocol. Initially, the replacement of the N-terminal His6 tag with an N-terminal glutathione-S- transferase (GST) tag was tested (Figure 2.2). The purification of this construct followed essentially the same steps as the His-tagged protein: affinity chromatography, anion exchange chromatography and finally gel filtration. The result of this purification is shown in Figure 2.2. In this case, the GST tag was not cleaved from the protein. Whilst this strategy still resulted in relatively pure material, the gel filtration profile indicated that the protein was not monodisperse. There are two possible reasons for this: the intrinsic ability of the GST tag to dimerise could be driving the formation of large oligomers which elute from the column at a lower retention volume, or the protein could simply be partially aggregated or insoluble and so be coming out in the void volume of the column. Whatever the reason, the protein seemed less well behaved than when His6-tagged. Moreover, the yield was only approximately 0.1 mg protein per litre of cell culture, which is significantly reduced in comparison to purification of the His6-tagged protein, and so this strategy was not pursued. An N-terminal twin StrepII tag was also tested. Again, the protein was captured via affinity chromatography on StrepTactin resin, before being eluted with desthiobiotin. The eluent was then further purified via ion exchange chromatography and size-exclusion chromatography. As for the GST-tagged material, the Strep tag was not cleaved from the protein. The results are shown in Figure 2.3. This purification strategy gave pure material at a reproducibly high yield (on average 1.5 mg protein per litre of cell culture), and so was subsequently used for the majority of later experiments (unless specified). Reconstitution of GCN2 Kinase Activation 52 Figure 2.2. The results of the purification for GST-tagged GCN2 (from 1.5 L of Sf9 cells). The gel filtration profile is shown above, with the fractions pooled indicated with a bracket. A gel showing the purity of the fractions pooled is shown below. Reconstitution of GCN2 Kinase Activation 53 Figure 2.3. The results of the purification for Strep-tagged GCN2 (from 1.5 L of Sf9 cells). The gel filtration profile is shown above, with the fractions pooled indicated with a bracket. A gel showing the purity of the fractions pooled is shown below. Reconstitution of GCN2 Kinase Activation 54 2.3.2. Biophysical Characterisation of Human GCN2 2.3.2.1. Buffer Optimisation by Differential Scanning Fluorimetry Initially, differential scanning fluorimetry (DSF) was used to optimise the buffer conditions in order to ascertain the optimal conditions of the protein. For these analyses, a Prometheus instrument was used to measure the intrinsic fluorescence of the tryptophan residues of the protein over the course of thermal denaturation. Due to the hydrophobic nature of their aromatic side chain, tryptophan residues are generally buried in the hydrophobic core of a folded protein. However, once the protein begins to denature, the tryptophan residues gradually become exposed to the aqueous environment and this causes their fluorescence to change. Specifically, their exposure results in a change to the ratio of fluorescence at 330 nm and 350 nm. Plotting these ratios against the temperature means therefore that a melting temperature (Tm) for the protein can be gained without the need for any extrinsic dye or fluorophore. The effect of different conditions on the melting temperature of the protein can then be interpreted to discern the conditions under which the protein is most stable. Firstly, the importance of the pH of the protein buffer was investigated. Figure 2.4. The first derivative of the ratio between GCN2’s intrinsic fluorescence at 350 and 330 nm plotted against the temperature for a range of different pH values. The means ± the standard deviations are plotted (n=3). Interestingly, this analysis showed that GCN2 undergoes a biphasic transition during denaturation (Figure 2.4). The first transition is between 45 ˚C and 50 ˚C, whilst the second is between 55 ˚C and 61 ˚C. The exact nature of these two unfolding events is unclear, but could represent the unfolding of two domains or domain clusters within the protein. Both transition Reconstitution of GCN2 Kinase Activation 55 points can be used for the analysis concerning the effect of the pH on protein stability (Figure 2.5). Figure 2.5. The transition point values for each sample plotted against the pH. On the left are the values from the first transition point and on the right are the values from the second transition point. Each transition point value was calculated from two independent experiments, and the means ± the standard deviations are plotted. In cases where the error bars are smaller than the symbol, the error bars are omitted. Data from the first transition point indicate that the optimum pH for this unfolding event appears to be between pH 7.6 and pH 7.8. The optimum stability for the second transition point seems to occur at pH values between 7.0 and 7.2. For this reason, in all subsequent experiments the pH was kept at a value between pH 7.0 and pH 7.8, most commonly pH 7.5. To ensure that aggregation was not occurring and skewing the data, scattering signal was also collected for each condition over the course of the thermal denaturation (Figure 2.6). These data showed no significant increase in scattering signal over the temperature course, indicating that no substantial aggregation occurred. Reconstitution of GCN2 Kinase Activation 56 Figure 2.6. Scattering signal for GCN2 at each pH tested plotted against the temperature. Each point represents the mean ± the standard deviations calculated from two independent experiments. Having determined an optimum pH range for GCN2 in solution, the next aim was to test the effects of the inclusion of a range of different detergents within the buffer. The detergents tested were Tween-20, lauryl maltose neopentyl glycol (LMNG), N-dodecyl β-D-maltoside (DDM) and 3-((3-cholamidopropyl)dimethylammonio)-1-propanesulfonate (CHAPS). As shown in Figure 2.7, whilst the scattering signals show that the addition of detergent does not affect GCN2’s propensity to aggregate, addition of any of these detergents results in a decrease in protein stability, demonstrated by a reduction in the temperature for both unfolding events in comparison to GCN2 in the absence of any detergents. For this reason, no detergent was included in the protein buffer for the majority of subsequent experiments. Reconstitution of GCN2 Kinase Activation 57 Figure 2.7. Differential scanning fluorimetry data for GCN2 in the presence of four different detergents, each at a range of concentrations. On the left, the first derivative of the ratio of fluorescence at 330 nm and 350 nm is shown across the thermal denaturation temperature range for each detergent. The profile for GCN2 in the absence of any detergent is shown in blue, and the profiles in the presence of detergent are identified in the key. In each case, the scattering signal across the temperature range is shown on the right. 2.3.2.2. SEC-MALS The next aim was to determine the oligomeric state of purified human GCN2. Size-exclusion chromatography in combination with multi-angle light scattering (SEC-MALS) was used, and gave an estimated molecular weight of 390 kDa for the protein in solution. The theoretical molecular weight of monomeric GCN2 is approximately 192 kDa, meaning that the SEC- MALS data suggest the protein exists as a dimer. Furthermore, the symmetrical differential Reconstitution of GCN2 Kinase Activation 58 refractive index profile and the approximately horizontal estimation of the molecular weight indicate that the protein is monodisperse (Figure 2.8). Figure 2.8. SEC-MALS data for purified human GCN2. The differential refractive index is shown with a continuous trace, and the light scattering data shown with a dashed line. The molecular weight of the protein was estimated by light scattering as approximately 390 kDa. 2.3.2.3. Surface Plasmon Resonance Having characterised the protein as a stable dimer, surface plasmon resonance (SPR) was then used to analyse the binding of deacylated tRNA to GCN2. GCN2 was immobilised on to a CM5 chip by amine coupling, and deacylated tRNA was then passed over the chip at a variety of concentrations. Binding was measured by a change in the refractive index (Figure 2.9). This titration was then used to estimate a binding coefficient of approximately 2 µM, indicating a relatively weak interaction between human GCN2 and deacylated tRNA. Figure 2.9. Surface plasmon resonance data for the interaction between deacylated tRNA and immobilised GCN2. Shown on the left are the SPR responses of different concentrations of tRNA being flowed over the chip. Each concentration was measured in three independent experiments, and the means of the data are plotted. The red lines indicate the fit for each profile. Shown on the right are the results of the titration, giving a KD value of approximately 2 µM. Reconstitution of GCN2 Kinase Activation 59 This has a few interesting implications for the proposed model in which deacylated tRNA binds to GCN2 to initiate the ISR. The intracellular concentration of total tRNA can be estimated at approximately 2 µM (Gurdon, 1986), suggesting that in order for GCN2 to be able to recognise an increase in the concentration of deacylated tRNA in this manner, the level of tRNA would have to rise to such a level that a significant fraction of the total cellular tRNA was deacylated. Given the sensitivity of GCN2’s response to amino acid starvation, in that it is able to recognise the absence of a single amino acid (which would only account for 5 % of the total tRNA), this seems contradictory. Some data indicate that the depletion of a single amino acid is able to affect the efficiency of the aminoacylation reaction for other tRNA species (Zaborske et al., 2009), which could be a potential mechanism for a rapid build up of deacylated tRNA, but it seems unlikely given the speed of GCN2’s response to the depletion. It therefore seems likely that in vivo this binding event, and therefore GCN2 activation, must require additional factors: a build up of deacylated tRNA may not be sufficient. Whether this represents an additional complexity that is restricted to mammalian GCN2 or whether this is also true for yeast Gcn2 remains to be seen. The interaction between yeast Gcn2 and deacylated tRNA was demonstrated using a gel shift assay in which 20 nM 32P radiolabelled deacylated tRNA was incubated with up to 80 nM GCN2 (Dong et al., 2000). If the tRNA KD is assumed to be similar between yeast Gcn2 and human GCN2 (i.e. ~ 2 µM), the fraction bound would be expected to be approximately 1 %. Given the sensitivity of detection techniques for autoradiography, and the clear excess of free tRNA in the published gel, it is possible that this is the case, however the qualitative nature of the image precludes further analysis. Alternatively, the affinity between yeast Gcn2 and tRNA could be significantly higher due to evolutionary divergence of the mammalian and yeast proteins, indicating that there may be important differences between the pathways of activation. To ensure the interaction between human GCN2 and deacylated tRNA is specific, an acylated equivalent is required. Due to the complex nature of producing stable acylated mammalian tRNA at high purity, this was not explored in this work, but would be an important component of future experiments. Reconstitution of GCN2 Kinase Activation 60 2.3.3. Testing Potential Interacting Partners of GCN2 2.3.3.1. Purification of GCN1 and GCN20 Both human GCN1 and human GCN20 (the most likely candidate; also known as ABC50) were cloned into separate pAceBac1 vectors with N-terminal twin StrepII tags and then expressed in Sf9 cells in the same manner as GCN2. Both proteins were purified through very similar protocols, involving affinity chromatography, anion exchange chromatography and finally gel filtration. The results of a typical purification for Strep-tagged GCN1 are shown in Figure 2.10. As shown, this strategy resulted in a pure sample of GCN1, and the gel filtration profile indicates the protein is monodisperse. The yield was consistently slightly less than for GCN2 (on average ~ 1 mg protein per litre of cell culture), however this is perhaps unsurprising given the extremely high molecular weight of the protein (just under 300 kDa). Similarly, good results were achieved for the purification of Strep-tagged GCN20 (Figure 2.11). The lower molecular weight (approximately 96 kDa) of this protein resulted in better expression and therefore an improved yield for the purification (~ 5 mg per litre of cell culture). The final peak on the gel filtration profile was less symmetrical than for either GCN1 or GCN2, and this could potentially be attributed to the presence of a contaminant with a molecular weight of approximately 45 kDa, coming just after the main GCN20 peak. The inclusion of this contaminant in the final sample was avoided as much as possible by pooling fractions from only a small part of the gel filtration peak (shown on the final gel in Figure 2.11), resulting in a very pure sample. Reconstitution of GCN2 Kinase Activation 61 Figure 2.10. The results of the purification of Strep-tagged GCN1. The gel filtration profile is shown above, and the pooled fractions indicated with a bracket. The gel below shows the purity of the corresponding fractions. Reconstitution of GCN2 Kinase Activation 62 Figure 2.11. The results of the purification of Strep-tagged GCN20. The gel filtration profile is shown above, and the pooled fractions indicated with a bracket. The gel below shows the purity of the corresponding fractions. 2.3.3.2. Identification of protein-protein interactions Once purified GCN1 and GCN20 had been obtained, they were then tested for the ability to interact with one another. This was tested through affinity capture of one component of the Reconstitution of GCN2 Kinase Activation 63 complex in the presence of an untagged potential binding partner, followed by washing to eliminate non-specific interactions and finally analysis via SDS-PAGE. Figure 2.12. Affinity capture analysis of interactions between GCN2, GCN1 and GCN20. The reaction totals are shown on the left-hand gel. The beads were then washed, and the remaining bound proteins eluted with sample buffer. The elutions are shown on the right-hand gel. The results of the pull-downs are shown in Figure 2.12. To control for non-specific binding of each protein to the StrepTactin resin, the first three reactions included untagged versions of GCN1, GCN2 and GCN20 alone. The elutions of these reactions show very little non-specific binding, indicating that the presence of any binding partner in the subsequent reactions is suggestive of a specific interaction. The next three reactions use Strep-GCN1 as bait, and show that GCN1 is able to pull down a very small amount of GCN2, but the affinity of the interaction appears to be low. Interestingly, in the reverse scenario (with Strep-tagged GCN2 and untagged GCN1) there is no observable presence of GCN1. This could be due to the fact that the N-terminal RWD domain of GCN2 is thought to be responsible for the interaction with GCN1, and so the immobilisation of this domain on the StrepTactin resin could prevent it from forming a productive interaction with GCN1. To test this hypothesis, it would require a C-terminally tagged version of GCN2, which would be expected to show similar pull-down results to the interaction between Strep-GCN1 and GCN2. Interestingly, in this pull-down analysis (Figure 2.12) no interaction between GCN1 and GCN20 could be observed, either between Strep-GCN1 and GCN20 or Strep-GCN20 and Reconstitution of GCN2 Kinase Activation 64 GCN1. (The inclusion of either ATP or GCN2 makes no difference to the interaction.) This is contradictory to data from yeast, where co-immunoprecipitation and yeast-two-hybrid experiments imply that GCN1 and GCN20 exist in a complex (Vazquez de Aldana et al., 1995). This implies a fundamental difference between the yeast and the human regulatory systems. It is possible that another protein in the human proteome fulfils the role of yeast GCN20, but given that human ABC50 (tested in this experiment) is the most likely candidate based upon sequence conservation, it seems likely that there could simply be key differences in the pathways. 2.3.4. Functional Analysis of Purified GCN2 2.3.4.1. Autophosphorylation Assays To initially ascertain whether the purified protein was functional, GCN2’s autophosphorylation activity was assayed in vitro. To do this, GCN2 was incubated in the presence of radiolabelled γ-33P-ATP at 32 ˚C for a time course of 10 minutes. Samples were taken at 0, 5 and 10 minutes, and the reaction quenched. The samples were then analysed by SDS-PAGE followed by autoradiography, and the results are shown in Figure 2.13. Figure 2.13. GCN2 autophosphorylation assay. The reactions were assembled as indicated above, and begun with the addition of γ-33P-ATP. Reactions were incubated at 32 ˚C and samples taken at 0, 5 and 10 minutes. The samples were then analysed by SDS-PAGE and visualised by autoradiography. This analysis allows the observation of the addition of a radiolabelled phosphate group to GCN2 in an in vitro reaction without the presence of any additional factors (Figure 2.13), indicating that the purified GCN2 is catalytically functional and is able to autophosphorylate. The next question was to determine whether the addition of deacylated tRNA stimulated the activity of GCN2, as suggested by the literature. Surprisingly, no significant stimulation of GCN2’s autophosphorylation activity was observed, even in the presence of 5 µM tRNA (Figure 2.13). Given the calculated KD of 2 µM, at a tRNA concentration of 5 µM, Reconstitution of GCN2 Kinase Activation 65 approximately 70 % of GCN2 should be bound to tRNA. This therefore indicates that tRNA binding alone is insufficient to activate the kinase, even if one hypothesises the concentration of deacylated tRNA could reach such levels, which seems unlikely. Given the fact that this system is ultimately one of translational regulation, the experiment also tested the effects of purified ribosomes on the autophosphorylation activity of GCN2. Importantly, the inclusion of a low concentration (50 nM) of ribosomes (purified from rabbit reticulocyte lysate and washed with high salt as described in Materials and Methods section 2.2.6) led to a significant increase in the autophosphorylation activity of GCN2 (Figure 2.13). The inclusion of tRNA in this reaction had no obvious effect (Figure 2.13). This finding correlates with data later published by Ishimura and colleagues, implicating stalled ribosomes in the in vivo activation of GCN2 (Ishimura et al., 2016). The purified ribosomes used in this assay are effectively empty, without any mRNA transcript, and so the specific functional state that this sample represents is unclear. The possibility that the purified ribosomes contain some form of deacylated tRNA species cannot be ruled out, and so it remains possible that tRNA is somehow acting as the major activator. However, given the concentration of ribosomes is 50 nM, the likelihood of tRNA being present at a sufficiently high concentration to lead to a significant activation is relatively small. If this is the case however, the data still imply a clear and necessary role for the ribosome in the activation mechanism, be it a direct role or a role in the potential transfer of tRNA to GCN2. To discriminate between these two possibilities with this assay, one would require a purified ribosomal sample in which there was no possibility of any deacylated tRNA present. 2.3.4.2. The effect of GCN1 and GCN20 To extend this analysis, the experiment was repeated with the inclusion of the potential regulatory partners GCN1 and GCN20, to determine whether despite their apparently weak affinity they have a role in the activation of GCN2 as described for their yeast counterparts (Figure 2.14). Reconstitution of GCN2 Kinase Activation 66 Figure 2.14. A GCN2 autophosphorylation assay to look at the effects of GCN1 and GCN20 on the activation of GCN2 by ribosomes. A gel showing the purified materials is shown on the left. Reaction components are indicated above the autoradiograph, and reactions were started by the addition of γ-33P- ATP. The reactions were incubated at 32 ˚C and samples taken at 0, 5 and 10 minutes. The samples were then analysed by SDS-PAGE and visualised by autoradiography. As shown in Figure 2.14, the activation of GCN2 by ribosomes is very reproducible. It should be pointed out that in this experiment, the concentration of deacylated tRNA was decreased to 500 nM, to make it more physiologically relevant. It is also interesting to note the increase in signal in this experiment compared to Figure 2.13, which is probably due to differences between the specific activity of the radiolabelled nucleotide. This increase in signal permits the clear observation of the appearance of phosphorylated ribosomal proteins upon incubation with GCN2 (Figure 2.14). This phosphorylation, as well as the phosphorylation of GCN2, is dependent on functional GCN2: replacement of the wild-type GCN2 with a catalytically dead mutant (GCN2 D848N (Padyana et al., 2005)) abolishes phosphorylation, eliminating the possibility that a protein associated with the ribosomes is responsible for the phosphorylation of either GCN2 or the ribosomal proteins. The functional implications of these phosphorylations could be very interesting, but more work needs to be done to establish whether or not they occur in vivo. The inclusion of GCN1 and GCN20 had no noticeable effect on the autophosphorylation activity of GCN2 (Figure 2.14), however it is possible to observe that GCN1 appears to be a substrate of GCN2’s kinase activity. This has not been previously reported, and it will be extremely interesting to determine whether this is also true in a physiological context. It is tempting to speculate that this represents a feedback mechanism, but this would have to involve other factors not present in this assay, as the presence of GCN1 does not seem to have an effect on GCN2. In contrast, GCN20 does not appear to be phosphorylated by GCN2, Reconstitution of GCN2 Kinase Activation 67 indicating that the phosphorylation of GCN1 is specific. It is important to note that the inclusion of GCN1 and GCN20 also have no observable effects on the activation of GCN2 by tRNA, dismissing the possibility that the lack of either of these proteins is preventing the proper activation of GCN2 by tRNA. Upon the addition of ribosomes, GCN2 is stimulated to the same extent as in the absence of GCN1 and GCN20. This stimulation extends to GCN2’s activity towards GCN1: there is a clear increase in GCN1 phosphorylation in the presence of ribosomes. 2.3.4.3. Identifying the phosphorylation sites The link between GCN2 activation and the autophosphorylation state of the two threonine residues in the active site was previously shown (Romano et al., 1998). For this reason, it was initially assumed that the presence of ribosomes was stimulating the autophosphorylation of these two residues. To test this assumption, a commercial antibody against threonine 899 was used to analyse the specific phosphorylation status of GCN2 during the reaction in the presence and absence of its regulators. The conditions for autophosphorylation were the same as for previous reactions, except for the replacement of radiolabelled γ-33P-ATP with non- radiolabelled ATP. The samples were then analysed by SDS-PAGE followed by Western blotting (Figure 2.15). Figure 2.15. Threonine 899 autophosphorylation assay. The reactions were assembled as indicated above the blot, and were started with the addition of ATP. Reactions were incubated at 32 ˚C and samples taken at 0, 5 and 10 minutes. The samples were then analysed by SDS-PAGE and visualised by Western blotting. Surprisingly, these results showed no significant change in the phosphorylation of threonine 899 upon the addition of either tRNA or ribosomes. It is possible that this is due to the detection method, although this seems unlikely due to the apparently different kinetics of the addition of the phosphate group between this analysis and the previous experiment (Figure 2.13). It therefore seems most likely that this discrepancy results from the fact that ribosomes are stimulating the autophosphorylation of a residue or residues other than threonine 899. Given the well-characterised association of GCN2 activation and the phosphorylation status Reconstitution of GCN2 Kinase Activation 68 of threonine 899, it is surprising to note that, in vitro, the incubation of the protein with ATP is sufficient to achieve phosphorylation (Figure 2.16). Figure 2.16. Threonine 899 autophosphorylation assay. ATP or just buffer was added to GCN2, and the reactions incubated at 32 ˚C. Samples were taken at 0, 5 and 10 minutes. The results were then visualised by SDS-PAGE followed by Western blotting. To address the question of which phosphorylation events may be stimulated by the addition of ribosomes, samples were prepared consisting of GCN2 alone (starting material), GCN2 after a 10 minute incubation with ATP, GCN2 after a 10 minute incubation with ATP and deacylated tRNA, and GCN2 after a 10 minute incubation with ATP and ribosomes. These samples were then each subjected to analysis by mass spectrometry to determine the phosphorylation sites on GCN2. [This mass spectrometry was performed by Mark Skehel.] A number of phosphorylation sites were detected, spanning the full-length of the protein. The presence or absence of each phosphorylation for each sample was then assessed, and the results are shown in Table 2.1. From this analysis, it is possible to observe that the starting material (purified GCN2) contains two phosphorylations: one on serine 264 and one on threonine 667. Serine 264 is part of the charged linker separating the RWD domain and pseudokinase domain of GCN2 and threonine 667 is part of the kinase domain. Neither of these sites have been previously characterised as phosphorylation sites, nor are they particularly well conserved (see sequence alignment Supplementary Figure 1). It is possible that their phosphorylation is an artefact of the expression and purification process, or they may play a role in vivo. The incubation of GCN2 with ATP leads to the phosphorylation of more sites. Pleasingly, it is possible to see the phosphorylation of threonine 899 and threonine 904 under these conditions, correlating with the previous data from the western blot (Figure 2.15). Reconstitution of GCN2 Kinase Activation 69 Reconstitution of GCN2 Kinase Activation 70 A number of other phosphorylations seem to occur upon the addition of ATP, including at serine 205, threonine 462 and serine 555, however their roles are as of yet unclear. The datasets for the phosphorylation sites on GCN2 after incubations with tRNA or ribosomes are unfortunately less complete than the previous two datasets, meaning that no peptides were detected for some regions of the protein and therefore precluding their analysis. Regrettably, the region containing threonine 899 and threonine 904 is subject to this problem, and so it is impossible to discern their phosphorylation status from the mass spectrometry data. However, the western blots in Figure 2.15 indicate threonine 899 at least is phosphorylated. Given the effect that ribosomes appear to have upon GCN2’s autophosphorylation activity, the phosphorylations that are induced upon incubation with ribosomes were of most interest. Serine 877 is located at the beginning of the activation loop of the kinase domain, and the mass spectrometry data indicated that its phosphorylation is dependent on the presence of ribosomes, which seemed interesting. In addition, the increase in phosphorylation of the C- terminal domain (serine 1504, serine 1515 and serine 1520) in the presence of ribosomes also suggested something may be happening in this area. Out of these three residues, serine 1504 is the most highly conserved, and so it was decided to test the importance of this phosphorylation event initially. 2.3.4.4. Testing the importance of specific phosphorylations In order to test the importance of the phosphorylation of serine 877 and serine 1504 for the activation of GCN2 by ribosomes, two constructs were produced in which these serine residues had been substituted with alanine residues. These constructs were expressed and purified, and then subjected to autophosphorylation analysis (Figure 2.17). Reconstitution of GCN2 Kinase Activation 71 Figure 2.17. A GCN2 autophosphorylation assay to look at the effects of the S877A and S1504A mutations on the activation of GCN2 by ribosomes. The gel on the left shows the purified material used for the assay. On the right is the autoradiograph. Reaction components are indicated above, and reactions were started by the addition of γ-33P-ATP. The reactions were incubated at 32 ˚C and samples taken at 0, 5 and 10 minutes. The samples were then analysed by SDS-PAGE and visualised by autoradiography. This autophosphorylation assay showed that the removal of either of the phosphorylation sites at serine 877 or serine 1504 does not prevent the stimulation of GCN2 by ribosomes, indicating neither phosphorylation is critical for GCN2 activation. The S877A mutation does cause a reduction in the total phosphorylation of GCN2 in both the presence and absence of ribosomes. The reduction in autophosphorylation activity observed for GCN2 S877A indicates that phosphorylation at this site could contribute to the protein’s kinase activity. The S1504A mutation appears to have no effect upon the phosphorylation of GCN2 either in the presence or absence of ribosomes, indicating that this phosphorylation does not seem to be important. To extend this analysis, the other phosphorylation sites identified in Table 2.1 could be similarly tested. It is of course possible that these phosphorylation sites are simply a result of GCN2 activation, and therefore that there is no stimulatory autophosphorylation site required for kinase activation. 2.3.4.5. Purification of eIF2α Whilst the autophosphorylation assay is informative, the activity of GCN2 against its physiological substrate was of the greatest interest. To this end, recombinant full-length eIF2α with an N-terminal His6 tag was cloned and expressed in bacterial culture. The protein was then purified by affinity chromatography, anion exchange chromatography and finally gel Reconstitution of GCN2 Kinase Activation 72 filtration (Figure 2.18). As shown in the final gel, this purification strategy produced very pure and monodisperse protein at a high yield (~ 3 mg per litre of cell culture). In a cellular context, eIF2α always exists as part of the trimeric complex eIF2. This poses a question concerning the importance of eIF2β and eIF2γ, and whether studying the α subunit alone will reflect its regulation when in the physiological complex. Whilst there is no definitive answer to this question, eIF2α has been used for this purpose many times in the literature, and there is no evidence that information concerning phosphorylation of eIF2α alone cannot be extrapolated to the whole complex. In fact, interestingly, only the N-terminus (residues 1 – 185) of the protein is often used as a substrate for eIF2α kinases in in vitro assays, as it has been shown via nuclear magnetic resonance that the N- and C-terminal domains of eIF2α are able to move independently of one another (Ito et al., 2004). Reconstitution of GCN2 Kinase Activation 73 Figure 2.18. The purification of recombinant His6-tagged eIF2α. The gel filtration profile is shown above the image, and a gel showing the purity of the corresponding fractions is shown below. The pooled fractions are indicated with brackets. Reconstitution of GCN2 Kinase Activation 74 2.3.4.6. eIF2α Phosphorylation Assay Once purified eIF2α was produced, the next step was to test the activity of GCN2 towards it. Initially, the recombinant eIF2α was simply included in the radiolabelled phosphorylation assay to see if this led to the appearance of a band of the correct molecular weight, indicating the addition of a radiolabelled phosphate group to eIF2α. Figure 2.19. Radiolabelled phosphorylation of eIF2α by GCN2. A gel showing the purified proteins is shown on the left. The reaction components are indicated above the autoradiograph. The reactions were started with the addition of γ-33P-ATP and incubated at 32 ˚C for 10 minutes. Samples were taken at 0, 5 and 10 minutes as indicated above. The positions of GCN1, GCN2 and eIF2α are indicated on the gel. As can be seen in Figure 2.19 there is very little, if any, phosphorylation of eIF2α in the absence of ribosomes. Similarly to the autophosphorylation of GCN2, upon the addition of purified ribosomes it is possible to see the appearance of a band with a molecular weight corresponding to that of eIF2α. However, quantitative analysis of this band is difficult due to the presence of a phosphorylated ribosomal protein with similar mobility in SDS-PAGE. It should be noted that the phosphorylation of eIF2α seems to be less efficient than the phosphorylation of GCN2 and GCN1, both in the absence and presence of ribosomes. Due to the difficulties in analysis, the detection method was then changed to make use of a specific antibody against the site of eIF2α phosphorylation (serine 52 in the human protein). This allowed clarification of the level of eIF2α phosphorylation, even in the presence of ribosomes. Given the fact that the inclusion of GCN1 and GCN20 seemed to make no significant difference to the phosphorylation of eIF2α, for the sake of simplicity they were Reconstitution of GCN2 Kinase Activation 75 excluded from further analysis. For each experiment, the total amount of eIF2α added to the reaction remained constant, however the detection of the total amount (via an antibody) demonstrated that the total amount remaining in each reaction varied considerably, both over the course of the experiment and between ribosome-containing and non-ribosome-containing samples. The reasons behind this were initially unclear, and many potential explanations were explored. It was eventually concluded that upon dilution of eIF2α, before assembling the reactions, the protein immediately began to stick to the eppendorf tube. This meant that a different amount of eIF2α was added to each reaction, depending on the order of assembly. Moreover, throughout the 10 minute reaction the same thing occurred, meaning that each sample taken from each reaction contained less and less total substrate. In the reactions which contained ribosomes, this occurred to a lesser extent, although was still prevalent. In order to combat these effects, it was ultimately required to pre-coat the reaction tubes in 100 mg/mL bovine serum albumin (BSA) the night before the experiment, and also to include 0.5 mg/mL BSA in the reactions. Figure 2.20. Results of an eIF2α phosphorylation assay. The reactions were assembled as indicated above the blots, and begun by the addition of ATP. The reactions were incubated at 32 ˚C and samples taken at 0, 5 and 10 minutes. The samples were analysed via SDS-PAGE visualised by Western blotting. As shown in Figure 2.20, this allows the conclusive observation that the addition of purified ribosomes is able to dramatically increase the phosphorylation of eIF2α by GCN2. Interestingly, this experiment also demonstrated that a high concentration of tRNA (5 µM) has a slight stimulatory effect, which was not as obvious in the GCN2 autophosphorylation experiment. However, in both cases the addition of 50 nM purified ribosomes increases the kinetics and efficiency of eIF2α phosphorylation by GCN2, demonstrating that ribosomes are a more potent activator of GCN2. As ribosomes apparently cause the phosphorylation of GCN2 residues other than threonine 899 (Figure 2.15), their ability to activate GCN2 implies that, contrary to what is often Reconstitution of GCN2 Kinase Activation 76 presumed, the phosphorylation of threonine 899 is not sufficient to drive activation of the kinase, though it may be closely associated with the process in vivo. 2.3.4.7. Identification of an interaction between GCN2 and Ribosomes An interaction between yeast Gcn2 and the ribosome has been described in the literature (He et al., 2014), but the conservation of this interaction for mammalian GCN2 was unclear, as the authors were unable to observe a similar interaction when using the murine protein. However, given the clear effects of ribosomes on the activity of GCN2, both towards itself and its substrates, it seemed very likely that some form of interaction was occuring. The stimulation by ribosomes of the autophosphorylation activity of the kinase was particularly persuasive, as an increase in eIF2α phosphorylation by GCN2 in the presence of ribosomes could be explained by an interaction occurring between eIF2α and the ribosome, which could theoretically result in eIF2α becoming a better substrate for the kinase activity. The stimulation of activity in the absence of the substrate, however, implies a direct interaction between GCN2 and the ribosome. To initially investigate this possibility, purified GCN2 was incubated in the presence and absence of cytosol (rabbit reticulocyte lysate, or RRL) before the sample was subjected to centrifugation through a sucrose gradient. The gradient was then fractionated, and each fraction was analysed by SDS-PAGE and Western blotting to assess the mobility of GCN2 through the gradient. Ribosomes have a well-characterised mobility through sucrose gradients that is very different from soluble proteins, so the aim was to see whether there was any GCN2 present in the deeper fractions of the gradient, indicating ribosomal association. Reconstitution of GCN2 Kinase Activation 77 Figure 2.21. The results of a co-migration assay looking at the mobility of full-length GCN2 or just the kinase domain through a gradient in the presence and absence of cytosol. The reactions indicated above were assembled and incubated at 32 ˚C for 15 minutes. The samples were then applied to a 10 – 50 % sucrose gradient and centrifuged for 30 minutes. The gradients were then fractionated, and the fractions analysed by SDS-PAGE and Western blotting with an anti-StrepII antibody. The graph below shows the A260 readings for each fraction of the gradient. The results of this experiment (shown in Figure 2.21) show that in the presence of cytosol, GCN2 migrates much more through the sucrose gradient, indicating that it is forming some form of higher-order association. The A260 readings for each fraction of the gradient demonstrate that ribosomes migrate to the lower part of the gradient, generally correlating with the migration of GCN2, indicating that GCN2 is likely to be interacting with ribosomes in some manner. To test the specificity of this interaction, the kinase domain alone of GCN2 (residues 585 – 1024) was cloned and purified, and subjected to the same co-migration assay. As can be clearly seen in Figure 2.21, the migration of the kinase domain alone is not affected by the presence of cytosol in the same way as the full-length protein, meaning that whilst GCN2 appears to form some form of ribosomal complex, the kinase domain alone does not. Given that this experiment was done with crude cytosol, this does not mean that there is a direct interaction occurring between GCN2 and ribosomes, as other factors could be required to mediate an indirect interaction. To address this, an affinity capture pull-down reaction was performed, in which GCN2 was incubated with purified ribosomes for 15 minutes at 32 ˚C, and then StrepTactin resin was used to capture GCN2 and any bound ribosomes. The beads Reconstitution of GCN2 Kinase Activation 78 were then extensively washed with buffer containing 0.1 % Triton X-100, and then the captured complexes were eluted with SDS-containing sample buffer. Figure 2.22. Pull-down analysis of GCN2’s interaction with the ribosome. Either full-length GCN2 or just the kinase domain was incubated with purified ribosomes (Rbs), and the resulting complexes were captured using StrepTactin resin. The beads were washed thoroughly and then the complexed proteins were eluted in sample buffer. The eluents were analysed via SDS-PAGE and Coomassie staining. The reaction totals are shown on the left and the elutions on the right. This pull-down analysis (Figure 2.22) demonstrates a direct interaction between GCN2 and ribosomes. Importantly, this interaction is only seen for the full-length protein and not the kinase domain alone, consistent with the co-migration assay in Figure 2.21. This implies a specific interaction, mediated by part of the protein sequence other than the catalytic domain. Analysis of the gel in ImageJ gave an estimated stoichiometric ratio of approximately 12 to 15 %, indicating just over a tenth of GCN2 was bound to ribosomes. Whilst it is possible to demonstrate a direct interaction between purified GCN2 and purified ribosomes, in a cellular context the interaction may be stabilised and/or regulated by additional factors. To try to identify other components of this complex in vivo, the pull-down analysis was performed again, except in the presence of cell lysate rather than purified ribosomes. The cell lysate (RRL) was gel filtered prior to incubation with the “bait” proteins Reconstitution of GCN2 Kinase Activation 79 to remove any endogenous biotin, which was found to reduce the efficiency of the pull-down. The resulting SDS-PAGE gel was then sectioned and subjected to protein identification by mass spectrometry. [This mass spectrometry was performed by Mark Skehel.] This protocol produced an extensive list of protein hits. To control for non-specific interactions, a sample with no Strep-tagged protein was included (referred to as no-bait), as well as the kinase domain of GCN2. Specific hits were then judged as those for which the spectral counts for a protein pulled down by GCN2 or the kinase domain were threefold or greater than the counts for the no-bait control. In addition, the total spectral count had to be greater than one in order for the protein hit to be included in the analysis. The final results of this analysis are shown in Table 2.2. Reconstitution of GCN2 Kinase Activation 80 Reconstitution of GCN2 Kinase Activation 81 Reconstitution of GCN2 Kinase Activation 82 It is clear that the vast majority of the GCN2-associated proteins are known ribosomal proteins (highlighted in yellow in Table 2.2), correlating with the data presented above indicating a direct interaction between GCN2 and the ribosome. Given this interaction, it is difficult to ascertain the proteins that may be associating with the complex specifically versus those that are simply associating with the ribosomes. All the protein hits observed are known to associate with ribosomes, and so it does not appear that the GCN2-ribosome interaction involves any obvious non-ribosomal proteins, although it is impossible to say whether or not these factors are affecting the interaction. For example SLFN14 (Schlafen family member 14), which shows the second highest spectral counts in the sample after GCN2, is known to be highly overexpressed in rabbit reticulocyte lysate (Pisareva et al., 2015) and predominantly ribosome-bound. The next question was whether the interaction between ribosomes and GCN2 is affected by changing the conditions, such as the addition of ATP, or the inclusion of GCN1 and GCN20. Furthermore, there was a faint possibility that the RRL could contain polysomes, i.e. mRNA transcripts with multiple ribosomes bound. The presence of these polysomes could have an impact on the apparent stoichiometry of the interaction as assessed by SDS-PAGE, as an interaction between GCN2 and a single ribosome within a polysome could result in the pull- down of all the ribosomes within the polysome. For this reason, the RRL was subjected to treatment with micrococcal nuclease to eliminate polysome structures, and the pull-down efficiency was compared to that of untreated RRL. Reconstitution of GCN2 Kinase Activation 83 Figure 2.23. A comparison between pull-down efficiency of ribosomes by GCN2 under different conditions. The first lane shows the results of incubating GCN2 in rabbit reticulocyte lysate (RRL) before capture on StrepTactin resin and washing. This complex can be eluted from the beads with 25 mM desthiobiotin (DTB; lane 2), but the majority of the complex is left on the beads (lane 3). The sample in lane 4 shows the effects of performing the entire procedure on ice (and so eliminating the 15 minute incubation at 32 ˚C). Lane 5 shows the pull-down performed in the presence of 0.5 mM ATP (the asterisk indicates the position of a newly appeared band). Lane 6 shows the effects of nucleasing the RRL before performing the pull-down to eliminate polysomes connected by mRNA, and lane 7 shows the pull-down in the presence of GCN1 and GCN20 as well as ATP. The non-specific binding is indicated in the last two lanes showing the lysate in the absence of any exogenous Strep-tagged protein. In this experiment (Figure 2.23), the interaction between GCN2 and ribosomes in cell lysate was tested and further characterised. Importantly, the GCN2-ribosomal complex can be eluted (albeit at a low efficiency) from the StrepTactin resin by desthiobiotin. It can also be observed that the 15 minute incubation at 32 ˚C of the complex is not strictly necessary for the interaction between GCN2 and ribosomes. Furthermore, it is interesting to note the effects of ATP upon the interaction. Whilst in general the presence of 0.5 mM ATP does not seem to affect the efficiency of the pull-down, the inclusion of ATP appears to result in the appearance of a band with an approximate molecular weight of 85 kDa (labelled with an Reconstitution of GCN2 Kinase Activation 84 asterisk in Figure 2.23). The identity of this band is unclear, but given that its appearance is dependent on ATP, it is tempting to speculate that it could be a regulator of GCN2 that is recruited upon GCN2 activation. Another possibility is that it is a phosphorylated version of a protein previously present (such as a ribosomal protein), and the phosphorylation (presumably by GCN2) affects its mobility through the gel. Further characterisation would need to be done to explore these possibilities, however. Figure 2.23 also shows the results of pre-treating the rabbit reticulocyte lysate with micrococcal nuclease to eliminate polysome structures. Fortunately, the treatment of the RRL with the nuclease does not affect the apparent pull- down efficiency, indicating that this is not a major problem. Finally, the effects of including GCN1 and GCN20 (both of which were Strep-tagged) alongside ATP were tested, and showed a slight increase in the pull-down efficiency, indicating that GCN1 and GCN20 are able to contribute to ribosome binding in some way. 2.3.4.8. Domain Mapping of GCN2 As described in the introduction, GCN2 is a multidomain protein consisting of five domains. At the N-terminus of the protein, there is an RWD domain, followed by a pseudokinase domain. Adjacent to this is the catalytically active kinase domain of the protein, followed by a HisRS-like domain (with sequence homology to the histidyl tRNA synthetase enzyme). Finally, there is a domain at the C-terminus of the enzyme known as the C-terminal domain (CTD). The contributions of each of these domains to the function of GCN2 has been the subject of much interest over recent years, but much of the work stems from genetic studies in yeast. It was therefore decided to perform a comprehensive truncation analysis to try to pin down the individual roles of each of the domains, and to try to identify the minimal fragment necessary and sufficient for GCN2 kinase activity, and its activation by ribosomes. 2.3.4.9. Purification of a GCN2 Truncation Library A library of domain truncations was constructed, and each construct cloned, expressed in Sf9 cells and purified with the aid of an N-terminal StrepII tag. The details of the truncation library are shown in Table 2.3. Reconstitution of GCN2 Kinase Activation 85 Table 2.3. A table describing the constructs that make up the GCN2 truncation library. The majority of the constructs purified well, and resulted in single peaks after gel filtration. However, there were a few exceptions. Unfortunately the constructs comprising the HisRS- like domain alone and the HisRS-like domain and the CTD [AIp48 and AIp45] were not expressed in Sf9 cells, and so could not be purified. Furthermore, the gel filtration profile for the mutant m2 protein (F1143L R1144L) was not symmetrical. The gel showed the existence of a higher molecular weight band, the identity of which was unclear. Mass spectrometry (performed by Mark Skehel) identified it as GCN2, and so it is most likely a higher order complex of GCN2. The deletion of the charged linker [AIp34] also produced an unusual gel filtration profile, indicating the protein was generally unhappy. As a result of these purifications, a library of constructs was assembled, as depicted in the schematic in Figure 2.24. Their final purity can be observed on the gel shown in Figure 2.25. Plasmid Name Construct Boundaries Description Expression Level Yield (mg/L cell culture) AIp23 1-1649 Full length Good 1.5 AIp47 585-1024 KD Good 0.5 AIp42 1-1492 RWD−YKD−KD−HisRS-like Good 1 AIp41 1-1024 RWD-YKD-KD Good 10 AIp40 192-1649 YKD−KD−HisRS-like−CTD Good 1 AIp44 192-1492 YKD−KD−HisRS-like Good 2 AIp58 260-1492 YKD−KD−HisRS-like Good 8 AIp49 149-539 YKD Good 1 AIp46 192-539 YKD Good 1 AIp50 260-539 YKD Good 1 AIp43 192-1024 YKD−KD Good 5 AIp52 260-1024 YKD−KD Good 4 AIp59 585-1492 KD−HisRS-like Good 1 AIp60 585-1649 KD−HisRS-like−CTD Good 2 AIp48 1057-1492 HisRS-like None None AIp45 1057-1649 HisRS-like−CTD None None AIp57 1-1649 (F1143L R1144L) m2 mutant OK 0.1 AIp34 1-1649 (Δ140-294) deletion of charged linker OK 0.1 AIp36 1-1649(D848N) kinase dead mutant Good 2 Reconstitution of GCN2 Kinase Activation 86 Figure 2.24. A schematic showing the GCN2 constructs that were produced. Figure 2.25. A gel showing the final purity of constructs A – O. 2.3.4.10. SEC-MALS Analysis of the Construct Library The next step was to determine the oligomeric state of each construct using SEC-MALS. The aim of this was to probe the regions of the protein that contribute to dimerisation, and to ascertain whether the CTD is predominantly responsible for dimerisation, as for yeast Gcn2. The results of the SEC-MALS and the estimated molecular weights are shown in Table 2.4. Reconstitution of GCN2 Kinase Activation 87 Table 2.4. A table showing the results of the SEC-MALS data, giving an estimated molecular weight for each construct. Given the theoretical molecular weight of each construct, the most likely oligomeric state is then given in the final column. Interestingly, the kinase and pseudokinase domains alone (constructs B and G/H) are both monomeric. A monomeric kinase domain stands in contrast to the crystal structure of the yeast kinase domain, which shows an anti-parallel dimeric arrangement (Padyana et al., 2005). Moreover, the authors state that their SEC-MALS analysis is consistent with a dimer in solution. This therefore implies a fundamental difference between the yeast and the mammalian proteins, validating the need for a thorough investigation of the human GCN2, as it is clear that data cannot be directly extrapolated from yeast Gcn2, despite the apparent conservation. Consistent with this, constructs containing both the pseudokinase and the kinase domains (constructs I and J) are also monomeric. The addition of the HisRS-like domain to the kinase domain (construct K), however, causes the protein to become dimeric, indicating that the HisRS-like domain is the major determinant for protein dimerisation, at least in vitro. Unfortunately, due to the lack of expression of constructs consisting of just the HisRS-like domain and the HisRS-like domain and the CTD, the oligomeric state of these constructs cannot be tested, but it seems unlikely that they would be anything other than dimeric. Finally, it is interesting to note that a construct lacking the C-terminal domain (construct C) is still dimeric. This is in contrast to the idea that the CTD is essential for dimerisation (Qiu et al., 1998). This could again be explained by differences between the yeast and the mammalian proteins, or the different results could be due to differences in measurement technique, as the abolition of dimerisation by deletion of the CTD was measured by co-immunoprecipitation. However, given the differences that can be observed in the kinase domain, it is possibly not surprising that other differences exist as well. Construct Construct Boundaries Theoretical MW (kDa) MALS MW (kDa) Oligomeric State A 1-1649 192.0 392.0 Dimer B 585-1024 54.3 59.4 Monomer C 1-1492 174.0 335.6 Dimer D 192-1649 170.0 405.2 Dimer E 192-1492 152.0 320.0 Dimer F 260-1492 144.5 294.9 Dimer G 192-539 44.1 48.4 Monomer H 260-539 36.6 37.8 Monomer I 192-1024 98.3 106.5 Monomer J 260-1024 90.3 96.7 Monomer K 585-1492 108.0 218.7 Dimer L 585-1649 126.0 263.7 Dimer M 1-1649 (F1143L R1144L) 191.9 346.7 Dimer O 1-1649 (D858N) 191.9 418.0 Dimer Reconstitution of GCN2 Kinase Activation 88 2.3.4.11. Autophosphorylation Activity of the Construct Library Once the constructs had been characterised, their functional activities were tested to try to ascertain the role and contribution of the individual domains to the activity and regulation of the kinase. This functional analysis was only performed on constructs containing the kinase domain. Figure 2.26. The autophosphorylation activity of each construct of GCN2 was tested in the presence and absence of ribosomes. Each construct was incubated with radiolabelled γ-33P-ATP for 5 minutes in either the absence or presence of ribosomes. After 5 minutes the reactions were stopped and the samples were analysed by SDS-PAGE before visualisation by autoradiography. The samples are identified by their construct ID (corresponding to the schematic in Figure 2.24), and the presence or absence of ribosomes is indicated by a plus or a minus. These results (Figure 2.26) imply many interesting facets of the kinase regulation. As shown before, the full-length kinase is stimulated by ribosomes. The lack of interaction between ribosomes and GCN2’s kinase domain (construct B) (Figure 2.22) means that it is unsurprising that the autophosphorylation activity of the kinase domain is not stimulated by ribosomes. There is a slight increase in band intensity, but it was hypothesised that this is simply due to an increase in the effective concentration of GCN2 because of the presence of ribosomes. This was demonstrated by a similar level of activation upon the inclusion of BSA. The deletion of the CTD (construct C) increases the activity of the protein both in the absence and presence of ribosomes. The increase in the basal level of autophosphorylation indicates that in the full-length protein the CTD has a role in the autoinhibition of the kinase, which has also been proposed for yeast Gcn2 (Lageix et al., 2015). It is interesting to note that the deletion of the RWD (construct D) also leads to the increase in autophosphorylation activity, but seemingly only in the presence of ribosomes. The implications of this are unclear, but it seems reasonable to conclude that the RWD also plays some role in kinase autoinhibition. Construct E, in which both the RWD domain and the CTD are deleted, is still more active than the wild-type protein, but interestingly is less active than either of the deletions in Reconstitution of GCN2 Kinase Activation 89 isolation. This implies that these two domains somehow coordinate and act together in their autoinhibitory roles, although the mechanism behind this is unclear. The deletion of the HisRS-like domain (constructs I and J) completely abrogates stimulation by ribosomes, indicating that this domain is essential for this form of stimulation. This is consistent with the m2 mutant protein (construct M), which also cannot be stimulated by ribosomes. This is a very interesting result, as it links ribosomal stimulation of the kinase with the model of activation by tRNA, which is known to be reliant upon the m2 motif. If one dismisses the unlikely possibility that tRNA is still somehow the activating signal in this assay, this demonstrates that the m2 motif does not simply function through binding tRNA. As much of the data in the literature demonstrating the importance of tRNA in the activation of GCN2 stems from analysis of an m2 mutant, the data may need to be reinterpreted if the m2 motif is important for the activation of GCN2 in the absence of tRNA. Interestingly, a construct containing only the kinase domain and the HisRS-like domain (construct K) appears to regain its ability to become stimulated by ribosomes; however, the basal level of autophosphorylation is still increased in comparison to the full-length protein (construct A). Adding the CTD to this construct (construct L) restores the normal basal level, again implicating the CTD in the autoinhibition of the kinase. Whilst these data are very interesting, the nature of the technique means they are difficult to interpret quantitatively. This is because it is unclear exactly which, and how many, sites are becoming phosphorylated on the full-length protein. It is therefore hard to ensure any differences in phosphorylation are simply due to changes in the activity level of the kinase domain rather than the removal of a subset of the substrate phosphorylation sites in certain truncated constructs. It was therefore decided to analyse the constructs’ ability to phosphorylate eIF2α, as this system allows the amount and quality of substrate to remain constant. 2.3.4.12. eIF2α Phosphorylation by the Construct Library The eIF2α phosphorylation assay was performed in the same way as the autophosphorylation assay, and again only looked at constructs that contained the catalytic kinase domain. The experiment was performed three times to allow quantification of three independent Reconstitution of GCN2 Kinase Activation 90 experiments. A representative replicate is shown in Figure 2.27, along with the quantification of the three independent repeats. Figure 2.27. The ability of each GCN2 construct to phosphorylate eIF2α was tested in the presence and absence of ribosomes. Each construct was incubated with ATP for 5 minutes at 32 ˚C in either the absence or presence of ribosomes. After 5 minutes the reactions were stopped and the samples were analysed by SDS-PAGE before visualisation by Western blotting. The samples are identified by their construct ID (corresponding to the schematic in Figure 2.24), and the presence or absence of ribosomes is indicated by a plus or a minus. The graph below shows a quantification of three independent experiments. The quantification was performed in ImageJ, and values are normalised to the phosphorylation of eIF2α by full- length GCN2 (A) in the presence of ribosomes. Plotted are the means ± the standard deviations (n=3). This quantification shows that upon the addition of purified ribosomes, the ability of the full- length protein (construct A) to phosphorylate eIF2α is stimulated by approximately 50 fold. Similarly to the results of the autophosphorylation assay, the kinase domain of GCN2 alone (construct B) is sufficient to phosphorylate eIF2α at a basal level, however no stimulation by ribosomes can be observed. Deletion of the CTD (construct C) results in a significant (approximately 20 fold) increase in the basal phosphorylation of eIF2α, consistent with the autophosphorylation data. Deletion of the RWD (construct D) has a similar effect, but is much less dramatic than for the CTD, indicating the autoinhibitory role of the CTD is more important for the prevention of promiscuous eIF2α phosphorylation. However, as previously suggested by the data from the autophosporylation assay, the deletion of the RWD domain as well as the CTD (construct E) dampens the promiscuous basal eIF2α phosphorylation, indicating their roles are cooperative. Reconstitution of GCN2 Kinase Activation 91 It is surprising to note the differences in basal eIF2α phosphorylation between constructs E and F. These constructs differ only in the presence or absence of the charged linker between the RWD and the pseudokinase domain of GCN2; however, the exclusion of the linker (in construct F) results in an approximately two-fold increase in basal eIF2α phosphorylation, indicating that this linker is also somehow implicated in the autoinhibition of the kinase. As expected, the deletion of the HisRS-like domain (constructs I and J) completely abrogates stimulation of the kinase by ribosomes. The minimal fragment that is necessary for a reasonable amount eIF2α phosphorylation in the presence of ribosomes appears to comprise the kinase domain and the HisRS-like domain (which correlates with the previous autophosphorylation data). However, to achieve maximal stimulation by the presence of ribosomes, GCN2 appears to require the pseudokinase domain (indicated by comparison of constructs K and L to constructs C, D, E and F). In summary, each domain of GCN2 appears to play some role in controlling its activity. The kinase domain of the enzyme is sufficient to basally phosphorylate the protein’s substrate eIF2α, however it is insensitive to ribosomal stimulation. The HisRS-like domain increases the protein’s intrinsic ability to phosphorylate eIF2α, and also enables the kinase to become stimulated by the addition of ribosomes. This stimulation is somehow further enhanced by the pseudokinase domain. The C-terminal domain is essential for the full inhibition of the kinase domain in the absence of an activating signal, and the RWD domain as well as the charged linker between the RWD domain and the pseudokinase domain are also involved in this autoinhibition. 2.3.4.13. Ribosome Binding by the Construct Library The next aim was to characterise the differential ribosome binding properties of each GCN2 construct to try to ascertain which domains of GCN2 are contributing to the interaction, and whether this correlates with the activation of each construct by ribosomes. To test this, each GCN2 construct was incubated with purified ribosomes, and then the complexes captured using StrepTactin resin as described previously. The results are shown in Figure 2.28. Reconstitution of GCN2 Kinase Activation 92 Figure 2.28. Analysis of the GCN2 construct library shown in Figure 2.24 for their ability to interact with ribosomes, as tested by pull-downs. Each construct was incubated with 100 nM purified ribosomes for 15 minutes at 32 ˚C, before bound proteins were captured on StrepTactin resin. After washing, the bound proteins were eluted from the beads in sample buffer, and the samples were analysed by SDS-PAGE. The reaction totals and flowthroughs are shown at the top of the image, and the elutions at the bottom left. A quantification of the elutions from three independent experiments is shown at the bottom right. The quantification was calculated in ImageJ, and used three different ribosomal proteins for each sample. Each of these was normalised to the pull-down by full-length GCN2, and then the differences averaged. Plotted are the means ± the standard deviations (n=3). This analysis indicates three separate domains of GCN2 contribute to the ribosomal binding by the protein. The first is the CTD, as when it is deleted (construct C) the pull-down efficiency drops by approximately 50 %. Second is the pseudokinase domain, which when eliminated (compare constructs D and L) causes a significant reduction in ribosomal binding. Finally, the HisRS-like domain appears to be essential for an interaction between GCN2 and ribosomes, as when it is deleted (constructs G – J) the constructs show very low affinity for the ribosome in comparison to construct E. Interestingly, the m2 mutant of GCN2 (construct Reconstitution of GCN2 Kinase Activation 93 M) shows a completely abrogated ability to bind to ribosomes, implying the m2 motif has some role in general RNA binding, rather than specifically tRNA. Truncation analysis is intrinsically limited by the fact that it is impossible to know how truncations or mutations are affecting the protein in unpredictable ways. It would be therefore beneficial to use a more precise technique to map the ribosomal binding sites of GCN2. This will be further addressed in Chapter Three. 2.3.4.14. Investigating specificity of activation Having comprehensively demonstrated that GCN2 is able to directly interact with ribosomes, and that this results in the activation of GCN2 towards its substrate eIF2α, the question arose concerning specificity. Ribosomes are always present in the cell, and yet GCN2 is only activated under conditions of nutrient stress. Therefore, there must be a layer of inbuilt specificity in terms of when ribosomes can activate GCN2. There are multiple options for how this could be achieved. Firstly, it is possible that in a cellular context, there exists a factor or factors that confer specificity on the system. The fact that GCN2 can bind to and become activated by ribosomes in an in vitro context argues against the requirement for a positive- acting factor. However, it is plausible that there is a factor (or factors) present on translating ribosomes that prevents GCN2 from stably or productively associating with ribosomes and therefore preventing the kinase from becoming activated. The possible identity of this factor (or factors) is still unclear, but it would be very interesting to test the ability of different components of the translational machinery for their ability to prevent ribosomal activation of GCN2. A second possibility is that GCN2 is able to recognise a specific ribosomal state, which is only stably present when translation has stalled. Throughout the translational cycle, the ribosome is known to adopt several different rotational states, but under normal conditions each of these states is transient and will quickly move on to the next one in the sequence as the tRNAs are processed and the nascent polypeptide chain is synthesised. However, under conditions that cause translation to stall, the lifetime of a specific rotated state can be extended. If GCN2 were able to recognise a specific rotated ribosomal state, this would constitute an interesting mechanism for sensing ribosomal processivity. To test this hypothesis, the ribosomes were pre-incubated with different antibiotics that cause the Reconstitution of GCN2 Kinase Activation 94 ribosome to become trapped in different rotational states. The ribosomes were then tested for their ability to stimulate the autophosphorylation activity of GCN2. Figure 2.29. Autophosphorylation assay. The purified ribosomes were pre-incubated with cycloheximide (CHX), emetine, anisomycin or didemnin B (DDB) before being added to GCN2. The reactions were then incubated with γ-33P-ATP at 32 ˚C for 5 minutes. The samples were then analysed by SDS-PAGE and visualised by autoradiography. As shown in Figure 2.29, the addition of the antibiotics tested to the reaction does not seem to change the ability of the ribosomes to stimulate GCN2. This implies that the rotational states that these antibiotics entrap do not affect the ribosome’s ability to activate GCN2. These four antibiotics are generally considered to represent a reasonably comprehensive summary of the four most stable rotational states of the ribosome (Lareau et al., 2014), but it is of course possible a rotational state that is not here tested has a large effect on GCN2 activation. It is also possible that the rotational state of the ribosome plays an indirect part in a cellular context. A third possibility would be the existence of a ribosomal modification that occurs as part of the cellular response to nutrient starvation. The importance and extent of these modifications has been recently characterised (Natchiar et al., 2017), and it would be an attractive mechanism. However, given the ribosomes have not been subjected to any form of in vivo amino acid starvation before purification, it seems relatively unlikely that they would have been specifically modified in some way. It should be noted that the production of the reticulocyte lysate in rabbits relies upon their prior treatment with acetylphenylhydrazine (to Reconstitution of GCN2 Kinase Activation 95 increase reticulocyte production), but there is no evidence to suggest this emulates the effects of amino acid starvation. Finally, recently evidence has emerged for a cellular pathway for the recognition of ribosomal collisions (Simms et al., 2017). During normal translation, there are multiple ribosomes on one mRNA transcript. In times of stress, when one ribosome on the mRNA transcript stalls, the other ribosomes on the message will continue until they reach the initial ribosome, creating a ribosomal collision. This collision event can be specifically recognised, and initiates a no-go decay pathway. It seems plausible that GCN2 could also recognise this collision event. However, much more work needs to be done to distinguish between the many options. Reconstitution of GCN2 Kinase Activation 96 2.4. Conclusions The role of GCN2 in the Integrated Stress Response has been the subject of much investigation. The physiological importance of the kinase in a wide range of cellular processes, and its position at the intersection of signalling networks orchestrating the assessment of nutrient availability and cell growth and proliferation, make it a potentially key therapeutic candidate. However, most of the current knowledge of the protein and the mechanisms behind its regulation are extrapolated from genetic studies in yeast. Due to the difficulties in determining precise molecular mechanisms through genetic studies, as well as the possibility that the systems are not entirely identical between yeast and mammals, this means that the understanding of the mammalian ISR is far from comprehensive. To begin to address this issue, in this work human GCN2 was initially expressed in Sf9 cells and purified to high homogeneity. A variety of biophysical techniques were then used to characterise the intrinsic properties of the protein: SEC-MALS showed the kinase to exist as a stable dimer, and SPR demonstrated its weak affinity to deacylated tRNA. The putative regulatory partners GCN1 and GCN20 were then also cloned, expressed and purified to test their ability to interact with GCN2 via pull-downs. This showed an apparently very low affinity between GCN2 and GCN1, and no interaction between GCN20 and GCN1 or GCN2. This implies significant intrinsic differences between the yeast and mammalian systems: yeast Gcn2 is thought to bind to Gcn1 with a relatively high affinity (as judged from pull-downs), and Gcn1 and Gcn20 appear to form a stable complex. Next, kinase assays were used to demonstrate the purified material was functional. Stable and tractable reconstituted systems were developed to allow investigation of both GCN2 autophosphorylation and eIF2α phosphorylation, allowing the effects of different potential regulatory inputs to be tested. Importantly, these studies showed that to achieve any stimulation of GCN2 by tRNA, the concentration of deacylated tRNA had to be increased to very high levels (5 µM). This seems to imply that a simple model whereby deacylated tRNA is able to bind to GCN2 and cause its activation is unlikely to explain the exquisite sensitivity that GCN2 is able to achieve in recognising amino acid depletion. However, the finding that a low concentration of purified ribosomes is able to dramatically stimulate GCN2’s kinase activity via a direct binding event fully supports the model put forward by Ishimura et al. (Ishimura et al., 2016). Given that the majority of previous experiments have been performed Reconstitution of GCN2 Kinase Activation 97 in vivo, it seems possible that the rise in deacylated tRNA could inhibit translation and lead to ribosomal stalling. It may be that these stalled ribosomes act as the direct activator of GCN2, rather than the deacylated tRNA, assigning GCN2 a role as a potential monitor of ribosomal processivity. In whole cell experiments, altering the balance between acylated and deacylated tRNA without impacting on the translational efficiency would be extremely challenging, and so it is perhaps unsurprising that the effects of these factors in isolation could be difficult to distinguish in vivo. It would be very interesting to monitor the translational processivity upon an addition of a large amount of deacylated tRNA to determine if the prevalence of stalled ribosomal complexes is increased. If so, one would then need to experimentally separate the effects of the two factors, for instance by engineering the bulk deacylation of a specific tRNA that is not used in translation under certain conditions, and determining the effects upon GCN2 activity. Of course, it is also possible that this phenomenon is specific to the human proteins, and represents a fundamentally different system of detection of amino acid depletion. However, the conservation of the kinase across the eukaryotic spectrum makes this seem unlikely. Given the existence of the histidyl tRNA synthetase-like domain of GCN2, and especially the conservation of the m2 tRNA-binding motif, it seems incongruous that deacylated tRNA plays no role in the process. However, the potential role for tRNA within this model is at the moment unclear. It has previously been suggested that deacylated tRNA could be transferred out of the ribosomal A site, in a manner potentially mediated by GCN1, to the HisRS-like domain of GCN2, where it would cause activation of its kinase activity. Whilst possible, this seems unlikely due to the specificity of the elongation factor eEF1A for acylated tRNA molecules over deacylated tRNA molecules. Therefore, the frequency of recruitment of deacylated tRNA to the ribosomal A site in a codon-specific manner seems likely to be very low, meaning that it is unclear how cognate deacylated tRNA could be present at the ribosome in this scenario. Conceptually, this would imply that an empty A site is the most likely ribosomal state that should activate GCN2, and so how tRNA is involved in this remains unclear. In order to ascertain the roles and contributions of each of GCN2’s five domains, a library of mutations and truncations was constructed, and subjected to testing for autophosphorylation activity, eIF2α phosphorylation and ribosome binding. A summary of the results is shown in Table 2.5. Reconstitution of GCN2 Kinase Activation 98 Table 2.5. A summary of the data obtained from the GCN2 construct library. The oligomeric state for each construct (except construct N) was determined by SEC-MALS. Autophosphorylation activity was tested by incubation with γ-33P-ATP in the presence and absence of ribosomes, and visualised by SDS-PAGE and autoradiography. eIF2α phosphorylation was tested by incubation with ATP in the presence and absence of ribosomes, and visualised by SDS-PAGE and western blotting. Ribosome binding was tested by affinity capture of the constructs in the presence of purified ribosomes, and visualised by SDS-PAGE and Coomassie staining. From these data, we can begin to characterise the function of each of the domains (summarised in Figure 2.30). Figure 2.30. A summary of the details of domain mapping for GCN2. The RWD domain appears to have a role in autoinhibition, as its deletion (construct D) leads to an increased level of autophosphorylation in the presence of ribosomes, although this effect is not extended to eIF2α phosphorylation by GCN2. The RWD domain is not important for ribosomal binding. Interestingly, the charged linker connecting the RWD domain and the pseudokinase domain also appears to be important in maintaining the latent state of the kinase domain, as deletion of the linker region (compare constructs E and F) causes an increase in - Rbs + Rbs - Rbs + Rbs A 1-1649 Full length Dimer + +++ + +++ ++++ B 585-1024 KD Monomeric + + + + none C 1-1492 RWD−YKD−KD−HisRS-like Dimer ++ ++++ ++ +++ ++ D 192-1649 YKD−KD−HisRS-like−CTD Dimer + ++++ + +++ ++++ E 192-1492 YKD−KD−HisRS-like Dimer + +++ + +++ ++ F 260-1492 YKD−KD−HisRS-like Dimer + +++ ++ +++ ++ G 192-539 YKD Monomeric none none none none none H 260-539 YKD Monomeric none none none none none I 192-1024 YKD−KD Monomeric ++ ++ + + none J 260-1024 YKD−KD Monomeric ++ ++ + + none K 585-1492 KD−HisRS-like Dimer ++ +++ ++ +++ none L 585-1649 KD−HisRS-like−CTD Dimer + +++ + +++ + M 1-1649 (F1143L F1144L) m2 mutant Dimer + + none none none N 1-1649 (Δ140-294) deletion of charged linker - + + none + none O 1-1649 (D848N) kinase dead mutant Dimer none none none none ++++ Construct Name Construct Boundaries Description Oligomeric State autoP eIF2αP Rbs Binding? Reconstitution of GCN2 Kinase Activation 99 basal GCN2 activity. Its position next to the RWD domain means that it seems likely that this is linked to the mechanism by which the RWD domain inhibits unstimulated GCN2. The contribution of the pseudokinase domain of GCN2 to the protein’s functionality appears relatively complex, and difficult to unpick. The pseudokinase domain is clearly involved in the autoinhibition of the kinase, indicating that the entire N-terminal half of the protein is involved in this somehow. This is demonstrated by the increase in basal kinase activity between constructs K and D, both in terms of autophosphorylation and eIF2α phosphorylation. The presence of the pseudokinase domain is also clearly important for ribosomal binding, as its deletion dramatically decreases ribosomal pull-down efficiency. Puzzlingly, despite the seemingly abrogated interaction with ribosomes, the kinase activity of this construct is still stimulated by the presence of ribosomes. One explanation for this could be that due to the increased basal activity even in the absence of ribosomes, the small amount of binding that is occurring for construct K is sufficient to achieve similar autophosphorylation and eIF2α phosphorylation levels as for the wild-type protein. The kinase domain alone is sufficient to basally phosphorylate eIF2α; however, it cannot bind to ribosomes, nor become stimulated in their presence, demonstrating the importance of the surrounding regulatory domains. The HisRS-like domain seems to orchestrate many of GCN2’s key functions. This domain is essential for the dimerisation of the protein, ribosome binding and stimulation of the kinase activity by ribosomes (as shown by comparing constructs E and I). Mutation of the m2 motif (construct M) completely abolishes any ribosomal binding and activation of GCN2 by ribosomes, indicating this motif is important for the protein’s function even in the absence of tRNA. Finally, the C-terminal domain of GCN2 plays a role in the maintenance of the kinase in an autoinhibited state. This is particularly obvious when comparing constructs K and L, as it is clear that the addition of the CTD to the construct is able to restore basal levels of activity to the kinase. There is much that can be done to extend this analysis of the mechanistic and regulatory details of the activation of GCN2. Whilst useful, the truncation analysis of which domains of GCN2 are important for ribosomal binding gives relatively general information, and it is difficult to infer any specifics concerning the interaction. This is especially the case as the Reconstitution of GCN2 Kinase Activation 100 evidence points to a multi-partite interaction involving the pseudokinase domain, the HisRS- like domain and the C-terminal domain. A more precise view of the interaction would allow the manipulation of individual facets to gain a more detailed picture of exactly how the activation is achieved, and this will be explored further in Chapter Three. Furthermore, the characterisation of a complex consisting of GCN2 bound to the ribosome would provide significant insights into the molecular mechanisms, and efforts towards this are detailed in Chapter Four. A very important part of this story will be to determine whether and why GCN2 becomes activated by stalled ribosomes instead of translating ribosomes. Probing different in vitro translation systems to this end could prove very useful, and it would be very interesting to see whether it is possible to observe changes in GCN2 activity by altering the level of nutrient availability, or changes to the mRNA transcript. If a particular ribosomal state or process could be identified to activate GCN2, it would then be interesting to try to replicate this state in cells and look for cellular activation of endogenous GCN2, to confirm any in vitro hypotheses. 101 Chapter Three – Characterisation of the interaction between GCN2 and the ribosome using HDX-MS 3.1. Introduction The functional analyses described in Chapter Two have shown the ribosome to have an important role in the activation of the kinase GCN2. Affinity capture pull-downs in combination with truncation analysis have characterised a direct interaction that is mediated by the pseudokinase domain, the HisRS-like domain and the C-terminal domain. However, it is challenging to get detailed information about the interaction from this sort of analysis. Furthermore, it is not possible to gain any information concerning which part of the ribosome GCN2 is binding to from these types of experiments. This information could have important implications for how GCN2 discriminates between translating and stalled ribosomes. For these reasons, the aim was to characterise the interaction between GCN2 and the ribosome in more detail using a technique originating in the 1950s based upon measuring rates of Hydrogen-Deuterium exchange between a protein and a deuterated solvent (Hvidt and Linderstrøm-Lang, 1954). Originally, the Hydrogen-Deuterium exchange was measured by ultracentrifugation through a density gradient, but it has subsequently been combined with both 2-dimensional nuclear magnetic resonance (HDX-NMR) (Otting and Wüthrich, 2009) and mass spectrometry (HDX-MS) (Zhang and Smith, 2008), allowing the incorporation of deuterium to be measured more accurately. Whilst able to theoretically generate single amino acid resolution data, HDX-NMR requires an isotopically labelled sample, which is often a bottleneck. Furthermore, NMR is intrinsically limited to the study of smaller proteins, meaning that HDX-MS is often the method of choice despite a lower nominal attainable resolution. Modern HDX-MS is a powerful technique that can be used to study many different facets of a protein’s structure-function relationship, including stability, folding and dynamics (Katta and Chait, 1991; Wales and Engen, 2006), and interactions with other molecules (Vadas et al., 2017). HDX-MS relies upon studying the ability of hydrogen atoms within a protein to exchange with deuterium present in solution. In a protein, there are generally three ‘types’ of hydrogen atoms: those bonded to carbons of the amino acid, those bonded to polar Characterisation of the GCN2-Ribosome Interaction by HDX-MS 102 atoms of the amino acid side chains and those bonded to nitrogen in an amide bond (Figure 3.1). Figure 3.1. The chemical structure of part of a polypeptide. The hydrogen atoms (H) part of methyl groups or bound to the α carbons (Cα) are circled in orange, the hydrogen atoms that are part of non-methyl terminal groups of amino acid side chains are shown in blue and the hydrogen atoms that are part of amide bonds are shown in green. The amino acid sequence is shown below. The isotopic exchange rate of each ‘type’ of proton is very different. Protons that are bonded to carbons have an exchange rate in the order of 1010 seconds under physiological conditions. This extremely slow exchange rate precludes the observation of the exchange reaction on any useful experimental timescale, and means they can be considered as essentially non- exchangeable. Conversely, the labile hydrogen atoms found in the terminal, polar or charged groups of amino acids exchange with solvent on a sub-millisecond timescale under physiological conditions, meaning that the deuterium atoms rapidly undergo back-exchange, in which the deuterium re-exchanges with protons in the non-deuterium containing aqueous buffers, for example in the fluidics system of the mass spectrometer. The exchange reaction is also less successfully quenched by standard techniques, making the retention of any peptide deuteration for analysis extremely challenging. It is interesting to note that techniques have been developed that are able to measure the exchange of labile hydrogens with ND3-labelling gas in the gas phase by mass spectrometry (Geller and Lifshitz, 2005; Rand et al., 2009a; Winger et al., 1992). Typically, however, the majority of HDX-MS-based studies cannot give meaningful information about the exchange of these labile hydrogens. The third ‘type’ of hydrogens are those atoms that are bonded to nitrogen and are part of the amide bond. These protons are able to exchange with the deuterium present in the solvent at measureable rates (in Characterisation of the GCN2-Ribosome Interaction by HDX-MS 103 the order to seconds to minutes under physiological conditions), meaning that differences in deuteration levels over time can be measured, and the rates of back exchange can be kept reasonably low. This therefore allows the analysis of the deuteration levels across the protein sequence (with the exception of proline residues), giving important information about the different regions of the protein. The rate of exchange (given by the intrinsic rate constant, or kch) is dependent on several different factors: 1. Temperature 2. pH 3. Surrounding residues 4. Involvement in secondary structures 5. Solvent accessibility An increase in temperature increases the rate of isotopic exchange according to the Arrhenius equation, and this will be discussed in more detail later. The exchange reaction is also pH- sensitive, with an exchange minimum occurring at approximately pH 2.5. This minimum occurs as a result of the fact that the exchange reaction can be either acid-catalysed or base- catalysed (Bai et al., 1993; Berger et al., 1959): At physiological pH (~ pH 7.5), the base-catalysed reaction dominates and k1 is predicted to be about 1010 M-1 min-1 (Bai et al., 1993; Molday et al., 1972). Above pH (or pD) 4, the rate rises by approximately an order or magnitude for each pH unit (Smith et al., 1997). This means that manipulation of the pH of the solution has dramatic effects on the rate of reaction, which allows the exchange reaction to be effectively quenched by decreasing the pH from ~ 7.5 to ~ 2.5. This causes a reduction in the rate of isotopic exchange of approximately four to five orders of magnitude. The effects of the surrounding amino acids must also be considered, as the steric and electrostatic properties of the nearby amino acids can also affect isotopic exchange rates. For example, it has been shown that an amide hydrogen between two isoleucine residues will have an exchange rate approximately 10-20 fold lower than an amide hydrogen between two alanine residues under the same condition (Bai et al., 1993). For a completely unfolded kch = k1[OD-] + k2[D3O+] + k3[D2O] (2) Characterisation of the GCN2-Ribosome Interaction by HDX-MS 104 peptide, involved in no secondary structural elements, the rate of isotopic exchange will be purely dependent on these three factors, and under physiological-like conditions (pH 7.0, 22 ˚C) the rate of exchange of a typical amide hydrogen is in the order of milliseconds to seconds (Bai et al., 1994). However, when part of a folded protein, amide hydrogens are often engaged in hydrogen-bonding networks as part of secondary structure elements such as α helices or β sheets. The involvement of an amide hydrogen in a hydrogen bond can theoretically slow the rate of isotopic exchange by up to a factor of 1010, making the presence or absence of secondary structure the major determinant of the rate of exchange (Coales et al., 2010; Milne et al., 1998; Skinner et al., 2012). It is important to note that even hydrogens involved in very stable hydrogen bonds are able to exchange with the solvent due to intrinsic structural fluctuations that can briefly disrupt the hydrogen-bonding network. This can be represented by the following equation, known as the Linderstrøm-Lang model (Berger and Linderstrøm- Lang, 1957): where N – Hcl represents an amide hydrogen engaged in a hydrogen bond, N – Hop represents an amide hydrogen not engaged in any hydrogen bonds and the rate constants kop and kcl describe the equilibrium between the two states. N – Dop represents the amide hydrogen having been replaced with deuterium and N – Dcl is when the hydrogen bond has reformed. The rate constant kch represents the intrinsic rate of exchange of the amide hydrogen in the absence of any secondary structure. The overall decrease in exchange rate due to the involvement of the amide proton in secondary structures and possibly a lack of solvent exposure can be described as a protection factor (P), where P is the ratio between the intrinsic rate constant kch and the experimentally determined rate constant kex. It is important to note that this reaction mechanism is based upon the assumption that no back-exchange is occurring, i.e. that kch is unidirectional due to the excess of D2O in the incubation buffer. There are two possible kinetic schemes for the exchange of amide hydrogens in folded proteins: EX2 and EX1 kinetics (Hvidt and Nielsen, 1966). EX2 kinetics describe situations in which the rate of exchange of the amide hydrogen (kch) is significantly slower than the opening and closing of the hydrogen bond (kch << kcl), and so the rate of exchange provides a Ν – Ηcl  Ν – Ηop → N – Dop  N – Dcl kop kcl kch D2O kcl kop (2) Characterisation of the GCN2-Ribosome Interaction by HDX-MS 105 reliable read out of the stability of the secondary structural elements in which the amide hydrogen is involved. This is the case for the majority of amide hydrogens in a folded protein, and can be identified by a binomial ion distribution and characteristic gradual increase in mass of a peptide over time. EX1 kinetics, on the other hand, occur when the intrinsic exchange of the amide hydrogen is fast in comparison to the opening and closing of the hydrogen bond (kch >> kcl). Regions with EX1 kinetics, therefore, are often areas of the protein that are not involved in secondary structures, such as flexible loops or hinges. Peptides containing amide hydrogens undergoing exchange by EX1 kinetics are characterised by a hallmark bimodal isotopic spectrum, representing multiple populations. To perform HDX-MS, the protein is incubated in deuterated buffer for a series of time points, allowing the amide hydrogens to exchange for deuterium. The exchange process is then quenched by the addition of formic acid to reduce the pH to ~ 2.5, thus quenching the reaction. Guanidinium chloride (GdHCl) is also added to denature the protein and the sample is then rapidly frozen. The sample is subsequently digested into short peptides by a protease (usually pepsin due to its functionality under acidic conditions and broad sequence preference, allowing for redundancy in the peptides produced) and the peptides are separated by ultra- performance liquid chromatography (UPLC), before being analysed by mass spectrometry. The incorporation of deuterium ions within each peptide can be measured as an increase in mass compared to the non-deuterated peptide, and thus over the time course a rate of exchange can be deduced. One of the most powerful facets of HDX-MS analysis is the comparison between the deuteration profiles of the same protein under different conditions, such as in the presence and absence of binding partners (Xiao et al., 2003). The chemical identity of these binding partners largely does not matter, so long as the affinity of the interaction is sufficient (Kd ≤ µM) and their presence does not affect the protein digestion and mass spectrometric analysis of the peptides. It is therefore possible to look at interactions between two proteins, between proteins and small molecules, and between proteins and lipid membranes (Anand et al., 2003; Chandramohan et al., 2016; Dautant et al., 2017; Vadas et al., 2017). A change in the rate of deuteration in the presence of a binding partner signifies the involvement of that particular peptide in some form of structural change that accompanies binding, and allows the localisation of a potential binding site (Figure 3.2). It is important to note that if there are allosteric changes that occur within the protein structure upon binding, these are also likely to Characterisation of the GCN2-Ribosome Interaction by HDX-MS 106 manifest as changes to the deuteration profile of the corresponding peptides (Rand et al., 2006). Whilst this gives important information concerning the conformational dynamics that accompany binding, it can occasionally lead to uncertainty about the specific site of binding, and so it is important to biochemically validate binding interfaces in cases where multiple regions show deuteration changes. Figure 3.2. Schematic of an HDX-MS experiment to define a binding site of one protein on to another. The protein of interest is incubated either alone or in the presence of excess binding partner in deuterated buffer for a series of time points. The reactions are then quenched by a drop in pH and rapid freezing. The protein of interest is digested into peptides by the acid-functional protease pepsin. The peptides are then separated by ultra-performance liquid chromatography (UPLC) before being analysed by mass spectrometry. Differences in the deuteration profiles of the two states indicate regions that may be involved in binding. Characterisation of the GCN2-Ribosome Interaction by HDX-MS 107 It is important to note that the differences in deuteration that can be detected are dependent on the labelling times that are tested. Different peptides will have different profiles of deuterium uptake, and so it is important to probe different labelling times to have the maximal chance of observing a difference between deuterium uptake rates of a peptide under different conditions. Very short labelling times are difficult to measure; however, various strategies can be employed to extend the labelling window (Coales et al., 2010). HDX-MS analysis allows the generation of information concerning the secondary structure and solvent exposure of the protein across the amino acid sequence, with the resolution usually limited by the length and number of peptides from the peptic digestion that can be identified by the mass spectrometer. Interestingly, this limit is starting to be pushed by the development of single amino acid resolution strategies involving further ion fragmentation within the mass spectrometer via electron transfer dissociation (Landgraf et al., 2012) or electron capture dissociation (Abzalimov et al., 2013; Landgraf et al., 2012; Pan et al., 2009; Rand et al., 2011; 2009b). This relies upon the ability to fragment the peptide ion without causing the internal rearrangement of labile hydrogen/deuterium atoms, known as scrambling, allowing an increased level of attainable resolution. Due to complexities in the optimisation of this technique, it is still only possible in some cases, meaning that it is typically only performed on individual peptides at the moment. For samples that are significantly larger or more complex, the quantity of peptides that can be detected becomes restricted by chromatographic limitations of the UPLC. The need to minimise back-exchange prevents the simple lengthening of the UPLC elution to increase separation. Therefore, to increase the potential peptide identification, the peptides must be separated further whilst in the gas phase of the mass spectrometer. To this end, the technique of ion mobility spectrometry (IMS) has recently been incorporated into the HDX-MS pipeline (Distler et al., 2014; Donohoe et al., 2015; Iacob et al., 2008). IMS allows the determination of the collisional cross-sectional size and charge of the ions in the gas phase, and thus enables greater separation between peptides during analysis without lengthening the experimental time. Given the versatility of HDX-MS, it was decided to use this technique to gain more information concerning the interaction between GCN2 and the ribosome. The extremely large molecular weight of the ribosome (approximately 2.4 MDa of protein and 3.2 MDa in total) means that this appears to be the largest complex that has been investigated by HDX-MS so Characterisation of the GCN2-Ribosome Interaction by HDX-MS 108 far. This meant that the sample preparation, data collection and data interpretation all had to be optimised to gain the most information possible. All HDX-MS experiments detailed in this chapter were performed and the data analysed in collaboration with Dr Glenn Masson. Characterisation of the GCN2-Ribosome Interaction by HDX-MS 109 3.2 . Materials and Methods 3.2.1. Initial GCN2-Ribosome Interaction Analysis by HDX-MS Initial Hydrogen-Deuterium exchange-mass spectrometry (HDX-MS) samples were prepared from two reactions: one consisting of ribosomes alone at 0.3 µM, and a second consisting of ribosomes at 0.3 µM and GCN2 at 0.5 µM. All reactions were assembled in RNC buffer [50 mM HEPES pH 7.5, 100 mM KOAc, 5 mM MgAc2, 1 mM DTT]. Both reactions were incubated on ice for 15 minutes and then aliquoted into 10 µL aliquots. To each reaction, 40 µL of deuterated RNC buffer [50 mM HEPES pH 7.5, 100 mM KOAc, 5 mM MgAc2, 94.8 % D2O (Acros Organics 7789-20)] was added and the reactions mixed rapidly, producing a final D2O concentration of 75.84 %. Initially, four incubation times were tested in triplicate: 3 seconds at 0 ˚C, 3 seconds at 22 ˚C, 30 seconds at 22 ˚C and 5 minutes at 22 ˚C. All reactions were quenched via the addition of 20 µL ice-cold Quench buffer [5 M GdCl, 8.4 % formic acid], followed by freezing in liquid nitrogen. The pH of the reaction after quenching was measured as pH 2.5 for a control sample. Each sample was thawed and injected on to the M-Class Acquity UPLC with HDX technology (Waters) kept at 0.1 ˚C. The proteins were digested on an Enzymate Pepsin Column (Waters, 186007233) at 15 ˚C for two minutes. The peptides were then eluted from the column on to an Acquity UPLC BEH C18 column (Waters, 186002346), equilibrated in Pepsin-A buffer [0.1 % formic acid] using a 3 - 43 % gradient of Pepsin-B buffer [0.1 % formic acid, 99.9 % acetonitrile] over 22 minutes. Data were collected on a Waters Synapt G2-Si with an electrospray ionisation source, from 50 to 2000 m/z. The spray voltage was 3.0 kV. Peptides were identified using ProteinLynx Global Server (Waters, 720001408EN). Three replicates of non-deuterated ribosomes were analysed by MSe (Silva et al., 2006) to allow peptide identification. To qualify for further analysis, peptides had to fulfil the following criteria: 1. Minimum intensity of 5,000 counts 2. Maximum length of 25 amino acids 3. Minimum number of three products 4. Minimum number of 0.05 products per amino acid 5. Maximum mass error of 10 ppm Characterisation of the GCN2-Ribosome Interaction by HDX-MS 110 Peptides that fulfilled the stated criteria were imported into DynamX (Waters, 720005145en) for data analysis. All data was subjected to automated data processing by DynamX, followed by manual inspection of the data. Peptides that showed an unacceptably low signal-to-noise ratio, precluding robust analysis, were eliminated at this stage. 3.2.2. Optimised GCN2-Ribosome Interaction Analysis by HDX-MS Optimised Hydrogen-Deuterium exchange-mass spectrometry (HDX-MS) samples were prepared from two reactions: one consisting of ribosomes alone at 0.5 µM, and a second consisting of ribosomes at 0.5 µM and GCN2 at 2.5 µM. All reactions were assembled in RNC buffer. Both reactions were incubated on ice for 15 minutes and then aliquoted into 10 µL aliquots. To each reaction, 40 µL of deuterated RNC buffer [50 mM HEPES pH 7.5, 100 mM KOAc, 5 mM MgAc2, 94.8 % D2O (Acros Organics 7789-20)] was added and the reactions mixed rapidly, producing a final D2O concentration of 75.84 %. Three incubation times were tested in triplicate: 5 minutes, 50 minutes and 500 minutes, all at 32 ˚C. To ensure the prolonged 500-minute time point was not having a detrimental effect on the structure of the ribosome, a pulse experiment was also performed in which the 10 µL protein aliquot was incubated at 32 ˚C for 495 minutes before 40 µL deuterated RNC buffer was added for 5 minutes. All reactions were quenched via the addition of 20 µL ice-cold Quench buffer [5 M GdCl, 8.4 % formic acid], followed by freezing in liquid nitrogen. The data were processed in the same way as the initial HDX-MS experiment described above. Following the processing of the dataset, changes in the deuteration profile for each peptide upon the addition of excess GCN2 were statistically analysed to determine the level of significance for the dataset. Peptides that showed changes greater than this threshold were therefore considered to be significantly affected by the presence of GCN2. Characterisation of the GCN2-Ribosome Interaction by HDX-MS 111 3.3. Results and Discussion 3.3.1. Optimisation of HDX-MS to allow the study of ribosomal interactors The aim was to detect the site on the ribosome to which GCN2 binds. HDX-MS was performed initially using the standard parameters, and then the process was optimised to improve the rigour of the data. Initially, 0.3 µM purified ribosomes were mixed with 0.5 µM GCN2 and the reactions were incubated on ice for 15 minutes, before being exposed to deuterated buffer at room temperature for a series of time points: 0.3 seconds, 3 seconds, 30 seconds, 5 minutes and 50 minutes. It is important to note that it is not possible to achieve a time point of 0.3 seconds manually, and so the effect of temperature on the exchange rate was used to generate a very short time point. Using the Arrhenius equation: where A is the frequency factor, Ea is the activation energy for the exchange of the amide hydrogen (equalling 17 kcal mol-1 (Bai et al., 1993)), R is the gas constant (1.985 x 10-3 kcal K-1 mol-1) and T is the temperature in Kelvin, it can be calculated that reducing the temperature from 22 ˚C to 0 ˚C reduces the rate of reaction by approximately ten fold (Coales et al., 2010). This means that performing the deuteration with all components on ice should lead to a tenfold reduction in the rate of deuteration. Given that the fastest time point that can be performed manually is 3 seconds, this generates a 0.3 second time point, allowing the analysis of the amide hydrogen atoms that exchange very quickly. The ribosome has an approximate molecular weight of 3.2 MDa. Proteins make up approximately 2.4 MDa of this, the rest being ribosomal RNA (rRNA). HDX-MS cannot detect deuteration of nucleic acids, and so only changes to protein surfaces could be detected. Given that the exterior of the eukaryotic ribosome is predominantly made up of protein, it was hoped that GCN2 was not bound to a site consisting of purely RNA. However, this was a potential source of concern. Ultimately, this meant that 2.4 MDa worth of peptides had to be analysed over a series of five timepoints, for three independent replicates. The data was initially processed automatically by the software DynamX, but the complexity of the spectra meant that each peptide had to be examined manually to ensure the processing was done kch = Ae -Ea RT (3) Characterisation of the GCN2-Ribosome Interaction by HDX-MS 112 correctly. This resulted in the generation of spectra for 742 peptides, representing coverage of approximately 56.95 % (i.e. the total number of amino acids present in one or more detected peptides out of the total number of amino acids present in all the ribosomal proteins searched for – for a list of ribosomal proteins please see Supplementary Table 1). It should be noted that this value represents the proportion of peptides that were searched for in the dataset. Typically, HDX-MS experiments are performed on recombinant samples, meaning that the protein components are known. The ribosomal sample, on the other hand, was purified from an endogenous source, meaning that it is impossible to be sure of the exact protein composition. The mean redundancy of the dataset (the average number of peptides in which each amino acid appears) was 1.43. This analysis allowed the identification of a single peptide from the ribosomal protein uL10 (previously known as P0 (Ban et al., 2014)) showing changes between the apo and GCN2- bound states (uL10 residues 121 – 137; sequence VTVPAQNTGLGPEKTSF) (Figure 3.3). Figure 3.3. The deuterium uptake plot for the uL10 peptide identified as showing differences between the apo state (blue) and the GCN2-bound state (red). Each point was measured in triplicate, and the means ± the standard deviations are plotted. The potential involvement of uL10 was extremely interesting, given its role in the recruitment and activation of translational regulators (Bargis-Surgey et al., 1999; Mochizuki et al., 2012; Naganuma et al., 2010). Furthermore, the identification of a role for uL10’s binding partners P1 and P2 in the specific activation of GCN2 means that GCN2 binding to uL10 would position it to be able to become activated by these factors (Figure 3.4) (Jiménez-Díaz et al., 2013). Characterisation of the GCN2-Ribosome Interaction by HDX-MS 113 Figure 3.4. The ribosomal peptide identified by HDX-MS. The peptide spanning residues from 121 – 137 in the ribosomal protein uL10 was identified as having a reduced isotopic exchange rate upon the addition of excess GCN2. The left image shows the structure of the ribosome (PDB code 3JAG) (Brown et al., 2015), with the peptide highlighted in pink. The approximate positions of the A, P and E sites are indicated. The boxed region is then expanded in the right-hand image, with a 90 ˚C rotation. The peptide is again shown in pink, and the structure of the rest of uL10 is shown in purple. However, there are several limitations to this result. The deuteration profiles for nearly every detected peptide of the ribosome show very low levels of total deuterium uptake, even after 50 minutes of incubation with deuterated buffer. This is probably due to the intrinsic stability of the ribosome, but this means that even a significant percentage change in the deuterium uptake of a peptide is still only a small absolute change in Daltons. For the peptide 121 – 137 of uL10, the addition of GCN2 causes a percentage deuteration change of 5 %, but an absolute mass change of only 0.5 Da. Unfortunately, there were no other overlapping peptides in this region to corroborate this observation. Given these factors in combination with the very large dataset, it is difficult to ascertain whether this change is significant, and therefore whether this region can be robustly assigned as a potential binding site for GCN2 on the ribosome. To address these issues, the sample preparation was optimised to increase total deuteration of the sample, to increase the amount of GCN2 bound and to increase the total coverage of the ribosome. To this end, the incubation times were altered to 5 minutes, 50 minutes and 500 minutes, all performed at 32 ˚C. This constitutes a rise in temperature of 10 ˚C, and from the Arrhenius equation (equation 2) it is possible to calculate that this corresponds to a three-fold increase in exchange rate. Under standard conditions (22 ˚C) these timepoints would therefore correspond to 15 minutes, 150 minutes and 1500 minutes. Importantly, to ensure the extended incubation at 32 ˚C did not have any adverse effects on the physiological state of the Characterisation of the GCN2-Ribosome Interaction by HDX-MS 114 ribosome, a pulse experimental control was also included. This control sample was incubated at 32 ˚C for 495 minutes in the absence of any deuterated buffer, and then deuterium- containing buffer was added and the sample incubated at 32˚C for a further 5 minutes. The sample was then quenched and the results compared to the 5 minutes at 32 ˚C time point. In addition to the increase in time and temperature of the deuterium exchange step, the concentration of ribosomes was also increased from 0.3 µM to 0.5 µΜ to try to increase total peptide coverage (the final concentration being limited by the relatively low maximum concentration of ribosomes after purification and the need to avoid saturating the mass spectrometer). Finally, the concentration of GCN2 was increased to 2.5 µM. Except for these three factors, the samples were prepared and the data collected and analysed in the same manner as previously. This analysis resulted in the identification of 1071 peptides, covering 8481 out of a total of 12804 amino acids, resulting in an overall coverage of 66.25 %. The only ribosomal protein that showed a total lack of coverage was uS14 (previously known as S29 (Ban et al., 2014)). The mean redundancy of the dataset was 1.78. A summary of both datasets is included in Supplementary Table 2. Encouragingly, the correction of the data to standard conditions (i.e. a deuterium incubation of 5 minutes at 32 ˚C corresponding to an incubation of 15 minutes at room temperature) seemed to fit well with the overall deuteration profile for a peptide from the two separate datasets (Figure 3.5). Figure 3.5. The deuterium uptake profile for a typical peptide over a time course of 0.3 seconds to 25 hours. Data points from the first dataset are indicated with pink stars and the data points from the second dataset are shown with blue stars. Characterisation of the GCN2-Ribosome Interaction by HDX-MS 115 In addition to this, the utilisation of the increased temperature was validated by the similarity between the 5-minute at 32 ˚C time point and the pulse experiment after a 495 minute incubation at 32 ˚C, indicating that the extended incubation is not having any significant effects on the structure of the ribosome. Overall, the data indicate that the optimised conditions have increased the total deuterium uptake of the ribosomal peptides. Comparison of the ribosome in the apo state versus the GCN2-bound state led to the identification of three overlapping peptides covering the same area as in the original dataset: uL10 residues 121 – 137 (sequence VTVPAQNTGLGPEKTSF), 138 – 155 (sequence FQALGITTKISRGTIEIL) and 148 – 157 (sequence SRGTIEILSD) (Figure 3.6). These three peptides were the only peptides in the entire dataset that showed a significant decrease in exchange upon addition of GCN2. An unpaired t test for each of these three peptides gave P values of < 0.00000001 compared to the entire dataset, indicating these changes are statistically significant. The data for every uL10 peptide observed is included in Supplementary Table 3. Characterisation of the GCN2-Ribosome Interaction by HDX-MS 116 Figure 3.6. The deuterium uptake plots for each of the three uL10 peptides identified as showing differences between the apo state (blue) and the GCN2-bound state (red). The residue numbers are shown above the plots. Each point was measured in triplicate, and the means ± the standard deviations are plotted. Characterisation of the GCN2-Ribosome Interaction by HDX-MS 117 It is important to note that in these experiments, the data are measuring uptake of deuterium per peptide over a time course, allowing the calculation of the rate of isotopic exchange for each peptide. This is not the same as directly mapping the loci of protein-protein interactions; however, it has been shown that areas with decreased isotopic exchange are very likely to represent regions that are found at the interface between the two binding partners. It is therefore very likely that the area identified by the HDX-MS data represents the GCN2 binding site on the ribosome (Figure 3.7). Figure 3.7. The area of the ribosome that showed changes in deuteration in the presence of GCN2. On the left, the affected region is highlighted in dark blue on the structure of the entire ribosome (PDB code 3JAG) (Brown et al., 2015), and the approximate locations of the A, P and E sites are indicated. The right-hand image shows the region with decreased isotopic exchange in the presence of GCN2 in greater detail. The ‘protected’ region is shown in dark blue, whilst the remainder of the uL10 structure is shown in cyan. Mapping this peptide on to the ribosomal structure demonstrates that this region consists of a loop protruding away from the ribosomal core, directly adjacent to the A site (Figure 3.7). Given the proposal that GCN2 could have a role in monitoring the processivity of translating ribosomes, this binding site would place the kinase in a perfect position to be able to monitor the state of the translation cycle, including the occupancy of the A site. These findings are based upon the comparison between two datasets: the ribosome alone and the ribosome in the presence of GCN2. It is important to note that the levels of deuterium incorporation are relative values, based on the maximal possible deuteration levels according to the final buffer composition, and therefore do not control for variations in the rate of back exchange across the primary structure of the protein. This means no conclusions can be drawn as to the absolute exchange rates of each peptide from these data. Given that the rate of back exchange should be equal for the same peptide under different conditions, however, for the Characterisation of the GCN2-Ribosome Interaction by HDX-MS 118 purposes of comparative data no fully deuterated or equilibrium labelled control sample was included. The use of HDX-MS to identify the binding site of GCN2 on the ribosome highlights the power of this technique to study large and extremely complex molecular structures. It should therefore be considered as highly complementary to structural biology, and represents a tool that could be utilised to investigate many more ribosomal interactions. It would be extremely interesting to look at a protein or molecule that interacts with the ribosome in a way that causes a conformational shift (such as intersubunit rotation), to see whether the peptides on the interface between the large and the small subunits show changes in isotopic exchange rates. The next aim was to characterise the area of GCN2 that is interacting with the ribosome, to try to extend the truncation analysis described in Chapter Two and provide molecular detail about the interaction. This could then inform the mutation of key residues within GCN2 to try to abrogate the interaction without affecting the general structure of the kinase (as was the case for the truncations). This would then allow a more precise analysis of the importance of ribosome binding in GCN2 activation. For this, HDX-MS was again tested. Due to the limited maximum concentration of purified ribosomes, and the need to use a molar excess of ribosomes to achieve significant occupancy levels, the concentration of GCN2 had to be reduced to 100 nM. Unfortunately, this was not sufficient for the detection of GCN2 peptides on the mass spectrometer, and so no data could be obtained from these experiments. Given the difficulty in obtaining higher yields of ribosomes, to extend this analysis it would be necessary to express and purify recombinant uL10 (ideally in the presence and absence of P1 and P2) (Abo et al., 2004). Purified uL10 could then be tested using a pull-down assay to determine whether it interacts with GCN2 in the absence of the remainder of the ribosome. If so, HDX-MS could be then used to determine the uL10 binding site on GCN2. This would allow the testing of molecular determinants of binding by mutation and analysis of the effects on ribosomal pull-down. Characterisation of the GCN2-Ribosome Interaction by HDX-MS 119 3.4. Conclusions This chapter describes the use of Hydrogen-Deuterium exchange-mass spectrometry to determine the site on the ribosome to which GCN2 binds. This is the first time this technique has been used to study such a large and complex molecule, and the optimisation of the analysis provides a framework for similar studies in the future. Given the panoply of proteins that interact with the ribosome in functionally important ways, and the information concerning dynamic changes arising from interactions that can be gained from HDX-MS, this constitutes an extremely useful development that is likely to prove invaluable. In this chapter, this optimised methodology has been utilised to identify the ribosomal protein uL10 as the binding site for GCN2. uL10 is part of the ribosomal P stalk, and sits directly adjacent to the A site of the ribosome, indicating that GCN2 would be perfectly positioned to monitor the translational cycle. This finding supports the model proposed in Chapter Two, as it is very plausible that GCN2’s ability to bind in this position could be regulated by the translational state, given uL10’s central role in the recruitment and activation of elongation factors. The statistical significance of the identification of this binding site is very high, indicating it is extremely unlikely the decreased deuteration of this region is entirely due to chance. Combined with the fact that the same region was identified, albeit less robustly, in the first, non-optimised dataset, this gives significance confidence in the identification of this site. However, it is possible that this is due to some sort of systemic error, and so this result needs to be independently validated by some other technique. The demonstration of an interaction between GCN2 and recombinant uL10 would be a powerful affirmation of this result, and would open the door to many different experiments, including the identification of the reciprocal binding site on GCN2 via HDX-MS. If this could be mapped, then mutations to abrogate this interaction could be tested via pull-down analysis. The identification of a mutation or combination of mutations that prevent this interaction could ultimately allow the in vivo analysis of the effect of disrupting the GCN2-ribosome interaction on a cell’s ability to respond to amino acid starvation. If this were to result in a decrease in the initiation of the ISR, this would be an extremely important result in favour of a model in which GCN2 acts as a monitor of translational stress. 120 Chapter Four – Structural Insights into GCN2 4.1. Introduction GCN2 is a key protein within the human proteome, yet detailed mechanistic information about its activation and regulation is so far relatively sparse. This is especially true in terms of structural insights, as there is no structural information available for any part of the human protein, and the only structural data on mammalian GCN2 concerns the murine C-terminal domain. As described in the introduction, there are structures of various domains of the yeast Gcn2, including the RWD domain, the kinase domain in multiple states, and the C-terminal domain. However, the findings in Chapter Two imply significant intrinsic differences between the yeast and the mammalian proteins, calling into question the extrapolation of structural insights from the yeast to the human protein. The identification of key roles for GCN2 in the development of many illnesses, including neurological degenerative disorders and cancer, has highlighted its potential as a therapeutic target. Particularly persuasive was the study done by Ye and colleagues in which the authors demonstrated that knocking out GCN2 in a K-Ras transformed cell line dramatically reduced tumour growth in mice (Ye et al., 2010). This suggests that GCN2 could be an important therapeutic target for cancer treatments. The structure of the kinase is therefore very desirable to inform the development of small molecules that could act specifically on GCN2. Structural Insights into GCN2 121 4.2. Materials and Methods 4.2.1. Crystallisation Trials 200 nL drops of purified GCN2 were dispensed into MRC 96-well plates and mixed with 200 nL of reservoir solution using the MOSQUITO (TTP Labtech). The plates were then left at 17 ˚C for up to three months. The full screen consisted of approximately 2000 different conditions (Gorrec, 2013; 2015; 2016; Gorrec et al., 2011; Stock et al., 2005). 4.2.2. Hydrogen-Deuterium Exchange-Mass Spectrometry Samples consisted of 10 µL aliquots of purified GCN2 at 5 µM in GF buffer [20 mM HEPES pH 7.5, 150 mM NaCl, 2 mM TCEP]. All sample preparation was performed on ice. 40 µL of ice-cold deuterated GF buffer [20 mM HEPES pH 7.5, 150 mM NaCl, 2 mM TCEP, 94.6 % D2O (Acros Organics 7789-20)] was added to each aliquot and the samples were mixed rapidly, producing a final D2O concentration of 75.6 %. The reactions were then immediately quenched by the addition of 20 µL ice-cold Quench buffer [5 M GdCl, 8.4 % formic acid], followed by freezing in liquid nitrogen. The samples were run on the mass spectrometer and the data processed in the same manner as described in the Materials and Methods for Chapter Three. The percentage deuteration for each peptide in comparison to a non-deuterated control was plotted against the midpoint residue number of the peptide (i). 4.2.3. Limited Proteolysis 10 µL reactions were assembled containing 1 mg/mL GCN2 in GF buffer [20 mM HEPES pH 7.5, 150 mM NaCl, 2 mM TCEP]. The proteases were from the protease kits Proti-Ace (Hampton Research HR2-429) and Proti-Ace 2 (Hampton Research HR2-432). Each protease was diluted to 5 µg/mL and 1 µg/mL in GF buffer. 10 µL of protease was then added to each aliquot of GCN2, and the reactions incubated on ice for 4 hours. The extent of proteolysis was then assessed by SDS-PAGE. For crystallography trials, 200 µL of either elastase at 1 µg/mL or endoproteinase Glu-C at 5 µg/mL was added to 200 µL of GCN2 at 1 mg/mL in GF buffer. The reactions were incubated on ice for four hours, and then used to set up a crystallisation screen. Structural Insights into GCN2 122 4.2.4. Nanobody Production Sample preparation A 500 µL reaction containing GCN2 and GCN1 at 6 µM and 3 µM respectively was assembled and incubated on ice for 20 minutes. The reaction was then subjected to cross- linking using the Coval K200 stabilisation kit (CovalX 2008k200) according to the manufacturer’s instructions. Cross-linking efficiency was judged through SDS-PAGE and Coomassie staining. The cross-linked samples were flash frozen in liquid nitrogen and shipped on dry ice to Els Pardon of Jan Steyeart’s laboratory in Vrije Universiteit Brussel in Brussels. Nanobody production and selection This procedure was carried out by Dr Els Pardon of Vrije Enuiversiteit Brussel. Nanobodies were produced through a previously described protocol (Pardon et al., 2014). Briefly, llamas were immunised with the cross-linked GCN1-GCN2 complex. After several rounds of immunisation over a period of approximately six weeks, a blood sample was taken and total cellular RNA was purified from the peripheral blood lymphocyte cells. The RNA was reverse transcribed to cDNA, and the nanobody-encoding regions were amplified by polymerase chain reaction (PCR) and inserted into a phage display vector. Phage display was then used to identify nanobodies that showed specific binding to the GCN2-GCN1 complex. Plasmids encoding these nanobodies were then sent by Dr Pardon. Nanobody expression 100 µL of chemically competent WK6 Su- cells were mixed with 1 µL of plasmid DNA and incubated on ice for 30 minutes. The cells were then subjected to 45 seconds at 42 ˚C, before returning to ice for another ten minutes. 150 µL of 2xTY media was then added and the cells were incubated at 37 ˚C for 30 minutes. The cells were finally plated on to tryptone yeast extract (TYE) plates containing 100 µg/mL ampicillin and left at 37 ˚C overnight. Five to six colonies were then inoculated into 20 mL lysogeny broth (LB) media containing 100 µg/mL ampicillin, 2 % glucose and 1 mM MgCl2. This preculture was then incubated at 37 ˚C, shaking at 220 r.p.m. After four hours, 15 mL of preculture was added to 900 mL LB media containing 100 µg/mL ampicillin, 2 % glucose and 1 mM MgCl2. This culture was incubated at 37 ˚C, shaking at 220 r.p.m. until the optical density reached an A600 of 0.7. At this point, Structural Insights into GCN2 123 expression of the nanobody was induced by the addition of 1 mM IPTG, and the temperature was lowered to 25 ˚C. The cells were grown for a further 20 hours at 25 ˚C, shaking at 220 r.p.m. The cells were harvested through centrifugation for 15 minutes at 7,000g in a H6000A rotor (ThermoFisher Scientific Sorvall 11250). The supernatant was carefully removed and the pellets were then immediately used for protein purification. Nanobody Purification The cell pellets were gently resuspended on ice in 15 mL TES buffer [200 mM Tris-HCl pH 8.0 (room temperature), 0.5 mM EDTA, 500 mM sucrose], before being incubated on a roller at 4 ˚C for one hour. 30 mL TES/4 buffer [50 mM Tris-HCl pH 8.0 (room temperature), 0.125 mM EDTA, 125 mM sucrose] was then added, and the cells were incubated at 4 ˚C for a further 45 minutes. The cells were then subjected to centrifugation for 30 minutes at 15000g at 4 ˚C in a Ti45 rotor (Beckman Coulter 339160), and the supernatant (representing the periplasmic fraction) was collected. 0.5 mL of Ni-NTA resin (Qiagen 30210), pre-equilibrated in PB1 buffer [50 mM sodium phosphate pH 7.0, 1 M NaCl], was added to the collected supernatant, and the solution was then incubated at 4 ˚C with gentle rotation for one hour. The sample was then applied to a BioRAD gravity flow column, and the beads were washed with 10 mL PB1 buffer, then with 10 mL PB2 buffer [50 mM sodium phosphate pH 7.0, 1 M NaCl, 15 mM imidazole pH 8.0]. The nanobody was then eluted from the resin in 7 mL PB3 buffer [50 mM sodium phosphate pH 7.0, 1 M NaCl, 200 mM imidazole pH 8.0], before being concentrated to 0.5 mL in a 15 mL 3K centrifugal filter (Amicon UFC900308). The sample was then injected on to a 10/300 Superdex 75 gel filtration column (GE Healthcare 17-5174-01), equilibrated in Nanobody buffer [20 mM Tris-HCl pH 7.5 (room temperature), 150 mM NaCl]. The peak fractions were then pooled and concentrated in a 15 mL 3K centrifugal filter as above. The final concentration was typically in the range of 0.4 mL at 2 – 4 mg/mL from 1 L cell culture. The protein sample was then aliquoted, flash frozen and stored at -80 ˚C. 4.2.5. Nanobody Interaction Analysis by Pull-Downs 100 µL reactions containing 2 µM Strep-tagged GCN2 or GCN1 and 3 µM purified nanobody were assembled in GF buffer [20 mM HEPES pH 7.5, 150 mM NaCl, 2 mM TCEP] and incubated at 4 ˚C for one hour. 40 µL of StrepTactin resin, equilibrated in GF buffer, was then added to each reaction and the samples were then rotated at 4 ˚C for 30 minutes. The beads Structural Insights into GCN2 124 were then sedimented by centrifugation at 500g for 1 minute at 4 ˚C, before the supernatant was removed and the beads resuspended in 1 mL GF-Wash buffer [20 mM HEPES pH 7.5, 150 mM NaCl, 2 mM TCEP, 0.1 % v/v Triton X-100]. The beads were washed in this way five times in total. All the supernatant was then aspirated, and the proteins eluted from the resin in 20 µL 2X LDS sample buffer. Samples were analysed via SDS-PAGE and binding was assessed using Coomassie stain. 4.2.6. Purification of GCN2-Nanobody Complexes In order to purify a stoichiometric complex of GCN2 and each nanobody, a threefold excess of a nanobody was added during the purification of GCN2. After ion exchange chromatography of GCN2, the protein concentration was estimated via UV spectroscopy, and a threefold excess of nanobody was then added. The reaction was then incubated on ice for 20 minutes before continuing with concentration for gel filtration. The binding of the nanobody could be checked via SDS-PAGE of the fractions from gel filtration. The GCN2-nanobody complexes were then concentrated in a 15 mL centrifugal filter (Amicon UFC905024) to approximately 0.2 mL at 2.5 mg/mL, and a limited crystallisation screen was set up for the complex. 4.2.7. Crystallisation of the Pseudokinase domain The construct AIp50 (residues 260 – 539) was purified and concentrated to 3 mg/mL. The purified protein was then used to set up a full MRC crystallisation screen. Crystals of AIp50 were initially seen in a variety of conditions, and grid screens were then set up using the Dragonfly (TTP Labtech). The best crystals were seen in the condition containing 564 mM LiSO4, 0.1 M 2-[(2-amino-2-oxoethyl)-(carboxymethyl)amino]acetic acid/NaOH (ADA/NaOH) pH 6.5, 14 % polyethylene glycol (PEG) 4K. The crystals were fished using 0.2 µM loops and transferred to a cryo-protectant solution containing 20 – 33 % trehalose before freezing in liquid nitrogen. Data were collected on the I03 beamline at Diamond Light Source and on the ID23-1 beamline at European Synchrotron Radiation Facility (ESRF) with the help of Minmin Yu. Synchrotron data were processed by Xia2 (Winter et al., 2013). Multiple data sets were merged using the BLEND (Aller et al., 2016) in the CCP4 package (Winn et al., 2011). Best statistics were obtained by merging three datasets from ESRF to a maximum resolution of 2.9 Å. Molecular replacement was carried out using PHASER (McCoy et al., 2007). An initial model of the domain was constructed using I-tasser (Yang Structural Insights into GCN2 125 and Zhang, 2015). This model was manually fit to the molecular replacement density and refined with REFMAC (Winn et al., 2003). To improve resolution, the N-terminal Strep tag was removed with TEV protease as described in the Materials and Methods for Chapter Two, and the protein was used to set up another full crystallisation screen. The best crystals were obtained in 1.2 M NaH2PO4, 0.8 M K2HPO4, 0.1 M 3-(Cyclohexylamino)-1-propanesulfonic acid (CAPS) pH 10.5, 0.2 M LiSO4 (final pH 6.1). Conditions were optimised using grid screens prepared using the Dragonfly. Crystal dehydration was performed through the addition of LiSO4 up to a final concentration of 1.3 M to the reservoir followed by incubation for up to 12 hours. Crystal cross-linking was performed by the addition of a 2 µL droplet of glutaraldehyde adjacent to the crystal-containing sitting drop, followed by incubation for up to 3 hours. 4.2.8. Negative Stain Electron Microscopy Negative stain electron microscopy (EM) samples were prepared using purified GCN2 at concentrations from 0.01 to 0.05 mg/mL. Gold R1.2/1.3 300 mesh grids (Quantifoil) were coated in a thin layer of amorphous carbon (produced using the Edwards Auto Evaporator E306A and estimated to be approximately 70 Å thick). The protocol for the synthesis of the amorphous carbon film and its deposition are detailed in (Passmore and Russo, 2016). Once dry, the grids were then glow discharged for 10 seconds at 6 kV using the Glow Discharger (Edwards Sputter Coater S150B). 3 µL of sample was then applied to the grid and left for 20 seconds. The grid was then twice gently touched against two droplets of water, followed by twice against two droplets of 2 % uranyl acetate. The excess liquid was then blotted with filter paper and the grids left to dry. The grids were screened and data were collected manually in Low Dose mode on a Tecnai G2 Spirit TEM (120 kV). The detector was a Gatan Orius SC200B CCD camera. The nominal magnification was 42,000x, corresponding to a pixel size of 2.37 Å. 4.2.9. Cryo-Electron Microscopy Sample Preparation Gold R1.2/1.3 300 mesh grids (Quantifoil) were glow discharged (Edwards Sputter Coater S150B) for 30 to 60 seconds at 6 kV. Manual freezing was performed at 4 ˚C. 3 µL sample was applied to the grid before a wait time of 30 seconds. The excess liquid was then manually blotted with filter paper for between 6 and 10 seconds. The grid was then plunged into liquid Structural Insights into GCN2 126 ethane at a temperature of ~ -170 ˚C. Automated freezing was performed on a Vitrobot MKIII (FEI), at 22 ˚C and 100 % humidity. 3 µL sample was applied to the grid followed by a wait time of 30 seconds. The sample was then blotted for 3 to 4 seconds and immediately plunged into liquid ethane. Amorphous Carbon Coated Grids A thin layer of amorphous carbon was synthesised using the Edwards Auto Evaporator E306A and deposited on to the surface of the grid as described in (Passmore and Russo, 2016). The grid was then glow discharged (Edwards Sputter Coater S150B) for 10 to 30 seconds at 6 kV, before the sample was applied and the grids were plunge frozen as described above. Graphene Oxide Coated Grids Before coating, the grids were glow discharged (Edwards Sputter Coater S150B) for 60 seconds at 7 kV. Graphene oxide (Sigma-Aldrich 763705) was diluted tenfold to 0.2 mg/mL in water. The solution was briefly spun down to remove any aggregates. 3 µL of graphene oxide was then applied to the surface of the glow-discharged grid followed by a wait time of 60 seconds. The excess liquid was then manually blotted with filter paper, and then the grid touched to a water droplet and blotted again. This was repeated with a second water droplet, followed by the inversion of the grid and a third repetition with a third water droplet. After the final blotting step, the grids were left for at least 24 hours before the sample was applied as described above. Multiple Sample Applications Multiple sample applications were performed exactly as described in (Snijder et al., 2017). 4.2.10. GraFix 12 mL gradients were assembled using a Gradient Master (Watson Marlow) in ultra clear centrifuge tubes (Beckman Coulter 344060) by mixing equal amounts of Light Buffer [20 mM HEPES pH 7.5, 150 mM NaCl, 2 mM TCEP, 10 % v/v glycerol] and Heavy Buffer [20 mM HEPES pH 7.5, 150 mM NaCl, 2 mM TCEP, 30 % v/v glycerol, 0.2 % glutaraldehyde (Sigma-Aldrich G5882)] at a pumping speed of 7 r.p.m. The gradients were left to sit at 4 ˚C for 1 – 2 hours before use. 400 µL of Cushion Buffer [20 mM HEPES pH 7.5, 150 mM NaCl, 2 mM TCEP, 5 % v/v glycerol] was then added to the top of the gradients, followed by 200 Structural Insights into GCN2 127 µL of GCN2 at 3 mg/mL. The gradients were then centrifuged at 283,807g in a SW40 Ti Rotor Bucket (Beckman-Coulter 331302) for 20 hours at 4 ˚C. The gradient was fractionated into 25 tubes each containing 20 µL of 1 M Tris-HCl pH 7.5 (room temperature) using a FC 203B Fraction Collector (Gilson) with a pump speed of 11 r.p.m. and a time of 0.32 minutes per tube. The migration of GCN2 through the gradient was assessed by SDS-PAGE and the peak fractions pooled and concentrated in a 15 mL centrifugal filter (Amicon UFC905024). The sample was then injected on to a Superdex 200 10/300 Increase column (GE Healthcare Life Sciences 28990944). The peak fractions were then pooled and concentrated in a 15 mL centrifugal filter (Amicon UFC905024) before being applied to EM grids as described above. 4.2.11. Assembly of GCN2-Ribosome Complexes The systems used for these assemblies are described in detail in (Feng and Shao, 2017). Protocol 1 The mRNA transcript (encoding a triple FLAG tag followed by the Ras open reading frame) was produced in a 110 µL in vitro transcription reaction. The transcription reaction contained 40 mM HEPES pH 7.5, 6 mM MgCl2, 2 mM spermidine (Sigma-Aldrich S0266), 10 mM glutathione, 0.5 mM adenosine triphosphate (ATP) (Sigma-Aldrich 10519979001), 0.5 mM cytidine triphosphate (CTP) (Sigma-Aldrich C1506), 0.5 mM uridine triphosphate (UTP) (Sigma-Aldrich U6875), 0.1 mM guanosine triphosphate (GTP) (Sigma-Aldrich 10106399001), 0.5 mM 7methyl diguanosine CAP analogue (New England Biolabs S1404L), 0.4 U/µL SP6 polymerase (New England Biolabs M0207S), 0.4 U/µL RNasin (Promega N2511) and 22 µL DNA template (provided by the Hegde lab). The reaction was incubated in a 37 ˚C waterbath for 1 hour, and then removed to ice. The transcription reaction was then immediately used in the translation reaction without need for purification of the transcript. The 2 mL translation reaction (split between four tubes) contained 35 % v/v nucleased rabbit reticulocyte lysate (RRL) (Green Hectares) (the nuclease treatment of RRL is described in the Materials and Methods of Chapter Two), 20 mM HEPES pH 7.5, 10 mM KOH, 50 mM KOAc, 1 mM ATP (Sigma-Aldrich 10519979001), 1 mM GTP (Sigma-Aldrich 10106399001), 12 mM creatine phosphate (Sigma-Aldrich 10621714001), 0.1 mg/mL tRNA (Sigma-Aldrich R4752), 40 µg/mL creatine kinase (Sigma-Aldrich 10127566001), 2 mM MgCl2, 1 mM glutathione, 0.3 mM spermidine (Sigma-Aldrich S0266), 40 µM of each of the amino acids except for methionine (Promega L9961) and a 5 % volume transcription reaction Structural Insights into GCN2 128 (in this case 100 µL). The tubes were incubated in a 32 ˚C waterbath for 8 minutes before being quenched by the addition of 500 µL ice-cold high salt buffer [50 mM HEPES pH 7.5, 750 mM KOAc, 15 mM MgAc2, 1 mM DTT] to each tube. The reactions were then layered on to two 800 µL high salt sucrose cushions [50 mM HEPES pH 7.5, 750 mM KOAc, 15 mM MgAc2, 1 mM DTT, 500 mM sucrose] and centrifuged at 539,511g in a TLA100.3 rotor (Beckman Coulter 349490) for 1 hour at 4 ˚C. The resultant pellets were each resuspended in 200 µL buffer [50 mM HEPES pH 7.5, 100 mM KOAc, 5 mM MgAc2, 1 mM DTT, 2.5 % sucrose] giving a total volume of 400 µL and a ribosome concentration of 100 nM. GCN1, GCN2 and deacylated tRNA were added to final concentrations of 1 µM, 1 µM and 16 µM respectively, increasing the reaction volume to 500 µL. The reaction was then incubated at 32 ˚C for 10 minutes and transferred to a spin column (Bio-Rad 7326008). 100 µL Anti-FLAG resin (Sigma-Aldrich A2220) was then added and the reaction rotated slowly at 4 ˚C for 1 hour. The beads were subsequently washed: first with 3 x 1 mL RNC buffer containing 0.1 % v/v Triton X-100 [50 mM HEPES pH 7.5, 100 mM KOAc, 5 mM MgAc2, 1 mM DTT, 0.1 % v/v Triton X-100], secondly with 3 x 1 mL RNC buffer containing higher salt and a higher detergent concentration [50 mM HEPES pH 7.5, 250 mM KOAc, 5 mM MgAc2, 1 mM DTT, 0.5 % v/v Triton X-100] and finally with 3 x 1 mL standard RNC buffer [50 mM HEPES pH 7.5, 100 mM KOAc, 5 mM MgAc2, 1 mM DTT]. All washes were performed by gravity flow. The complex was then eluted from the resin by incubation with 200 µL RNC buffer containing 3X FLAG peptide (Sigma-Aldrich S4799) at 0.1 mg/mL at room temperature for 25 minutes, followed by a further 200 µL rinse. Extra GCN1, GCN2 and deacylated tRNA were added to the reaction to increase the concentration to 1 µM. The resulting sample was centrifuged at 539,511g in a TLA100.3 rotor (Beckman Coulter 349490) for 40 minutes at 4 ˚C, and the pellet resuspended in 30 µL RNC buffer. Finally a further 1 µM GCN1, GCN2 and deacylated tRNA were added. The sample was then diluted to give a ribosomal concentration of approximately 120 nM (on the assumption that 1 A260 unit corresponds to a concentration of approximately 20 nM), and immediately used to make EM grids. For all ribosomal samples, Cu R2/2 300 mesh grids (Quantifoil) were coated in a layer of amorphous carbon (details described above) and then glow discharged (Edwards Sputter Coater S150B) for 25 seconds at 6 kV. Freezing was performed using the Vitrobot MKIII (FEI). 3 µL sample (with a ribosome concentration of 120 nM) was applied to the grid, Structural Insights into GCN2 129 followed by a wait period of 30 seconds. The grid was blotted for 3 seconds and immediately plunged into liquid ethane. Protocol 2 50 nM Strep-GCN2 and 50 nM untagged GCN1 were added to 4 mL gel filtered rabbit reticulocyte lysate (gel filtration was performed according to the protocol outlined in the Materials and Methods of Chapter Two). The reaction was then incubated at 32 ˚C for 15 minutes before being removed to ice. (To ensure the total reaction volume was warmed, the reaction was split into four 1 mL aliquots, and mixed gently after 7 minutes of incubation.) 50 µL of StrepTactin Sepharose High Performance resin (GE Healthcare Life Sciences 28-9355- 99) equilibrated in RNC-Wash buffer [RNC buffer + 0.1 % v/v Triton X-100] was then added to each tube of the reaction, and the mixtures slowly rotated at 4 ˚C for 1 hour. The beads were then sedimented through centrifugation at 500g for 1 minute at 4 ˚C, before the supernatant was removed and the beads were resuspended in 1 mL RNC-Wash buffer. This washing step was repeated seven times in total and the beads were then pooled into a single tube. The beads were then rinsed with RNC buffer to remove the Triton X-100, and then incubated with 200 µL RNC-Elute buffer [RNC buffer + 25 mM desthiobiotin (IBA 2-1000- 005)] for 10 minutes. The beads were then transferred into a spin column (Pierce 69705) and the elution isolated from the beads by a brief centrifugation. The resulting sample was concentrated by centrifugation at 539,511g in a TLA100.3 rotor (Beckman Coulter 349490) for 40 minutes at 4 ˚C, and the pellet resuspended in 30 µL RNC buffer. The sample was then diluted to give a ribosomal concentration of approximately 120 nM and immediately used to make EM grids as described above. Protocol 3 2 mL RRL was layered upon an 800 µL high salt sucrose cushion [50 mM HEPES pH 7.5, 750 mM KOAc, 15 mM MgAc2, 1 mM DTT, 500 mM sucrose] and centrifuged at 539,511g in a TLA100.3 rotor (Beckman Coulter 349490) for 1 hour at 4 ˚C. The resultant pellet was resuspended in 50 µL buffer [50 mM HEPES pH 7.5, 100 mM KOAc, 5 mM MgAc2, 1 mM DTT, 2.5 % sucrose] giving a ribosomal concentration of 120 nM. Final concentrations of 600 nM GCN2, 200 nM deacylated tRNA and 50 µM adenylyl imidodiphosphate (AMPPNP) (Roche 10102547001) were then added to the sample. The sample was then immediately used to make EM grids as described above. Structural Insights into GCN2 130 Protocol 4 Protocol 4 was performed in the same way as Protocol 3, except the final additions consisted of 1 µM GCN1, 1 µM GCN2, 200 nM deacylated tRNA and 50 µM AMPPNP (Roche 10102547001). Protocol 5 The mRNA substrate (encoding a triple FLAG tag followed by the autonomously-folding villin headpiece domain (VHP) followed by the open reading frame of Sec61β containing a UAG stop codon after Val68) was in vitro transcribed as described in Protocol 1 (Shao et al., 2016). The transcription product was then used in a 3 mL in vitro translation reaction, assembled as described in Protocol 1, except with the addition of 0.5 µM purified recombinant eRF1AAQ (Hegde lab, purified according to (Brown et al., 2015)). The reaction was split into three 1 mL aliquots. The reaction was incubated at 32 ˚C for 6.5 minutes. At this point, a dominant negative version of Hbs1L (H384A) (Hegde lab, purified according to (Shao et al., 2013) was added to a final concentration of 0.5 µM. The translation reaction was then continued for another 18.5 minutes (25 minutes in total). KOAc and MgAc2 were then added to increase the final concentrations to 750 mM and 15 mM respectively. The reaction was layered on to three 1.6 mL high salt sucrose cushions [50 mM HEPES pH 7.5, 750 mM KOAc, 15 mM MgAc2, 1 mM DTT, 500 mM sucrose] and centrifuged at 539,511g in a TLA100.3 rotor (Beckman Coulter 349490) for 1 hour at 4 ˚C. The resultant pellets were each resuspended in 200 µL buffer [50 mM HEPES pH 7.5, 100 mM KOAc, 5 mM MgAc2, 1 mM DTT, 2.5 % sucrose] and then pooled into a spin column (Bio-Rad 7326008), giving a total volume of 600 µL. 75 µL Anti-FLAG resin (Sigma-Aldrich A2220) was then added and the reaction rotated slowly at 4 ˚C for 1 hour. The beads were subsequently washed: first with 3 x 1 mL RNC buffer containing 0.1 % Triton X-100 [50 mM HEPES pH 7.5, 100 mM KOAc, 5 mM MgAc2, 1 mM DTT, 0.1 % v/v Triton X-100], secondly with 3 x 1 mL RNC buffer containing higher salt and a higher detergent concentration [50 mM HEPES pH 7.5, 250 mM KOAc, 5 mM MgAc2, 1 mM DTT, 0.5 % v/v Triton X-100] and finally with 3 x 1 mL standard RNC buffer [50 mM HEPES pH 7.5, 100 mM KOAc, 5 mM MgAc2, 1 mM DTT]. All washes were performed by gravity flow. The ribosome-nascent chain complex was then eluted from the resin by incubation with 200 µL RNC buffer containing 3X FLAG peptide (Sigma-Aldrich S4799) at 0.1 mg/mL at room Structural Insights into GCN2 131 temperature for 25 minutes, followed by a further 200 µL rinse. The elution was centrifuged at 539,511g in a TLA100.3 rotor (Beckman Coulter 349490) for 40 minutes at 4 ˚C, and the pellet resuspended in 30 µL RNC buffer. This resulted in a ribosomal concentration of 230 nM. A 40 µL reaction was then assembled containing 150 nM ribosome-nascent chain complexes, 40 µM deacylated tRNA (including supressor tRNA) and 1.25 µM apramycin. The reaction was incubated at 32 ˚C for 5 minutes. 10 µL of a mixture containing 3 µM GCN2, 3 µM GCN1, 5 µM AMPPNP (Roche 10102547001) and 5 µM GTP (Sigma-Aldrich 10106399001) was then added, and the reaction incubated for a further 5 minutes. The reaction was transferred to ice, and then immediately used to make EM grids as described in Protocol 1. The final concentrations of the sample were 120 nM ribosome-nascent chains, 32 µM tRNA, 1 µM apramycin, 600 nM GCN1, 600 nM GCN2, 1 µM AMPPNP and 1 µM GTP. Protocol 6 The mRNA substrate encoded a triple FLAG tag followed by a VHP domain, followed by the open reading frame of Sec61β containing three UUG codons at amino acid position 127 (Feng and Shao, 2017). The transcript was in vitro transcribed as described in Protocol 1. The product was then used in a 5 mL in vitro translation as described in Protocol 1, except with the omission of the exogenous tRNA. The reaction was incubated at 32 ˚C for 15 minutes and then removed to ice. KOAc and MgAc2 were then added to increase the final concentrations to 750 mM and 15 mM respectively. The reaction was then layered on to five 1.6 mL high salt sucrose cushions [50 mM HEPES pH 7.5, 750 mM KOAc, 15 mM MgAc2, 1 mM DTT, 500 mM sucrose] and centrifuged at 539,511g in a TLA100.3 rotor (Beckman Coulter 349490) for 1 hour at 4 ˚C. The ribosome-nascent chain complexes were affinity purified via the triple FLAG tag as described in Protocol 4. The elution was centrifuged at 539,511g in a TLA100.3 rotor (Beckman Coulter 349490) for 40 minutes at 4 ˚C, and the pellet resuspended in 30 µL RNC buffer. 2 µM GCN1, 2 µM GCN2, 1 µM AMPPNP (Roche 10102547001) and 1 µM apramycin were then added to the reactions. 10 µL StrepTactin Sepharose High Performance resin (GE Healthcare Life Sciences 28-9355-99) equilibrated in RNC-Wash buffer [RNC buffer + 0.1 % v/v Triton X-100] was then added, and the mixture slowly rotated at 4 ˚C for 1 hour. The beads were then sedimented through centrifugation at 500g for 1 minute at 4 ˚C, before the supernatant was Structural Insights into GCN2 132 removed and the beads were resuspended in 1 mL RNC-Wash buffer. This washing step was repeated three times in total and the beads transferred to a fresh tube. The beads were rinsed with RNC buffer to remove the Triton X-100, and then incubated with 200 µL RNC-Elute buffer [RNC buffer + 25 mM desthiobiotin (IBA 2-1000-005)] for 10 minutes. The beads were then transferred into a spin column (Pierce 69705) and the elution isolated from the beads by a brief centrifugation. The resulting sample was concentrated by centrifugation at 539,511g in a TLA100.3 rotor (Beckman Coulter 349490) for 40 minutes at 4 ˚C and the pellet resuspended in 30 µL RNC buffer. The sample was then diluted to give a ribosomal concentration of approximately 120 nM and immediately used to make EM grids as described above. Protocol 7 300 nM purified ribosomes (the purification procedure is described in the Materials and Methods of Chapter Two) were mixed with 5 µM GCN2 in RNC buffer (final volume 500 µL) and the reactions were incubated at 32 ˚C for 15 minutes. 40 µL StrepTactin Sepharose High Performance resin (GE Healthcare Life Sciences 28-9355-99) equilibrated in RNC- Wash buffer [RNC buffer + 0.1 % v/v Triton X-100] was then added to the reaction and the tubes slowly rotated at 4 ˚C for 1 hour. The beads were then sedimented through centrifugation at 500g for 1 minute at 4 ˚C, before the supernatant was removed and the beads were resuspended in 1 mL RNC-Wash buffer. This washing step was repeated three times in total and the beads transferred to a fresh tube. The beads were then rinsed with RNC buffer to remove the Triton X-100, and then incubated with 160 µL RNC-Elute buffer [RNC buffer + 25 mM desthiobiotin (IBA 2-1000-005)] for 10 minutes. The beads were then transferred into a spin column (Pierce 69705) and the elution isolated from the beads by a brief centrifugation. The resulting sample was concentrated by centrifugation at 539,511g in a TLA100.3 rotor (Beckman Coulter 349490) for 40 minutes at 4 ˚C and the pellet resuspended in 30 µL RNC buffer. The concentration was then measured, and the sample diluted to a ribosomal concentration of 120 nM. Glutaraldehyde (Sigma-Aldrich G5882) was added to a final concentration of 0.05 % and the sample was incubated on ice for 30 minutes. The cross- linking reaction was then quenched by the addition of 100 mM Tris-HCl pH 7.5 (room temperature), and the samples were immediately used to make EM grids as described above. Structural Insights into GCN2 133 4.2.12. Cryo-Electron Microscopy Data Collection All the data were collected using the automated EPU software (FEI) on the Titan Krios FEG cryo-TEM microscope with an X-FEG (field emission gun) as the electron source and an acceleration voltage of 300 kV. The images were captured on a Falcon II direct electron detector (FEI), with a dose rate of approximately 35 electrons per Å2 per second and a nominal magnification of 75,000x, corresponding to a pixel size of 1.04 Å. The rate of movie frame collection was 16 frames per second, with total exposure of ~ 1.1 second per image. 4.2.13. Cryo-Electron Microscopy Data Processing The movie frames were aligned and averaged using MotionCorr (Li et al., 2013). Poor micrographs (with large amounts of contamination, poor contrast, astigmatism or charging) were discarded. Information concerning the contrast transfer function for each averaged micrograph was obtained using Gctf (Zhang, 2016). The particles were then selected from the images using either the autopicking function of RELION (Scheres, 2012; 2015) or the e2boxer.py function of EMAN2 (Tang et al., 2007). The particles were extracted from the micrographs and subjected to reference-free 2-dimensional class averaging in RELION. Good 2D classes were then selected, and the particles from these classes were subjected to 3- dimensional refinement using a 30 Å low-pass filtered cryo-EM reconstruction of a ribosome (from PDB code 3JAH) as a reference. The particles were then subjected to 3D classification to try to isolate a subset of the particles containing significant extra density. At this stage the density maps were assessed to look for the presence of any extra density. Multiple rounds of classification with different subsets of particles were performed to increase the chances of identifying regions of extra density. Focused classification with signal subtraction (Bai et al., 2015) was used to further improve classification strategies. Particle subsets that looked promising were subjected to statistical particle-based movie correction in RELION, in which radiation damage is compensated for by the B-factor weighting of each movie frame within a stack. These particles were then re-refined. The resolution values reported are based on the Fourier Shell Correlation (FSC) 0.143 criterion (Rosenthal and Henderson, 2003). Structural Insights into GCN2 134 4.3. Results and Discussion 4.3.1. Crystallography of full-length GCN2 The initial aim was to try to crystallise the full-length protein to gain a comprehensive structural understanding of the kinase. To this end, a full crystallisation screen was set up to test over 2000 conditions. The concentration of GCN2 was varied between 2 mg/mL and 6 mg/mL, using the presence or absence of precipitation and phase separation to qualitatively assess each concentration. The addition of a stoichiometric amount of the small molecule inhibitor staurosporine was also tested, however none of these strategies resulted in the production of protein crystals. 4.3.1.1. Intrinsic Disorder Analysis by HDX-MS It was reasoned that the lack of crystallisation could be due to regions of flexibility within GCN2. To characterise the presence and stability of secondary structure across the length of the protein, Hydrogen-Deuterium exchange-mass spectrometry (HDX-MS) was used. The level of deuteration of GCN2 after a 0.3 second incubation was measured by mass spectrometry, and the results are plotted against the midpoint of the peptic peptide (i) in Figure 4.1. Figure 4.1. A global deuteration profile for full-length GCN2. For each peptide, the percentage deuteration at the 0.3 second time point is plotted against the peptide midpoint (i). The data plotted are the means ± the standard deviations (n=3). The dotted line denotes 50 % deuteration. The overall coverage value for this analysis was approximately 55 %, meaning that information was only available for just over half of the protein sequence. This is lower than 0 500 1000 1500 0 20 40 60 80 100 i Pe rc en ta ge D eu te ra tio n Data 1 Structural Insights into GCN2 135 expected, but does provide some information concerning areas of intrinsic disorder within the protein. From previous work in the group it has been shown that peptides that exhibit percentage deuteration levels of 50 % or greater at the 0.3 s time point are likely to be intrinsically disordered (Burke et al., 2014; Fowler et al., 2016). This therefore suggests that full-length human GCN2 contains three regions of intrinsic disorder: residues 560 – 575, residues 698 – 705 and residues 1508 – 1522. These three regions are found in the linker between the pseudokinase domain and the kinase domain, the linker between the N- and C- lobes of the kinase domain and the linker between the HisRS-like domain and the C-terminal domain respectively. It has been shown that this information can be used to aid crystallisation by the rational design of constructs in which the regions of disorder are eliminated (Spraggon et al., 2004), and so constructs with these regions deleted were cloned, expressed and purified. It should be noted that this data is lacking a control for the rate of back exchange, i.e. the re- protonation of peptides during chromatographic separation in the mass spectrometer. The rate of back exchange varies across the primary structure of the protein, and areas with very high rates of deuteration are also likely to have higher rates of back exchange, although due to the low pH and low temperature this should still be significantly slower than the rate of initial deuteration. The system typically shows back exchange rates of approximately 15 % on average, but ultimately it is not possible to draw definitive conclusions on the absolute rates of exchange without taking the individual rates of back exchange for each peptide. If this type of experiment were to be repeated, therefore, this important control should be included. Deletion of either of the second two regions resulted in very poor expression of the protein. Deletion of the first region (GCN2 Δ140 – 294) resulted in some expression; however, it was not monodisperse after gel filtration, and the concentration was low. A limited crystallisation screen was set up with the available material, but did not result in any crystals. 4.3.1.2. Limited Proteolysis In a further effort to reduce flexibility within GCN2, a limited proteolysis screen was set up. GCN2 was incubated with a range of different proteases, and the cleavage pattern was assessed by SDS-PAGE (Figure 4.2). Structural Insights into GCN2 136 Figure 4.2. SDS-PAGE of the results of limited proteolysis. GCN2 was incubated in the presence of either a high concentration (5 µg/mL) or a low concentration (1 µg/mL) of protease on ice for four hours, and then the cleavage was assessed by SDS-PAGE. The proteases are indicated above the lanes. 5 and 1 refer to 5 µg/mL and 1 µg/mL respectively. From this, elastase (at 1 µg/mL) and endoproteinase Glu-C (at 5 µg/mL) were identified as promising candidates as their cleavage resulted in a single major species with a lower molecular weight than the full-length protein. These cleavage reactions were then scaled up and used to set up a crystallisation screen, but again unfortunately these screens yielded no crystals. 4.3.1.3. Nanobodies To try to induce GCN2 crystallisation, it was decided to develop specific nanobodies to bind to the complex. Nanobodies (depicted in Figure 4.3) are small antibody fragments that are able to bind specifically to proteins, and they have been shown to be powerful crystallisation chaperones. They are able to stabilise regions of flexibility within the protein, as well as contributing towards improved crystal contacts (Desmyter et al., 1996; Rostislavleva et al., 2015). Structural Insights into GCN2 137 Figure 4.3. The structure of a typical nanobody (PDB code 1MEL) (Desmyter et al., 1996). The fold is a β sandwich consisting of ten β strands. The specificity of binding is predominantly dependent on the sequence of the variable loops, which are indicated at the top of the image. The structure is coloured from the N- (blue) to the C-terminus (red). In nature, nanobodies exist as part of so-called heavy-chain only antibodies, which are present in the biological family of Camelidae. The antigen-binding site of most antibodies consists of two variable domains, one from each of the heavy and the light chains, whereas in these heavy-chain only antibodies the antigen-binding site is made up of a single variable domain (from a single chain) (Figure 4.4). This domain, when isolated, is known as a nanobody, and represents a small and very stable protein fragment that is able to bind tightly and specifically to its cognate antigen. Figure 4.4. On the left is a schematic showing the structure of a conventional antibody. The two heavy chains are shown in light and dark green, and the two light chains are shown in mid and light blue. CH denotes constant domains of the heavy chain and CL denotes constant domains of the light chain. VH and VL denote variable domains of the heavy and light chains respectively. The middle panel shows the basic structure of a heavy chain-only antibody found in Camelidae. The nanobody (right) is a single variable domain, containing the full binding capability of the entire heavy chain-only antibody. Structural Insights into GCN2 138 Nanobodies are generated through the immunisation of llamas with the protein of interest (Pardon et al., 2014). The immune system of the llama undergoes an immune response against the injected antigen, resulting in the production of both conventional and heavy chain-only antibodies that specifically target the immunogen. A blood sample is then taken, and a cDNA library can be produced. Polymerase chain reaction (PCR) is used to amplify the nanobody gene fragments from the cDNA. The nanobodies that bind tightly to the antigen can then be selected for by phage display, and these selected nanobodies can be expressed recombinantly in Escherichia coli. They can thus be produced in high quantities and used for crystallisation screening, amongst other things. For these reasons, it was decided to develop nanobodies against GCN2 to aid crystallisation. To prepare the sample, GCN2 and GCN1 were cross-linked in the hope that this could result in the production of nanobodies that bind to GCN2, GCN1 and potentially also some that stabilise the complex between the two proteins. GCN2 and GCN1 were cross-linked to completion by the cross-linker CovalX (Figure 4.5), before being sent to collaborators in Jan Steyaert’s laboratory. The production and selection of nanobodies was performed by Dr Els Pardon. Figure 4.5. SDS-PAGE of the cross-linked sample before being used as an antigen to provoke the generation of nanobodies against the complex. GCN2 and GCN1 were mixed and then the cross-linker CovalX was added. After cross-linking, the reactions were analysed by SDS-PAGE and Coomassie staining. Structural Insights into GCN2 139 The immunisation and selection resulted in the production of twenty nanobodies. These nanobodies could be subcategorised into thirteen families, based on the similarities between the complementarity-determining regions, shown in Table 4.1. Table 4.1. The twenty nanobodies that were produced against the GCN2-GCN1 complex. Given the extremely high levels of sequence conservation within the nanobody families, one nanobody from each family was chosen giving a total of 13 nanobodies. Each of these thirteen nanobodies was expressed in E. coli and purified from the periplasm via the C-terminal His6 tag. Finally, each nanobody was subjected to gel filtration. A single, monodisperse peak was achieved for each of the thirteen nanobodies, and SDS-PAGE analysis demonstrated the purity of the material produced (Figure 4.6). The yield of each nanobody was between 1 and 4 mg from 1 L of bacterial culture. Nanobody ID Original ID Family NB01 CA11161 NB02 CA11162 NB03 CA11166 NB04 CA11171 NB05 CA11218 2 NB06 CA11219 3 NB07 CA11165 NB08 CA11170 NB09 CA11172 NB10 CA11224 NB11 CA11225 NB12 CA11222 5 NB13 CA11223 6 NB14 CA11226 7 NB15 CA11217 8 NB16 CA11168 9 NB17 CA11220 10 NB18 CA11221 11 NB19 CA11160 12 NB20 CA11173 13 1 4 Structural Insights into GCN2 140 Figure 4.6. The results of the purification of the nanobodies. A typical gel filtration chromatogram is shown on the left, and the gel on the right shows the yield and final purity of each nanobody. Having produced purified nanobodies, their ability to bind GCN2 or GCN1 was tested via pull-down analysis. The nanobody was incubated with either Strep-tagged GCN2 or Strep- tagged GCN1, before StrepTactin resin was added to capture any bound proteins. The results from the first nanobody (NB01) were assessed by SDS-PAGE and are shown in Figure 4.7. Figure 4.7. Pull-down analysis to determine to which protein each nanobody binds. The data shown are for NB01. NB01 was incubated with Strep-GCN2, Strep-GCN1 or Strep-GCN2 and untagged GCN1, and the reactions were incubated on ice for one hour (totals). StrepTactin resin was then added, and the reactions were incubated for a further 30 minutes. The beads were spun down and the supernatant removed (flowthroughs). The beads were then washed extensively, and the bound proteins eluted with sample buffer (elutions). The positions of the bands corresponding to GCN1, GCN2 and NB01 are indicated on the right of the gel. Structural Insights into GCN2 141 This analysis of the binding properties of NB01 shows that the nanobody is able to bind to GCN2 but not GCN1. It also does not appear to dramatically increase the ability of GCN2 to bind to GCN1. This analysis was repeated for each of the thirteen purified nanobodies, and produced the following results (shown in Table 4.2). (None of the nanobodies improved complex formation.) Table 4.2. A table showing the results of the pull-down analysis. Whilst this analysis represents an initial and qualitative characterisation of the interactions between GCN2 and each of the nanobodies, there is much scope to gain a more comprehensive analysis of these interactions, for example by techniques such as surface plasmon resonance or bio-layer interferometry. These could give quantitative data concerning the affinities of the different nanobodies for GCN2 or GCN1, and this information could prove useful for future studies. After initial confirmation of binding, the ability of each nanobody to induce crystallisation of GCN2 was tested. Purified GCN2 was mixed with a threefold molar excess of a nanobody, and the reaction was incubated on ice for 20 minutes before the complex was subjected to gel filtration. The peak fractions were then analysed by SDS-PAGE to determine their protein composition (Figure 4.8). Nanobody ID Binds to GCN2? Binds to GCN1? NB01 Yes No NB05 Yes No NB06 Yes No NB07 No No NB12 Yes No NB13 Yes No NB14 Yes No NB15 Yes No NB16 Yes No NB17 Yes No NB18 Yes No NB19 No Yes NB20 Yes No Structural Insights into GCN2 142 Figure 4.8. The production of GCN2-nanobody complexes. The gel filtration profiles for NB07 and NB12 are shown on the left, and the SDS-PAGE analysis of the peak fractions is shown on the right. This analysis correlates with the information gained from the pull-down assay shown in Table 4.2. NB07 does not associate with GCN2, and elutes from gel filtration separately from the kinase. NB12, on the other hand, elutes with GCN2, indicating that a stable complex has been formed. The stoichiometry of this interaction is interesting, as given that a threefold molar excess was added, one would expect to see a larger amount of free nanobody eluting after the GCN2-nanobody complex. There is a small peak eluting at approximately 18 mL, but the absorbance value seems unexpectedly low, and this was the case reproducibly for the majority of the nanobodies. The explanation for this is unclear, but could be due to inexact estimation of the concentrations of GCN2 or the nanobodies, or could be due to some sort of minor aggregation occurring, as there does seem to be a shoulder on the left-hand side of the GCN2- nanobody complex peak. It could also imply that more than one copy of the nanobody is binding to a single GCN2 molecule. To investigate this further, the stoichiometry of the complex as suggested by the gel could be analysed. Unfortunately, the gel shown in Figure 4.8 is effectively overloaded with GCN2 and so an accurate estimation of the stoichiometry cannot be achieved, requiring a gel containing lower amounts of the complex. Structural Insights into GCN2 143 After gel filtration, the GCN2-nanobody complexes were concentrated and crystallisation screens were set up. Unfortunately, this strategy did not result in the production of any crystals. 4.3.2. Crystallography of the Pseudokinase domain As described above, the crystallisation of full-length GCN2 proved challenging. For this reason, crystallisation screens were set up for shorter constructs of the protein. The structures of individual domains would provide less information than the full-length protein structure, but knowledge of their architecture would still be a useful addition to the field, given that there is currently no structural information concerning the human protein. Furthermore, the structures of individual domains of proteins have previously been shown to be very useful for the interpretation of lower resolution structures from electron microscopy, and so a structure of a truncated version or single domain of GCN2 could potentially inform electron microscopy studies in the future. After multiple attempts at the crystallisation of different domains, crystals of the pseudokinase domain (without the charged linker; residues 260 – 539 [AIp50]) appeared in several conditions of the full LMB crystallisation screen, two examples of which are shown in Figure 4.9. (All crystallisation trials are summarised in Table 4.3.) Structural Insights into GCN2 144 Structural Insights into GCN2 145 Figure 4.9. Images of the initial crystals hits in two conditions. Grid screens around these two conditions were set up to optimise crystal quality, and the best crystals were obtained from the screen around condition B shown in Figure 4.9 (564 mM LiSO4, 0.1 M ADA/NaOH pH 6.5, 14 % PEG 4K). Several cryoprotectants were tested by analysing crystal diffraction on the in-house X-ray source, including PEG 4K, ethylene glycol, glycerol, high concentrations of LiSO4 and trehalose. The best of these proved to be 21.5 % trehalose, and this was used to freeze crystals. The crystals were shot at Diamond Light Source, and initially diffracted to approximately 3.5 Å. The crystals then underwent several rounds of optimisation to try to improve the resolution. Batch and streak seeding, crystal dehydration and crystal cross-linking with glutaraldehyde were tried to improve the resolution, but none of these strategies resulted in significant improvements in resolution. Further grid screening (including pH and additive screens) also did not result in an improvement in resolution, and nor did the pursuit of alternative initial hits. Initial analysis of the diffraction data indicated that there was significant flexibility in the N-terminal region, and so it was hypothesised that the removal of the N-terminal Strep tag could improve the stability of this region and so improve resolution. The Strep-tag was cleaved using TEV protease during protein purification, and the cleaved protein then set up for a full crystallisation screen. Hits were obtained in a variety of conditions, the best of which consisted of 1.2 M NaH2PO4, 0.8 M K2HPO4, 0.1 M CAPS pH 10.5 and 0.2 M LiSO4, giving a final pH of 6.1. A grid screen around this condition was set up, and crystals appeared in the majority of drops. Another round of cryo-protectant testing was performed, and glycerol gave the best results. The crystals were frozen and shot at Diamond Light Source and the European Synchrotron Radiation Facility (ESRF) with the help of Minmin Yu. The best diffracted to approximately 2.9 Å. Structural Insights into GCN2 146 The data were initially phased using molecular replacement, and then experimental phasing was attempted using tantalum bromide and sodium bromide soaks. However, experimental phasing was inadequate, and so molecular replacement was ultimately used, using the structure of Ste20 protein kinase SPAK as an initial model (PDB code 5D9H). This allowed the construction of an initial model for the GCN2 pseudokinase domain, which was then manually fit into the electron density before refinement using REFMAC (Winn et al., 2003). The data processing and model refinement were performed by Dr Roger Williams. The resulting structure of the human pseudokinase domain is shown in Figure 4.10, and a table of the corresponding statistics is shown in Table 4.4. Figure 4.10. The structure of the pseudokinase domain (residues 260 – 539) of human GCN2, solved by X-ray crystallography to a resolution of 2.6 Å. The structure is coloured from the N-terminus in blue to the C-terminus in red. Structural Insights into GCN2 147 Table 4.4. A table giving the statistics for the X-ray crystallography dataset and model shown in Figure 4.10 As shown in Figure 4.10, the structure of the pseudokinase domain of human GCN2 looks like a typical bilobal kinase domain, as could be expected. Kinase domains typically are more highly conserved through the N-lobes and less so within the C-lobes. Interestingly however, comparison between the structure of the pseudokinase domain from the human protein and the kinase domain from the yeast protein shows that there is a higher level of structural conservation between the C-lobes than between the N-lobes (Figure 4.11). Merged native data (three crystals) Data collection Space group P422 Cell dimensions a, b, c (Å) 121.22, 121.22, 62.87 a, b, g (°) 90.0, 90.0, 90.0 Resolution (Å) 2.9 (3.08-2.90) Rsym or Rmerge 0.091(12.84) I / sI 24.8(0.6) Completeness (%) 100.0(99.9) Redundancy 35.5(36.7) Refinement Resolution (Å) 2.9 No. reflections 10455 Rwork / Rfree 0.32/0.41 No. atoms Protein 1665 Ligand/ion 0 Water 0 B-factors Protein 138.5 Ligand/ion 0 Water 0 R.m.s. deviations Bond lengths (Å) 0.0142 Bond angles (°) 2.00 Structural Insights into GCN2 148 Figure 4.11. A side-by-side comparison of the structures of the human pseudokinase domain (left) and the yeast kinase domain (PDB code 1ZYD) (Padyana et al., 2005) (right). This strongly suggests that a gene duplication event has occurred at some stage through the evolution of the kinase, leading to the production of two highly similar domains adjacent to one another. 4.3.3. Electron Microscopy of GCN2 The field of electron microscopy has undergone a transformation during the last two decades. This is largely due to the hugely influential technological advances that have revolutionised the resolution obtainable and the size of protein that can be studied, such as the development of direct electron detectors (Kuhlbrandt, 2014) and advanced data processing strategies (Fernandez-Leiro and Scheres, 2016; Scheres, 2012). Given the challenges associated with attempts to crystallise the full-length GCN2, electron microscopy (EM) was used to try to gain structural insights into GCN2 without the need for crystallisation. Negative stain EM was used to gain an initial insight into the sample. Negative stain EM involves applying the sample to an EM grid coated with a thin, amorphous carbon layer, and then depositing a solution of uranyl acetate to the grid to increase the contrast of the sample in the electron microscope. The uranyl acetate stain surrounds the particles on the grid, but is excluded from their volume. An electron beam passing through the sample interacts with the stain, and can give a high contrast image of the specimen. The resolution of this technique is limited to approximately 20 Å, and in general can only give information about the surface topology of the protein due to the ‘encasing’ of the particles in a heavy metal shell. However, it provides a relatively quick way to assess sample quality and gain important insights into the overall architecture of the protein (Figure 4.12). Structural Insights into GCN2 149 Figure 4.12. Negative stain images of GCN2. The images were taken on a Tecnai G2 Spirit cryo-TEM, with a nominal magnification of 42,000x (corresponding to a pixel size of 2.37 Å). A 100 nm scale bar is shown in the bottom right-hand corner of each image. A negative stain dataset was collected on a Tecnai G2 Spirit microscope with a pixel size of 2.37 Å. A total of 246 images were collected, containing approximately 31,000 particles. The images were processed using RELION (Scheres, 2012), and generated a set of 2D classes shown in Figure 4.13. Figure 4.13. 2D classes of negative stain images of GCN2. A subset of these 2D classes were used to generate an initial model via the Stochastic Gradient Descent algorithm (Punjani et al., 2017) in RELION, which was then used as a Structural Insights into GCN2 150 reference to reconstruct a 3D volume showing the low-resolution (21.7 Å) architecture of GCN2 (Figure 4.14). Figure 4.14. Three views of the 3D volume reconstruction for GCN2 from negative stain data. The resolution was calculated as 21.6 Å by Fourier Shell Correlation analysis in RELION. The approximate dimensions of the volume are indicated on the image. These results indicate GCN2 to be a relatively globular structure, with dimensions of approximately 165 Å x 120 Å x 180 Å. At this very low resolution it is not possible to decipher any details concerning the arrangement of the domains, although this could be attempted through the addition of bulky tags or insertions at different positions within the primary structure. However, this analysis gives important information about the size and shape of the particles that can be used to inform future structural work. Having initially characterised the overall architecture of GCN2, cryo-electron microscopy was then used to try to gain higher resolution information about the structure. Unfortunately, sample preparation proved very challenging, and it was difficult to observe any particles on the EM grid, even when the protein concentration was increased to 6 mg/mL. As shown in Figure 4.15, some particles were visible occasionally, but the contrast was always very low and they were very difficult to reproduce. Structural Insights into GCN2 151 Figure 4.15. A cryo-EM image of GCN2. The protein concentration was 3 mg/mL. A 100 nm scale bar is shown in the bottom right-hand corner. Initially, the buffer conditions were varied extensively to try to improve the quality, homogeneity and distribution of particles on the EM grid. This included varying the pH, the salt concentration and testing the inclusion of different detergents (listed in Table 4.5), however none of these changes resulted in an improvement. Next, the preparation of the EM grid was varied, including replacing copper Quantifoil grids with gold Quantifoil grids, covering the grid in either a thin layer of amorphous carbon or in a thin layer of graphene oxide flakes, and varying the method and conditions of plunge freezing (Table 4.5). However, no improvements were seen from any of these changes. To try to improve particle quality, a cross-linking protocol known as GraFix was used (Kastner et al., 2008; Stark, 2010). The aim of this technique is to decrease conformational heterogeneity and prevent denaturation of the protein by exposing the sample to a low concentration of cross-linker whilst being sedimented through a glycerol density gradient. This has the advantage over conventional cross-linking techniques, in which the cross-linker is simply added to the sample in solution, as it promotes intramolecular cross-links over intermolecular cross-links, reducing the propensity for aggregation of the sample. Freshly purified GCN2 was applied to a glycerol-glutaraldehyde gradient and subjected to ultra-centrifugation. The gradient was then fractionated, and the peak fractions were injected on to a gel filtration column. Structural Insights into GCN2 152 Figure 4.16. The results of the GraFix cross-linking procedure. The gel shows gel filtration fractions of GCN2 prior to the sedimentation-cross-linking step, and then the gel filtration fractions for GCN2 after GraFix cross-linking. The numbers above the gel correspond to the fraction numbers from the gel filtration chromatogram shown on the right. As shown in Figure 4.16, the GraFix procedure resulted in a reasonably monodisperse sample. SDS-PAGE analysis of the peak fractions from the gel filtration chromatogram in comparison to gel filtration analysis of purified GCN2 that had not been subjected to GraFix showed a clear increase in molecular weight, consistent with the fact that GCN2 forms a stable dimer in solution. This sample was then concentrated and used for EM grid preparation, but unfortunately did not result in any improvements to the EM images. Multiple sample applications to the grid before plunge-freezing was another strategy explored, with the aim of increasing the density of particles. This has been shown to yield significantly improved grids for proteins that show an unexpectedly low number of particles, and cannot be further concentrated (Snijder et al., 2017). Up to three sample applications were tested, but none of the grids showed any improvements, and typically showed more contamination on the grid. A full list of the tested conditions is tabulated in Table 4.5. Structural Insights into GCN2 153 Structural Insights into GCN2 154 4.3.4. Electron Microscopy of the GCN2-Ribosome Complex Ribosomes have proved very amenable to cryo-electron microscopy, and there exist many EM structures of ribosomes in different states and in the presence of different binding partners. Having identified a direct interaction between human GCN2 and the ribosome (as described in Chapter Two), and given the challenges in obtaining a good cryo-EM sample for GCN2 alone, it was therefore decided to try to study the GCN2-ribosome complex by cryo-EM. This would not only provide information about the structure of GCN2, but theoretically could give significant insights into the manner by which ribosomes activate GCN2. All of the sample preparation and data processing for GCN2-ribosomal complexes was done in close collaboration with Dr Sichen Shao. The initial strategy (protocol 1 in Materials and Methods section 4.2.11) consisted of running an in vitro translation reaction, in which the rabbit reticulocyte ribosomes translated a transcript containing a triple FLAG tag followed by the Ras open reading frame (Feng and Shao, 2017). The reaction was quenched with ice-cold high salt buffer after 8 minutes (when the majority of active ribosomes should be undergoing translation elongation), and the ribosomes were then pelleted by centrifugation. The ribosomes were resuspended, and the concentration measured as ~ 100 nM. A tenfold molar excess (1 µM) of GCN2 and GCN1 was then added to the resuspended ribosomes, and the complex was captured with Anti- FLAG beads. These ribosomes were then eluted and concentrated. Analysis by Western blot (Figure 4.17) showed that GCN1 and to a lesser extent GCN2 were present in the elution. The final ribosome concentration was 120 nM, and a further 1 µM of GCN2 and GCN1 was added, before the sample was applied to EM grids and frozen. Figure 4.17. Western blots of the sample preparation. The lanes correspond to (from left to right): the initial translation reaction, the ribosomes after pelleting and resuspension, the ribosomes after addition of excess GCN1 and GCN2, the flowthrough after incubation with Anti-FLAG resin and finally the elution from the Anti-FLAG resin. The samples were blotted with Anti-Strep and Anti-FLAG antibodies to determine the recovery of GCN1/GCN2 and the translated substrate respectively. Structural Insights into GCN2 155 A dataset containing 1463 images was collected on a Polara Tecnai G2 FEG TEM, and was processed with RELION. This yielded a 3D reconstruction of the ribosome at 6.9 Å, but unfortunately there was no extra electron density present that could be attributable to GCN2. The data processing consisted of particle selection and 2D classification, followed by multiple rounds of 3D refinement and 3D classification. The reference model used for initial 3D refinements was an EM map of an empty ribosome, and subsequently maps from previous refinements or classifications were used. Different subsets of particles (based on the rotational state of the ribosome, the presence or absence of tRNA in the A, P or E sites, or the presence or absence of potential regions of extra density) were combined in multiple different ways to try to isolate a class of particles containing any clear extra density. Unfortunately, none of the different processing pathways tried resulted in any density that could be attributed to GCN2. To increase the occupancy of GCN2 bound to the ribosome, the strategy was altered to use StrepTactin resin to capture the Strep tag of GCN2 rather than the FLAG tag (protocol 2 in Materials and Methods section 4.2.11). Given that this method of tagging the translating ribosomes was no longer required, the ribosomes were not subjected to an in vitro translation reaction first, but were simply incubated with 50 nM GCN2 and 50 nM GCN1. The complex was then captured on StrepTactin resin and washed. The complex was eluted with desthiobiotin, and the concentration measured. The estimated ribosomal concentration was 40 nM, so the reaction was centrifuged and the pellet resuspended. The ribosomal concentration was then measured as 190 nM, and this sample was applied to grids. SDS-PAGE analysis demonstrated that the sample contained GCN1, GCN2 and ribosomal proteins (Figure 4.18). Figure 4.18. A Coomassie-stained gel showing the elution of the GCN2-GCN1-ribosomal complex from StrepTactin resin. Structural Insights into GCN2 156 A dataset containing 1233 images was collected on this sample, but the processing again yielded no electron density that could be attributed to GCN2, despite numerous rounds of re- classification and re-refinement. In both these datasets, there was a class of ribosomes (consisting of approximately 11 - 15 % of the particles in the dataset) that contained electron density resembling the eukaryotic elongation factor 2 (eEF2), which binds in the ribosomal A site (Voorhees et al., 2014) (Figure 4.19). Figure 4.19. Electron density maps from the initial GCN2-ribosome dataset. Two classes are shown. The class shown in grey represents an apparently empty ribosome and the class in blue contains extra density in the A site (11 % of the dataset). The density appears very consistent with the translation factor eEF2 (mapped on to the density map in dark blue). The structure of eEF2 comes from PDB code 3J7P (Voorhees et al., 2014). It was reasoned that the presence of eEF2 could be obscuring the density for GCN2, as the HDX-MS data presented in Chapter Two showed that GCN2 is binding in a very similar region and so the particles could be difficult for RELION to distinguish during classification. For this reason, samples were prepared in which the ribosomes had been washed in high salt buffer prior to being incubated with 600 nM or 1 µM GCN2 to remove any bound factors (protocols 3 and 4 in Materials and Methods section 4.2.11). The affinity purification was also omitted for this sample, in case the extended sample preparation time was not compatible with the maintenance of the complex. Two datasets containing 900 and 1600 images respectively were collected, but again despite extensive processing, neither dataset resulted in the observation of any clear electron density for GCN2. Structural Insights into GCN2 157 This presents a few possibilities: GCN2 could be present on the ribosome, but in a sufficiently flexible manner that it is not possible to clearly observe in an averaged reconstruction, or alternatively the complex could be dissociating during vitrification. A lack of complex stability could plausibly be due to the ribosomes not being in the correct state for GCN2 to recognise and bind to with the highest affinity. For this reason, different methods of stalling the ribosomes were tested to see whether they gave better results. Firstly, an in vitro translation reaction was set up in which the ribosomes were translating a transcript consisting of a triple FLAG tag followed by an open reading frame in which an amber stop codon (UAG) was inserted at residue 69 (protocol 5 in Materials and Methods section 4.2.11). The translation reaction contained an inactive mutant of eRF1 (eRF1AAQ), causing the ribosomes to stall upon reaching the stop codon. Partway through the translation reaction, a dominant negative version of the ribosome release factor Hbs1 (Hbs1 H348A) was added to prevent the stalled ribosomal complexes being dismantled and recycled (Brown et al., 2015). After 25 minutes, the ribosomes were spun through a high salt sucrose cushion to remove the bound factors and then were affinity purified via the FLAG tag. A mixture of 40 µM deacylated tRNA (containing amber-suppressor tRNA), 1.25 µM of the antibiotic apramycin (to try to stabilise the complex as used for RelA (Brown et al., 2016)), 3 µM GCN2, 3 µM GCN1, 5 µM of an ATP analogue and 5 µM GTP was then added, and the reaction was incubated on ice. This sample was used for EM grid preparation. Data collection on this sample was performed on the Titan Krios FEG cryo-TEM and resulted in 1553 images, which were subsequently processed using RELION. The data initially looked promising, as the initial refinement showed increased levels of noise adjacent to the A site. The data were subjected to multiple rounds of classification using different subsets of particles, but the putative extra density observed never became any clearer. A processing technique called focused classification with signal subtraction (Bai et al., 2015) has recently been developed in which it is possible to classify particles based solely on a small region of the map. This prevents the region of interest being effectively outweighed during classification by the experimental noise and mass of the remainder of the complex. This is particularly relevant when dealing with ribosomal particles; most classifications will inherently focus on classifying the particles according to differences in the ribosome rather than according to the existence of a potentially low-occupancy binding partner. The translation factor eEF2 was thus used to make a mask around the A site, including the area Structural Insights into GCN2 158 directly adjacent to the uL10 binding site mapped in Chapter Three. The signal outside of this mask was subtracted from each particle and then the subtracted particles were classified based on the density inside the mask. Unfortunately, none of the resulting classes contained any reliable extra density. Secondly, samples were prepared using a leucine stalling system (protocol 6 in Materials and Methods section 4.2.11). This system is based upon the fact that rabbit reticulocytes are adapted for the optimal translation of haemoglobin, which makes up approximately 90 % of the total protein produced by the reticulocytes (Lodish and Desalu, 1973). Haemoglobin contains very few leucine residues, and so it appears as though leucyl-tRNAs are limiting when translating other proteins (Smith and McNamara, 1974). Generally, this can be easily combated for the translation of other proteins by the addition of exogenous tRNA to the in vitro translation reaction (Pelham and Jackson, 1976). However, it means that the insertion of a sequence of two or more leucine-encoding codons into a transcript in the absence of exogenous tRNA causes ribosomal stalling and translational arrest (Feng and Shao, 2017). Theoretically, this situation would closely resemble a physiological state of amino acid starvation, assuming the deacylated tRNA is not recruited to the ribosome in a sequence- dependent manner. To test whether ribosomes stalled in this manner could be used to make better EM samples, an in vitro translation reaction was run using a transcript substrate containing three leucine- encoding UUG codons. Ribosome-bound factors were then removed by centrifugation through a high salt-containing sucrose cushion and the ribosomes were affinity purified through the triple FLAG tag on the nascent protein. 2 µM GCN2, 2 µM GCN1, 1 µM AMPPNP and 1 µM of the antibiotic apramycin were then added and the reactions incubated before being used to make grids. 1444 images were collected on the Titan Krios microscope and the data processed in the same way as described above, using multiple rounds of classification or focused classification with signal subtraction. Unfortunately, all the classes obtained showed apparently empty ribosomes with no obvious density attributable to GCN2. Given that stalling the ribosomes in different states did not have any positive effect on the electron density for GCN2, it was therefore decided to test the effect of cross-linking the sample in the hope that this could stabilise the complex and allow observation of GCN2 Structural Insights into GCN2 159 (protocol 7 in Materials and Methods section 4.2.11). To this end, high salt-purified ribosomes were incubated with an excess of GCN2, and then the complexes captured on StrepTactin resin. The beads were then washed and the GCN2-ribosome complex eluted. 0.05 % glutaraldehyde was added to the elution, and the reaction was incubated on ice for 30 minutes. The cross-linking reaction was quenched with Tris buffer, and then the sample used to prepare grids. Data were collected using the Titan Krios, and resulted in 1705 images. The data were processed using RELION as previously described. Focused classification with signal subtraction gave one class of six (44229 particles) that showed extra density within the masked region. These particles (with the full signal values restored) were subjected to 3D refinement, resulting in the density map shown in Figure 4.20, which shows a clear region of extra density directly adjacent to uL10 above the A site. Figure 4.20. The electron density map for a GCN2-ribosomal complex at 7.2 Å. The 60S subunit is coloured in blue, the 40S subunit is in yellow and the E and P site tRNAs are in lilac and red respectively. The area of extra density attributable to GCN2 is shown in green. The structure of uL10 is shown within the density (from PDB code 3JAG) (Brown et al., 2015), and the protected region identified by HDX-MS analysis in Chapter Three is highlight in dark blue. As depicted in Figure 4.20, the position of this extra density matches the GCN2-binding site as identified by HDX-MS in Chapter Three extremely well, indicating that this density could represent GCN2. However, several different processing strategies including further rounds of 2-dimensional and 3-dimensional classification, as well as further focused classification with signal subtraction, were not able to improve the resolution. The size of the region of density would not account for the entirety of monomeric GCN2 (according to the negative stain reconstruction); however, it is approximately the same size as the dimeric kinase domain (PDB code 1ZYD) (Padyana et al., 2005). This indicates the remainder of the protein could be Structural Insights into GCN2 160 flexible or could have been denatured during vitrification. A table of statistics for the dataset is shown in Table 4.6. Table 4.6. A table giving the statistics for the dataset that resulted in the map shown in Figure 4.20. To try to improve the resolution, more data were collected on samples prepared in the same manner. The data were initially processed following the same pipeline, except that the map shown in Figure 4.20 was used as the initial reference for 3D refinement. Despite this, no classes in these datasets showed any similar regions of extra density, and further rounds of classification and refinement gave similar results. The reasons behind this are unclear, and suggest that there may have been small differences in the sample preparation that had large effects on the occupancy of the complex (such as small variations in the length of incubation with glutaraldehyde, or differences in the protein purification steps). An alternative possibility is that the density observed in Figure 4.20 is an artefact that does not represent GCN2 binding. This is an area of significant concern, and implies that the sample still requires significant optimisation to increase the GCN2 occupancy on the ribosome, and to decrease the conformational heterogeneity of the sample to improve the resolution before any conclusions can be drawn. All datasets collected on GCN2-ribosomal samples are listed in Table 4.7. Dataset XL_GCN Microscope FEI Titan Krios Nominal magnification 75,000x Voltage (kV) 300 Detector Falcon II Mode Integrating Energy filter slit width (eV) 20 Data collection method FEI EPU Number of Frames 20 Exposure time (s) 1.1 Dose (e-/Å2) 45 Number of micrographs 1705 Pixel size (Å) 1.04 Box size (pixels) 480 Number of raw particles 72629 Number of particles after 2D classification 47562 Number of particles after focused classification 44229 Resolution of refined map (Å) 7.2 Structural Insights into GCN2 161 Structural Insights into GCN2 162 4.4. Conclusions Structural insights into GCN2 have long been sought to fully understand the mechanistic basis behind the activation of GCN2 under conditions of amino acid starvation. The identification of the role of GCN2 in the development of multiple disorders, including pulmonary veno- occlusive disease, cancer and neurological degenerative conditions, has further enhanced interest in the structural characterisation of the protein, as detailed molecular information could be used to inform the rational design of small molecules with potentially important therapeutic implications. The yeast Gcn2 has been partially structurally characterised: specifically, structures of the RWD domain, the kinase domain and the C-terminal domain have been solved. The manner in which these domains interact with one another, and the conformational rearrangement that is proposed to occur upon kinase activation, is still elusive. Furthermore, as noted previously, there exist significant intrinsic differences between human GCN2 and yeast Gcn2, meaning that structural information cannot necessarily be extrapolated from the yeast protein to the human. For these reasons, this chapter describes extensive efforts to gain structural insights into the human GCN2. Initially, attempts were made to produce diffracting crystals of the full-length protein. A wide range of crystallisation conditions were tested, but unfortunately no crystals were obtained. To try to aid crystallisation, flexible regions of the protein were removed, either in a rational manner, as informed by disorder analysis by HDX-MS, or by limited proteolysis, but these approaches did not yield any crystallisation. Next, small antibody fragments known as nanobodies were produced to act as potential crystallisation chaperones. The nanobodies were produced through the immunisation of llamas with a cross-linked complex of GCN1 and GCN2, followed by rounds of selection by phage display. These nanobodies were expressed, purified and characterised, before being used to set up crystallisation screens with the full-length protein. Nanobodies have been shown to be powerful tools in the crystallisation of large proteins, due to their ability to bind to and reduce the flexibility of certain regions, as well as improving crystal contacts between molecules in the crystal. However, no improvements in crystallisation of the protein were seen upon the inclusion of nanobodies. Having been unsuccessful in the crystallisation of the full-length protein, the next aim was to try to crystallise truncated versions of GCN2. Crystallisation trays were set up for many GCN2 constructs (summarised in Table 4.3), and crystals were obtained for the pseudokinase Structural Insights into GCN2 163 domain. After several rounds of optimisation, crystals that diffracted to high resolution were obtained, leading to a 2.9 Å structure of the human GCN2 pseudokinase domain. This is the first structure of the human GCN2 protein, and is the first time the pseudokinase domain of GCN2 has been structurally characterised in any organism. The pseudokinase domain in the yeast protein is thought to have a stimulatory role towards the kinase domain (Lageix et al., 2014), and in Chapter Two of this thesis it is identified as having an important role in GCN2’s ability to bind ribosomes. The molecular details of the mechanisms behind these effects are unknown, but the structure of the pseudokinase domain could inform future experiments and models. The structure of the domain resembles a typical bilobal kinase domain, with a high level of structural similarity between the C-lobes of the human pseudokinase domain and the yeast kinase domain. This suggests that GCN2 may have been subject to a gene duplication event throughout evolution, producing two kinase-like domains adjacent to one another. In a further attempt to characterise the full-length GCN2, substantial efforts were made using electron microscopy. Initially, negative stain electron microscopy indicated that the protein sample was reasonably monodisperse and amenable to electron microscopy. The 2D classes and resulting 3D model gave low-resolution insights into the structure of the protein, indicating that it is a globular structure with dimensions of approximately 165 Å x 120 Å x 180 Å. Subsequently, cryo-EM was explored in an attempt to gain high resolution information about GCN2 without the need for crystallisation, as that appeared to be a major roadblock. GCN2 seemed a good candidate for cryo-EM, as it has a reasonably high molecular weight (390 kDa as a dimer), purifies extremely well, and produced good quality negative stain EM data. However, despite trialling many different conditions and sample preparation techniques, no reproducible, high quality particles could be observed. The panoply of electron microscopy structures of the ribosome in the literature combined with the identification of a direct and functionally important interaction between GCN2 and the ribosome meant that studying the GCN2-ribosome complex by cryo-electron microscopy was a very exciting avenue to explore. The benefits of this strategy were threefold: firstly, the EM sample preparation for ribosomal samples is relatively robust and the particles are more distinct, making sample preparation and screening simpler. Secondly, an interaction with the ribosome could make GCN2 more stable and therefore increase the resolution obtainable. Finally, a structure of GCN2 bound to the ribosome could be very informative, as it could Structural Insights into GCN2 164 potentially provide insights into how the interaction leads to the activation of the kinase domain of GCN2. For these reasons, many attempts were made to produce an EM sample containing GCN2 bound to the ribosome. Unfortunately, this proved very challenging and most of the time resulted in an electron density map containing no extra density that could be attributable to GCN2, despite multiple attempts at optimisation. The reasons behind this are not clear, but could simply be that the affinity of the reaction is too low to give significant occupancy at the tested concentrations. An important aspect of future experiments should be the calculation of the KD of the interaction between GCN2 and the ribosome, as this would allow an accurate prediction of the fraction of the ribosomes that should be bound to GCN2. The KD could either be determined using a binding assay such as surface plasmon resonance or microscale thermophoresis, or estimated by titrating ribosomes into the in vitro activity assay described in Chapter Two. This would give an EC50 rather than a KD, but would at least give an idea of the concentration range. If the KD is sufficiently low that the assembly strategies tried would be expected to result in significant GCN2 occupancy, alternative explanations must be considered. The lack of observable GCN2 could feasibly be due to either the flexibility of the protein when bound to the ribosome, or to the fracturing of the complex during the vitrification procedure. Given the difficulty in observing GCN2 particles alone in cryo specimens compared to negative stained specimens, it seems plausible to speculate that GCN2 does not easily survive plunge-freezing. One common method to increase protein stability in these experiments is to include a glutaraldehyde cross-linking step. Cross-linking the GCN2-ribosome sample with glutaraldehyde led to the generation of an electron density map in which there is clear extra density adjacent to the A site. Encouragingly, this extra density does not resemble any known translation factors. Moreover, the position of the extra density perfectly matches the binding site identified by HDX-MS in Chapter Three. This therefore seems to represent the first look at GCN2 bound to the ribosome. It is important to note the limitations of these data however. Firstly, the use of a cross-linker can generate artefacts, and so it will be important to validate these findings using alternative techniques. This could be done by collecting a dataset on a sample consisting of cross-linked ribosomes, in the absence of GCN2. However, the correlation with the binding site identified in solution by HDX-MS makes this possibility unlikely. Secondly, despite the use of many Structural Insights into GCN2 165 different processing strategies (including different methods of 2D and 3D classification, different reference models and focused classification with signal subtraction) the resolution remained at approximately 7.2 Å. This means that the factor remains essentially unidentifiable, although the ribosomes were washed in high salt buffer before incubation with GCN2, which should remove the majority of other bound factors, indicating that it is very likely to be GCN2. Again, the correlation with the HDX-MS data emphasises this point. The observable density is not sufficiently large to represent the entirety of GCN2. This therefore implies that the observed density could only represent a single domain of GCN2 (the dimeric kinase domain would approximately fit into the region), with the remainder of the protein invisible (due to flexibility or denaturation during vitrification). The putative identity of this domain is unclear, especially as the truncation analysis described in Chapter Two demonstrated that the pseudokinase domain, the HisRS-like domain and the C-terminal domain are all important for ribosomal binding. To address this question, the resolution of the EM map would have to be significantly improved, or alternatively an HDX-MS approach could be taken, using purified recombinant uL10 to probe the ribosomal binding site on GCN2. It is important to note that the repetition of the sample preparation and data collection did not yield any similar extra density. The reasons behind this are unknown and present a cause for concern. This means that much more work needs to be done before one can say for sure if this truly is GCN2 bound to the ribosome If the observable extra density depicted in Figure 4.20 does represent GCN2, the low- resolution insights obtainable in this EM map still preclude any analysis of the molecular details, meaning that an improvement in resolution will be of the utmost interest. To obtain a higher resolution map, it seems likely that the biochemical characterisation of the system must be improved. In particular, the manner by which GCN2 achieves ribosomal selectivity will likely prove crucial. Identification of a specific ribosomal state to which GCN2 may bind with highest affinity will inform sample preparation and data processing strategies and hopefully increase GCN2 occupancy within the dataset. One possibility in this vein, briefly discussed in Chapter Two, is that GCN2 recognises collided ribosomes on a single mRNA transcript, a phenomenon that has recently been characterised as an important stress signal (Simms et al., 2017). If GCN2 could be shown to have increased affinity towards collided ribosomes compared with single ribosomes, this could be very interesting. The data processing strategies Structural Insights into GCN2 166 performed typically select against ‘aggregated’ ribosomes, and so if GCN2 were to bind to multiple ribosomes then it is plausible to suggest that these particles are being filtered out at the particle selection stage. However, much more biochemical and structural work must be done before any conclusions can be drawn. 167 Chapter Five – Investigation of the conformational remodelling of kinases by the Hsp90 co-chaperone Cdc37 using HDX- MS 5.1. Introduction Protein kinases are an extremely important part of the cell’s regulatory machinery, representing ~ 1.7 % of the human genome (Endicott et al., 2012). Approximately one third of all proteins are regulated through a phosphorylation event of some sort (Taylor et al., 2012). The proper regulation of kinases is therefore of crucial importance to the cell, as dysregulation can have severe consequences for both the cell and the organism. Generally, protein kinase domains exist in two (or more) conformational states: an inactive form and an active form (in which the catalytic DFG triad is able to orient the ATP substrate properly). Their proper regulation therefore depends on a controlled conformational switch between the two states according to a variety of regulatory inputs. Such inputs can include allosteric regulatory interactions, post-translational modifications and changes in oligomeric state. These changes result in small alterations to the structure of the kinase, which can then be transferred via a network of intramolecular interactions to the active site, where they result in a shift from the inactive to the active conformation, or vice versa. Kinase activity can also be subject to regulation by the action of the cellular chaperone Hsp90 (Karagöz and Rüdiger, 2015). Chaperones are a major component of the cellular proteostasis machinery, and guard against protein misfolding and aggregation (Kim et al., 2013; Saibil, 2013; Winkler et al., 2012). There are many different chaperones within the cell, differing in structure, function and client selectivity (Bukau et al., 2006; Mayer, 2010). For example, the Hsp70 family of chaperones typically recognise short and highly hydrophobic protein sequences and so are generally more involved at the early stages of protein folding, whilst the Hsp90 family generally interact with protein intermediates at later stages of folding. Whilst the importance of chaperones is often only considered during the initial folding of a nascent protein, Hsp90 chaperones are also essential for the maintenance and activation of approximately 60 % of protein kinases over the entirety of their lifetime (Karagöz and Investigation of Kinase Domain Remodelling by Cdc37 168 Rüdiger, 2015; Taipale et al., 2012). Inhibition of Hsp90 causes rapid ubiquitylation and the subsequent degradation of its ‘client’ proteins (Miyata et al., 2013). The Hsp90 family of chaperones are ATP-dependent, meaning that they rely upon progression through a conformationally dynamic ATP binding and hydrolysis cycle (Ali et al., 2006; Pearl et al., 2008). Hsp90 is an extremely well conserved homodimeric protein, and each monomer consists of three distinct domains (Ali et al., 2006). At the C-terminus of the protein is a dimerisation domain containing a conserved pentapeptide sequence (MEEVD), which is the primary binding motif for Hsp90 co-chaperones. The middle portion of Hsp90 interacts with both client proteins and co-chaperones, whilst the N-terminal domain is responsible for the binding and hydrolysis of ATP. In the absence of ATP Hsp90 generally adopts an open conformation; however, ATP binding promotes a conformational shift to a closed arrangement in which the N-terminal domains also dimerise allowing ATP hydrolysis to occur (Krukenberg et al., 2011; Mayer and Le Breton, 2015). The group of proteins which bind to Hsp90 chaperones are a select and eclectic group, containing many proteins with important roles in signalling such as telomerase, nitric oxide synthase, and a panoply of key protein kinases including AKT (Basso et al., 2002; Fontana et al., 2002; Sato et al., 2000), B-Raf (Grammatikakis et al., 1999; Schulte et al., 1995; Stancato et al., 1993) and Cyclin-dependent kinase 4 (Cdk4) (Pearl, 2005; Pearl and Prodromou, 2006; Stepanova et al., 1996). Many of the proteins identified as Hsp90 clients are known oncogenes, and are involved in the establishment of some of the hallmark physiological changes associated with cancer progression (Neckers and Workman, 2012). Typically, many of the tumourigenic mutations that occur in these proteins cause destabilisation of the protein, therefore making them more reliant on the Hsp90 chaperone system. This therefore means that cancer cells are often significantly more sensitive to Hsp90 inhibitors than healthy cells, making Hsp90 a promising target for anti-cancer therapeutic development (Neckers and Workman, 2012; Whitesell and Lindquist, 2005). The specificity behind Hsp90’s interactions with this distinct group of clients is relatively unclear. For the protein kinase clients, specificity appears to rely on the co-chaperone cell division cycle protein 37 (Cdc37). Cdc37 was originally discovered as a component of a complex including Hsp90 and the Rous sarcoma viral protein pp60v-src (Brugge, 1986), but has subsequently been observed in complexes alongside Hsp90 with a variety of other proteins including Cdk4 (Dai et al., 1996) and the receptor tyrosine kinase Epidermal Growth Investigation of Kinase Domain Remodelling by Cdc37 169 Factor Receptor (EGFR) (Lavictoire et al., 2003). These studies have led to the identification of Cdc37 as the protein kinase-specific targeting subunit of Hsp90 (Pearl, 2005), and it has been shown that Hsp90 cannot interact with kinase clients in the absence of Cdc37 (Hunter and Poon, 1997). Cdc37 has been shown to bind to Hsp90 and to client kinases via distinct molecular sites. The N-terminal portion of the protein (residues 1 – 126) is responsible for binding to the client kinase, whilst Hsp90 binding is dependent on the central portion of the protein (residues 128 – 282) (Shao et al., 2003a). The function of the extended C-terminal region is currently unknown. However, some mutations in the N-terminal region of Cdc37 have been shown to dramatically affect Hsp90 binding, indicating that the two processes are not entirely separate (Miyata and Nishida, 2004; Shao et al., 2003a; 2003b). A co-crystal structure of a complex between the C-terminal and middle domains of Cdc37 and the N-terminal domain of Hsp90 was solved in 2004 (Figure 5.1) (Roe et al., 2004), but this gave no real insights into how the complex may bind to client kinases. The specificity of this interaction is a topic of much interest, as many client kinome studies have failed to identify any common structural motifs, and kinase dependence on the Hsp90 chaperone can vary considerably even between two closely related kinases. The only trend that has emerged from these studies is based on the thermal stability of the protein, with less stable kinases more likely to be an Hsp90 client (Taipale et al., 2012) Figure 5.1. The crystal structure of the core Hsp90-Cdc37 complex (PDB code 1US7) (Roe et al., 2004). The N-terminal domain of Hsp90 (residues 1 – 208; gold) was crystallised bound to the middle and C- terminal domains of Cdc37 (residues 138 – 378; red). The residue numbers on the structure correspond to the residues built into the density (2 – 208 and 148 – 348 for Hsp90 and Cdc37 respectively). Investigation of Kinase Domain Remodelling by Cdc37 170 One client kinase which has not been well studied in relation to its interactions with Cdc37 and Hsp90 is the Fibroblast Growth Factor Receptor 3 (FGFR3) (Acquaviva et al., 2014; Jin et al., 2011; Laederich et al., 2011; Taipale et al., 2012). The FGFR family of kinases are a key component of many cellular signalling pathways. The FGFRs are a group of receptor tyrosine kinases that are able to specifically recognise fibroblast growth factors (FGFs), a group of 18 distinct ligands. There are four main members of the FGFR family, known as FGFR1, FGFR2, FGFR3 and FGFR4 (Turner and Grose, 2010). Interestingly, a fifth related receptor (FGFR5) has also been discovered, but the protein does not have a tyrosine kinase domain, and has been postulated to be involved in negative signalling (Wiedemann and Trueb, 2000). FGFR3, and to a lesser extent FGFR1, have been found to be reliant upon the Hsp90 chaperone system, whilst FGFR2 and FGFR4 show no such dependence (Taipale et al., 2012). The FGFRs consist of three extracellular immunoglobulin domains, a transmembrane helix and an intracellular domain with tyrosine kinase activity (Figure 5.2). Interaction with the FGFR ligand causes receptor dimerisation and transautophosphorylation of the tyrosine kinase domains. The intracellular FGFR substrate (FRS2) can then be recruited to the phosphorylated intracellular FGFR domain via a phosphotyrosine-binding domain. This allows the FRS2 to be phosphorylated by the FGFR. Phosphorylated FRS2 can then recruit adaptor proteins, such as son of sevenless (SOS) and growth factor receptor-bound 2 (GRB2) (Gotoh, 2008). The presence of these factors leads to the activation of the membrane-bound Ras, which then initiates the downstream activation of the mitogen-activated protein kinase (MAPK) pathway, ultimately resulting in the promotion of cell growth and proliferation. In parallel to the activation of the MAPK pathway, the presence of GRB2 at the membrane also results in the recruitment of the GRB2-associated binding protein 1 (GAB1) (Figure 5.2). GAB1 is then phosphorylated by the FGFR and subsequently is able to recruit the lipid kinase phosphatidylinositol-3-kinase (PI3K) to the plasma membrane. Through the production of phosphatidylinositol-3,4,5-trisphosphate (PIP3) at the membrane, PI3K causes the activation of the protein kinase AKT, which results in the initiation of numerous downstream pathways involved in the promotion of progression through the cell cycle and cell survival. Independently of these two signalling cascades, the N-terminal Src homology 2 (SH2) domain of the enzyme phospholipase Cγ (PLCγ) is also able to bind to phosphorylated intracellular FGFR domains (Figure 5.2), via a phosphotyrosine towards the C-terminus of the FGFR Investigation of Kinase Domain Remodelling by Cdc37 171 protein (Bae et al., 2009; Peters et al., 1992). The PLCγ is subsequently phosphorylated and this causes its activation, allowing it to hydrolyse phosphatidylinositol-4,5-bisphosphate (PIP2) to produce inositol-1,4,5-trisphosphate (IP3) and diacylglycerol (DAG). The production of DAG results in the activation of protein kinase C (PKC), which reinforces the activation of the MAPK pathway via the phosphorylation of Raf. Figure 5.2. The signalling pathways downstream of fibroblast growth factor receptors (FGFRs). When the fibroblast growth factor (FGF) binds, the receptors dimerise and transautophosphorylate. The FGFR substate (FRS2) can bind to the phosphorylated tyrosine residues on the FGFR intracellular domain, and this allows it to be phosphorylated by the FGFR. This then allows the recruitment of the proteins GRB2 and SOS, which acts as a guanine exchange factor for the small G protein Ras. This initiates the MAPK signalling cascade, resulting in the promotion of cell growth and proliferation. Upon GRB2 recruitment, its binding partner GAB1 is also recruited, and can then be phosphorylated by the FGFR. This enables it to activate the lipid kinase PI3K, which results in the activation of the protein kinase AKT. The activation of AKT initiates a downstream signalling pathway parallel to the MAPK cascade, again resulting in the promotion of cell growth and proliferation. Finally, the enzyme phospholipase Cγ (PLCγ) is also able to bind to the phosphorylated tyrosine residues of the FGFR, and upon binding hydrolyses phosphatidylinositol-4,5-bisphosphate (PIP2) lipids to produce inositol-1,4,5-trisphosphate (IP3) and diacylglycerol (DAG). DAG can then activate protein kinase C (PKC), which further activates the initiation of the MAPK cascade. Given the multitude of growth and proliferation-promoting pathways that are inititated upon the activation of FGFRs, it is unsurprising that their dysregulation often results in malignancy. Many tumours have been demonstrated to show largescale overexpression or constitutively activating mutations within the FGFRs. The increase in awareness of the oncogenic role of FGFRs has prompted a rapid increase in efforts to target them for therapeutic purposes (Dieci et al., 2013; Katoh, 2016). A full and comprehensive understanding of their structural and Investigation of Kinase Domain Remodelling by Cdc37 172 functional characteristics is therefore of crucial importance to inform these efforts so as to better diagnose mechanisms for specifically targeting oncogenic versions. The manner in which FGFR3 interacts with Cdc37/Hsp90 is unknown, as is the reason why FGFR3 depends on the chaperone but FGFR1, FGFR2 and FGFR4 do not. Given the power of Hydrogen-Deuterium exchange-mass spectrometry (HDX-MS), as described in Chapter Three, it was decided to utilise this technique to study the FGFR3 kinase domain-Cdc37 interaction, and to compare the interaction with that of other known Cdc37/Hsp90 client kinase domains. This work was performed as a collaboration with Dr Tom Bunney and Professor Matilda Katan at University College London, and was published in February 2018 (Bunney et al., 2018). The biochemical work was performed by Dr Tom Bunney and colleagues, and the HDX-MS work was performed by myself with the help of Dr Glenn Masson. Investigation of Kinase Domain Remodelling by Cdc37 173 5.2. Materials and Methods All Materials and Methods for the experiments presented in this chapter are described in detail in Bunney et al., 2018. Included here are the HDX-MS methods, as performed by myself. 5.2.1. HDX-MS Hydrogen-Deuterium exchange-mass spectrometry (HDX-MS) samples were prepared containing FGFR3 kinase domain (FGFR3 KD: residues 455 – 768) (wild-type, E466K or I538F) or B-Raf kinase domain V600E (B-Raf KD V600E: residues 448-723) at 5 µM either alone or in the presence of Cdc37 at 7.5 µM in reaction buffer [25 mM Tris-HCl pH 8.0 (room temperature), 150 mM NaCl, 1 mM TCEP]. For the reciprocal study, to map the molecular effects of FGFR3 binding on Cdc37, the reactions contained 5 µM Cdc37 alone or in the presence of 7.5 µM FGFR3 KD E466K (all purified protein samples were obtained from Dr Tom Bunney at University College London). The reactions were incubated on ice for 15 minutes and then 10 µL aliquots of the reaction were mixed rapidly with 40 µL deuterated reaction buffer [25 mM Tris-HCl pH 8.0 (room temperature), 150 mM NaCl, 1 mM TCEP, 94.8 % v/v D2O (Acros Organics 7789-20)], producing a final D2O concentration of 75.4 %. Four incubation conditions were performed: 3 seconds on ice (equivalent to 0.3 seconds at room temperature), 3 seconds at 22 ˚C, 30 seconds at 22 ˚C and 5 minutes at 22 ˚C. All the reactions were quenched by the addition of 20 µL of ice-cold Quench buffer [5 M GdCl, 8.4 % formic acid] and then immediately frozen in liquid nitrogen. Each sample was thawed and injected on to an ultra-performance liquid chromatography (UPLC) system submerged in ice. The sample was then passed over two Poroszyme immobilised pepsin columns (Applied Bioscience) in series at 0.15 mL/minute for 3 minutes. The peptides were then passed on to a 1.7 µM Acquity UPLC BEH C18 pre-column (Waters, 186002346), equilibrated in Pepsin-A buffer [0.1 % formic acid], and eluted using a 5 – 45 % gradient of Pepsin-B buffer [0.1 % formic acid, 99.9 % acetonitrile] over 20 minutes. The column was then washed for 4 minutes with Pepsin-Wash buffer [0.1 % formic acid, 76 % acetonitrile]. Data were collected on a Xevo quadrupole time of flight mass spectrometer (Waters) over a time period of 30 minutes from 350 to 1500 m/z. The electrospray ionisation source operated at 225 ˚C, and the spray voltage was 3.5 kV. Investigation of Kinase Domain Remodelling by Cdc37 174 Peptides were identified from non-deuterated control samples analysed by MSe (Silva et al., 2006), with a tolerance of 15 parts per million for MS observations and 0.2 Da for tandem MS/MS observations. The peptides were then analysed using Mascot Distiller (Matrix Science), using a minimum Mascot score of 10. The peptides were then imported into HD- Examiner (Sierra Analytics) and the data were processed automatically, followed by manual inspection. Peptides showing an unacceptably low signal-to-noise ratio precluding robust analysis were eliminated at this stage. All mentions of deuteration percentages in the text are relative values, as no entirely deuterated sample could be obtained. Data are expressed as means ± the standard deviations. Investigation of Kinase Domain Remodelling by Cdc37 175 5.3. Results and Discussion 5.3.1. Reconstitution of a FGFR3-Cdc37-Hsp90 complex The ability of wild-type, full-length FGFR3 to bind to Cdc37 and form a ternary complex with Hsp90 has been previously documented (Taipale et al., 2012). This interaction is known to involve the intracellular kinase domain of FGFR3 (residues 455 – 768; hereafter denoted as FGFR3 KD) (Laederich et al., 2011). To test this, recombinant Hsp90, Cdc37 and FGFR3 KD were expressed in E. coli, each with an N-terminal His6 tag. The proteins were then purified using HisTrap affinity chromatography, anion exchange chromatography and finally gel filtration. Once purified, the individual components were then mixed at a ratio of 2:1:1 for Hsp90, Cdc37 and FGFR3 KD respectively and incubated for 15 minutes. Both the individual components and the mixture were then analysed by gel filtration. The results are shown in Figure 5.3. Figure 5.3. Gels showing the results of the size-exclusion chromatography. The top three panels show the elutions of Hsp90, Cdc37 and FGFR3 KD alone. The bottom panel shows the results of mixing the three components before injection on to the size-exclusion column. TC indicates the fractions containing the ternary complex, and the fractions containing free FGFR3 KD are also indicated. This figure is adapted with permission from Bunney et al., 2018. As shown in Figure 5.3, combining Hsp90, Cdc37 and FGFR3 KD causes the components to elute together, indicating the ternary complex has been reconstituted in vitro. It is important to note that FGFR3 is a relatively weak Hsp90 client, meaning that only a minor proportion (approximately 12 %) is incorporated into the ternary complex. The interaction between Cdc37 and FGFR3 KD was then further characterised using an affinity pull-down experiment. Cdc37 was immobilised on Talon resin, and then increasing Investigation of Kinase Domain Remodelling by Cdc37 176 amounts of FGFR3 KD were added to the resin. The samples were incubated with gentle mixing for 1 hour at 4 ˚C, and then the beads were washed to remove any non-bound protein. The bound proteins were then eluted from the beads with SDS-containing sample buffer and analysed by SDS-PAGE (Figure 5.4). In order to compare the ability of FGFR3 to bind to Cdc37 with other Cdc37/Hsp90 clients, a recombinant version of the kinase domain of B-Raf was also purified and subjected to this analysis. B-Raf is a serine/threonine protein kinase that is unrelated to FGFR3, but is a well-known Hsp90 client. The B-Raf KD construct used contained the well-characterised oncogenic mutation V600E, as well as a number of solubilising mutations (Tsai et al., 2008). Figure 5.4. Pull-down analysis of the Cdc37-FGFR3 KD interaction compared to the Cdc37-B-Raf KD interaction. Cdc37 was immobilised on Talon resin via the N-terminal His tag, and the beads washed to remove any unbound Cdc37. Different amounts of the FGFR3 KD (left gel; black bars) or B-Raf KD V600E (right gel; grey bars) was then added to each reaction, and the samples incubated for 1 hour at 4 ˚C. The beads were then washed and the bound proteins eluted with sample buffer. The amount of kinase bound in each case is quantified below the gels. This figure is adapted with permission from Bunney et al., 2018. As shown in Figure 5.4, immobilised Cdc37 is clearly able to pull down FGFR3 KD, and adding in more FGFR3 KD increases the amount that is pulled down. In comparison, B-Raf KD V600E showed significantly higher occupancy under similar conditions, indicating that the affinity between B-Raf KD V600E and Cdc37 is significantly higher than for FGFR3 KD and Cdc37. 5.3.2. Characterisation of different FGFR3 mutations A recent study analysed a series of cancer-associated mutations within the kinase domain of FGFR3 to identify those which have a direct impact on FGFR3’s kinase activity and those which are most likely simply passenger mutations (Patani et al., 2016). The same panel of mutations were tested for their ability to bind to Cdc37 using pull-down analysis. Investigation of Kinase Domain Remodelling by Cdc37 177 Figure 5.5. Quantification of an affinity pull-down reaction between immobilised Cdc37 and the FGFR3 KD variants indicated above. The data plotted are the fold differences compared to the wild-type protein (wt: 1). Mutations that resulted in significantly increased binding are indicated with pink stars. This analysis (Figure 5.5) identified three mutations which significantly increase FGFR3 KD’s ability to bind to Cdc37: E466K, I538F and N540K. Each of these three mutations also enhances FGFR3 KD’s ability to be incorporated into a ternary complex with both Cdc37 and Hsp90, as judged by size-exclusion chromatography (Figure 5.6). Figure 5.6. A quantification of the amount of FGFR3 KD wild-type (wt), E466K, I538F or N540K that is incorporated into a ternary complex with Cdc37 and Hsp90. The proteins were incubated together for 15 minutes at 4 ˚C and then injected on to a gel filtration column. The resulting fractions were then analysed by SDS-PAGE and the bands quantified using ImageJ. This figure is adapted with permission from Bunney et al., 2018. Investigation of Kinase Domain Remodelling by Cdc37 178 The reasons behind these mutations’ increased affinity for Cdc37 and Hsp90 were unclear. The original study (Patani et al., 2016) demonstrated that out of these three mutations, only N540K caused an increase in kinase activity levels: E466K retained similar activity levels to the wild-type protein and I538F showed slightly lower activity than the wild-type. This indicates that the dependence on Hsp90/Cdc37 is not directly linked to constitutively activating mutations. Given the proposed link between the thermostability of the kinase and its dependence on Hsp90/Cdc37, the thermal stability of each of the three constructs that showed increased binding to Cdc37 was tested and compared to the wild-type FGFR3 KD, the other non-client FGFR kinase domains and the unrelated client B-Raf. The thermal stability was measured using a thermofluor assay: the dye SYPRO orange was added to each sample and the samples slowly heated from 10 ˚C to 90 ˚C. Whilst the protein remains folded, the fluorescence of SYPRO Orange is largely quenched by surrounding water molecules. As the protein unfolds, the fluorescent dye can bind non-specifically to the newly exposed hydrophobic regions of the protein. This binding effectively excludes water molecules, resulting in an increase in fluorescence of the dye. The unfolding process can thus be monitored by the increase in fluorescence, allowing the calculation of an apparent melting temperature (Tm). Figure 5.7. The melting temperature for the FGFR3 KD mutations in comparison to wild-type FGFR3 KD, B-Raf KD V600E, and the other FGFR KDs. The Tm values were calculated from the midpoint of the fluorescence change reflecting the denaturation of the protein in a thermofluor assay. Green stars indicate kinases that are not dependent on Hsp90, and pink stars denote Hsp90 clients. The data plotted are the means ± the standard deviations (n=3). This figure is adapted with permission from Bunney et al., 2018. This analysis provides a very interesting insight into what makes a particular kinase into an Hsp90 client. In Figure 5.7, it can clearly be seen that the three mutations that cause an increase in the construct’s ability to bind to Cdc37 also consistently lead to a significant Investigation of Kinase Domain Remodelling by Cdc37 179 reduction in protein stability. The thermal stability therefore seems to correlate well with the protein’s ability to bind to Cdc37 and Hsp90, indicating that this could be a defining characteristic of Hsp90 clients. To test this hypothesis, the thermostability of the mutations that did not cause an increase in Cdc37 binding was tested using the same method. Figure 5.8. The melting temperature for the FGFR3 KD mutations that did not alter the protein’s ability to interact with Cdc37. The wild-type kinase domain and E466K mutation were included as positive controls. The Tm values were calculated from the midpoint of the fluorescence change reflecting the denaturation of the protein in a thermofluor assay. Data plotted are the means ± the standard deviations (n=3). This figure is adapted with permission from Bunney et al., 2018. Figure 5.8 clearly shows that none of the other mutations cause a similar drop in thermostability, implying that the protein’s stability is a crucial factor in determining how strongly a protein interacts with the co-chaperone Cdc37 and by extension with Hsp90. It is interesting to note that the three mutations that do increase the affinity of the interaction are all found within the N-lobe of the kinase domain. Closer inspection of the available crystal structures show that all three of these mutations are located in key regulatory features of the kinase domain, for example I538 is part of the DFG latch and N540 is located within the previously described molecular brake (Chen et al., 2017). This suggests that these three residues have crucial roles in maintaining the stability of the FGFR3 kinase domain, and therefore explains why their substitution has such detrimental effects upon the stability of the kinase domain. To further characterise the effects of these mutations, HDX-MS was used to investigate the way in which the mutations affected the isotopic exchange rates across the protein sequence. Investigation of Kinase Domain Remodelling by Cdc37 180 Samples containing either wild-type FGFR3 KD, FGFR3 KD E466K or FGFR3 KD I538F were therefore prepared and exposed to deuterated buffer for 0.3 seconds (the manner in which such a short deuteration period is achieved is discussed in more detail in Chapter Three). These two mutations were chosen for further investigation due to their slightly better expression and increased stability compared to N540K. The deuteration reaction was then quenched, and the proteins digested into peptic peptides before being analysed on the mass spectrometer. Deuterated peptides were then compared to a non-deuterated control to give a percentage deuteration, and this is plotted against the peptide midpoint in Figure 5.9. Figure 5.9. The global deuteration profile for FGFR3 KD wild-type (wt: blue), E466K (red) and I538F (green). The percentage deuteration after a 0.3 second incubation with deuterated buffer is plotted against the midpoint of each peptide (i). The percentage deuteration is corrected for the length of the peptide and the maximal level of deuteration that could be achieved based on the final buffer composition. A schematic depicting the secondary structure elements of the kinase domain is shown below and the positions of the mutations are indicated with red and green dashed lines. The data were collected in triplicate and the means ± the standard deviations for each peptide are plotted. Error bars that are smaller than the graphical icon are omitted. This figure is adapted with permission from Bunney et al., 2018. This analysis prompts some interesting observations, although it is important to reiterate the limitations of this experiment given the lack of a control for the rate of back exchange for each individual peptide (as discussed on page 134). Again, if this type of experiment were to be repeated, a fully deuterated or equilibrium labelled control should be included. Given that the rate of isotopic exchange for an amide proton is predominantly dependent on the higher order structure, previous work from the group has suggested that deuteration levels above 50 % after 0.3 seconds are indicative of intrinsic disorder, or very dynamic secondary structure elements (Burke et al., 2014; Fowler et al., 2016). This implies that the kinase domain of Investigation of Kinase Domain Remodelling by Cdc37 181 FGFR3 contains areas with high flexibility, most notably at the N-terminus of the domain, amino acids 570 – 590, and around the activation loop. Interestingly, the E466K substitution causes a significant increase in the percentage deuteration of peptides at the N-terminus of the domain, indicating that it has caused an increase in the flexibility of this region of the kinase domain. This makes sense given the position of the mutation, and could be explained by the location of E466 at the centre of an allosteric network within the N-lobe of the kinase domain. The effects of mutating residue I538 are much less pronounced, and not obvious through this analysis. However, the resultant drop in Tm clearly implies that I538F also causes a significant decrease in stability. Why this is not observed using HDX-MS is not clear but could be due to smaller decreases in stability across the protein rather than a significant drop in stability of one particular region, as for E466K. Increasing the temperature or duration of the labelling reaction could help with the detection of changes in the mutant protein. To extend the analysis, the FGFR3 KD (wild-type) and the mutant I538F were also characterised by nuclear magnetic resonance (NMR) spectroscopy (Bunney et al., 2018). As expected, this showed chemical perturbations in the spectra for I538F in comparison to the wild-type protein. Specifically, changes in the chemical shifts for the adjacent residues K536 and N537 were observed, which was predictable given the insertion of the aromatic phenylalanine residue. Further afield, there were also observable changes to the resonances for residues within the DFG latch and the activation loop, as well as changes to residues involved in the allosteric network of the N-lobe. This therefore gives insights into how the I538F mutation causes the destabilisation of the kinase, leading to the reduced Tm value observed in the thermofluor assay (Figure 5.7). 5.3.3. Investigation of the effects of Cdc37 binding to FGFR3 To gain more information about how the kinase domain of FGFR3 interacts with Cdc37, HDX-MS was again utilised. The manner in which HDX-MS can be used to give information concerning a potential binding site has been discussed in detail in Chapter Three. Briefly, peptides with reduced isotopic exchange rates in the presence of a binding partner compared to the protein alone can be surmised to represent the binding ‘footprint’ of the interacting partner. The wild-type FGFR3 KD, FGFR3 E466K, FGFR3 I538F and B-Raf V600E were all tested to try to identify a common mode of interaction between different kinase domains, and Investigation of Kinase Domain Remodelling by Cdc37 182 in the hope of explaining the differences between a strong client-Cdc37 interaction and a weak one. The initial aim was to characterise the effects of Cdc37 binding on the wild-type kinase domain of FGFR3. FGFR3 KD at 5 µM was therefore mixed with 7.5 µM Cdc37, and the samples then incubated on ice for 30 minutes. The samples were mixed rapidly with deuterated buffer for 0.3 seconds, 3 seconds, 30 seconds or 300 seconds before the reactions were quenched and rapidly frozen in liquid nitrogen. The samples were then digested into peptides and the deuterium incorporation of each peptide measured by mass spectrometry. Figure 5.10. The difference in the percentage deuteration of peptides from FGFR3 KD (wild-type) upon addition of the interacting partner Cdc37. Percentage deuterations are corrected for the length of each peptide and the maximal deuteration of the peptide based on the deuterium composition of the buffer. The percentage deuteration change was then calculated by subtracting the percentage deuteration of each peptide in the FGFR3 KD-Cdc37 complex sample from the same peptide in the FGFR3 KD apo sample. The differences at each time point (0.3, 3, 30, 300 seconds) were then summed for each peptide. The data plotted are the summed means ± the standard deviations (n=3). Error bars that are smaller than the graphical icon are omitted. A schematic depicting the secondary structure elements of the kinase domain is shown below. This figure is adapted with permission from Bunney et al., 2018. The results shown in Figure 5.10 show the differences in percentage deuteration of FGFR3 KD peptides when in complex with Cdc37 and in the apo state. A decrease in the percentage deuteration upon the addition of Cdc37 (corresponding to a positive value on the y axis of Figure 5.10) would indicate ‘protection’, i.e. a reduced rate of isotopic exchange in the presence of the binding partner, whilst an increase in percentage deuteration upon addition of Cdc37 (a negative value in Figure 5.10) would indicate ‘exposure’, representing an increased rate of isotopic exchange in the presence of the complex partner. No major changes can be Investigation of Kinase Domain Remodelling by Cdc37 183 observed upon the addition of a stoichiometric excess of Cdc37, making the characterisation of the binding site difficult. Interestingly, there appears to be an area of increased isotopic exchange upon Cdc37 binding (residues 505-525), which could represent an area of localised unfolding. However, the observed changes are relatively small, probably due to the relatively low affinity of the interaction. Both the E466K and I538F mutations caused the FGFR3 KD to be able to pull down more Cdc37 (Figure 5.5), indicating a higher affinity interaction. The interaction between Cdc37 and the mutant proteins was therefore characterised using HDX-MS as described above to try to gain more insights into the manner in which Cdc37 binds to Hsp90-client kinases. The results of these experiments are shown in Figure 5.11. Figure 5.11. The difference in the percentage deuteration of peptides from FGFR3 KD E466K (red) and FGFR3 KD I538F (green) upon addition of the interacting partner Cdc37. The percentage deuteration change was calculated by subtracting the percentage deuteration of each peptide in the FGFR3 KD-Cdc37 complex sample from the same peptide in the FGFR3 KD apo sample. The differences at each time point (0.3, 3, 30, 300 seconds) were then summed for each peptide. The data plotted are the summed means ± the standard deviations (n=3). A schematic depicting the secondary structure elements of the kinase domain is shown below and the positions of the mutations are indicated with red and green dashed lines. This figure is adapted with permission from Bunney et al., 2018. As shown in Figure 5.11, there are extensive changes in the deuteration profile of both FGFR3 KD E466K and I538F upon the inclusion of Cdc37. In both cases, the addition of Cdc37 leads to a significant increase in percentage deuteration of the peptides from the N- terminal region of the kinase domain, and a decrease in the percentage deuteration of the peptides of several regions of the C-lobe of the kinase domain. Encouragingly, these changes correlate with the small changes seen in the wild-type FGFR3 KD, indicating that the manner Investigation of Kinase Domain Remodelling by Cdc37 184 of binding is conserved. Figure 5.12 shows these changes mapped on to the structure of the kinase domain to allow further analysis. Figure 5.12. The structure of the FGFR3 kinase domain (PDB code 4K33) (Huang et al., 2013) with the Hydrogen-Deuterium exchange differences for E466K and I538F constructs painted on to the structure. Peptides with changes > 5 % upon addition of excess Cdc37 are coloured according to the key on the right. The timepoint with the greatest change for each peptide was used for the colour scheme. This figure is adapted with permission from Bunney et al., 2018. This analysis clearly shows that Cdc37 binding leads to a region of the C-lobe showing a decrease in the isotopic exchange rate, implying that this region is likely to be at the protein- protein interface (Konermann et al., 2011). Importantly, there is also a significant increase in the rate of isotopic exchange for the majority of the N-lobe of the kinase. This implies a major restructuring or unfolding event is occurring as a result of Cdc37 binding, which has interesting implications when one considers the role of thermostability in defining ‘clientness’. To determine whether this effect was conserved across other client kinases, the interaction between B-Raf KD V600E and Cdc37 was also analysed using HDX-MS. The results of the analysis are shown in Figure 5.13. Investigation of Kinase Domain Remodelling by Cdc37 185 Figure 5.13. The difference in the percentage deuteration of peptides from B-Raf KD V600E upon addition of the interacting partner Cdc37. The percentage deuteration change was calculated by subtracting the percentage deuteration of each peptide in the B-Raf KD V600E-Cdc37 complex sample from the same peptide in the B-Raf KD V600E apo sample. The differences at each time point (0.3, 3, 30, 300 seconds) were then summed for each peptide. The data plotted are the summed means ± the standard deviations (n=3). A schematic depicting the secondary structure elements of the kinase domain is shown below. This figure is adapted with permission from Bunney et al., 2018. This characterisation of the B-Raf KD-Cdc37 interaction shows many key similarities to the interaction of FGFR3 KD and Cdc37. Upon Cdc37 binding, the N-lobe of the kinase domain of B-Raf undergoes a dramatic increase in isotopic exchange in a very similar manner to FGFR3 KD. There are also similar regions of decreased isotopic exchange in the C-lobe, indicating the position and mode of Cdc37 binding seems to be conserved across multiple, unrelated kinase clients. This can also be illustrated more clearly by mapping the observed changes on to the structure of the B-Raf kinase domain (Figure 5.14). Figure 5.14. The structure of the B-Raf kinase domain (PDB code 4R5Y) (Tang et al., 2015) with the Hydrogen-Deuterium exchange differences painted on to the structure. Peptides with changes > 5 % upon addition of excess Cdc37 are coloured according to the key on the right. The timepoint with the greatest change for each peptide was used for the colouring. This figure is adapted with permission from Bunney et al., 2018. Investigation of Kinase Domain Remodelling by Cdc37 186 These HDX-MS studies have thus identified a potential mode of interaction between Cdc37 and client kinases that causes a dramatic restructuring of the N-lobe of the kinase. To corroborate these conclusions, the FGFR3 KD I538F-Cdc37 interaction was studied using NMR spectroscopy (Bunney et al., 2018). Titrating Cdc37 into NMR samples containing FGFR3 KD I538F led to the appearance of sharp, well-defined resonances in the random-coil area of the spectrum. By combining 3D techniques and selective labelling, these resonances were assigned to the N-terminal region of the N-lobe of FGFR3 KD. The appearance of these sharp peaks is indicative of increased mobility and localised unfolding in this region upon this addition of Cdc37, consistent with the results obtained through HDX-MS. 5.3.4. Investigation of the effects of FGFR3 binding to Cdc37 To look at the reciprocal aspects of the interaction, HDX-MS was also used to study the molecular changes to Cdc37 that occur upon binding to FGFR3 KD. Given the relatively weak interaction between the wild-type kinase domain and Cdc37, the E466K mutation was used for this analysis. The strong similarities between the effects of Cdc37 binding to the different kinase variants tested imply the mode of binding is conserved, so the effects on Cdc37 upon FGFR3 KD E466K binding are likely to be representative of all kinase clients. The experiment was done in the same manner as described previously, except that the protein mixtures contained a 1.5-fold molar excess of FGFR3 KD E466K over Cdc37. The results are shown in Figure 5.15. Investigation of Kinase Domain Remodelling by Cdc37 187 Figure 5.15. The difference in the percentage deuteration of peptides from Cdc37 upon addition of FGFR3 KD E466K. The percentage deuteration change was calculated by subtracting the percentage deuteration of each peptide in the Cdc37-FGFR3 KD E466K complex sample from the same peptide in the Cdc37 apo sample. The differences at each time point (0.3, 3, 30, 300 seconds) were then summed for each peptide. The data plotted are the summed means ± the standard deviations (n=3). The brackets indicate areas of decreased solvent exposure upon FGFR3 KD E466K binding. This figure is adapted with permission from Bunney et al., 2018. Unfortunately, the pepsin digestion of Cdc37 was non-optimal, meaning that the analysis was unable to detect peptides covering the entirety of the protein sequence. Specifically, no peptides were observed representing parts of the N-terminus of the protein, leading to a region with no information between residues 30 – 120 (Figure 5.15). Despite this, the results still allowed the identification of two regions that show dramatically reduced isotopic exchange rates in the presence of FGFR3 KD E466K: residues 5 – 28 and residues 302 – 341 (indicated with brackets in Figure 5.15). Whilst these regions are distant in the protein sequence, mapping the changes on to a structural model of Cdc37 shows that they could be spatially adjacent, and so could form a single binding site (Figure 5.16). Investigation of Kinase Domain Remodelling by Cdc37 188 Figure 5.16. A model of the structure of the co-chaperone Cdc37 (based on numerous structural fragments in combination with SAXS data (SASDBP9)) (Bunney et al., 2018) with the Hydrogen-Deuterium exchange differences painted on to the structure. Peptides with changes > 5 % upon addition of excess FGFR3 KD E466K are coloured according to the key on the right. The timepoint with the greatest change for each peptide was used for the colouring. The N- and C-termini are indicated on the structure, and the region missing from the HDX-MS data is demarcated with a dashed line. This figure is adapted with permission from Bunney et al., 2018. Interestingly, a recent NMR study looking at the interaction between Cdc37 and B-Raf V600E also identified these two regions as the potential interacting sites (Keramisanou et al., 2016). Moreover, the small changes that are visible in the middle of the protein (residues 219 – 251 and 270 – 293) in Figure 5.15 were also noted by NMR, and tentatively attributed to conformational changes as a result of binding. The HDX-MS data presented here are therefore in close agreement with those NMR data, indicating that the FGFR3 kinase domain and the B- Raf kinase domain interact with Cdc37 in a very similar manner. Together, these data suggest the existence of a common Cdc37-kinase client interaction interface and mode of binding. The apparently multipartite nature of this interaction made defining the spatial organisation of the complex from these data alone challenging. To aid understanding of the architecture of this binary complex, small-angle X-ray scattering (SAXS) was therefore used to obtain a low- resolution envelope. Scattering data were collected for both components individually (FGFR3 KD I538F and Cdc37) as well as for the complex, and ultimately allowed the generation of a model for the two interacting proteins (Bunney et al., 2018). Models produced via different methodologies (either ab initio or rigid body approaches, in both cases using constraints from the HDX-MS and NMR studies) converged at low resolutions to give an asymmetric structure with distinctive characteristics, shown in Figure 5.17. Investigation of Kinase Domain Remodelling by Cdc37 189 Figure 5.17. The calculated SAXS envelope for the complex is shown above, with the region thought to represent the kinase domain coloured in light grey and the region thought to represent Cdc37 shown in dark grey. The dimensions of the model are indicated. Below is the best fit for the model, with the kinase domain in green and Cdc37 in purple. The key regions, including those identified by HDX-MS and NMR, are indicated on the structure. The model shown in Figure 5.17 depicts the extreme N-terminus and some C-terminal helices of Cdc37 interacting mostly with the C-lobe of the client kinase (in this case FGFR3 KD I538F). The N-terminal region of Cdc37 also extends outwards towards the N-lobe of the kinase domain, and could have a role in the destabilisation of the N-lobe. Investigation of Kinase Domain Remodelling by Cdc37 190 5.4. Conclusions The Hsp90 chaperone system is known to have an important role in the regulation of approximately 60 % of the protein kinases within the cell. Despite this, the molecular mechanisms behind these interactions have been unclear for many years. Furthermore, there is a fundamental lack of understanding concerning what makes one kinase dependent on Hsp90, whilst another very similar kinase domain is not. The fact that so many of the kinase clients of this chaperone system are known and important oncogenes makes a comprehensive understanding of the system very desirable, especially given the recent development of Hsp90 inhibitors as potential therapeutic candidates. This chapter describes efforts to address these outstanding questions in the field through the combination of biochemical characterisation, Hydrogen-Deuterium exchange-mass spectrometry (HDX-MS), nuclear magnetic resonance (NMR) spectroscopy and small-angle X-ray scattering (SAXS). Initially, the kinase domain of FGFR3, the chaperone Hsp90 and the co-chaperone Cdc37 were expressed and purified. Their ability to assemble into both binary (containing FGFR3 KD and Cdc37) and ternary complexes was demonstrated using pull-down reactions as well as size-exclusion chromatography. A panel of cancer-linked mutations in the kinase domain of FGFR3 were then analysed, and each mutation was tested for its ability to bind to Cdc37. This led to the identification of three mutations that significantly increased the affinity of the interaction, causing a large increase in the amount of FGFR3 KD pulled down by immobilised Cdc37. The reason why these three mutations had such a large effect on the affinity of the FGFR3 KD-Cdc37 interaction was initially unclear, and showed no correlation to the kinase activity of each construct. However, analysis of the thermal stability of each of the mutants using a thermofluor assay led to the observation that each mutation that showed increased affinity for Cdc37 also resulted in a significant decrease in the thermostability of the protein. This suggested that the defining characteristic of ‘clientness’, i.e. the dependence of a protein kinase on the Hsp90 chaperone system, could be a lack of intrinsic stability in the protein. To gain more information about this, HDX-MS was then used to probe the effects of two mutations that caused an increase in affinity towards Cdc37 and a decrease in protein stability. This analysis showed that the E466K mutation caused a significant destabilisation of the N- terminal region of the N-lobe of the kinase domain. The effects of the I538F mutation were Investigation of Kinase Domain Remodelling by Cdc37 191 less obvious and were not discernable through this HDX-MS analysis. Consideration of the positions of the mutated residues within the structure of the FGFR3 kinase domain showed that they make up part of an allosteric network that is likely to help stabilise the structure of the kinase domain N-lobe. It is therefore logical that substitution of these residues has a deleterious and destabilising effect. The next aim was to use HDX-MS to characterise the interaction between Cdc37 and client kinases. Comparison of isotopic exchange rates across the structure identified a region of decreased deuteration in the presence of Cdc37 within the C-lobe of the kinase domain. Whilst HDX-MS does not directly map the sites of protein-protein interactions, areas of decreased isotopic exchange are very often located at the interface, and so it is likely that Cdc37 binds to FGFR3 KD through the C-lobe of the kinase domain. Importantly, the HDX- MS analysis also identified a region with significantly increased isotopic exchange rates at the N-terminus of the kinase domain. Analysis of multiple client kinases showed that this appears to be a common phenomenon, indicating that this could be a biologically important consequence of the interaction. This conclusion was supported by NMR data, lending further weight to this hypothesis. The observation that the localised unfolding of the N-lobe of the kinase domain occurs upon Cdc37 binding has interesting implications, especially given the previous description of the correlation between ‘clientness’ and thermostability of the protein. If the interaction with Cdc37 relies upon the unfolding of the kinase domain, it is reasonable to suppose that this interaction is precluded in cases in which the kinase domain is too stable to be easily disrupted. This provides a potential molecular mechanism by which kinase clients could be distinguished from non-clients by Cdc37. Once bound to Cdc37, the unfolded or disordered N-lobe of the kinase domain could then be recognised and bound by Hsp90, forming the ternary complex (Figure 5.18). Investigation of Kinase Domain Remodelling by Cdc37 192 Figure 5.18. A schematic depicting the manner by which Cdc37 and Hsp90 could distinguish between client and non-client kinases. HDX-MS was also used to characterise the molecular effects of the interaction on Cdc37, and allowed the identification of two potential interacting sites. The extreme N-terminus of Cdc37 showed significantly reduced rates of isotopic exchange, as did a helical region towards the C- terminus of the protein, indicating a bipartite interaction. Combining the experimentally determined constraints from NMR spectroscopy and HDX-MS with the study of the complex via SAXS allowed the generation of a model for the binary complex between FGFR3 KD and Cdc37. This model indicates that the C-terminal region of Cdc37 interacts with the C-lobe of the kinase domain, and the extreme N-terminal section of Cdc37 extends towards the N-lobe, likely promoting the local unfolding. This is the first model of the binary structure that has been generated, and therefore offers an important insight into the architecture of the complex. A recent cryo-electron microscopy structure of a ternary complex between Hsp90, Cdc37 and cyclin-dependent kinase 4 (Cdk4) shows many similarities with the model presented here (Verba et al., 2016). The structure shows Hsp90 in a dimeric, ‘closed’ conformation, with a ring-like structure, and Cdc37 appears to be in an open hairpin conformation that allows it to Investigation of Kinase Domain Remodelling by Cdc37 193 wrap around Hsp90. Interestingly, the two lobes of Cdk4 are highly separate, and much of the N-lobe is unfolded and threaded through the lumen of the Hsp90 ring. As predicted by the HDX-MS data described in this chapter, the C-lobe of the kinase is apparently intact and interacts with the C-terminal region of Cdc37. The N-terminus of Cdc37 interacts with the N- lobe, and accurately mimics the typical interactions between the N- and C-lobes, thus likely stabilises this partially unfolded state. This structure therefore correlates with the data presented here extremely well. This implies that the key conformational rearrangements that are necessary for the formation of the ternary complex are in the most part coordinated by the action of Cdc37, which is able to prepare the kinase domain in the open or unfolded state to allow Hsp90 to bind. 194 Conclusions and Future Directions The work presented in this thesis represents the first attempt to develop a system to reconstitute the in vitro activation of the human nutrient sensing kinase GCN2. This protein is implicated in numerous processes in health and disease, yet much of the current knowledge is extrapolated from genetic studies on the yeast homologue. This means the molecular mechanisms that underlie how GCN2 is able to recognise and respond to amino acid starvation are not well understood, hindering the development of therapeutic strategies aimed towards this protein. Human GCN2 was expressed and purified to high homogeneity, before being characterised by a wide range of biophysical techniques, including size-exclusion chromatography in combination with multi-angle light scattering, Hydrogen-Deuterium exchange-mass spectrometry and differential scanning fluorimetry. Purified, recombinant human GCN2 was shown to autophosphorylate in the presence of ATP, and could phosphorylate a recombinant version of its physiological substrate eIF2α. Unexpectedly, these experiments demonstrated that, contrary to the prevalent model for GCN2 activation, physiologically relevant concentrations of deacylated tRNA do not stimulate phosphorylation of eIF2α by GCN2. Measuring the affinity of the interaction between GCN2 and deacylated tRNA via surface plasmon resonance showed the interaction to be relatively weak, calling into question the proposition that GCN2 is able to recognise a simple rise in the concentration of deacylated tRNA and indicating that there must be other factors involved in this process. Purified ribosomes were shown to be a potent activator of GCN2, and a direct interaction between GCN2 and the ribosome was demonstrated. This interaction was initially characterised by truncation analysis, which showed the pseudokinase domain, the histidyl tRNA synthetase- like domain and the C-terminal domain all to have a role in the interaction. These analyses allow the proposal of a model for GCN2 activation. In an amino acid starvation state, the intracellular availability of aminoacylated tRNA molecules would decrease. Thus, during translation, if the ribosome encountered an mRNA codon encoding an amino acid that is unavailable, the likelihood of efficient recruitment of the cognate aminoacylated tRNA would be decreased. This could lead to a situation in which the ribosome is stalled on the mRNA transcript, most likely with an empty A site. GCN2 could Conclusions and Future Directions 195 somehow recognise this state and become activated, thus enabling the phosphorylation of eIF2α and the initiation of the Integrated Stress Response. Whilst conceptually logical, there is much evidence missing that would confirm this model. The key outstanding question concerns the specificity of this interaction: why GCN2 is not constitutively activated by the ribosome, given what is seen in the in vitro assay. There are many potential explanations for this – for example it is plausible that in cells, GCN2 is simply outcompeted for ribosome binding by the many proteins that constitute the translational machinery, and are continually being recruited to and from the ribosome during the elongation cycle. To test this, one could add in various ribosome binding proteins to the in vitro activity assay, and see whether the presence of any factor or combination of factors leads to a down-regulation of GCN2 in the presence of ribosomes. If such a factor were identified, this would be an extremely promising lead in terms of how GCN2 achieves specificity. An alternative possibility is that GCN2 recognises a specific ribosomal state that signals a stalled elongation complex. An attractive possibility for this centres around the recent characterisation of ribosomal collisions, and their role in the initiation of quality control pathways (Simms et al., 2017). If GCN2 were able to specifically bind to di- or tri-ribosomes, this would confer specificity to the reaction that the experiments done with ‘bulk ribosomes’ may miss. It is possible to separate these collided ribosomes from monosomes via gradient fractionation, and so testing their ability to differentially stimulate GCN2 would be very interesting. If any difference were to exist, the further characterisation of the interaction could yield a wealth of structural and functional information. A common facet of these models is the stalling-dependent recruitment of GCN2 to the ribosome. For this reason, it would be extremely valuable to be able to ascertain the intracellular location of GCN2 under starvation and non-starvation conditions. If it were possible to fluorescently tag endogenous GCN2 and observe changes in the intracellular localisation of the kinase upon starvation, this would be ideal. The success of this would depend upon the ability to identify ribosomal localisation of the kinase, potentially to the rough endoplasmic reticulum. However, this may be difficult due to the existence of cytoplasmic ribosomes. Alternatively, it may be possible to judge changes in ribosomal localisation via co-migration of GCN2 and ribosomes through a sucrose gradient under different conditions; however, this analysis could result in artefacts as the treatment could lead to complex dissociation. It is also possible that GCN2 is generally able to bind to Conclusions and Future Directions 196 ribosomes, but upon amino acid starvation the interaction is modulated in some way causing the activation of the kinase. The Hydrogen-Deuterium exchange-mass spectrometry (HDX-MS) study described in Chapter Three resulted in the identification of the GCN2 binding site on the ribosome. The use of HDX-MS to identify a binding site on such a large and complex molecule shows the power of this technique, and the optimisation of the method could have important implications for future studies of other ribosome-interacting proteins. This analysis showed that GCN2 binds to part of the P stalk protein uL10, which sits directly above the A site of the ribosome. This suggests that GCN2 binds next to the A site, meaning that it would be in a plausible position to monitor the translational cycle. For example, the translation factor eukaryotic elongation factor 2 (eEF2) binds to uL10 after each peptide bond formation as part of its role promoting the GTP-dependent translocation of the ribosome. It is plausible to suggest that upon translational stalling, eEF2 would no longer be recruited to uL10, potentially allowing GCN2 to bind to the ribosome and thus become activated. Of course, there are many factors that associate with this region of the ribosome, and so there are many possibilities in this vein. The identification of this binding site in two separate HDX-MS experiments provides confidence in this result; however, it is possible that there is some unknown systemic bias towards this region of uL10 in the analysis. To confirm this binding site, it would be ideal to perform an independent validation. The simplest way would be to purify a recombinant version of uL10 (as demonstrated in (Abo et al., 2004) and show that this protein is able to bind to GCN2. This would have the further benefit of allowing the performance of HDX-MS on the GCN2-uL10 interaction with the aim of identifying the region of GCN2 that is involved in the interaction. Knowledge of which GCN2 residues are important for the interaction would allow the substitution of these residues to test whether binding could be inhibited. Moreover, any mutations with an observable effect (that did not affect basal activity) could then be tested in vivo to determine whether the GCN2-ribosome interaction is crucial for the initiation of the ISR in response to amino acid starvation. This would be an extremely important result, and would show that the GCN2-ribosome interaction is crucial for GCN2 to be able to respond to amino acid starvation. Chapter Four of this work describes efforts to gain structural information about GCN2 alone or as part of a complex. Negative stain electron microscopy resulted in a low-resolution Conclusions and Future Directions 197 structure of the protein, allowing initial characterisation of the dimensions and general architecture. However, the inherent inability of this technique to reach resolutions higher than approximately 20 Å precludes in-depth analysis. The full-length protein was then studied by cryo-electron microscopy, but no reproducible particles were observed, and so no structural information could be gleaned. X-ray crystallography efforts were also ongoing in parallel, and many different techniques to improve crystallisation outcomes were screened, including disorder analysis by HDX-MS, limited proteolysis and the production of nanobodies. Ultimately, these endeavours resulted in a 2.6 Å structure of the pseudokinase domain, which is the first structure of the human protein. The ability to purify large amounts of the protein and to produce a variety of construct variants mean that there are many other avenues that could be explored in this area. For example, the nanobodies were unable to provoke crystallisation of the full-length protein, but it is possible they would be useful in crystallising various GCN2 truncations. For this, HDX-MS could be initially used to map the binding site of each nanobody on to the full-length protein. The nanobodies could then be used rationally with different truncations in crystallography trials. Having already produced this library of nanobodies, their uses could be extended beyond simple crystallisation chaperones. For instance, they could also be tested to see if their presence affects the activity of GCN2 at all. For example, if a nanobody bound to a similar site as the ribosome or the substrate eIF2α, then its inclusion would prevent GCN2 from becoming activated by ribosomes, or from phosphorylating eIF2α at all. Any nanobody that showed any effect on binding could become a useful tool in unpicking the molecular intricacies of the mechanism of the kinase. A cryo-EM structure of a GCN2-ribosome complex would be extremely desirable, as this could potentially give insights into how GCN2 recognises stalled ribosomes, and how this interaction leads to the activation of GCN2. Many different strategies were tried in order to achieve this goal, as described in Chapter Four, including different approaches to isolating the ribosomes, different combinations of proteins and the inclusion of cross-linkers. Upon the inclusion of the cross-linker glutaraldehyde, it was possible to observe some low-resolution extra density adjacent to the binding site on uL10. Whilst encouraging, the inability to reproduce this density alongside the difficulties improving the resolution sufficiently to be able to confirm that it does not belong to another ribosome-binding protein mean that there is much work still to be done in this area. Given the apparently low levels of occupancy Conclusions and Future Directions 198 achieved by any sample preparation method, it seems as though improvements must be made in the understanding of the biochemistry of the system to inform future structural studies. For example, if it could be shown that GCN2 binds to a specific ribosomal state (such as a collided ribosome, for example), then efforts to stabilise that state would likely prove fruitful. Alternatively, if GCN2 is denaturing somehow during the vitrification procedure, trying shorter constructs of the protein could be helpful. This would benefit from determining the regions of GCN2 to which the ribosome binds, as this knowledge could potentially inform the choice of construct used for this study. Finally, Chapter Five of this thesis describes the characterisation of another system of kinase regulation. The chaperone Hsp90 and its co-chaperone Cdc37 are essential for the stability and maintenance of approximately 60 % of the human kinome, yet the mechanisms behind how client kinases are recognised and bound were unclear. Using a combination of biochemical techniques and comprehensive analysis by HDX-MS, this work demonstrated that a kinase’s dependence on Hsp90/Cdc37 correlates with the thermal stability of the protein. Furthermore, the HDX-MS showed that Cdc37 binding causes a dramatic remodelling of the N-lobe of client kinases, providing molecular insights into the complex assembly and the defining characteristics of ‘clientness’. As the majority of oncogenic kinases are Hsp90 clients, structural and functional insights into the client-chaperone interaction could prove extremely valuable for the development of inhibitors targeting this chaperone system. Together, the insights described in this thesis have contributed towards a full and comprehensive understanding of how cells recognise and respond to their environment. 199 References Abo, Y., Hagiya, A., Naganuma, T., Tohkairin, Y., Shiomi, K., Kajiura, Z., Hachimori, A., Uchiumi, T., and Nakagaki, M. (2004). Baculovirus-mediated expression and isolation of human ribosomal phosphoprotein P0 carrying a GST-tag in a functional state. Biochemical and Biophysical Research Communications 322, 814–819. Abzalimov, R.R., Bobst, C.E., and Kaltashov, I.A. (2013). A new approach to measuring protein backbone protection with high spatial resolution using H/D exchange and electron capture dissociation. Anal. Chem. 85, 9173–9180. Acquaviva, J., He, S., Zhang, C., Jimenez, J.-P., Nagai, M., Sang, J., Sequeira, M., Smith, D.L., Ogawa, L.S., Inoue, T., et al. (2014). FGFR3 translocations in bladder cancer: differential sensitivity to HSP90 inhibition based on drug metabolism. Mol. Cancer Res. 12, 1042–1054. Agirrezabala, X., Fernández, I.S., Kelley, A.C., Cartón, D.G., Ramakrishnan, V., and Valle, M. (2013). The ribosome triggers the stringent response by RelA via a highly distorted tRNA. EMBO Reports 14, 811–816. Ali, M.M.U., Roe, S.M., Vaughan, C.K., Meyer, P., Panaretou, B., Piper, P.W., Prodromou, C., and Pearl, L.H. (2006). Crystal structure of an Hsp90-nucleotide-p23/Sba1 closed chaperone complex. Nature 440, 1013–1017. Aller, P., Geng, T., Evans, G., and Foadi, J. (2016). Applications of the BLEND Software to Crystallographic Data from Membrane Proteins. Adv. Exp. Med. Biol. 922, 119–135. Anand, G.S., Law, D., Mandell, J.G., Snead, A.N., Tsigelny, I., Taylor, S.S., Eyck, Ten, L.F., and Komives, E.A. (2003). Identification of the protein kinase A regulatory RIalpha-catalytic subunit interface by amide H/2H exchange and protein docking. Proc Natl Acad Sci USA 100, 13264–13269. Andersen, C.B.F., Becker, T., Blau, M., Anand, M., Halic, M., Balar, B., Mielke, T., Boesen, T., Pedersen, J.S., Spahn, C.M.T., et al. (2006). Structure of eEF3 and the mechanism of transfer RNA release from the E-site. Nature 443, 663–668. Anthony, T.G., McDaniel, B.J., Byerley, R.L., McGrath, B.C., Cavener, D.R., McNurlan, M.A., and Wek, R.C. (2004). Preservation of Liver Protein Synthesis during Dietary Leucine Deprivation Occurs at the Expense of Skeletal Muscle Mass in Mice Deleted for eIF2 Kinase GCN2. J. Biol. Chem. 279, 36553–36561. B'chir, W., Maurin, A.-C., Carraro, V., Averous, J., Jousse, C., Muranishi, Y., Parry, L., Stepien, G., Fafournoux, P., and Bruhat, A. (2013). The eIF2α/ATF4 pathway is essential for stress-induced autophagy gene expression. Nucleic Acids Res 41, 7683–7699. Bae, J.H., Lew, E.D., Yuzawa, S., Tomé, F., Lax, I., and Schlessinger, J. (2009). The selectivity of receptor tyrosine kinase signaling is controlled by a secondary SH2 domain binding site. Cell 138, 514–524. References 200 Bai, X.-C., Rajendra, E., Yang, G., Shi, Y., and Scheres, S.H.W. (2015). Sampling the conformational space of the catalytic subunit of human γ-secretase. eLIFE 4, 1485. Bai, Y., Milne, J.S., Mayne, L., and Englander, S.W. (1993). Primary structure effects on peptide group hydrogen exchange. Proteins 17, 75–86. Bai, Y., Milne, J.S., Mayne, L., and Englander, S.W. (1994). Protein stability parameters measured by hydrogen exchange. Proteins 20, 4–14. Ban, N., Beckmann, R., Cate, J.H.D., Dinman, J.D., Dragon, F., Ellis, S.R., Lafontaine, D.L.J., Lindahl, L., Liljas, A., Lipton, J.M., et al. (2014). A new system for naming ribosomal proteins. Curr. Opin. Struct. Biol. 24, 165–169. Bar-Peled, L., Schweitzer, L.D., Zoncu, R., and Sabatini, D.M. (2012). Ragulator is a GEF for the rag GTPases that signal amino acid levels to mTORC1. Cell 150, 1196–1208. Bargis-Surgey, P., Lavergne, J.P., Gonzalo, P., Vard, C., Filhol-Cochet, O., and Reboud, J.P. (1999). Interaction of elongation factor eEF-2 with ribosomal P proteins. Eur. J. Biochem. 262, 606–611. Basso, A.D., Solit, D.B., Chiosis, G., Giri, B., Tsichlis, P., and Rosen, N. (2002). Akt forms an intracellular complex with heat shock protein 90 (Hsp90) and Cdc37 and is destabilized by inhibitors of Hsp90 function. J. Biol. Chem. 277, 39858–39866. Berger, A., Loewenstein, A., and Meiboom, S. (1959). Nuclear Magnetic Resonance Study of the Protolysis and Ionization of N-Methylacetamide 1. J. Am. Chem. Soc. 81, 62–67. Berger, A., and Linderstrøm-Lang, K. (1957). Deuterium exchange of poly-dl-alanine in aqueous solution. Archives of Biochemistry and Biophysics 69, 106–118. Berlanga, J.J., Santoyo, J., and De Haro, C. (1999). Characterization of a mammalian homolog of the GCN2 eukaryotic initiation factor 2alpha kinase. Eur. J. Biochem. 265, 754– 762. Bertolotti, A., Zhang, Y., Hendershot, L.M., Harding, H.P., and Ron, D. (2000). Dynamic interaction of BiP and ER stress transducers in the unfolded-protein response. Nat Cell Biol 2, 326–332. Brown, A., Fernández, I.S., Gordiyenko, Y., and Ramakrishnan, V. (2016). Ribosome- dependent activation of stringent control. Nature 534, 277–280. Brown, A., Shao, S., Murray, J., Hegde, R.S., and Ramakrishnan, V. (2015). Structural basis for stop codon recognition in eukaryotes. Nature 524, 493–496. Brugarolas, J., Lei, K., Hurley, R.L., Manning, B.D., Reiling, J.H., Hafen, E., Witters, L.A., Ellisen, L.W., and Kaelin, W.G. (2004). Regulation of mTOR function in response to hypoxia by REDD1 and the TSC1/TSC2 tumor suppressor complex. Genes Dev. 18, 2893–2904. Brugge, J.S. (1986). Interaction of the Rous sarcoma virus protein pp60src with the cellular proteins pp50 and pp90. Curr. Top. Microbiol. Immunol. 123, 1–22. References 201 Bukau, B., Weissman, J., and Horwich, A. (2006). Molecular chaperones and protein quality control. Cell 125, 443–451. Bunney, T.D., Inglis, A.J., Sanfelice, D., Farrell, B., Kerr, C.J., Thompson, G.S., Masson, G.R., Thiyagarajan, N., Svergun, D.I., Williams, R.L., et al. (2018). Disease Variants of FGFR3 Reveal Molecular Basis for the Recognition and Additional Roles for Cdc37 in Hsp90 Chaperone System. Structure 26, 446–458.e448. Burke, J.E., Inglis, A.J., Perisic, O., Masson, G.R., McLaughlin, S.H., Rutaganira, F., Shokat, K.M., and Williams, R.L. (2014). Structures of PI4KIIIβ complexes show simultaneous recruitment of Rab11 and its effectors. Science 344, 1035–1038. Cambiaghi, T.D., Pereira, C.M., Shanmugam, R., Bolech, M., Wek, R.C., Sattlegger, E., and Castilho, B.A. (2014). Evolutionarily conserved IMPACT impairs various stress responses that require GCN1 for activating the eIF2 kinase GCN2. Biochemical and Biophysical Research Communications 443, 592–597. Carrara, M., Prischi, F., Nowak, P.R., and Ali, M.M. (2015). Crystal structures reveal transient PERK luminal domain tetramerization in endoplasmic reticulum stress signaling. The EMBO Journal 34, 1589–1600. Chandramohan, A., Krishnamurthy, S., Larsson, A., Nordlund, P., Jansson, A., and Anand, G.S. (2016). Predicting Allosteric Effects from Orthosteric Binding in Hsp90-Ligand Interactions: Implications for Fragment-Based Drug Design. PLoS Comput Biol 12, e1004840. Chen, H., Marsiglia, W.M., Cho, M.-K., Huang, Z., Deng, J., Blais, S.P., Gai, W., Bhattacharya, S., Neubert, T.A., Traaseth, N.J., et al. (2017). Elucidation of a four-site allosteric network in fibroblast growth factor receptor tyrosine kinases. eLIFE 6, 983. Chesnokova, E., Bal, N., and Kolosov, P. (2017). Kinases of eIF2a Switch Translation of mRNA Subset during Neuronal Plasticity. Int J Mol Sci 18, 2213. Coales, S.J., E, S.Y., Lee, J.E., Ma, A., Morrow, J.A., and Hamuro, Y. (2010). Expansion of time window for mass spectrometric measurement of amide hydrogen/deuterium exchange reactions. Rapid Commun. Mass Spectrom. 24, 3585–3592. Costa-Mattioli, M., Gobert, D., Harding, H., Herdy, B., Azzi, M., Bruno, M., Bidinosti, M., Ben Mamou, C., Marcinkiewicz, E., Yoshida, M., et al. (2005). Translational control of hippocampal synaptic plasticity and memory by the eIF2alpha kinase GCN2. Nature 436, 1166–1173. Cui, W., Li, J., Ron, D., and Sha, B. (2011). The structure of the PERK kinase domain suggests the mechanism for its activation. Acta Crystallogr. D Biol. Crystallogr. 67, 423–428. Dai, K., Kobayashi, R., and Beach, D. (1996). Physical interaction of mammalian CDC37 with CDK4. J. Biol. Chem. 271, 22030–22034. Dar, A.C., Dever, T.E., and Sicheri, F. (2005). Higher-order substrate recognition of eIF2alpha by the RNA-dependent protein kinase PKR. Cell 122, 887–900. References 202 Dautant, A., Meyer, P., and Georgescauld, F. (2017). Hydrogen/Deuterium Exchange Mass Spectrometry Reveals Mechanistic Details of Activation of Nucleoside Diphosphate Kinases by Oligomerization. Biochemistry 56, 2886–2896. Delforge, J., Messenguy, F., and Wiame, J.M. (1975). The regulation of arginine biosynthesis in Saccharomyces cerevisiae. The specificity of argR- mutations and the general control of amino-acid biosynthesis. Eur. J. Biochem. 57, 231–239. Deng, J., Harding, H.P., Raught, B., Gingras, A.-C., Berlanga, J.J., Scheuner, D., Kaufman, R.J., Ron, D., and Sonenberg, N. (2002). Activation of GCN2 in UV-irradiated cells inhibits translation. Curr. Biol. 12, 1279–1286. Desmyter, A., Transue, T.R., Ghahroudi, M.A., Thi, M.H., Poortmans, F., Hamers, R., Muyldermans, S., and Wyns, L. (1996). Crystal structure of a camel single-domain VH antibody fragment in complex with lysozyme. Nat. Struct. Biol. 3, 803–811. Dey, M., Cao, C., Dar, A.C., Tamura, T., Ozato, K., Sicheri, F., and Dever, T.E. (2005a). Mechanistic link between PKR dimerization, autophosphorylation, and eIF2alpha substrate recognition. Cell 122, 901–913. Dey, M., Trieselmann, B., Locke, E.G., Lu, J., Cao, C., Dar, A.C., Krishnamoorthy, T., Dong, J., Sicheri, F., and Dever, T.E. (2005b). PKR and GCN2 kinases and guanine nucleotide exchange factor eukaryotic translation initiation factor 2B (eIF2B) recognize overlapping surfaces on eIF2alpha. Molecular and Cellular Biology 25, 3063–3075. Dhaliwal, S., and Hoffman, D.W. (2003). The crystal structure of the N-terminal region of the alpha subunit of translation initiation factor 2 (eIF2alpha) from Saccharomyces cerevisiae provides a view of the loop containing serine 51, the target of the eIF2alpha-specific kinases. J. Mol. Biol. 334, 187–195. Di Como, C.J., and Arndt, K.T. (1996). Nutrients, via the Tor proteins, stimulate the association of Tap42 with type 2A phosphatases. Genes Dev. 10, 1904–1916. Diallinas, G., and Thireos, G. (1994). Genetic and biochemical evidence for yeast GCN2 protein kinase polymerization. Gene 143, 21–27. Dieci, M.V., Arnedos, M., Andre, F., and Soria, J.C. (2013). Fibroblast growth factor receptor inhibitors as a cancer treatment: from a biologic rationale to medical perspectives. Cancer Discov 3, 264–279. Distler, U., Kuharev, J., Navarro, P., Levin, Y., Schild, H., and Tenzer, S. (2014). Drift time- specific collision energies enable deep-coverage data-independent acquisition proteomics. Nat. Methods 11, 167–170. Dong, J., Qiu, H., Garcia-Barrio, M., Anderson, J., and Hinnebusch, A.G. (2000). Uncharged tRNA Activates GCN2 by Displacing the Protein Kinase Moiety from a Bipartite tRNA- Binding Domain. Molecular Cell 6, 269–279. Donohoe, G.C., Arndt, J.R., and Valentine, S.J. (2015). Online deuterium hydrogen exchange and protein digestion coupled with ion mobility spectrometry and tandem mass spectrometry. Anal. Chem. 87, 5247–5254. References 203 Düvel, K., Santhanam, A., Garrett, S., Schneper, L., and Broach, J.R. (2003). Multiple roles of Tap42 in mediating rapamycin-induced transcriptional changes in yeast. Molecular Cell 11, 1467–1478. Eleftheriadis, T., Tsogka, K., Pissas, G., Antoniadi, G., Liakopoulos, V., and Stefanidis, I. (2016). Activation of general control nonderepressible 2 kinase protects human glomerular endothelial cells from harmful high-glucose-induced molecular pathways. Int Urol Nephrol 48, 1731–1739. Endicott, J.A., Noble, M.E.M., and Johnson, L.N. (2012). The structural basis for control of eukaryotic protein kinases. Annu. Rev. Biochem. 81, 587–613. Eyries, M., Montani, D., Girerd, B., Perret, C., Leroy, A., Lonjou, C., Chelghoum, N., Coulet, F., Bonnet, D., Dorfmüller, P., et al. (2014). EIF2AK4 mutations cause pulmonary veno- occlusive disease, a recessive form of pulmonary hypertension. Nat. Genet. 46, 65–69. Feng, Q., and Shao, S. (2017). In vitro reconstitution of translational arrest pathways. Methods. Fernandez-Leiro, R., and Scheres, S.H.W. (2016). Unravelling biological macromolecules with cryo-electron microscopy. Nature 537, 339–346. Fontana, J., Fulton, D., Chen, Y., Fairchild, T.A., McCabe, T.J., Fujita, N., Tsuruo, T., and Sessa, W.C. (2002). Domain mapping studies reveal that the M domain of hsp90 serves as a molecular scaffold to regulate Akt-dependent phosphorylation of endothelial nitric oxide synthase and NO release. Circ. Res. 90, 866–873. Fowler, M.L., McPhail, J.A., Jenkins, M.L., Masson, G.R., Rutaganira, F.U., Shokat, K.M., Williams, R.L., and Burke, J.E. (2016). Using hydrogen deuterium exchange mass spectrometry to engineer optimized constructs for crystallization of protein complexes: Case study of PI4KIIIβ with Rab11. Protein Sci. 25, 826–839. Garcia-Barrio, M., Dong, J., Cherkasova, V.A., Zhang, X., Zhang, F., Ufano, S., Lai, R., Qin, J., and Hinnebusch, A.G. (2002). Serine 577 Is Phosphorylated and Negatively Affects the tRNA Binding and eIF2α Kinase Activities of GCN2. J. Biol. Chem. 277, 30675–30683. Garcia-Barrio, M., Dong, J., Ufano, S., and Hinnebusch, A.G. (2000). Association of GCN1±GCN20 regulatory complex with the N-terminus of eIF2 kinase GCN2 is required for GCN2 activation. The EMBO Journal 19, 1887–1899. Gardner, B.M., and Walter, P. (2011). Unfolded proteins are Ire1-activating ligands that directly induce the unfolded protein response. Science 333, 1891–1894. Geller, O., and Lifshitz, C. (2005). A fast flow tube study of gas phase H/D exchange of multiply protonated ubiquitin. J Phys Chem A 109, 2217–2222. Gorrec, F. (2013). The current approach to initial crystallization screening of proteins is under-sampled. J Appl Crystallogr 46, 795–797. Gorrec, F. (2015). The MORPHEUS II protein crystallization screen. Acta Crystallogr F Struct Biol Commun 71, 831–837. References 204 Gorrec, F. (2016). Protein crystallization screens developed at the MRC Laboratory of Molecular Biology. Drug Discov. Today 21, 819–825. Gorrec, F., Palmer, C.M., Lebon, G., and Warne, T. (2011). Pi sampling: a methodical and flexible approach to initial macromolecular crystallization screening. Acta Crystallogr. D Biol. Crystallogr. 67, 463–470. Gotoh, N. (2008). Regulation of growth factor signaling by FRS2 family docking/scaffold adaptor proteins. Cancer Sci. 99, 1319–1325. Grammatikakis, N., Lin, J.H., Grammatikakis, A., Tsichlis, P.N., and Cochran, B.H. (1999). p50(cdc37) acting in concert with Hsp90 is required for Raf-1 function. Molecular and Cellular Biology 19, 1661–1672. Gromadski, K.B., Schümmer, T., Strømgaard, A., Knudsen, C.R., Kinzy, T.G., and Rodnina, M.V. (2007). Kinetics of the interactions between yeast elongation factors 1A and 1Balpha, guanine nucleotides, and aminoacyl-tRNA. J. Biol. Chem. 282, 35629–35637. Guo, F., and Cavener, D.R. (2007). The GCN2 eIF2α Kinase Regulates Fatty-Acid Homeostasis in the Liver during Deprivation of an Essential Amino Acid. Cell Metabolism 5, 103–114. Gurdon, J.B. (1986). Nuclear transplantation in eggs and oocytes. J. Cell Sci. Suppl. 4, 287– 318. Hagiwara, Y., Hirai, M., Nishiyama, K., Kanazawa, I., Ueda, T., Sakaki, Y., and Ito, T. (1997). Screening for imprinted genes by allelic message display: identification of a paternally expressed gene impact on mouse chromosome 18. Proc Natl Acad Sci USA 94, 9249–9254. Han, A.-P., Yu, C., Lu, L., Fujiwara, Y., Browne, C., Chin, G., Fleming, M., Leboulch, P., Orkin, S.H., and Chen, J.-J. (2001). Heme-regulated eIF2 kinase (HRI) is required for translational regulation and survival of erythroid precursors in iron deficiency. The EMBO Journal 20, 6909–6918. Hanks, S.K., Quinn, A.M., and Hunter, T. (1988). The protein kinase family: conserved features and deduced phylogeny of the catalytic domains. Science 241, 42–52. Hao, S., Sharp, J.W., Ross-Inta, C.M., McDaniel, B.J., Anthony, T.G., Wek, R.C., Cavener, D.R., McGrath, B.C., Rudell, J.B., Koehnle, T.J., et al. (2005). Uncharged tRNA and sensing of amino acid deficiency in mammalian piriform cortex. Science 307, 1776–1778. Hara, K., Yonezawa, K., Weng, Q.P., Kozlowski, M.T., Belham, C., and Avruch, J. (1998). Amino acid sufficiency and mTOR regulate p70 S6 kinase and eIF-4E BP1 through a common effector mechanism. J. Biol. Chem. 273, 14484–14494. Hara, K., Maruki, Y., Long, X., Yoshino, K.-I., Oshiro, N., Hidayat, S., Tokunaga, C., Avruch, J., and Yonezawa, K. (2002). Raptor, a binding partner of target of rapamycin (TOR), mediates TOR action. Cell 110, 177–189. Harding, H.P., Zhang, Y., and Ron, D. (1999). Protein translation and folding are coupled by References 205 an endoplasmic-reticulum-resident kinase. Nature 397, 271–274. He, H., Singh, I., Wek, S.A., Dey, S., Baird, T.D., Wek, R.C., and Georgiadis, M.M. (2014). Crystal Structures of GCN2 Protein Kinase C-terminal Domains Suggest Regulatory Differences in Yeast and Mammals. J. Biol. Chem. 289, 15023–15034. Hinnebusch, A.G. (1985). A Hierarchy of trans-acting Factors Modulates Translation of an Activator os Amino Acid Biosynthetic Genes in Saccharomyces cerevisiae. Molecular and Cellular Biology 5, 2349–2360. Hinnebusch, A.G. (2005). Translational Regulation of GCN4 and the General Amino Acid Control of Yeast. Annual Review of Microbiology 59, 407–450. Hinnebusch, A.G. (2014). The scanning mechanism of eukaryotic translation initiation. Annu. Rev. Biochem. 83, 779–812. Huang, Z., Chen, H., Blais, S., Neubert, T.A., Li, X., and Mohammadi, M. (2013). Structural mimicry of a-loop tyrosine phosphorylation by a pathogenic FGF receptor 3 mutation. Structure 21, 1889–1896. Hunter, S., Jones, P., Mitchell, A., Apweiler, R., Attwood, T.K., Bateman, A., Bernard, T., Binns, D., Bork, P., Burge, S., et al. (2012). InterPro in 2011: new developments in the family and domain prediction database. Nucleic Acids Res 40, D306–D312. Hunter, T., and Poon, R.Y. (1997). Cdc37: a protein kinase chaperone? Trends Cell Biol. 7, 157–161. Hvidt, A., and Linderstrøm-Lang, K. (1954). Exchange of hydrogen atoms in insulin with deuterium atoms in aqueous solutions. Biochimica Et Biophysica Acta 14, 574–575. Hvidt, A., and Nielsen, S.O. (1966). Hydrogen exchange in proteins. Adv. Protein Chem. 21, 287–386. Iacob, R.E., Murphy, J.P., and Engen, J.R. (2008). Ion mobility adds an additional dimension to mass spectrometric analysis of solution-phase hydrogen/deuterium exchange. Rapid Commun. Mass Spectrom. 22, 2898–2904. Imataka, H., Gradi, A., and Sonenberg, N. (1998). A newly identified N-terminal amino acid sequence of human eIF4G binds poly(A)-binding protein and functions in poly(A)-dependent translation. The EMBO Journal 17, 7480–7489. Inoki, K., Li, Y., Xu, T., and Guan, K.-L. (2003a). Rheb GTPase is a direct target of TSC2 GAP activity and regulates mTOR signaling. Genes Dev. 17, 1829–1834. Inoki, K., Li, Y., Zhu, T., Wu, J., and Guan, K.-L. (2002). TSC2 is phosphorylated and inhibited by Akt and suppresses mTOR signalling. Nat Cell Biol 4, 648–657. Inoki, K., Zhu, T., and Guan, K.-L. (2003b). TSC2 mediates cellular energy response to control cell growth and survival. Cell 115, 577–590. Ishimura, R., Nagy, G., Dotu, I., Chuang, J.H., and Ackerman, S.L. (2016). Activation of References 206 GCN2 kinase by ribosome stalling links translation elongation with translation initiation. eLIFE 1–22. Ishimura, R., Nagy, G., Dotu, I., Zhou, H., Yang, X.-L., Schimmel, P., Senju, S., Nishimura, Y., Chuang, J.H., and Ackerman, S.L. (2014). RNA function. Ribosome stalling induced by mutation of a CNS-specific tRNA causes neurodegeneration. Science 345, 455–459. Ito, T., Marintchev, A., and Wagner, G. (2004). Solution structure of human initiation factor eIF2alpha reveals homology to the elongation factor eEF1B. Structure 12, 1693–1704. Jaspart, A., Calmels, C., Cosnefroy, O., Bellecave, P., Pinson, P., Claverol, S., Guyonnet- Dupérat, V., Dartigues, B., Benleulmi, M.S., Mauro, E., et al. (2017). GCN2 phosphorylates HIV-1 integrase and decreases HIV-1 replication by limiting viral integration. Scientific Reports 7, 2283. Jia, M.H., Larossa, R.A., Lee, J.M., Rafalski, A., Derose, E., Gonye, G., and Xue, Z. (2000). Global expression profiling of yeast treated with an inhibitor of amino acid biosynthesis, sulfometuron methyl. Physiol. Genomics 3, 83–92. Jiang, H.-Y., and Wek, R.C. (2005). Phosphorylation of the alpha-subunit of the eukaryotic initiation factor-2 (eIF2alpha) reduces protein synthesis and enhances apoptosis in response to proteasome inhibition. J. Biol. Chem. 280, 14189–14202. Jiang, Y., and Broach, J.R. (1999). Tor proteins and protein phosphatase 2A reciprocally regulate Tap42 in controlling cell growth in yeast. The EMBO Journal 18, 2782–2792. Jiménez-Díaz, A., Remacha, M., Ballesta, J.P.G., and Berlanga, J.J. (2013). Phosphorylation of Initiation Factor eIF2 in Response to Stress Conditions Is Mediated by Acidic Ribosomal P1/P2 Proteins in Saccharomyces cerevisiae. PLoS ONE 8, e84219–11. Jin, Y., Zhen, Y., Haugsten, E.M., and Wiedlocha, A. (2011). The driver of malignancy in KG-1a leukemic cells, FGFR1OP2-FGFR1, encodes an HSP90 addicted oncoprotein. Cell. Signal. 23, 1758–1766. Karagöz, G.E., and Rüdiger, S.G.D. (2015). Hsp90 interaction with clients. Trends Biochem. Sci. 40, 117–125. Kastner, B., Fischer, N., Golas, M.M., Sander, B., Dube, P., Boehringer, D., Hartmuth, K., Deckert, J., Hauer, F., Wolf, E., et al. (2008). GraFix: sample preparation for single-particle electron cryomicroscopy. Nat. Methods 5, 53–55. Katoh, M. (2016). Therapeutics Targeting FGF Signaling Network in Human Diseases. Trends Pharmacol. Sci. 37, 1081–1096. Katta, V., and Chait, B.T. (1991). Conformational changes in proteins probed by hydrogen- exchange electrospray-ionization mass spectrometry. Rapid Commun. Mass Spectrom. 5, 214–217. Keil, M., Sonner, J.K., Lanz, T.V., Oezen, I., Bunse, T., Bittner, S., Meyer, H.V., Meuth, S.G., Wick, W., and Platten, M. (2016). General control non-derepressible 2 (GCN2) in T cells controls disease progression of autoimmune neuroinflammation. J. Neuroimmunol. 297, References 207 117–126. Keramisanou, D., Aboalroub, A., Zhang, Z., Liu, W., Marshall, D., Diviney, A., Larsen, R.W., Landgraf, R., and Gelis, I. (2016). Molecular Mechanism of Protein Kinase Recognition and Sorting by the Hsp90 Kinome-Specific Cochaperone Cdc37. Molecular Cell 62, 260–271. Kim, D.-H., Sarbassov, D.D., Ali, S.M., King, J.E., Latek, R.R., Erdjument-Bromage, H., Tempst, P., and Sabatini, D.M. (2002). mTOR interacts with raptor to form a nutrient- sensitive complex that signals to the cell growth machinery. Cell 110, 163–175. Kim, D.-H., Sarbassov, D.D., Ali, S.M., Latek, R.R., Guntur, K.V.P., Erdjument-Bromage, H., Tempst, P., and Sabatini, D.M. (2003). GβL, a positive regulator of the rapamycin- sensitive pathway required for the nutrient-sensitive interaction between raptor and mTOR. Molecular Cell 11, 895–904. Kim, Y.E., Hipp, M.S., Bracher, A., Hayer-Hartl, M., and Hartl, F.U. (2013). Molecular chaperone functions in protein folding and proteostasis. Annu. Rev. Biochem. 82, 323–355. Konermann, L., Pan, J., and Liu, Y.-H. (2011). Hydrogen exchange mass spectrometry for studying protein structure and dynamics. Chem Soc Rev 40, 1224–1234. Krukenberg, K.A., Street, T.O., Lavery, L.A., and Agard, D.A. (2011). Conformational dynamics of the molecular chaperone Hsp90. Q. Rev. Biophys. 44, 229–255. Kubota, H., Sakaki, Y., and Ito, T. (2000). GI Domain-mediated Association of the Eukaryotic Initiation Factor 2α Kinase GCN2 with Its Activator GCN1 Is Required for General Amino Acid Control in Budding Yeast. J. Biol. Chem. 275, 20243–20246. Kuhlbrandt, W. (2014). The Resolution Revolution. Science 343, 1443–1444. Kuhle, B., Eulig, N.K., and Ficner, R. (2015). Architecture of the eIF2B regulatory subcomplex and its implications for the regulation of guanine nucleotide exchange on eIF2. Nucleic Acids Res 43, 9994–10014. Laederich, M.B., Degnin, C.R., Lunstrum, G.P., Holden, P., and Horton, W.A. (2011). Fibroblast Growth Factor Receptor 3 (FGFR3) Is a Strong Heat Shock Protein 90 (Hsp90) Client. J. Biol. Chem. 286, 19597–19604. Lageix, S., Rothenburg, S., Dever, T.E., and Hinnebusch, A.G. (2014). Enhanced Interaction between Pseudokinase and Kinase Domains in Gcn2 stimulates eIF2α Phosphorylation in Starved Cells. PLoS Genet 10, e1004326. Lageix, S., Zhang, J., Rothenburg, S., and Hinnebusch, A.G. (2015). Interaction between the tRNA-Binding and C-Terminal Domains of Yeast Gcn2 Regulates Kinase Activity In Vivo. PLoS Genet 11, e1004991. Landgraf, R.R., Chalmers, M.J., and Griffin, P.R. (2012). Automated hydrogen/deuterium exchange electron transfer dissociation high resolution mass spectrometry measured at single- amide resolution. J. Am. Soc. Mass Spectrom. 23, 301–309. References 208 Lareau, L.F., Hite, D.H., Hogan, G.J., and Brown, P.O. (2014). Distinct stages of the translation elongation cycle revealed by sequencing ribosome-protected mRNA fragments. eLIFE 3, e01257. Lavictoire, S.J., Parolin, D.A.E., Klimowicz, A.C., Kelly, J.F., and Lorimer, I.A.J. (2003). Interaction of Hsp90 with the nascent form of the mutant epidermal growth factor receptor EGFRvIII. J. Biol. Chem. 278, 5292–5299. Lee, A.S.Y., Kranzusch, P.J., and Cate, J.H.D. (2015). eIF3 targets cell-proliferation messenger RNAs for translational activation or repression. Nature 522, 111–114. Lee, A.S., Kranzusch, P.J., Doudna, J.A., and Cate, J.H.D. (2016). eIF3d is an mRNA cap- binding protein that is required for specialized translation initiation. Nature 536, 96–99. Lee, K.M., Yusa, K., Chu, L.O., Yu, C.W.H., Oono, M., Miyoshi, T., Ito, K., Shaw, P.C., Wong, K.B., and Uchiumi, T. (2013). Solution structure of human P1*P2 heterodimer provides insights into the role of eukaryotic stalk in recruiting the ribosome-inactivating protein trichosanthin to the ribosome. Nucleic Acids Res 41, 8776–8787. Lee, K.-M., Yu, C.W.-H., Chan, D.S.-B., Chiu, T.Y.-H., Zhu, G., Sze, K.-H., Shaw, P.-C., and Wong, K.-B. (2010). Solution structure of the dimerization domain of ribosomal protein P2 provides insights for the structural organization of eukaryotic stalk. Nucleic Acids Res 38, 5206–5216. Lee, K.-M., Yu, C.W.-H., Chiu, T.Y.-H., Sze, K.-H., Shaw, P.-C., and Wong, K.-B. (2011). Solution structure of the dimerization domain of the eukaryotic stalk P1/P2 complex reveals the structural organization of eukaryotic stalk complex. Nucleic Acids Res 40, 3172–3182. Leib, D.E., and Knight, Z.A. (2015). Re-examination of Dietary Amino Acid Sensing Reveals a GCN2-Independent Mechanism. Cell Rep 13, 1081–1089. Li, X., Mooney, P., Zheng, S., Booth, C.R., Braunfeld, M.B., Gubbens, S., Agard, D.A., and Cheng, Y. (2013). Electron counting and beam-induced motion correction enable near- atomic-resolution single-particle cryo-EM. Nat. Methods 10, 584–590. Llácer, J.L., Hussain, T., Marler, L., Aitken, C.E., Thakur, A., Lorsch, J.R., Hinnebusch, A.G., and Ramakrishnan, V. (2015). Conformational Differences between Open and Closed States of the Eukaryotic Translation Initiation Complex. Molecular Cell 59, 399–412. Lodish, H.F., and Desalu, O. (1973). Regulation of synthesis of non-globin proteins in cell- free extracts of rabbit reticulocytes. J. Biol. Chem. 248, 3420–3427. Ma, T., Trinh, M.A., Wexler, A.J., Bourbon, C., Gatti, E., Pierre, P., Cavener, D.R., and Klann, E. (2013). Suppression of eIF2α kinases alleviates Alzheimer's disease-related plasticity and memory deficits. Nat. Neurosci. 16, 1299–1305. Ma, Y., Lu, Y., Zeng, H., Ron, D., Mo, W., and Neubert, T.A. (2001). Characterization of phosphopeptides from protein digests using matrix-assisted laser desorption/ionization time- of-flight mass spectrometry and nanoelectrospray quadrupole time-of-flight mass spectrometry. Rapid Commun. Mass Spectrom. 15, 1693–1700. References 209 Manning, B.D., Tee, A.R., Logsdon, M.N., Blenis, J., and Cantley, L.C. (2002). Identification of the tuberous sclerosis complex-2 tumor suppressor gene product tuberin as a target of the phosphoinositide 3-kinase/akt pathway. Molecular Cell 10, 151–162. Marciniak, S.J., Garcia-Bonilla, L., Hu, J., Harding, H.P., and Ron, D. (2006). Activation- dependent substrate recruitment by the eukaryotic translation initiation factor 2 kinase PERK. J Cell Biol 172, 201–209. Martin, F., Barends, S., Jaeger, S., Schaeffer, L., Prongidi-Fix, L., and Eriani, G. (2011). Cap- assisted internal initiation of translation of histone H4. Molecular Cell 41, 197–209. Marton, M.J., Vazquez de Aldana, C.R., Qiu, H., Chakraburtty, K., and Hinnebusch, A.G. (1997). Evidence that GCN1 and GCN20, Translational Regulators of GCN4, Function on Elongating Ribosomes in Activation of eIF2a Kinase GCN2. Molecular and Cellular Biology 17, 4474–4489. Maurin, A.-C., Chaveroux, C., Lambert-Langlais, S., Carraro, V., Jousse, C., Bruhat, A., Averous, J., Parry, L., Ron, D., Alliot, J., et al. (2012). The amino acid sensor GCN2 biases macronutrient selection during aging. Eur J Nutr 51, 119–126. Maurin, A.-C., Jousse, C., Averous, J., Parry, L., Bruhat, A., Cherasse, Y., Zeng, H., Zhang, Y., Harding, H.P., Ron, D., et al. (2005). The GCN2 kinase biases feeding behavior to maintain amino acid homeostasis in omnivores. Cell Metabolism 1, 273–277. Mayer, M.P. (2010). Gymnastics of molecular chaperones. Molecular Cell 39, 321–331. Mayer, M.P., and Le Breton, L. (2015). Hsp90: breaking the symmetry. Molecular Cell 58, 8– 20. McCoy, A.J., Grosse-Kunstleve, R.W., Adams, P.D., Winn, M.D., Storoni, L.C., and Read, R.J. (2007). Phaser crystallographic software. J Appl Crystallogr 40, 658–674. Mellor, H., and Proud, C.G. (1991). A synthetic peptide substrate for initiation factor-2 kinases. Biochemical and Biophysical Research Communications 178, 430–437. Merritt, E.A., Arakaki, T.L., Gillespie, J.R., Larson, E.T., Kelley, A., Mueller, N., Napuli, A.J., Kim, J., Zhang, L., Verlinde, C.L.M.J., et al. (2010). Crystal structures of trypanosomal histidyl-tRNA synthetase illuminate differences between eukaryotic and prokaryotic homologs. J. Mol. Biol. 397, 481–494. Milne, J.S., Mayne, L., Roder, H., Wand, A.J., and Englander, S.W. (1998). Determinants of protein hydrogen exchange studied in equine cytochrome c. Protein Sci. 7, 739–745. Miyata, Y., and Nishida, E. (2004). CK2 controls multiple protein kinases by phosphorylating a kinase-targeting molecular chaperone, Cdc37. Molecular and Cellular Biology 24, 4065– 4074. Miyata, Y., Nakamoto, H., and Neckers, L. (2013). The therapeutic target Hsp90 and cancer hallmarks. Curr. Pharm. Des. 19, 347–365. Mochizuki, M., Kitamyo, M., Miyoshi, T., Ito, K., and Uchiumi, T. (2012). Analysis of References 210 chimeric ribosomal stalk complexes from eukaryotic and bacterial sources: structural features responsible for specificity of translation factors. Genes Cells 17, 273–284. Molday, R.S., Englander, S.W., and Kallen, R.G. (1972). Primary structure effects on peptide group hydrogen exchange. Biochemistry 11, 150–158. Murchie, M.-J., and Leader, D.P. (1978). Codon-specific interaction of uncharged transfer- RNA with eukaryotic ribosomes. Biochimica Et Biophysica Acta 520, 233–236. Naganuma, T., Nomura, N., Yao, M., Mochizuki, M., Uchiumi, T., and Tanaka, I. (2010). Structural Basis for Translation Factor Recruitment to the Eukaryotic/Archaeal Ribosomes. J. Biol. Chem. 285, 4747–4756. Nakamura, A., and Kimura, H. (2017). A new role of GCN2 in the nucleolus. Biochemical and Biophysical Research Communications 485, 484–491. Nameki, N., Yoneyama, M., Koshiba, S., Tochio, N., Inoue, M., Seki, E., Matsuda, T., Tomo, Y., Harada, T., Saito, K., et al. (2004). Solution structure of the RWD domain of the mouse GCN2 protein. Protein Sci. 13, 2089–2100. Narasimhan, J., Staschke, K.A., and Wek, R.C. (2004). Dimerization Is Required for Activation of eIF2 Kinase Gcn2 in Response to Diverse Environmental Stress Conditions. J. Biol. Chem. 279, 22820–22832. Natarajan, K., Meyer, M.R., Jackson, B.M., Slade, D., Roberts, C., Hinnebusch, A.G., and Marton, M.J. (2001). Transcriptional profiling shows that Gcn4p is a master regulator of gene expression during amino acid starvation in yeast. Molecular and Cellular Biology 21, 4347– 4368. Natchiar, S.K., Myasnikov, A.G., Kratzat, H., Hazemann, I., and Klaholz, B.P. (2017). Visualization of chemical modifications in the human 80S ribosome structure. Nature 372, 1716. Naveau, M., Lazennec-Schurdevin, C., Panvert, M., Dubiez, E., Mechulam, Y., and Schmitt, E. (2013). Roles of yeast eIF2α and eIF2β subunits in the binding of the initiator methionyl- tRNA. Nucleic Acids Res 41, 1047–1057. Neckers, L., and Workman, P. (2012). Hsp90 molecular chaperone inhibitors: are we there yet? Clin. Cancer Res. 18, 64–76. Niederberger, P., Miozzari, G., and Hütter, R. (1981). Biological role of the general control of amino acid biosynthesis in Saccharomyces cerevisiae. Molecular and Cellular Biology 1, 584–593. Nika, J., Rippel, S., and Hannig, E.M. (2001). Biochemical analysis of the eIF2beta gamma complex reveals a structural function for eIF2alpha in catalyzed nucleotide exchange. J. Biol. Chem. 276, 1051–1056. Nomura, T., Nakano, K., Maki, Y., Naganuma, T., Nakashima, T., Tanaka, I., Kimura, M., Hachimori, A., and Uchiumi, T. (2006). In vitro reconstitution of the GTPase-associated centre of the archaebacterial ribosome: the functional features observed in a hybrid form with References 211 Escherichia coli 50S subunits. Biochem. J. 396, 565–571. Ogle, J.M., Carter, A.P., and Ramakrishnan, V. (2003). Insights into the decoding mechanism from recent ribosome structures. Trends Biochem. Sci. 28, 259–266. Ortega, L.G., McCotter, M.D., Henry, G.L., McCormack, S.J., Thomis, D.C., and Samuel, C.E. (1996). Mechanism of interferon action. Biochemical and genetic evidence for the intermolecular association of the RNA-dependent protein kinase PKR from human cells. Virology 215, 31–39. Otting, G., and Wüthrich, K. (2009). Heteronuclear filters in two-dimensional [1H, 1H]-NMR spectroscopy: combined use with isotope labelling for studies of macromolecular conformation and intermolecular interactions. Q. Rev. Biophys. 23, 39. Padyana, A.K., Qiu, H., Roll-Mecak, A., Hinnebusch, A.G., and Burley, S.K. (2005). Structural Basis for Autoinhibition and Mutational Activation of Eukaryotic Initiation Factor 2α Protein Kinase GCN2. J. Biol. Chem. 280, 29289–29299. Pan, J., Han, J., Borchers, C.H., and Konermann, L. (2009). Hydrogen/deuterium exchange mass spectrometry with top-down electron capture dissociation for characterizing structural transitions of a 17 kDa protein. J. Am. Chem. Soc. 131, 12801–12808. Pardon, E., Laeremans, T., Triest, S., Rasmussen, S.O.R.G.F., nig, A.W.O., Ruf, A., Muyldermans, S., Hol, W.G.J., Kobilka, B.K., and Steyaert, J. (2014). A general protocol for the generation of Nanobodies for structural biology. Nature Protocols 9, 674–693. Passmore, L.A., and Russo, C.J. (2016). Specimen Preparation for High-Resolution Cryo-EM. Meth. Enzymol. 579, 51–86. Patani, H., Bunney, T.D., Thiyagarajan, N., Norman, R.A., Ogg, D., Breed, J., Ashford, P., Potterton, A., Edwards, M., Williams, S.V., et al. (2016). Landscape of activating cancer mutations in FGFR kinases and their differential responses to inhibitors in clinical use. Oncotarget 7, 24252–24268. Pavitt, G.D., Ramaiah, K.V., Kimball, S.R., and Hinnebusch, A.G. (1998). eIF2 independently binds two distinct eIF2B subcomplexes that catalyze and regulate guanine- nucleotide exchange. Genes Dev. 12, 514–526. Pearl, L.H. (2005). Hsp90 and Cdc37 -- a chaperone cancer conspiracy. Curr. Opin. Genet. Dev. 15, 55–61. Pearl, L.H., and Prodromou, C. (2006). Structure and mechanism of the Hsp90 molecular chaperone machinery. Annu. Rev. Biochem. 75, 271–294. Pearl, L.H., Prodromou, C., and Workman, P. (2008). The Hsp90 molecular chaperone: an open and shut case for treatment. Biochem. J. 410, 439–453. Pelham, H.R., and Jackson, R.J. (1976). An efficient mRNA-dependent translation system from reticulocyte lysates. Eur. J. Biochem. 67, 247–256. Pereira, C.M., Sattlegger, E., Jiang, H.-Y., Longo, B.M., Jaqueta, C.B., Hinnebusch, A.G., References 212 Wek, R.C., Mello, L.E.A.M., and Castilho, B.A. (2005). IMPACT, a protein preferentially expressed in the mouse brain, binds GCN1 and inhibits GCN2 activation. J. Biol. Chem. 280, 28316–28323. Perkins, D.N., Pappin, D.J., Creasy, D.M., and Cottrell, J.S. (1999). Probability-based protein identification by searching sequence databases using mass spectrometry data. Electrophoresis 20, 3551–3567. Peters, K.G., Marie, J., Wilson, E., Ives, H.E., Escobedo, J., Del Rosario, M., Mirda, D., and Williams, L.T. (1992). Point mutation of an FGF receptor abolishes phosphatidylinositol turnover and Ca2+ flux but not mitogenesis. Nature 358, 678–681. Peterson, T.R., Laplante, M., Thoreen, C.C., Sancak, Y., Kang, S.A., Kuehl, W.M., Gray, N.S., and Sabatini, D.M. (2009). DEPTOR is an mTOR inhibitor frequently overexpressed in multiple myeloma cells and required for their survival. Cell 137, 873–886. Pisareva, V.P., Muslimov, I.A., Tcherepanov, A., and Pisarev, A.V. (2015). Characterization of Novel Ribosome-Associated Endoribonuclease SLFN14 from Rabbit Reticulocytes. Biochemistry 54, 3286–3301. Potter, C.J., Pedraza, L.G., and Xu, T. (2002). Akt regulates growth by directly phosphorylating Tsc2. Nat Cell Biol 4, 658–665. Preedy, V.R., and Garlick, P.J. (1986). The response of muscle protein synthesis to nutrient intake in postabsorptive rats: the role of insulin and amino acids. Biosci. Rep. 6, 177–183. Punjani, A., Rubinstein, J.L., Fleet, D.J., and Brubaker, M.A. (2017). cryoSPARC: algorithms for rapid unsupervised cryo-EM structure determination. Nat. Methods 14, 290–296. Qiu, H., Dong, J., Hu, C., Francklyn, C.S., and Hinnebusch, A.G. (2001). The tRNA-binding moiety in GCN2 contains a dimerization domain that interacts with the kinase domain and is required for tRNA binding and kinase activation. The EMBO Journal 20, 1425–1438. Qiu, H., Garcia-Barrio, M., and Hinnebusch, A.G. (1998). Dimerization by Translation Initiation Factor 2 Kinase GCN2 Is Mediated by Interactions in the C-Terminal Ribosome- Binding Region and the Protein Kinase Domain. Molecular and Cellular Biology 18, 2697– 2711. Ramirez, M., Wek, R.C., and Hinnebusch, A.G. (1991). Ribosome association of GCN2 protein kinase, a translational activator of the GCN4 gene of Saccharomyces cerevisiae. Molecular and Cellular Biology 11, 3027–3036. Rand, K.D., Jørgensen, T.J.D., Olsen, O.H., Persson, E., Jensen, O.N., Stennicke, H.R., and Andersen, M.D. (2006). Allosteric activation of coagulation factor VIIa visualized by hydrogen exchange. J. Biol. Chem. 281, 23018–23024. Rand, K.D., Pringle, S.D., Morris, M., Engen, J.R., and Brown, J.M. (2011). ETD in a traveling wave ion guide at tuned Z-spray ion source conditions allows for site-specific hydrogen/deuterium exchange measurements. J. Am. Soc. Mass Spectrom. 22, 1784–1793. Rand, K.D., Pringle, S.D., Murphy, J.P., Fadgen, K.E., Brown, J., and Engen, J.R. (2009a). References 213 Gas-phase hydrogen/deuterium exchange in a traveling wave ion guide for the examination of protein conformations. Anal. Chem. 81, 10019–10028. Rand, K.D., Zehl, M., Jensen, O.N., and Jørgensen, T.J.D. (2009b). Protein hydrogen exchange measured at single-residue resolution by electron transfer dissociation mass spectrometry. Anal. Chem. 81, 5577–5584. Ravindran, R., Khan, N., Nakaya, H.I., Li, S., Loebbermann, J., Maddur, M.S., Park, Y., Jones, D.P., Chappert, P., Davoust, J., et al. (2014). Vaccine activation of the nutrient sensor GCN2 in dendritic cells enhances antigen presentation. Science 343, 313–317. Ravindran, R., Loebbermann, J., Nakaya, H.I., Khan, N., Ma, H., Gama, L., Machiah, D.K., Lawson, B., Hakimpour, P., Wang, Y.-C., et al. (2016). The amino acid sensor GCN2 controls gut inflammation by inhibiting inflammasome activation. Nature 531, 523–527. Roe, S.M., Ali, M.M.U., Meyer, P., Vaughan, C.K., Panaretou, B., Piper, P.W., Prodromou, C., and Pearl, L.H. (2004). The Mechanism of Hsp90 regulation by the protein kinase-specific cochaperone p50(cdc37). Cell 116, 87–98. Roffé, M., Hajj, G.N.M., Azevedo, H.F., Alves, V.S., and Castilho, B.A. (2013). IMPACT Is a Developmentally Regulated Protein in Neurons That Opposes the Eukaryotic Initiation Factor 2α Kinase GCN2 in the modulation of Neurite Outgrowth. J. Biol. Chem. 288, 10860– 10869. Rojas, M., Arias, C.F., and López, S. (2010). Protein kinase R is responsible for the phosphorylation of eIF2alpha in rotavirus infection. J. Virol. 84, 10457–10466. Romano, P.R., Garcia-Barrio, M., Zhang, X., Wang, Q., Taylor, D.R., Zhang, F., Herring, C., Matthews, M.B., Qin, J., and Hinnebusch, A.G. (1998). Autophosphorylation in the Activation Loop Is Required for Full Kinase Activity In Vivo of Human and Yeast Eukaryotic Initiation Factor 2a kinases PKR and GCN2. Molecular and Cellular Biology 18, 2282–2297. Rosenthal, P.B., and Henderson, R. (2003). Optimal determination of particle orientation, absolute hand, and contrast loss in single-particle electron cryomicroscopy. J. Mol. Biol. 333, 721–745. Rostislavleva, K., Soler, N., Ohashi, Y., Zhang, L., Pardon, E., Burke, J.E., Masson, G.R., Johnson, C., Steyaert, J., Ktistakis, N.T., et al. (2015). Structure and flexibility of the endosomal Vps34 complex reveals the basis of its function on membranes. Science 350, aac7365. Ruff, M., Krishnaswamy, S., Boeglin, M., Poterszman, A., Mitschler, A., Podjarny, A., Rees, B., Thierry, J.C., and Moras, D. (1991). Class II aminoacyl transfer RNA synthetases: crystal structure of yeast aspartyl-tRNA synthetase complexed with tRNA(Asp). Science 252, 1682– 1689. Saibil, H. (2013). Chaperone machines for protein folding, unfolding and disaggregation. Nat. Rev. Mol. Cell Biol. 14, 630–642. Sancak, Y., Bar-Peled, L., Zoncu, R., Markhard, A.L., Nada, S., and Sabatini, D.M. (2010). Ragulator-Rag Complex Targets mTORC1 to the Lysosomal Surface and Is Necessary for Its References 214 Activation by Amino Acids. Cell 141, 290–303. Sancak, Y., Peterson, T.R., Shaul, Y.D., Lindquist, R.A., Thoreen, C.C., Bar-Peled, L., and Sabatini, D.M. (2008). The Rag GTPases bind raptor and mediate amino acid signaling to mTORC1. Science 320, 1496–1501. Sancak, Y., Thoreen, C.C., Peterson, T.R., Lindquist, R.A., Kang, S.A., Spooner, E., Carr, S.A., and Sabatini, D.M. (2007). PRAS40 is an insulin-regulated inhibitor of the mTORC1 protein kinase. Molecular Cell 25, 903–915. Sandbaken, M., Lupisella, J.A., DiDomenico, B., and Chakraburtty, K. (1990). Isolation and characterization of the structural gene encoding elongation factor 3. Biochimica Et Biophysica Acta 1050, 230–234. Sato, S., Fujita, N., and Tsuruo, T. (2000). Modulation of Akt kinase activity by binding to Hsp90. Proc Natl Acad Sci USA 97, 10832–10837. Sattlegger, E., and Hinnebusch, A.G. (2000). Separate domains in GCN1 for binding protein kinase GCN2 and ribosomes are required for GCN2 activation in amino acid-starved cells. The EMBO Journal 19, 6622–6633. Sattlegger, E., and Hinnebusch, A.G. (2005). Polyribosome Binding by GCN1 Is Required for Full Activation of Eukaryotic Translation Initiation Factor 2α Kinase GCN2 during Amino Acid Starvation. J. Biol. Chem. 280, 16514–16521. Sattlegger, E., Barbosa, J.A.R.G., Moraes, M.C.S., Martins, R.M., Hinnebusch, A.G., and Castilho, B.A. (2011). Gcn1 and Actin Binding to Yih1. The Journal of Biological Chemistry 286, 10341–10355. Sattlegger, E., Swanson, M.J., Ashcraft, E.A., Jennings, J.L., Fekete, R.A., Link, A.J., and Hinnebusch, A.G. (2004). YIH1 Is an Actin-binding Protein That Inhibits Protein Kinase GCN2 and Impairs General Amino Acid Control When Overexpressed. J. Biol. Chem. 279, 29952–29962. Scheres, S.H.W. (2012). RELION: implementation of a Bayesian approach to cryo-EM structure determination. Journal of Structural Biology 180, 519–530. Scheres, S.H.W. (2015). Semi-automated selection of cryo-EM particles in RELION-1.3. Journal of Structural Biology 189, 114–122. Schmitt, E., Blanquet, S., and Mechulam, Y. (2002). The large subunit of initiation factor aIF2 is a close structural homologue of elongation factors. The EMBO Journal 21, 1821– 1832. Schulte, T.W., Blagosklonny, M.V., Ingui, C., and Neckers, L. (1995). Disruption of the Raf- 1-Hsp90 molecular complex results in destabilization of Raf-1 and loss of Raf-1-Ras association. J. Biol. Chem. 270, 24585–24588. Shao, J., Irwin, A., Hartson, S.D., and Matts, R.L. (2003a). Functional dissection of cdc37: characterization of domain structure and amino acid residues critical for protein kinase binding. Biochemistry 42, 12577–12588. References 215 Shao, J., Prince, T., Hartson, S.D., and Matts, R.L. (2003b). Phosphorylation of serine 13 is required for the proper function of the Hsp90 co-chaperone, Cdc37. J. Biol. Chem. 278, 38117–38120. Shao, S., Malsburg, von der, K., and Hegde, R.S. (2013). Listerin-Dependent Nascent Protein Ubiquitination Relies on Ribosome Subunit Dissociation. Molecular Cell 50, 637–648. Shao, S., Murray, J., Brown, A., Taunton, J., Ramakrishnan, V., and Hegde, R.S. (2016). Decoding Mammalian Ribosome-mRNA States by Translational GTPase Complexes. Cell 167, 1229–1240.e15. Sharma, A., Mariappan, M., Appathurai, S., and Hegde, R.S. (2010). In Vitro Dissection of Protein Translocation into the Mammalian Endoplasmic Reticulum. In Protein Secretion, (Totowa, NJ: Humana Press), pp. 339–363. Shatkin, A.J. (1976). Capping of eucaryotic mRNAs. Cell 9, 645–653. Shi, Y., Vattem, K.M., Sood, R., An, J., Liang, J., Stramm, L., and Wek, R.C. (1998). Identification and Characterization of Pancreatic Eukaryotic Initiation Factor 2 a-Subunit Kinase, PEK, Involved in Translational Control. Molecular and Cellular Biology 18, 7499– 7509. Sievers, F., Wilm, A., Dineen, D., Gibson, T.J., Karplus, K., Li, W., Lopez, R., McWilliam, H., Remmert, M., Söding, J., et al. (2011). Fast, scalable generation of high-quality protein multiple sequence alignments using Clustal Omega. Mol. Syst. Biol. 7, 539–539. Silva, J.C., Gorenstein, M.V., Li, G.-Z., Vissers, J.P.C., and Geromanos, S.J. (2006). Absolute Quantification of Proteins by LCMS E. Molecular & Cellular Proteomics 5, 144–156. Simms, C.L., Yan, L.L., and Zaher, H.S. (2017). Ribosome Collision Is Critical for Quality Control during No-Go Decay. Molecular Cell 68, 361–373.e365. Skinner, J.J., Lim, W.K., Bédard, S., Black, B.E., and Englander, S.W. (2012). Protein dynamics viewed by hydrogen exchange. Protein Sci. 21, 996–1005. Skogerson, L., and Wakatama, E. (1976). A ribosome-dependent GTPase from yeast distinct from elongation factor 2. Proc Natl Acad Sci USA 73, 73–76. Smith, D.L., Deng, Y., and Zhang, Z. (1997). Probing the non-covalent structure of proteins by amide hydrogen exchange and mass spectrometry. J Mass Spectrom 32, 135–146. Smith, D.W., and McNamara, A.L. (1974). The distribution of transfer ribonucleic acid in rabbit reticulocytes. Levels of aminoacylation and ribosomal attachment during hemoglobin synthesis. J. Biol. Chem. 249, 1330–1334. Smith, E.M., Finn, S.G., Tee, A.R., Browne, G.J., and Proud, C.G. (2005). The tuberous sclerosis protein TSC2 is not required for the regulation of the mammalian target of rapamycin by amino acids and certain cellular stresses. J. Biol. Chem. 280, 18717–18727. Snijder, J., Borst, A.J., Dosey, A., Walls, A.C., Burrell, A., Reddy, V.S., Kollman, J.M., and Veesler, D. (2017). Vitrification after multiple rounds of sample application and blotting References 216 improves particle density on cryo-electron microscopy grids. Journal of Structural Biology 198, 38–42. Spraggon, G., Pantazatos, D., Klock, H.E., Wilson, I.A., Woods, V.L., and Lesley, S.A. (2004). On the use of DXMS to produce more crystallizable proteins: structures of the T. maritima proteins TM0160 and TM1171. Protein Sci. 13, 3187–3199. Stancato, L.F., Chow, Y.H., Hutchison, K.A., Perdew, G.H., Jove, R., and Pratt, W.B. (1993). Raf exists in a native heterocomplex with hsp90 and p50 that can be reconstituted in a cell- free system. J. Biol. Chem. 268, 21711–21716. Stark, H. (2010). GraFix: Stabilization of Fragile Macromolecular Complexes for Single Particle Cryo-EM. In Cryo-EM Part a Sample Preparation and Data Collection, (Elsevier), pp. 109–126. Stepanova, L., Leng, X., Parker, S.B., and Harper, J.W. (1996). Mammalian p50Cdc37 is a protein kinase-targeting subunit of Hsp90 that binds and stabilizes Cdk4. Genes Dev. 10, 1491–1502. Stock, D., Perisic, O., and Löwe, J. (2005). Robotic nanolitre protein crystallisation at the MRC Laboratory of Molecular Biology. Prog. Biophys. Mol. Biol. 88, 311–327. Suraweera, A., Münch, C., Hanssum, A., and Bertolotti, A. (2012). Failure of amino acid homeostasis causes cell death following proteasome inhibition. Molecular Cell 48, 242–253. Taipale, M., Krykbaeva, I., Koeva, M., Kayatekin, C., Westover, K.D., Karras, G.I., and Lindquist, S. (2012). Quantitative analysis of HSP90-client interactions reveals principles of substrate recognition. Cell 150, 987–1001. Tang, G., Peng, L., Baldwin, P.R., Mann, D.S., Jiang, W., Rees, I., and Ludtke, S.J. (2007). EMAN2: an extensible image processing suite for electron microscopy. Journal of Structural Biology 157, 38–46. Tang, Z., Yuan, X., Du, R., Cheung, S.-H., Zhang, G., Wei, J., Zhao, Y., Feng, Y., Peng, H., Zhang, Y., et al. (2015). BGB-283, a Novel RAF Kinase and EGFR Inhibitor, Displays Potent Antitumor Activity in BRAF-Mutated Colorectal Cancers. Mol. Cancer Ther. 14, 2187–2197. Tarun, S.Z., and Sachs, A.B. (1996). Association of the yeast poly(A) tail binding protein with translation initiation factor eIF-4G. The EMBO Journal 15, 7168–7177. Taylor, S.S., Keshwani, M.M., Steichen, J.M., and Kornev, A.P. (2012). Evolution of the eukaryotic protein kinases as dynamic molecular switches. Philos. Trans. R. Soc. Lond., B, Biol. Sci. 367, 2517–2528. Thomis, D.C., and Samuel, C.E. (1995). Mechanism of interferon action: characterization of the intermolecular autophosphorylation of PKR, the interferon-inducible, RNA-dependent protein kinase. J. Virol. 69, 5195–5198. Tong, H., Hateboer, G., Perrakis, A., Bernards, R., and Sixma, T.K. (1997). Crystal structure of murine/human Ubc9 provides insight into the variability of the ubiquitin-conjugating system. J. Biol. Chem. 272, 21381–21387. References 217 Triana-Alonso, F.J., Chakraburtty, K., and Nierhaus, K.H. (1995). The elongation factor 3 unique in higher fungi and essential for protein biosynthesis is an E site factor. J. Biol. Chem. 270, 20473–20478. Tsai, J., Lee, J.T., Wang, W., Zhang, J., Cho, H., Mamo, S., Bremer, R., Gillette, S., Kong, J., Haass, N.K., et al. (2008). Discovery of a selective inhibitor of oncogenic B-Raf kinase with potent antimelanoma activity. Proc. Natl. Acad. Sci. U.S.a. 105, 3041–3046. Turner, N., and Grose, R. (2010). Fibroblast growth factor signalling: from development to cancer. Nat. Rev. Cancer 10, 116–129. Tyzack, J.K., Wang, X., Belsham, G.J., and Proud, C.G. (2000). ABC50 Interacts with Eukaryotic Initiation Factor 2 and Associates with the Ribosome in an ATP-dependent Manner. J. Biol. Chem. 275, 34131–34139. Tzamarias, D., Roussou, I., and Thireos, G. (1989). Coupling of GCN4 mRNA translational activation with decreased rates of polypeptide chain initiation. Cell 57, 947–954. Ung, T.L., Cao, C., Lu, J., Ozato, K., and Dever, T.E. (2001). Heterologous dimerization domains functionally substitute for the double-stranded RNA binding domains of the kinase PKR. The EMBO Journal 20, 3728–3737. Vadas, O., Jenkins, M.L., Dornan, G.L., and Burke, J.E. (2017). Using Hydrogen-Deuterium Exchange Mass Spectrometry to Examine Protein-Membrane Interactions. Meth. Enzymol. 583, 143–172. Van de Velde, L.-A., Guo, X.-Z.J., Barbaric, L., Smith, A.M., Oguin, T.H., Thomas, P.G., and Murray, P.J. (2016). Stress Kinase GCN2 Controls the Proliferative Fitness and Trafficking of Cytotoxic T Cells Independent of Environmental Amino Acid Sensing. Cell Rep 17, 2247– 2258. Vasudevan, D., Clark, N.K., Sam, J., Cotham, V.C., Ueberheide, B., Marr, M.T., and Ryoo, H.D. (2017). The GCN2-ATF4 Signaling Pathway Induces 4E-BP to Bias Translation and Boost Antimicrobial Peptide Synthesis in Response to Bacterial Infection. Cell Rep 21, 2039– 2047. Vattem, K.M., and Wek, R.C. (2004). Reinitiation involving upstream ORFs regulates ATF4 mRNA translation in mammalian cells. Proc Natl Acad Sci USA 101, 11269–11274. Vazquez de Aldana, C.R., Marton, M.J., and Hinnebusch, A.G. (1995). GCN20, a novel ATP binding cassette protein, and GCN1 reside in a complex that mediates activation of the eIF2a kinase GCN2 in amino acid-starved cells. The EMBO Journal 14, 3184–3199. Velankar, S., McNeil, P., Mittard-Runte, V., Suarez, A., Barrell, D., Apweiler, R., and Henrick, K. (2005). E-MSD: an integrated data resource for bioinformatics. Nucleic Acids Res 33, D262–D265. Verba, K.A., Wang, R.Y.-R., Arakawa, A., Liu, Y., Shirouzu, M., Yokoyama, S., and Agard, D.A. (2016). Atomic structure of Hsp90-Cdc37-Cdk4 reveals that Hsp90 traps and stabilizes an unfolded kinase. Science 352, 1542–1547. References 218 Visweswaraiah, J., Lageix, S., Castilho, B.A., Izotova, L., Kinzy, T.G., Hinnebusch, A.G., and Sattlegger, E. (2011). Evidence that eukaryotic translation elongation factor 1A (eEF1A) binds the Gcn2 protein C terminus and inhibits Gcn2 activity. J. Biol. Chem. 286, 36568– 36579. Visweswaraiah, J., Lee, S.J., Hinnebusch, A.G., and Sattlegger, E. (2012). Overexpression of eukaryotic translation elongation factor 3 impairs Gcn2 protein activation. J. Biol. Chem. 287, 37757–37768. Voorhees, R.M., Fernández, I.S., Scheres, S.H.W., and Hegde, R.S. (2014). Structure of the mammalian ribosome-Sec61 complex to 3.4 Å resolution. Cell 157, 1632–1643. Wales, T.E., and Engen, J.R. (2006). Hydrogen exchange mass spectrometry for the analysis of protein dynamics. Mass Spectrom Rev 25, 158–170. Waller, T., Lee, S.J., and Sattlegger, E. (2012). Evidence that Yih1 resides in a complex with ribosomes. FEBS Journal 279, 1761–1776. Wang, X., Campbell, L.E., Miller, C.M., and Proud, C.G. (1998). Amino acid availability regulates p70 S6 kinase and multiple translation factors. Biochem. J. 334, 261–267. Wek, R.C., Ramirez, M., Jackson, B.M., and Hinnebusch, A.G. (1990). Identification of positive-acting domains in GCN2 protein kinase required for translational activation of GCN4 expression. Molecular and Cellular Biology 10, 2820–2831. Wek, R.C., Jackson, B.M., and Hinnebusch, A.G. (1989). Juxtaposition of domains homologous to protein kinases and histidyl-tRNA synthetases in GCN2 protein suggests a mechanism for coupling GCN4 expression to amino acid availability. Proc Natl Acad Sci USA 86, 4579–4583. Wek, S.A., Zhu, S., and Wek, R.C. (1995). The Histidyl-tRNA Synthetase-Related Sequence in the eIF-2 Protein Kinase GCN2 Interacts with tRNA and Is Required for Activation in Response to Starvation for Different Amino Acids. Molecular and Cellular Biology 15, 4497– 4506. Whitesell, L., and Lindquist, S.L. (2005). HSP90 and the chaperoning of cancer. Nat. Rev. Cancer 5, 761–772. Wiedemann, M., and Trueb, B. (2000). Characterization of a novel protein (FGFRL1) from human cartilage related to FGF receptors. Genomics 69, 275–279. Winger, B.E., Light-Wahl, K.J., Rockwood, A.L., and Smith, R.D. (1992). Probing qualitative conformation differences of multiply protonated gas-phase proteins via hydrogen/deuterium isotopic exchange with water-d2. J. Am. Chem. Soc. 114, 5897–5898. Winkler, J., Tyedmers, J., Bukau, B., and Mogk, A. (2012). Chaperone networks in protein disaggregation and prion propagation. Journal of Structural Biology 179, 152–160. Winn, M.D., Ballard, C.C., Cowtan, K.D., Dodson, E.J., Emsley, P., Evans, P.R., Keegan, R.M., Krissinel, E.B., Leslie, A.G.W., McCoy, A., et al. (2011). Overview of the CCP4 suite and current developments. Acta Crystallogr. D Biol. Crystallogr. 67, 235–242. References 219 Winn, M.D., Murshudov, G.N., and Papiz, M.Z. (2003). Macromolecular TLS refinement in REFMAC at moderate resolutions. Meth. Enzymol. 374, 300–321. Winter, G., Lobley, C.M.C., and Prince, S.M. (2013). Decision making in xia2. Acta Crystallogr. D Biol. Crystallogr. 69, 1260–1273. Wolfner, M., Yep, D., Messenguy, F., and Fink, G.R. (1975). Integration of amino acid biosynthesis into the cell cycle of Saccharomyces cerevisiae. J. Mol. Biol. 96, 273–290. Wu, S., and Kaufman, R.J. (1997). A model for the double-stranded RNA (dsRNA)- dependent dimerization and activation of the dsRNA-activated protein kinase PKR. J. Biol. Chem. 272, 1291–1296. Xiao, F., Huang, Z., Li, H., Yu, J., Wang, C., Chen, S., Meng, Q., Cheng, Y., Gao, X., Li, J., et al. (2011). Leucine deprivation increases hepatic insulin sensitivity via GCN2/mTOR/S6K1 and AMPK pathways. Diabetes 60, 746–756. Xiao, H., Kaltashov, I.A., and Eyles, S.J. (2003). Indirect assessment of small hydrophobic ligand binding to a model protein using a combination of ESI MS and HDX/ESI MS. J. Am. Soc. Mass Spectrom. 14, 506–515. Xu, X., Hu, J., McGrath, B.C., and Cavener, D.R. (2013). GCN2 in the brain programs PPARγ2 and triglyceride storage in the liver during perinatal development in response to maternal dietary fat. PLoS ONE 8, e75917. Yamagata, K., Sanders, L.K., Kaufmann, W.E., Yee, W., Barnes, C.A., Nathans, D., and Worley, P.F. (1994). Rheb, a growth factor- and synaptic activity-regulated gene, encodes a novel Ras-related protein. J. Biol. Chem. 269, 16333–16339. Yan, G., Shen, X., and Jiang, Y. (2006). Rapamycin activates Tap42-associated phosphatases by abrogating their association with Tor complex 1. The EMBO Journal 25, 3546–3555. Yang, J., and Zhang, Y. (2015). Protein Structure and Function Prediction Using I-TASSER. Curr Protoc Bioinformatics 52, 5.8.1–.8.15. Yang, R., Wek, S.A., and Wek, R.C. (2000). Glucose limitation induces GCN4 translation by activation of Gcn2 protein kinase. Molecular and Cellular Biology 20, 2706–2717. Yatime, L., Mechulam, Y., Blanquet, S., and Schmitt, E. (2007). Structure of an archaeal heterotrimeric initiation factor 2 reveals a nucleotide state between the GTP and the GDP states. Proc. Natl. Acad. Sci. U.S.a. 104, 18445–18450. Ye, J., Kumanova, M., Hart, L.S., Sloane, K., Zhang, H., De Panis, D.N., Bobrovnikova- Marjon, E., Diehl, J.A., Ron, D., and Koumenis, C. (2010). The GCN2-ATF4 pathway is critical for tumour cell survival and proliferation in response to nutrient deprivation. The EMBO Journal 29, 2082–2096. Zaborske, J.M., Narasimhan, J., Jiang, L., Wek, S.A., Dittmar, K.A., Freimoser, F., Pan, T., and Wek, R.C. (2009). Genome-wide analysis of tRNA charging and activation of the eIF2 kinase Gcn2p. J. Biol. Chem. 284, 25254–25267. References 220 Zhang, K. (2016). Gctf: Real-time CTF determination and correction. Journal of Structural Biology 193, 1–12. Zhang, P., McGrath, B.C., Reinert, J., Olsen, D.S., Lei, L., Gill, S., Wek, S.A., Vattem, K.M., Wek, R.C., Kimball, S.R., et al. (2002). The GCN2 eIF2alpha kinase is required for adaptation to amino acid deprivation in mice. Molecular and Cellular Biology 22, 6681–6688. Zhang, Z., and Smith, D.L. (2008). Determination of amide hydrogen exchange by mass spectrometry: A new tool for protein structure elucidation. Protein Sci. 2, 522–531. Zheng, Y., and Jiang, Y. (2005). The yeast phosphotyrosyl phosphatase activator is part of the Tap42-phosphatase complexes. Mol. Biol. Cell 16, 2119–2127. Zhu, S., and Wek, R.C. (1998). Ribosome-binding Domain of Eukaryotic Initiation Factor-2 Kinase GCN2 Facilitates Translation Control. The Journal of Biological Chemistry 273, 1808–1814. Zhu, S., Sobolev, A.Y., and Wek, R.C. (1996). Histidyl-tRNA Synthetase-related Sequences in GCN2 Protein Kinase Regulate in Vitro Phosphorylation of eIF-2*. The Journal of Biological Chemistry 271, 24989–24994. Zoncu, R., Bar-Peled, L., Efeyan, A., Wang, S., Sancak, Y., and Sabatini, D.M. (2011). mTORC1 senses lysosomal amino acids through an inside-out mechanism that requires the vacuolar H(+)-ATPase. Science 334, 678–683. Supplementary Materials Supplementary Figure 1. Sequence alignment for GCN2 from mouse, human, S. pombe, D. melanogaster, C. elegans, M. bruma, C. diphtheriae, S. japonicus and S. cerevisiae. Alignment was performed using Clustal Omega (Sievers et al., 2011). The sequence is annotated with information in the literature. Supplementary Table 1. A table of the proteins that make up the rabbit ribosome. In the cases where the rabbit sequence is not available, the human sequence is given. Name Uniprot Reference Protein Sequence uS2 XP_002716404.1 MSGALDVLQMKEEDVLKFLAAGTHLGGTNLDFQMEQYIYKRKSDGIYI INLKRTWEKLLLAARAIVAIENPADVSVISSRNTGQRAVLKFAAATGAT PIAGRFTPGTFTNQIQTAFREPRLLVVTDPRADHQPLTEASYVNLPTIAL CNTDSPLRYVDIAIPCNNKGAHSVGLMWWMLAREVLRMRGTISREHP WEVMPDLYFYRDPEEIEKEEQAAAEKAVTKEEFQGEWTAPAPEFTATQ PEVADWSEGVQVPSVPIQQFPTEDWSAQPATEDWSAAPTAQATEWVG TTTEWS eS1 XP_002711475.1 MAVGKNKRLTKGGKKGAKKKVVDPFSKKDWYDVKAPAMFNIRNIGK TLVTRTQGTKIASDGLKGRVFEVSLADLQNDEVAFRKFKLITEDVQGK NCLTNFHGMDLTRDKMCSMVKKWQTMIEAHVDVKTTDGYLLRLFCV GFTKKRNNQIRKTSYAQHQQVRQIRKKMMEIMTREVQTNDLKEVVNK LIPDSIGKDIEKACQSIYPLHDVFVRKVKMLKKPKFELGKLMELHGEGS SSGKATGDETGAKVERADGYEPPVQESV uS5 XP_013847858.1 MADDAGAAGGPGDPGGPGIGGRGGFRGGFGSGVRGRGRGRGRG RGRGRGARGGKAEDKEWLPVTKLGRLVKDMKIKSLEEIYLFSLPI KESEIIDFFLGASLKDEVLKIMPVQKQTRAGQRTRFKAFVAIGDYN GHVGLGVKCSKEVATAIRGAIILAKLSIVPVRRGYWGNKIGKPHT VPCKVTGRCGSVLVRLIPAPRGTGIVSAPVPKKLLMMAGIDDCYT SARGCTATLGNFAKATFDAISKTYSYLTPDLWKETVFTKSPYQEF TDHLVKTHTRVSVQRTQAPAVATT uS3 XP_002708735.1 MAVQISKKRKFVADGIFKAELNEFLTRELAEDGYSGVEVRVTPTRTEIII LATRTQNVLGEKGRRIRELTAVVQKRFGFPEGSVELYAEKVATRGLCAI AQAESLRYKLLGGLAVRRACYGVLRFIMESGAKGCEVVVSGKLRGQR AKSMKFVDGLMIHSGDPVNYYVDTAVRHVLLRQGVLGIKVKIMLPWD PSGKIGPKKPLPDHVSIVEPKDEILPTTPISEQKGGKPEPPAMPQPVPTA eS4 NP_000998.1 MARGPKKHLKRVAAPKHWMLDKLTGVFAPRPSTGPHKLRECLPLIIFL RNRLKYALTGDEVKKICMQRFIKIDGKVRTDITYPAGFMDVISIDKTGE NFRLIYDTKGRFAVHRITPEEAKYKLCKVRKIFVGTKGIPHLVTHDARTI RYPDPLIKVNDTIQIDLETGKITDFIKFDTGNLCMVTGGANLGRIGVITN RERHPGSFDVVHVKDANGNSFATRLSNIFVIGKGNKPWISLPRGKGIRL TIAEERDKRLAAKQSSG uS7 XP_002721942.1 MTEWETAAPAVAETPDIKLFGKWSTDDVQINDISLQDYIAVKEKYAKY LPHSAGRYAAKRFRKAQCPIVERLTNSMMMHGRNNGKKLMTVRIVKH AFEIIHLLTGENPLQVLVNAIINSGPREDSTRIGRAGTVRRQAVDVSPLR RVNQAIWLLCTGAREAAFRNIKTIAECLADELINAAKGSSNSYAIKKKD ELERVAKSNR eS6 XP_002708126.1 MKLNISFPATGCQKLIEVDDERKLRTFYEKRMATEVAADALGEEWKG YVVRISGGNDKQGFPMKQGVLTHGRVRLLLSKGHSCYRPRRTGERKR KSVRGCIVDANLSVLNLVIVKKGEKDIPGLTDTTVPRRLGPKRASRIRK LFNLSKEDDVRQYVVRKPLNKEGKKPRTKAPKIQRLVTPRVLQHKRRR IALKKQRTKKNKEEAAEYAKLLAKRMKEAKEKRQEQIAKRRRLSSLRA STSKSESSQK eS7 NP_001272763.1 MFSSSAKIVKPNGEKPDEFESGISQALLELEMNSDLKAQLRELNITAAKE IEVGGGRKAIIIFVPVPQLKSFQKIQVRLVRELEKKFSGKHVVFIAQRRIL PKPTRKSRTKNKQKRPRSRTLTAVHDAILEDLVFPSEIVGKRIRVKLDGS RLIKVHLDKAQQNNVEHKVETFSGVYKKLTGKDVNFEFPEFQL eS8 XP_002708870.1 MGISRDNWHKRRKTGGKRKPYHKKRKYELGRPAANTKIGPRRIHTVG VRGGNKKYRALRLDVGNFSWGSECCTRKTRIIDVVYNASNNELVRTKT LVKNCIVLIDSTPYRQWYESHYALPLGRKKGAKLTPEEEEILNKKRSKK IQKKYDERKKNAKISSLLEEQFQQGKLLACIASRPGQCGRADGYVLEG KELEFYLRKIKARKGK uS4 NP_001164865.1 MPVARSWVCRKTYVTPRRPFEKSRLDQELKLIGEYGLRNKREVWRVK FTLAKIRKAARELLTLDEKDPRRLFEGNALLRRLVRIGVLDEGKMKLD YILGLKIEDFLERRLQTQVFKLGLAKSIHHARVLIRQRHIRVRKQVVNIP SFIVRLDSQKHIDFSLRSPYGGGRPGRVKRKNAKKGQGGAGAGDDEEE D eS10 XP_002714657.1 MLMPKKNRIAIYELLFKEGVMVAKKDVHMPKHPELADKNVPNLHVM KAMQSLKSRGYVKEQFAWRHFYWYLTNEGIQYLRDYLHLPPEIVPATL RRSRPETGRPRPKGLEGERPARLTRGEADRDTYRRSAVPPGADKKA EAGAGSATEFQFRGGFGRGRGQPPQ uS17 XP_002709911.1 MADIQTERAYQKQPTIFQNKKTVLLGETGKEKLPRYYKNIGLGFKTPKE AIEGTYIDKKCPFTGNVSIRGRILSGVVTKMKMQRTIVIRRDYLHYIRKY NRFEKRHKNMSVHLSPCFRDVQIGDIVTVGECRPLSKTVRFNVLKVTK AAGTKKQFQKF eS12 XP_002714875.1 MAEEGIAAGGVMDVNTALQEVLKTALIHDGLARGIREAAKALDKRQA HLCVLASNCDEPMYVKLVEALCAEHQINLIKVDDNKKLGEWVGLCKI DREGKPRKVVGCSCVVVKDYGKESQAKDVIEEYFKCKK uS15 XP_002721403.1 MGRMHAPGKGLSQSALPYRRSVPTWLKLTSDDVKEQIYKLAKKGLTP SQIGVILRDSHGVAQVRFVTGNKILRILKSKGLAPDLPEDLYHLIKKAVA VRKHLERNRKDKDAKFRLILIESRIHRLARYYKTKRVLPPNWKYESSTA SALVA uS11 XP_002710354.1 MAPRKGKEKKEEQVISLGPQVAEGENVFGVCHIFASFNDTFVHVTDLS GKETICRVTGGMKVKADRDESSPYAAMLAAQDVAQRCKELGITALHIK LRATGGNRTKTPGPGAQSALRALARSGMKIGRIEDVTPIPSDSTRRKGG RRGRRL uS19 XP_008259356.1 MAEVEQKKKRTFRKFTYRGVDLDQLLDMSYEQLMQLYSARQRRRLSR GLRRKQHSLLKRLRKAKKEAPPMEKPEVVKTHLRDMIILPEMVGSMV GVYNGKTFNQVEIKPEMIGHYLGEFSITYKPVKHGRPGIGATHSSRFIPL K uS9 XP_002711497.1 MPSKGPLQSVQVFGRKKTATAVAHCKRGNGLIKVNGRPLEMIEPRTLQ YKLLEPVLLLGKERFAGVDIRVRVKGGGHVAQIYAIRQSISKALVAYYQ KYVDEASKKEIKDILIQYDRTLLVADPRRCESKKFGGPGARARYQKSY R eS17 XP_002713874.1 MGRVRTKTVKKAARVIIEKYYTRLGNDFHTNKRVCEEIAIIPSKKLRNKI AGYVTHLMKRIQRGPVRGISIKLQEEERERRDNYVPEVSALDQEIIEVDP DTKEMLKLLDFGSLSNLQVTQPTVGMNFKTPRGAV uS13 XP_002714578.1 MSLVIPEKFQHILRVLNTNIDGRRKIAFAITAIKGVGRRYAHVVLRKADI DLTKRAGELTEDEVERVITIMQNPRQYKIPDWFLNRQKDVKDGKYSQV LANGLDNKLREDLERLKKIRAHRGLRHFWGLRVRGQHTKTTGRRGRT VGVSKKK eS19 XP_002716238.1 MPGVTVKDVNQQEFVRALAAFLKKSGKLKVPEWVDTVKLAKHKELA PYDENWFYTRAASTARHLYLRGGAGVGSMTKIYGGRQRNGVMPSHFS RGSKSVARRVLQALEGLKMVEKDQDGGRKLTPQGQRDLDRIAGQVAA AKKKH uS10 NP_001240663.1 MAFKDTGKTPVEPEVAIHRIRITLTSRNVKSLEKVCADLIRGAKEKNLK VKGPVRMPTKTLRITTRKTPCGEGSKTWDRFQMRIHKRLIDLHSPSEIV KQITSISIEPGVEVEVTIADA eS21 NP_001015.1 MQNDAGEFVDLYVPRKCSASNRIIGAKDHASIQMNVAEVDKVTGRFN GQFKTYAICGAIRRMGESDDSILRLAKADGIVSKNF uS8 XP_008256009.1 MVRMNVLADALKSINNAEKRGKRQVLIRPCSKVIVRFLTVMMKHGYI GEFEIIDDHRAGKIVVNLTGRLNKCGVISPRFDVQLKDLEKWQNNLLPS RQFGFIVLTTSAGIMDHEEARRKHTGGKILGFFF uS12 XP_002713980.1 MGKCRGLRTARKLRSHRRDQKWHDKQYKKAHLGTALKANPFGGASH AKGIVLEKVGVEAKQPNSAIRKCVRVQLIKNGKKITAFVPNDGCLNFIE ENDEVLVAGFGRKGHAVGDIPGVRFKVVKVANVSLLALYKGKKERPR S eS24 XP_008268135.1 MNDTVTIRTRKFMTNRLLQRKQMVIDVLHPGKATVPKTEIREKLAKMY KTTPDVIFVFGFRTHFGGGKTTGFGMIYDSLDYAKKNEPKHRLARHGL YEKKKTSRKQRKERKNRMKKVRGTAKANVGAGKK eS25 XP_002722751.1 MPPKDDKKKKDAGKSAKKDKDPVNKSGGKAKKKKWSKGKVRDKLN NLVLFDKATYDKLCKEVPNYKLITPAVVSERLKIRGSLARAALQELLSK GLIKLVSKHRAQVIYTRNTKGGDAPAAGEDA eS26 XP_008254857.1 MTKKRRNNGRAKKGRGHVQPIRCTNCARCVPKDKAIKKFVIRNIVEAA AVRDISEASVFDAYVLPKLYVKLHYCVSCAIHSKVVRNRSREARKDRT PPPRFRPAGAAPRPPPKPM eS27 XP_002708254.1 MPLAKDLLHPSPEEEKRKHKKKRLVQSPNSYFMDVKCPGCYKITTVFS HAQTVVLCVGCSTVLCQPTGGKARLTEGCSFRRKQH eS28 NP_001022.1 MDTSGVQPIKLARVTKVIGKTGSQGQCTQVRVEFMDDTSRSIIRNVKGP VREGDVLTLLESEREARRLR uS14 NP_001276741.1 MGHQQLYWSHPRKFGQGSRSCRVCSNRHGLIRKYGLNMCRQCFRQYA KDIGFIKLD eS30 XP_002723447.1 MQLFVRAQELHTLEVTGRETVAQIKAHVASLEGIAPEDQVVLLAGTPL EDEATLGQCGVEALSTLEVAGRMLGGKVHGSLARVGKVRGQTLKVA KQEKKKKRTGRAKRRMQYNRRFVNVVPTFGKKKGPNANS eS31 XP_002724420.1 MQIFVKTLTGKTITLEVEPSDTIENVKAKIQDKEGIPPDQQRLIFAGKQLE DGRTLSDYNIQKESTLHLVLRLRGGAKKRKKKSYTTPKKNKHKRKKV KLAVLKYYKVDENGKISRLRRECPSDECGAGVFMASHFDRHYCGKCC LTYCFNKPEDK uL2 XP_002723515.1 MGRVIRGQRKGAGSVFRAHVKHRKGAARLRAVDFAERHGYIKGIVKD IIHDPGRGAPLAKVVFRDPYRFKKRTELFIAAEGIHTGQFVYCGKKAQL NIGNVLPVGTMPEGTIVCCLEEKPGDRGKLARASGNYATVISHNPETKK TRVKLPSGSKKVISSANRAVVGVVAGGGRIDKPILKAGRAYHKYKAKR NCWPRVRGVAMNPVEHPFGGGNHQHIGKPSTIRRDAPAGRKVGLIAAR RTGRLRGTKTVQEKEN uL3 XP_002711434.1 MSHRKFSAPRHGSLGFLPRKRSSRHRGKVKSFPKDDPSKPVHLTAFLGY KAGMTHIVREVDRPGSKVNKKEVVEAVTIVETPPMVVVGIVGYVETPR GLRTFKTVFAEHISDECKRRFYKNWHKSKKKAFTKYCKKWQDDAGKR QLDKDFSSMKKYCQVIRVLAHTQMRLLPLRQKKAHLMEIQVNGGTVA EKLDWARERLEQQVPVSQVFGQDEMIDVIGVTKGKGYKGVTSRWHTK KLPRKTHRGLRKVACIGAWHPARVAFSVARAGQKGYHHRTEINKKIY KIGQGYLIKDGKLIKNNASTDYDLSDKSINPLGGFVHYGEVTNDFVML KGCVVGTKKRVLTLRKSLLVQTKRRALEKIDLKFIDTTSKFGHGRFQTV EEKKAFMGPLKKDRIAKEEGA uL4 NP_001182746.1 MACARPLISVYSEKGESSGKNVTLPAVFKAPIRPDIVNFVHTNLRKNNR QPYAVSELAGHQTSAESWGTGRAVARIPRVRGGGTHRSGQGAFGNMC RGGRMFAPTKTWRRWHRRVNTTQKRYAICSALAASALPALVMSKGH RIEEVPELPLVVEDKVEGYKKTKEAVLLLKKLKAWNDIKKVYASQRM RAGKGKMRNRRRIQRRGPCVIYNEDNGIVKAFRNIPGITLLNVTKLNIL KLAPGGHVGRFCIWTESAFRKLDDLYGTWRKAASLKSNYNLPMHKML NTDLSRILKSPEIQRALRAPRKKIHRRVLKKNPLKNLRIMLKLNPYAKT MRRNTILRQARNHKLRVERAAAALAAKSDPKEAPAKKKPVVGKKVK KPRAVGIKQKKKPVVGRKAAAAKKPAADKKAADKRAGPEDKKPAA uL18 NP_001182608.1 MGFVKVVKNKAYFKRYQVKFRRRREGKUDYYARKRLVIQDKNKYNU PKYRMIVRVUNRDIICQIAYARIEGDMIVCAAYAHELPKYGVKVGLUN YAAAYCUGLLLARRLLNRFGMDKIYEGQVEVUGDEYNVESIDGQPGA FUCYLDAGLARUUUGNKVFGALKGAVDGGLSIPHSUKRFPGYDSESKE FNAEVHRKHIMGQNVADYMRYLMEEDEDAYKKQFSQYIKNNVUPDM MEEMYKKAHAAIRENPVYEKKPKREVKKKRWNRPKMSLAQKKDRV AQKKASFLRAQERAAES eL6 XP_002719802.1 MAGEKAPAAKPDAUKKSPAKKADHARGKAKKKULAEKKPKKGKPHC SRNPVLVRGIGRYSRSAMYSRKALYKRKYAAPKSRIERKKKREKVLAU VUKPVGGDKNGGURVVKLRKMPRYYPUEDVPRKLLSHGKKPFSQHV RKLRASIUPGUILIILUGRHRGKRVVFLKQLSSGLLLVUGPLSLNRVPLR RUHQKFVIAUSUKIDISGVKIPKHLUDAYFKKKKLRKPRHQEGEIFDUE KEKYEIUEQRKVDQKAVDSQILPKIKAVPQLQGYLRSVFALUNGVYPH KLVF uL30 XP_002710681.1 MEGAEEKKKVPAVPETLKKKRRNFAELKIKRLRKKFAQKMLRKARRK LIYEKAKHYHKEYRQMYRTEIRMARMARKAGNFYVPAEPKLAFVIRIR GINGVSPKVRKVLQLLRLRQIFNGTFVKLNKASINMLRIVEPYIAWGYP NLKSVNELIYKRGYGKINKKRIALTDNTLIARSLGKYNIICMEDLIHEIYT VGKHFKEANNFLWPFKLSSPRGGMKKKTTHFVEGGDAGNREDQINRLI RRMN eL8 XP_008273621.1 MSSYRLGYCMKEERHNLVLCLWSQSPGILNSKCLWPFTNIHLLVGALP REGAGGAWGGGRSEQLPTCSTTHHDFTWDKKVVNPLFEKRPKNFGIG QDIQPKRDLTRFVKWPRYIRLQRQRAILYKRLKVPPAINQFTQVLDRQT ATQLLKLAHKYRPETKQEKKQRLLARAEKKAAGKGDVPTKRPPVLRA GVNTVTTLVENKKAQLVVIAHDVDPIELVVFLPALCRKMGVPYCILKG KARLCRLVHRKTCTTVAFTQVNSEDKGALAKLVEAIRTNYNDRYDEIR RHWGGNVLGPKSVARIAKLEKAKAKELATKLG uL6 XP_002709438.1 MKTILSNQTVDIPENVDISLKGRTVIVKGPRGTLRRDFNHINVELSLLGK KKKRLRVDKWWGNRKELATVRTICSHVQNMIKGVTLGFRYKMRSVY AHFPINVVIQENGSLVEIRNFLGEKYIRRVRMRPGVACSVSQAQKDELV LEGNDIELVSNSAALIQQATTVKNKDIRKFLDGIYVSEKGTVQQADE uL16 NP_001164845.1 MGRRPARCYRYCKNKPYPKSRFCRGVPDAKIRIFDLGRKKAKVDEFPL CGHMVSDEYEQLSSEALEAARICANKYMVKSCGKDGFHIRVRLHPFHV IRINKMLSCAGADRLQTGMRGAFGKPQGTVARVHIGQVIMSIRTKLQN KEHVVEALRRAKFKFPGRQKIHISKKWGFTKFNADEFEDMVAEKRLIP DGCGVKYIPNRGPLDKWRALHS uL5 XP_008263912.1 MAQDQGEKENPMRELRIRKLCLNICVGESGDRLTRAAKVLEQLTGQTP VFSKARYTVRSFGIRRNEKIAVHCTVRGAKAEEILEKGLKVREYELRKN NFSDTGNFGFGIQEHIDLGIKYDPSIGIYGLDFYVVLGRPGFSIADKKRRT GCIGAKHRISKEEAMRWFQQKYDGIILPGK eL13 XP_002710580.1 MAPSRNGMILKPHFHKDWQRRVATWFNQPARKIRRRKARQARARRIA PRPAAGPIRPIVRCPTVRYHTKVRAGRGFSLEELRVAGIHKKVARTIGIS VDPRRRNKSTESLQANVQRLKEYRSKLVLFPRKPSAPKKGDSSAEELKL ATQLTGPVMPIRNVFKKEKARVITEEEKNFKAFASLRMARANARLFGIR AKRAKEAAEQDVEKKK eL14 XP_002713132.1 MVFRRFVEVGRVAYVSFGPHAGKLVAIVDVIDQNRALVDGPCTRVRR QAMPFKCMQLTDFILKFPHSARQKYVRKAWEKADINTKWAATRWAK KIEARERKAKMTDFDRYKVMKAKKMRNRIIKNEVKKLQRAALLKASP KKAPVAKGAVAAAAAAAKVPAKKATAAGKKAAAQKAPAQKAPAQK AAGQKAAQPPKAQKGQKPPAQKAPA PKASGKKA eL15 XP_002716262.1 MGAYKYIQELWRKKQSDVMRFLLRVRCWQYRQLSALHRAPRPTRPD KARRLGYKAKQGYVIYRIRVRRGGRKRPVPKGATYGKPVHHGVNQLK FARSLQSVAEERAGRHCGALRVLNSYWVGEDSTYKFFEVILIDPFHKAI RRNPDTQWITKPVHKHREMRGLTSAGRKSRGLGKGHKFHHTIGGSRR AAWRRRNTLQLHRYR uL13 XP_002723961.2 MAEGQVLVLDGRGHLLGRLAAIVAKQVLLGRKVVVVRCEGINISGNFY RNKLKYLAFLRKRMNTNPSRGPYHFRAPSRIFWRTVRGMLPHKTKRG QAALDRLKVFDGIPPPYDKKKRMVVPAALKVVRLKPTRKFAYLGRLA HEVGWKYQAVTATLEEKRKEKAKIHYRKKKQLMRLRKQAEKNVEKK ISKFTDVLKTHGLLV uL22 XP_002713574.1 MVRYSLDPENPTKSCKSRGSNLRVHFKNTRETAQAIKGMHIRKATKYL KDVTLKKQCVPFRRYNGGVGRCAQAKQWGWTQGRWPKKSAEFLLH MLKNAESNAELKGLDVDSLVIEHIQVNKAPKMRRRTYRAHGRINPYM SSPCHIEMILTEKEQIVPKPEEEVAQKKKISQKKLKKQKLMARE eL18 XP_008250645.1 MGVDIRHNKDRKVRRKEPKSQDIYLRLLVKLYRFLARRTNSTFNQVVL KRLFMSRTNRPPLSLSRMIRKMKLPGRENKTAVVVGTVTDDVRVQEVP KLKVCALRVTSRARSRILKAGGKILTFDQLALDSPKGRGTVLLSGPRKG REVYRHFGKAPGTPHSHTKPYVRSKGRKFERARGRRASRGYKN eL19 XP_008269617.1 MSMLRLQKRLASSVLRCGKKKVWLDPNETNEIANANSRQQIRKLIKDG LIIRKPVTVHSRARCRKNTLARRKGRHMGIGKRKGTANARMPEKVTW MRRMRILRRLLRRYRESKKIDRHMYHSLYLKVKGNVFKNKRILMEHIH KLKADKARKKLLAXVAEARRSKTKEARKRREERLQAKKEEIIKTLSKE EETKK eL20 XP_004595602.1 MKASGTLREYKVVGRCLPTPKCRTPPLYRMRIFAPNHVVAKSRFWYFV SQLKKMKKSSGEIVYCGQVFEKSPLRVKNFGIWLRYDSRSGTHNMYRE YRDLTTAGAVTQCYRDMGARHRARAHSIQIMKVEEIAASKCRRPAVK QFHDSKIKFPLPHRVLRRQHKPRFTTKRPNTFF eL21 XP_002710693.1 MTNTKGKRRGTRYMFSRPFRKHGVVPLATYMRIYKKGDIVDIKGMGT VQKGMPHKCYHGKTGRVYNVTQHAVGIVVNKQVKGKILAKRINVRIE HIKHSKSRDSFLKRVKENDQKKKEAKEKGTWVQLKRQPAPPREAHF VRTNGKEPELLEPIPYEFMA eL22 XP_004596146.1 MAPVKKLVAKGGKKKKQLLKFTLDCTHPVEDGIMDAANFEQFLQERI KVNGKAGNLGGGVVSLERSKSKITVTSEVPFSKRYLKYLTKKYLKKNN LRDWLRVVANTKESYELRYFQINQDEEEEEEED uL14 XP_002719380.2 MSKRGRGGSSGAKFRISLGLPVGAVINCADNTGAKNLYIISVKGIKGRL NRLPAAGVGDMVMATVKKGKPELRKKVHPAVVIRQRKSYRRKDGVF LYFEDNAGVIVNNKGEMKGSAITGPVAKECADLWPRIASNAGSIA eL24 XP_002716775.1 MKVELCSFSGYKIYPGHGRRYARTDGKVFQFLNAKCESAFLSKRNPRQ INWTVLYRRKHKKGQSEEIQKKRTRRAVKFQRAITGASLADIMAKRNQ KPEVRKAQREQAIRAAKEAKKAKQASKKTAMAAAKAPTKAAPKQKIV KPVKVSAPRVGGKR uL23 NP_001192153.1 MAPKAKKEAPAPPKVEAKAKALKAKKAVLKGVHSHKKKKIRTSPTFR RPKTLRLRRQPKYPRKSAPRRNKLDHYAIIKFPLTTESAMKKIEDNNTL VFIVDVKANKHQIKQAVKKLYDIDVAKVNTLIRPDGEKKAYVRLAPDY DALDVANKIGII uL24 XP_002719046.1 MKFNPFVTSDRSKNRKRHFNAPSHIRRKIMSSPLSKELRQKYNVRSMPI RKDDEVQVVRGHYKGQQIGKVVQVYRKKYVIYIERVQREKANGTTVH VGIHPSKVVITRLKLDKDRKKILERKAKSRQVGKEKGKYKEETIEKMQE eL27 XP_002714215.1 MGKFMKPGKVVLVLAGRYSGRKAVIVKNIDDGTSDRPYSHALVAGID RYPRKVTAAMGKKKIAKRSKIKSFVKVYNYNHLMPTRYSVDIPLDKTV VNKDVFRDPALKRKARREAKVKFEERYKTGKNKWFFQKLRF uL15 XP_002708842.1 MPSRLRKTRKLRGHVSHGHGRIGKHRKHPGGRGNAGGMHHHRINFDK YHPGYFGKVGMRHYHLKRNQSFCPTVNLDKLWTLVSEQTRVNAAKN KTGAAPIIDVVRSGYYKVLGKGKLPKQPVIVKAKFFSRRAEEKIKGVGG ACVLVA eL29 XP_002713279.1 MAKSKNHTTHNQSRKWHRNGIKKPRSQRYESLKGVDPKFLRNMRFAK KHNKKGLKKMQANNAKAMAARAEAIKALVKPKEVKPTIPKGVSRKL HRLAYIAHPKLGRRARARIARGLRLSRPQTKAKAKTEPQIKGKVKAQIK AQAQAQIKSKGKGKAQAETKPKAQAETKPKAQAQAKPKAQAQGKPK AQAQGKPKAQAQAKPKAQAQAKPKAQAQTKPKAQATPAAPVPAQAP PKGAQPPAKAP eL30 XP_002710752.1 MVAAKKTKKSLESINSRLQLVMKSGKYVLGYKQSLKMIRQGKAKLVI LANNCPALRKSEIEYYAMLAKTGVHHYSGNNIELGTACGKYYRVCTLS IIDPGDSDIIRSMPEQTGEK eL31 XP_002709965.1 MAPAKKGGEKKKGRSAINEVVTREYTINIHKRIHGVGFKKRAPRALKEI RKFAMKEMGTPDVRIDTRLNKAVWAKGIRNVPYRIRVRLSRKRNEDE DSPNKLYTLVTYVPVTTFKNLQTVNVDEN eL32 XP_002708876.1 MAALRPLVKPKIVKKRTKKFIRHQSDRYVKIKRNWRKPRGIDNRVRRR FKGQILMPNIGYGSNKKTKHMLPSGFRKFLVHNVKELEVLLMCNKSYC AEIAHNVSSKNRKAIVERAAQLAIRVTNPNARLRSEENE eL33 NP_001192154.1 MSGRLWCKAIFAGYKRGLRNQREHTALLKIEGVYARDETEFYLGKRC AYVYKAKNNTVTPGGKPNKTRVIWGKVTRAHGNSGMVRAKFRSNLP AKAIGHRIRVMLYPSRI eL34 XP_002718179.1 MVQRLTYRRRLSYNTASNKTRLSRTPGNRIVYLYTKKVGKAPKSACGV CPGRLRGVRAVRPKVLMRLSKTKKHVSRAYGGSMCAKCVRDRIKRAF LIEEQKIVVKVLKAQAQSQKAK uL29 XP_008271551.1 MAKIKARDLRGKKKEELLKQLDDLKVELSQLRVAKVTGGAASKLSKIR VVRKSIARVLTVINQTQKENLRKFYKGKKYKPLDLRPKKTRAMRRRLN KHEESLKTKKQQRKERLYPLRKYAVKA eL36 XP_002708713.1 MALRYPMAVGLNKGHKVTKNVSKPRHSRRRGRLTKHTKFVRDMIREV CGFAPYERRAMELLKVSKDKRALKFIKKRVGTHIRAKRKREELSSVLA AMRKAAAKKD eL37 XP_002714044.1 MTKGTSSFGKRRNKTHTLCRRCGSKAYHLQKSTCGKCGYPAKRKRKY NWSAKAKRRNTTGTGRMRHLKIVYRRFRHGFREGTTPKPKRAAVAAS SSS eL38 XP_002711541.1 MPRKIEEIKDFLLTARRKDAKSVNIKKNKDNVKFKVRCSRYLYTLVITD KEKAEKLKQSLPPGLAVKELK eL39 XP_008246941.1 MGLCPGWMGLSSPLSWCAEWTPALAMSSHKTFRIKRFLAKKQKQNRP IPQWIWMKTGNKIRYNSKRRHWR RTKLGL eL40 XP_004595836.1 MGDPESGGCIPPDQQRLIFAGKQLEDGRTLSDYNIQKESTLHLVLRLRG GIIEPSLRQLAQKYNCDKMICRKCYARLHPRAVNCRKKKCGHTNNLRP KKKVK eL41 A0A087WNH4 MRAKWRKKRMRRLKRKRRKMRQRSK eL42 XP_002708510.1 MVNVPKTRRTFCKKCGKHQPHKVTQYKKGKDSLYAQGKRRYDRKQS GYGGQTKPIFRKKAKTTKKIVLRLECVEPNCRSKRMLAIKRCKHFELGG DKKRKGQVIQF eL43 XP_002708723.1 MAKRTKKVGIVGKYGTRYGASLRKMVKKIEISQHAKYTCSFCGKTKM KRRAVGIWHCGSCMKTVAGGAWT YNTTSAVTVKSAIRRLKELKDQ eL28 XP_002713247.1 MSAHLQWMVVRNCSSFLIKRNKQTYSTEPNNLKARNSFRYNGLIHRKT VGVEPAADGKGVVVVMKRRSGQRKPATSYVRTTINKNARATLSSIRH MIRKNKYHPDLRMAAIRRASAILRSQKPVMVKRKRTRPTKSS uL10 XP_002719840.1 MPREDRATWKSNYFLKIIQLLDDYPKCFIVGADNVGSKQMQQIRMSLR GKAVVLMGKNTMMRKAIRGHLENNPALEKLLPHIRGNVGFVFTKEDL TEIRDMLLANKVPAAARAGAIAPCEVTVPAQNTGLGPEKTSFFQALGIT TKISRGTIEILSDVQLIKTGDKVGASEATLLNMLNISPFSFGLIIQQVFDN GSIYNPEVLDITEDTLHSRFLEGVRNVASVCLQIGYPTVASVPHSIINGY KRVLALSVETEYTFPLAEKVKAFLADPSAFVAAAPVAAASTAAPAAAA AAPAKVEAKEESEESDEDMGFGLFD uL11 XP_002720524.1 MPPKFDPNEIKVVYLRCTGGEVGATSALAPKIGPLGLSPKKVGDDIAKA TGDWKGLRITVKLTIQNRQAQIEVVPSASALIIKALKEPPRDRKKQKNIK HSGNITFDEIVNIARQMRHRSLARELSGTIKEILGTAQSVGCNVDGRHPH DIIDDINSGAVECPAS RPLP1 XP_002722423.1 MASVSELACIYSALILHDDEVTVTEDKINALIKAAGVNVEPFWPGLFAK ALANVNIGSLICNVGAGGAAPAAGAAPAGGPAPAAAAAPAEEKKVEA KKEESEESDDDMGFGLFD RPLP2 XP_002724382.1 FLLPALGGNSLPSAKDIKKILDSMGIEADDDRLNKVISELNGKKIEDVIA QGIGKLASVPAGMAVAVSATAGSVAPAARSAPTTSIFWLNLTQIPQQFI RIHWDERRNI Su pp le m en ta ry T ab le 2 . A su m m ar y of th e tw o H D X -M S da ta se ts th at w er e co lle ct ed . Supplementary Table 3. A table giving the % deuteration for each peptide of uL10. The data shown are means of three technical replicates, and the standard deviations are also given. The values are coloured according to the key in the bottom right.