Gain-Of-Function and Dominant-Negative Effects of Distinct P53 Mutations in Lung Tumours

Gain-of-function and dominant-negative effects of distinct p53 mutations in lung tumours

Frances Kathryn Turrell

Clare College, University of Cambridge

This dissertation is submitted for the degree of Doctor of Philosophy

December 2017

2 Gain-of-function and dominant-negative effects of distinct p53 mutations in lung tumours

Frances Kathryn Turrell

Lung cancer is the most common cause of cancer-related mortality worldwide with current treatments providing limited therapeutic benefit in most cases. TP53 (Trp53, p53) mutations occur in approximately 50% of lung adenocarcinoma cases and are associated with poor prognosis and so novel therapies that target these p53 mutant lung tumours are urgently needed. Despite the high frequency of p53 mutations in lung tumours, the impact these mutations have on response to therapy remains unclear in this cancer type. The aim of my project is to characterise the gain-of-function and dominant- negative effects of p53 mutations in lung tumours and to identify ways of therapeutically targeting these p53 mutant tumours based on dependencies and susceptibilities that our analysis uncovers.

To characterise the gain-of-function and dominant-negative effects of p53 mutations I compared p53 mutant murine lung tumour cells that endogenously express either a contact (R270H, equivalent to R273H in humans) or conformational (R172H, equivalent to R175H in humans) p53 mutant protein and p53 null lung tumour cell lines; both in the presence and absence of wild-type p53. Interestingly, transcriptional and functional analysis uncovered metabolic gain-of-functions that are specific to the type of p53 mutation. Upregulation of mevalonate pathway expression was observed only in R270H lung tumours and consequently R172H and R270H lung tumours displayed distinct sensitivities to simvastatin, a mevalonate pathway inhibitor widely used in the clinic. Furthermore, the transcriptional signature underlying this sensitivity to simvastatin was also present in human lung tumours with contact p53 mutations, indicating that these findings may be clinically relevant.

On the other hand, our analysis of the potential dominant-negative effects of the p53 mutants on wild-type p53 demonstrated that wild-type p53 was able to induce typical p53 target genes to a similar level in p53 null and mutant cells. Furthermore, wild-type p53 restoration resulted in comparable tumour suppressive responses in p53 mutant and null tumours and thus, p53-restoration therapy will likely be of benefit to patients with p53 mutations in lung cancer. Hence, I have demonstrated that lung tumours harbouring contact and conformational p53 mutations display common and distinct therapeutic susceptibilities.

3 Preface

Declaration of originality

This dissertation is the result of my own work and includes nothing which is the outcome of work done in collaboration except where indicated in the text.

I confirm that no part of this dissertation has been submitted for any other degree or qualification.

Statement of length

This dissertation does not exceed 60,000 words in length.

i Publications

Turrell FK, Kerr EM, Gao M, Thorpe H, Doherty G, Cridge J, Shorthouse D, Speed A, Samarajiwa S, Hall BA, Meryl Griffiths M and Martins CP (2017). Lung tumors with distinct p53 mutations respond similarly to p53-targeted therapy but exhibit genotype- specific statin sensitivity. Genes and Development 2017 Aug 8.

Kerr EM, Gaude E, Turrell FK, Frezza C, Martins CP. Mutant Kras copy number defines metabolic reprogramming and therapeutic susceptibilities. Nature. 2016 March 03; (531):110–113 (work not included in the thesis).

Talks and poster presentations

F. K. Turrell, E. M. Kerr, M. Gao, H. Thorpe and C. P. Martins. Lung tumours with distinct p53 mutations display genotype-specific metabolic profiles and differential sensitivities to statins. Short ‘poster-taster’ talk and poster presentation - Beatson International Cancer Conference, 2nd-5th July 2017.

F. K. Turrell Determining the impact of distinct p53 mutations on lung tumour evolution and therapy. Talk - MRC Cancer Unit annual symposium, 2nd December 2016

F. K. Turrell, M. Gao, E. M. Kerr and C. P. Martins Determining the impact of distinct p53 mutations in lung tumours. Poster - MRC Cancer Unit annual symposium, 20th November 2015.

F. K. Turrell, M. Gao, E. Gaude, E. M. Kerr, C. Frezza and C. P. Martins. Determining the metabolic impact of p53 mutations in lung cancer. Poster - MRC Cancer Unit annual symposium, 21st November 2014.

ii Acknowledgements

Firstly, I would like to thank my supervisor Dr Carla Martins for the opportunity to pursue a PhD in her laboratory and her continued guidance. I would also like to thank Dr Emma Kerr for all of her scientific input and support over the years, which have been instrumental to this work, and for being a great friend during my time here in Cambridge. I would like to thank Dr Meiling Gao for providing the samples for the microarray analysis and therefore, her work towards initiating the project.

I am also grateful to all members of the Martins and Shields laboratories, past and present, who have all helped me in some way over the years. In particular, I would like to thank Dr Angela Riedel, Dr Luisa Pedro, Dr Ilya Kotov, Dr Gary Doherty, Jennifer Harris, Sarah Davidson and Alyson Speed for their continuous friendship and support. I would also like to thank Dr Emma Kerr, Dr Hannah Thorpe, Dr Gary Doherty and Jake Cridge for their help addressing reviewer comments. Furthermore, I am grateful to Ellie Gregson, who has been pursuing her PhD alongside me from interview day until submission, for supporting me through it.

I thank all the staff at the Biological Resources Unit and Cambridge Biomedical Services for their help with the in vivo work. In particular, I am grateful to Alyson Speed and Mike Mitchell for their invaluable help with the in vivo experiments. In addition, I would like to thank Dr Christian Frezza and members of his laboratory, particularly Edo Gaude, for running the mass spectrometry and for their advice on the metabolism side of the project. I am also grateful to Dr Shamith Samarajiwa and Dr David Shorthouse for their work on the microarray data normalisation and advice on the analysis, respectively.

Finally, I would like to thank all my family and friends as without their support none of this would have been possible.

iii Abstract

Lung cancer is the most common cause of cancer-related mortality worldwide with current treatments providing limited therapeutic benefit in most cases. TP53 (Trp53, p53) mutations occur in approximately 50% of lung adenocarcinoma cases and are associated with poor prognosis and so novel therapies that target these p53 mutant lung tumours are urgently needed. Despite the high frequency of p53 mutations in lung tumours, the impact these mutations have on response to therapy remains unclear in this cancer type. The aim of my project is to characterise the gain-of- function and dominant-negative effects of p53 mutations in lung tumours and to identify ways of therapeutically targeting these p53 mutant tumours based on dependencies and susceptibilities that our analysis uncovers.

On the other hand, our analysis of the potential dominant-negative effects of the p53 mutants on wild-type p53 demonstrated that wild-type p53 was able to induce typical p53 target genes to a similar level in p53 null and mutant cells. Furthermore, wild- type p53 restoration resulted in comparable tumour suppressive responses in p53 mutant and null tumours and thus, p53-restoration therapy will likely be of benefit to patients with p53 mutations in lung cancer. Hence, I have demonstrated that lung tumours harbouring contact and conformational p53 mutations display common and distinct therapeutic susceptibilities.

iv List of abbreviations

2D/3D Two-dimensional/ three-dimensional 4OHT 4-hydroxytamoxifen 14-3-3 Tyrosine 3-monooxygenase/tryptophan 5-monooxygenase activation protein sigma Acat2 Acetyl-coA acetyltransferase 2 Actb Beta-actin Adeno-Cre Cre recombinase-expressing adenovirus Aif Apoptosis-inducing factor ALK Anaplastic lymphoma kinase AMPK 5’-adenosine monophosphate-activated protein kinase ANOVA Analysis of variance Apaf1 Apoptotic peptidase activating factor 1 APC Allophycocyanin ATP Adenosine triphosphate AWERB Animal welfare and ethical review body Bax Bcl-2-associated X protein Bbc3 Bcl2 binding component 3 BCA Bicinchoninic acid Birc5 Baculoviral inhibitor of apoptosis repeat-containing 5 BrdU Bromodeoxyuridine BSA Bovine serum albumin Btg2 B cell translocation gene 2 CCCP Carbonyl cyanide m-chlorophenyl hydrazine Ccng1 Cyclin G1 Cdkn1a Cyclin-dependent kinase inhibitor 1A cDNA Complementary DNA ChIP-seq Chromatin immunoprecipitation sequencing Cpt1a Carnitine palmitoyltransferase 1A Crot Carntine-O-octanoyltransferase Ctrl Control Cyp51a1 Cytochrome P450 family 51 subfamily A member 1 DAPI 4',6-diamidino-2-phenylindole DBD DNA-binding domain v Ddb2 Damage-specific DNA binding protein 2 Ddit4 DNA damage inducible transcript 4 Dgka Diacylglycerol kinase alpha dH2O Deionised water Dhcr7 7-Dehydrocholesterol reductase Dhcr24 24-Dehydrocholesterol reductase DMEM Dulbecco’s modified eagle medium DMSO Dimethyl sulfoxide DN Dominant-negative DNA Deoxyribonucleic acid Dnaja1 DnaJ Heat Shock Protein Family Member A1 DNase Deoxyribonuclease Dr5 Death receptor 5 E2f7 E2F transcription factor 7 ECAR Extracellular acidification rate ECL Enhanced chemiluminescence EDTA Ethylenediaminetetraacetic acid EGFR Epidermal growth factor receptor Ei24 Autophagy associated transmembrane protein EMEM Eagle’s minimum essential media EpCAM Epithelial cell adhesion molecule Ephx1 Epoxide hydrolase 1 ER Endoplasmic reticulum ER Estrogen receptor Ercc5 Excision repair cross-complementing rodent repair deficiency complementation group 5 Erk Extracellular signal-regulated kinase-1 ETC Electron transport chain Euk Eukaryotic FACS Fluorescence-activated cell sorting FAO Fatty acid oxidation Fas Fas cell surface death receptor FC Fold-change FCS Fetal calf serum Fdps Farnesyl diphosphate synthase

vi FITC Fluorescein isothiocyanate FPP Farnesyl pyrophsophate G6PD Glucose-6-phosphate dehydrogenase Gadd45 Growth arrest and DNA-damage-inducible 45 Gapdh Glyceraldehyde-3-phosphate dehydrogenase gDNA Genomic DNA GFP Green fluorescent protein Gls2 Glutaminase 2 Glut Glucose transporter GOF Gain-of-function GGPP Geranylgeranyl pyrophosphate GSEA Gene set enrichment analysis GSH Reduced glutathione GSSG Oxidised glutathione GTP Guanosine triphosphate GTPase Guanosine triphosphatase Gtse1 G2 and S-phase expressed 1 Gy Gray H&E Haemotoxylin and Eosin HBSS Hank’s balanced salt solution Hmgcr 3-Hydroxy-3-Methylglutaryl-CoA Reductase Hmgcs1 3-Hydroxy-3-Methylglutaryl-CoA Synthase 1 Hsp40 Heat-shock protein 40 i.d. Internal diameter Idi2 Isopentenyl-diphosphate delta isomerase 2 IgG Immunoglobulin G i.p. Intraperitoneal injection IPA Ingenuity pathway analysis LC3 Microtubule-associated protein 1A/1B-light chain 3 Ldlr Low Density Lipoprotein Receptor Lipa Lipase A, lysosomal acid type LOH Loss-of-heterozygosity Lpin1 Lipin 1 LSL Lox-Stop-Lox Lss Lanosterol synthase

vii Kras Kirten rat sarcoma viral oncogene homolog MAPK Mitogen-activated protein kinase Mcd Malonyl-CoA decarboxylase Mdm2 Mouse Double Minute 2 MET Hepatocyte growth factor receptor MeV Multiple experiment viewer MS Mass spectrometry Msmol Methylsterol monooxygenase mTOR Mammalian target of rapamycin complex 1 MVA Mevalonate Mvd Mevalonate diphosphate decarboxylase Mvk Mevalonate kinase NAPDH Nicotinamide adenine dinucleotide phosphate-oxidase NaPyr Sodium pyruvate NGF Nerve growth factor Nr2f2 Nuclear Receptor Subfamily 2 Group F Member 2 NS Non-significant NSCLC Non-small cell lung cancer NTC Non-template control OCR Oxygen consumption rate OXPHOS Oxidative phosphorylation (T)p53 Tumour protein p53 p53R2 Ribonucleotide Reductase Regulatory TP53 Inducible Subunit M2B p63 Tumour protein p53 p73 Tumour protein p53 PAGE Polyacrylamide gel electrophoresis Pai1 Plasminogen activator inhibitor PBS(-T) Phosphate-buffered saline (-Tween) PCR Polymerase chain reaction PDGFR Platelet-derived growth factor receptor Pdk2 Pyruvate dehydrogenase kinase 2 Perp p53 apoptosis effector PFA Paraformaldehyde PFU Plaque-forming unit

viii Pggt1b Protein geranylgeranyltransferase type I subunit beta Pgm Phosphoglycerate mutase Phlda3 Pleckstrin homology like domain family A member 3 PI Propidium iodide PI3K Phosphoinositide-3 kinase Pig3 p53 inducible gene 3 Pilra Paired immunoglobin like type 2 receptor alpha PM Plasma membrane Pmaip1 Phorbol-12-myristate-13-acetate-induced protein 1 Pml Promyelocytic leukaemia Pmvk Phosphomevalonate kinase Polk DNA polymerase  PPP Pentose phosphate pathway PR Proline-rich domain Pten Phosphatase and tensin homologue Ptprv Protein tyrosine phosphatase receptor type V Rabggta Rab geranylgeranyltransferase alpha subunit Rho Ras Homolog Family Member A RIPA Radioimmunoprecipitation RNA Ribonucleic acid RNAse Ribonuclease ROS Reactive oxygen species Rprm Reprimo TP53 dependent G2 arrest mediator homolog rRNA Ribosomal RNA S6k Ribosomal protein S6 kinase SAM Significance analysis for microarrays SCLC Small cell lung cancer Sco2 Synthesis of cytochrome c oxidase 2 SD Standard deviation SDS Sodium dodecyl sulfate SEM Standard error of the mean Sesn Sestrin Slc2a1/4/9 Solute Carrier Family 2 Member 1/4/9 Slc19a2 Solute carrier family 19 member 2 shRNA Short-hairpin RNA

ix Soat2 Sterol O-acyltransferase 2 Spf Specific-pathogen free Sqle Squalene epoxidase Srebf Sterol regulatory element binding transcription factor Stn Simvastatin TAD Transactivation domain Tam Tamoxifen TAZ Transcriptional coactivator with PDZ-binding motif TBS-T Tris-buffered saline-Tween TCGA The cancer genome atlas Tet Tetramerisation domain Tigar TP53-induced glycolysis and apoptosis regulator Tm7sf2 Transmembrane 7 superfamily member 2 Trp53inp1 Transformation related protein 53 inducible nuclear protein 1 Tsc2 Tuberous sclerosis complex 2 TUNEL Terminal deoxynucleotidyl transferase dUTP nick end labeling Upp1 Uridine Phosphorylase 1 qRT-PCR Quantitative real-time polymerase chain reaction Xpc Xeroderma pigmentosum complementation group c WT Wild-type YAP Yes associated protein 1 Zfp365 Zinc-finger protein 365 Zmat3 Zinc Finger Matrin-Type 3

x List of figures and tables

1. Introduction

Figure 1.1. The structure of p53…………………………………………………………..2 Figure 1.2. Effector functions of p53 induced in response to cellular stresses……………………………………………………………………………………....3 Figure 1.3. Examples of genes regulated by p53 and their functional distribution across the ‘classical’ p53-mediated responses………………………………………….5 Figure 1.4. Metabolic regulation by p53………………………………………………….7 Figure 1.5. Mutations in the p53 gene in cancer……………………………………….11 Figure 1.6. Impact of p53 mutations in cancer………………………………………....12 Figure 1.7. Mevalonate pathway in cancer……………………………………...……...16 Figure 1.8. TP53 mutations in human lung adenocarcinoma………………………....18 Figure 1.9. p53ERTAM mouse model……………………………………………………..22

2. Material and Methods

Table 2.1. List of primary antibodies used for Western blotting……………………….43

3. Generation and transcriptional profiling of murine lung tumour cells

Figure 3.1. p53ER and p53Fx lung tumour models…………………………….….….....49 Figure 3.2. Lung tumour burden………………………………………………………….51 Figure 3.3. Validation of genotypes of lung tumour mice……………………………...52 Figure 3.4. Generation and characterisation of lung tumour cell lines………...….....54 Figure 3.5. Characterisation of the growth of lung tumour cell lines…………………55 Figure 3.6. p53 mutations display transcriptional GOFs when compared to p53 null and also p53 mutant-specific effects……………………………………………………..57 Figure 3.7. Wild-type p53 restoration similarly induces or represses a subset of genes in p53 mutant and null cells………………………………………...... 58 Figure 3.8. p53 mutants display DN effects on wild-type p53 transcriptional function……………………………………………………………………………………....60

xi Figure 3.9. Model predicting the effects of p53 mutations in lung tumours based on transcriptional data………………………………………………………………..……….62

4. Defining mutant p53 gain-of-functions in lung tumours

Figure 4.1. Analysis of growth-factor signalling in p53 mutant cells (ER model).....66 Figure 4.2. p53 mutant cells do not exhibit increased invasiveness compared to p53 null cells…………………………………………………………………………………….68 Figure 4.3. Transplantation of p53 mutant cells compared to p53 null, did not reduce host survival………………………………………………………………………………..69 Figure 4.4. Endogenous p53 mutant and null lung tumours display comparable tumour burden and proliferation………………………………………………………….71

5. Distinct transcriptional and metabolic profiles in p53 R172H and R270H mutant cells

Figure 5.1. Metabolic genes are differentially expressed in p53 R172H and R270H cells………………………………………………………………………………………....76 Figure 5.2. Altered expression of genes involved in lipid, protein and mitochondrial respiration in p53 R172H and R270H cells……………………………………………..77 Figure 5.3. p53 R172H and R270H cells display altered metabolic profiles…….....78 Figure 5.4. p53 R270H cells display higher levels of glycolysis and a trend for increased mitochondrial respiratory capacity…………………………………………..80 Figure 5.5. Analysis of protein levels and expression of metabolic regulators in p53 mutant and null cells………………………………………………………………………81 Figure 5.6. p53 R270H cells display higher levels of intracellular cholesterol than R172H cells………………………………………………………………………………...83 Figure 5.7. Higher MVA pathway gene expression in p53 R270H cells and lung tumours………….…….…..………………………………………………….…………….85 Figure 5.8. Srebf2 regulates MVA pathway gene expression in the lung tumour cells……………………………..…………………………………………………….……..87

xii 6. Characterisation of the dominant-negative effects of p53 mutations in lung tumour cells

Figure 6.1. Wild-type p53 restoration altered the expression of genes involved in various tumour-suppressive functions of p53…………………………………………...92 Figure 6.2. Typical p53 targets are similarly induced by wild-type p53 in p53 mutant and null cells………………………………..……………………………………..…….....93 Figure 6.3. p53 target genes are upregulated at later time points in p53 null and mutant cells…………………………………………………………………….………...... 95 Figure 6.4. Validation of DN and ‘wild-type-like’ transcriptional effects...... 96 Figure 6.5. Similar reduction in proliferation observed in p53 null and mutant cell lines after wild-type p53 restoration……………………………………………………...98 Figure 6.6. p53 restoration induces apoptosis, but not senescence, in p53 mutant and null cell cultures……………………………………………………………………….99

7. Targeting p53 mutant lung tumours

Figure 7.1. Similar anti-proliferative effects in p53 null and mutant lung tumours following wild-type p53 restoration…………………………………………...... 103 Figure 7.2. Wild-type p53 restoration induces similar apoptotic responses in p53 null and mutant lung tumours………………………………………………………………..104 Figure 7.3. p53 R270H cells display increased sensitivity to treatment with simvastatin compared to p53 null and R172H cells…………………………………..106 Figure 7.4. Determining why p53 R270H cells are more sensitive to simvastatin..108 Figure 7.5. Statin treatment induces a cytostatic and cytotoxic response in p53 R270H lung tumours but not in R172H and null tumours…………………………….110 Figure 7.6. Simvastatin treatment reduces lung tumour burden and may improve host survival in a p53 R270H lung tumour model…………………………….…….…111 Table 7.1. Classification of TP53 mutations in lung adenocarcinoma cases (TCGA, Nature 2014)………………………………………………………………...…..113 Figure 7.7. Increased MVA gene expression occurs in human lung tumours that express a contact p53 mutation compared to a conformational p53 mutation……..114

xiii Table of contents

Preface ...... i

Acknowledgements……………………………………………………………………………………...iii

Abstract ...... iv

List of abbreviations ...... v

List of figures and tables ...... xi

Table of contents ...... xiii

1. Introduction ...... 1 1.1. The role of the transcription factor p53 ...... 1 1.1.1. ‘Classical’ functions of p53 ...... 1 1.1.2. p53-mediated regulation of metabolism ...... 6 1.1.3. Tumour-suppressive functions of p53 ...... 9 1.2. TP53 mutations in cancer ...... 10 1.2.1. Different classes of p53 mutations ...... 11 1.2.2. Mutant p53 gain-of-functions ...... 13 1.2.3. Metabolic impact of p53 mutations ...... 14 1.2.4. Mutant p53 dominant-negative effects ...... 15 1.3. Lung cancer ...... 17 1.3.1. Role of p53 mutations in lung cancer ...... 18 1.3.2. Therapeutic strategies against p53 mutant lung tumours ...... 19 1.3.3. Murine models of lung cancer ...... 21 1.4. Aims of thesis ...... 24

2. Materials and Methods ...... 25 2.1. In vivo work ...... 25 2.1.1. Colony management ...... 25 2.1.2. Induction of endogenous tumours ...... 26 2.1.3. Murine experimental cohorts ...... 26 2.1.4. Genotyping and sequencing ...... 26 2.1.5. Tamoxifen and simvastatin treatments ...... 27 2.1.6. Transplantation of lung tumour cell lines, imaging and survival studies .. 28 2.1.7. Tissue sampling and processing ...... 28 2.2. Immunohistochemistry ...... 29

xiv 2.2.1. Ki67 staining and quantification ...... 29 2.2.2. TUNEL staining and quantification ...... 29 2.3. Isolation and maintenance of lung tumour cell lines ...... 30 2.3.1. Isolation of lung tumour cell lines ...... 30 2.3.2. Genotyping of cell lines ...... 31 2.3.3. Maintenance of lung tumour cell lines ...... 31 2.3.4. Generation of luciferase-expressing lung tumour cell lines...... 31 2.4. In vitro treatments...... 32 2.5. 2D proliferation assays ...... 32 2.6. 3D cultures ...... 33 2.7. Microarray analysis ...... 33 2.8. qRT-PCR ...... 34 2.9. FACS ...... 35 2.9.1. EpCAM/PDGFR ...... 35 2.9.2. BrdU/PI ...... 36 2.10. Cell death analysis...... 36 2.10.1. Ethidium homodimer/Calcein AM kit ...... 36 2.10.2. Cleaved caspase-3 immunofluorescence ...... 37 2.11. Senescence assay ...... 37 2.12. Migration and invasion assays ...... 38 2.12.1. Scratch assay ...... 38 2.12.2. Invasive growth ...... 38 2.13. Metabolomics ...... 39 2.13.1. Metabolite extraction ...... 39 2.13.2. Mass spectrometry ...... 39 2.13.3. Analysis ...... 39 2.14. Extracellular flux assays ...... 40 2.14.1. Plate preparation ...... 40 2.14.2. Running the assay ...... 40 2.15. Measurement of total protein and cell size ...... 41 2.16. Intracellular lipid measurements ...... 41 2.16.1. Oil red O staining ...... 41 2.16.2. Cholesterol staining and quantification ...... 41 2.17. SDS-PAGE and Western blotting...... 42 2.18. Immunoprecipitation and binding partner analysis ...... 44

xv 2.19. Srebf2 knockdown ...... 45 2.20. Analysis of publically available datasets ...... 45 2.20.1. ChIP datasets ...... 45 2.20.2. TCGA lung adenocarcinoma dataset ...... 45 2.21. Statistical analysis ...... 46

3. Generation and transcriptional profiling of murine lung tumour cells ...... 47 3.1. Introduction ...... 47 3.2. Results ...... 48 3.2.1. Generation of p53 mutant tumours of different p53 deficiencies ...... 48 3.2.2. Isolation and characterisation of p53 mutant and null lung tumour cell lines ...... 50 3.2.3. Transcriptional impact of p53 mutations in lung tumour cultures ...... 53 3.3. Discussion ...... 61

4. Defining mutant p53 gain-of-functions in lung tumours ...... 63 4.1. Introduction ...... 64 4.2. Results ...... 65 4.2.1. Functional profiling of p53 mutant and p53 null lung tumour cells ...... 65 4.2.2. p53 null, R172H and R270H lung tumours are similarly aggressive ...... 67 4.3. Discussion ...... 70

5. Distinct transcriptional and metabolic profiles in p53 R172H and p53 R270H mutant lung tumour cells ...... 73 5.1. Introduction ...... 73 5.2. Results ...... 74 5.2.1. Mutant-type specific transcriptional and metabolic profiles ...... 74 5.2.2. Mevalonate pathway gene upregulation is a gain-of-function specific to p53R270H lung tumours ...... 84 5.3. Discussion ...... 89

6. Characterisation of the dominant-negative effects of p53 mutants in lung tumour cells ...... 90 6.1 Introduction ...... 90 6.2 Results ...... 91 6.2.1 Transcriptional analysis of wild-type p53 activity in p53 mutant lung tumour cells...... 91 6.2.2. Effect of p53 mutant on wild-type p53 tumour-suppressive functions ...... 94

xvi 6.3 Discussion ...... 100

7. Targeting p53 mutant lung tumours ...... 101 7.1. Introduction ...... 101 7.2. Results ...... 101 7.2.1. Similar tumour-suppressive responses are induced by wild-type p53 restoration in p53 mutant and null tumours ...... 101 7.2.2 p53R270H lung tumours display enhanced sensitivity to mevalonate pathway inhibition ...... 102 7.3 Discussion ...... 112

8. Discussion ...... 115 8.1. Summary ...... 115 8.2. Impact of p53 mutations in lung cancer ...... 115 8.3. Clinical Implications ...... 119

Bibliography……………………………………………………………………………...122

Appendices ...... 136

xvii 1. Introduction

1.1. The role of the transcription factor p53

The transcription factor p53 regulates a multitude of genes which impact on numerous cellular functions (Bieging et al., 2014). The p53 protein is comprised of amino-terminal transactivation domains, a proline-rich domain, a DNA-binding domain (DBD), a tetramerisation domain and a carboxy-terminal regulatory domain (Figure 1.1; Bullock and Fersht, 2001). The transcription factor functions as a homotetramer, binding to two decameric half-sites in the DNA and inducing the transcription of its target genes (Figure 1.1; McLure and Lee, 1998). p53 can also function to repress the transcription of genes through interfering with other transcription factors or recruiting of histone-modifying enzymes (Ho and Benchimol, 2003).

1.1.1. ‘Classical’ functions of p53

Under normal conditions, the p53 protein is maintained at low levels by the E3 ubiquitin ligase mouse double minute 2 (Mdm2; Kubbutat et al., 1997). Various types of cellular stress, including DNA damage, oncogenic stress, ribosomal stress, hypoxia and nutrient deprivation, can result in p53 activation, often through interfering with the Mdm2-p53 interaction (Figure 1.2; Bieging et al., 2014). For example, the stress sensor P19ARF stabilises p53 by interacting with Mdm2 and blocking Mdm2- mediated degradation of p53 in response to stress stimuli such as oncogene activation (Horn and Vousden, 2007; Tao and Levine, 1999). Various posttranslational modifications of p53, such as phosphorylation by kinases activated in response to DNA damage, also function to stabilise p53 by blocking Mdm2 from binding (Xu, 2003). Once activated p53 can elicit many types responses, primarily through transcriptional activation of its target genes. These p53-mediated responses include, but are not limited to, cell cycle arrest, DNA repair, senescence, apoptosis and regulation of cellular metabolism. Mdm2 is a direct transcriptional target of p53 and so it supports a negative-feedback loop to control the extent of the p53 response (Barak et al., 1993; Wu et al., 1993). Importantly, the type and duration of the p53 mediated response depends on the context and the nature of the cellular stress (Horn and Vousden, 2007).

p53 TAD PR DBD Tet Reg

Figure 1.1. The structure of p53.

Upper, the p53 protein consists of amino-terminal transactivation domains (TAD), a proline-rich (PR) domain, a DNA-binding domain (DBD), a tetramerisation domain (Tet) and a carboxy-terminal regulatory domain (Reg) (adapted from (Bullock and Fersht, 2001). Lower, structure of 4 independent ,monomers of the core domain of human

p53, with the tetrameric p53 binding to the two DNA half-sites (Kitayner et al., 2006). The DNA is shown in purple and the four core domains of p53 shown in blue, orange, green and red.

The first p53 target gene to be identified was Cdkn1a, a gene which encodes the cyclin-dependent kinase inhibitor, p21 (el-Deiry et al., 1993; Harper et al., 1993). Loss of p21 has been shown to compromise p53-mediated G1-arrest in mouse models following DNA damage (Brugarolas et al., 1995; Deng et al., 1995), demonstrating the importance of p21 in the p53-mediated arrest of proliferation. However, loss of p21 in these studies did not enhance predisposition to tumours (Brugarolas et al., 1995; Deng et al., 1995), although it has been shown result in tumorigenic effects in other models (Cole et al., 2010; Jackson et al., 2003; Jones et al., 1999). Moreover, p21 has been shown to be involved in another type of proliferative arrest, termed senescence, induced again by stress signals such as DNA damage and oncogenic stress. Senescent cells differ from quiescent cells, displaying a larger, flatter morphology. They also have distinct markers, for example β-galactosidase activity, or secrete specific factors such as immune modulators (Munoz-Espin and Serrano, 2014; Xue et al.,2007).

Ribosomal stress DNA damage Hypoxia Telomere erosion

Oncogenic Metabolic

stress stress WT p53

Cell cycle

arrest Metabolism

Senescence Apoptosis DNA repair

Figure 1.2. Effector functions of p53 induced in response to cellular stresses.

The scheme depicts examples of cellular stresses that can activate p53 (upper, green) and the responses downstream of p53 that can be subsequently engaged (lower, blue). Adapted from Bieging et al., 2014.

Importantly, whether p21 mediates senescence or a more transient arrest of proliferation depends on the context, the cellular stress and the activity of other pathways such as mammalian target of rapamycin (mTOR; (Demidenko and Blagosklonny, 2008; Demidenko et al., 2009). In addition, other p53 targets are likely required to ensure that these responses are implemented.

Importantly, the findings from studies on the p53-mediated p21 response, in addition to investigations into the responses mediated by other p53 target genes, demonstrate that the p53-mediated response is dependent on the type, severity and duration of the cellular stress and the cellular context (Bieging and Attardi, 2012; Bieging et al., 2014). Moreover, specific p53 targets genes can be involved in different functions downstream of p53 and which response they help elicit depends on the context. A recent study compiled a list of around 350 p53 target genes based on evidence that p53 bound and regulated the gene (Fischer, 2017). Furthermore, data from chromatin immunoprecipitation sequencing (ChIP-seq) experiments

3 provide evidence for p53 binding many more genes, which suggests that a multitude of additional p53 target genes exist (http://chip-atlas.org). Therefore, the cellular responses induced by p53 are very complex and likely involve the regulation of a subset of target genes or for particular genes to be induced above a threshold to activate a particular response at that precise time. A summary of some of the known p53 target genes and responses that they are primarily involved in is shown in Figure 1.3.

In addition to p21, many other p53 targets have been shown to contribute to a p53- induced proliferative arrest. For instance, the p53 target gene, growth arrest and DNA-damage-inducible 45 (Gadd45a), is important for the control of G2/M progression, with Gadd45a-/- and p53-/- cells displaying similar chromosomal defects (Hollander et al., 1999; Kastan et al., 1991). Also, p53 can regulate the expression of E2F transcription factor 7 (E2f7; Aksoy et al., 2012; Carvajal et al., 2012) and promyelocytic leukemia (Pml) in response to oncogenic-stress which can induce cellular senescence (de Stanchina et al., 2004; Ferbeyre et al., 2000; Pearson et al., 2000). Further examples of p53 targets can be involved in the induction of the anti- proliferative response of p53 are shown in Figure 1.3.

In the context of DNA damage, the p53-mediated arrest in the cell cycle allows for repair of the genome, reducing the propagation of damaged DNA (Kastan et al., 1992; Williams and Schumacher, 2016). This led to p53 being known as the ‘Guardian of the genome’ (Lane, 1992). In addition to mediating the induction of cell cycle arrest in response to DNA damage, p53 can induce genes involved in DNA repair pathways to help repair the damage to the DNA (Williams and Schumacher, 2016). Nucleotide excision repair, base excision repair and non-homologous end- joining can all be induced by p53 depending on the type of DNA damage present via the p53-mediated transcriptional regulation of specific DNA repair genes. (Seo et al., 2002; Tang et al., 1999; Tron et al., 1998). Examples of p53 target genes involved in these DNA repair programs are shown in Figure 1.3.

The induction of apoptosis is another well-characterised function of p53, first shown in brain choroid plexus epithelial tumours (Symonds et al., 1994). In the same model, Bcl-2-associated X protein (Bax) was identified as a p53 target gene required for

Figure 1.3. Examples of genes regulated by p53 and their functional distribution across the ‘classical’ p53-mediated responses.

Examples of p53 target genes involved in cell cycle arrest, senescence, DNA repair and apoptosis, based on (Aksoy et al., 2012; Bersani et al., 2014; Bieging et al., 2014; Collavin et al., 2000; Kannan et al., 2000; Kawase et al., 2009; Zhang et al., 2013b; Zhou et al., 2002). 14-3-3, tyrosine 3-monooxygenase/tryptophan 5-monooxygenase activation protein sigma, Ywhas; Apaf1, apoptotic peptidase activating factor 1; Bax, Bcl-2-associated X protein; Birc5, baculoviral inhibitor of apoptosis repeat-containing 5, survivin; Btg2, B cell translocation gene 2; Ddb2, damage-specific DNA binding protein 2; Dr5, death receptor 5; E2f7, E2F transcription factor 7; Ercc5, excision repair cross-complementing rodent repair deficiency, complementation group 5; Fas, Fas cell surface death receptor, Gadd45, growth arrest and DNA-damage-inducible 45; Gtse1, G2 and S-phase expressed 1, B99; Noxa, otherwise known as Pmaip1, phorbol-12-myristate-13-acetate-induced protein 1; p21, also known as Cdkn1a, cyclin-dependent kinase inhibitor 1A; Pai1, plasminogen activator inhibitor; Perp, p53 apoptosis effector; Phlda3, pleckstrin homology like domain family A member 3; Pig3, p53 inducible gene 3; Pml, promyelocytic leukaemia; Polk, DNA polymerase ; Ptprv, protein tyrosine phosphatase receptor type V; Puma also known as Bbc3, bcl2 binding component 3; Rprm, reprimo TP53 dependent G2 arrest mediator homolog; Xpc, xeroderma pigmentosum, complementation group c; Zfp365, zinc-finger protein 365; Zmat3, Zinc finger matrin-type 3.

p53-mediated apoptosis (Yin et al., 1997). However, like other p53-mediated functions, apoptosis can be regulated by a number of different p53 target genes. Apoptosis can be triggered by p53 via either the intrinsic or extrinsic cell death pathway (Bieging and Attardi, 2012). The intrinsic pathway can be induced via p53- mediated expression of targets such as p53 upregulated modulator of apoptosis (Puma) and Noxa, in addition to Bax. For instance, cells which are deficient for Puma display an almost complete lack of p53-mediated apoptosis following DNA damage

5 (Jeffers et al., 2003; Villunger et al., 2003), indicating Puma is an essential mediator of p53-dependent apoptosis. In contrast, Puma-deficient mice are not prone to developing tumours spontaneously, again indicating that the p53-mediated response is context-dependent (Jeffers et al., 2003; Villunger et al., 2003). On the other hand, the extrinsic cell death pathway can be induced through p53 target genes such as Fas and death receptor 5 (Dr5), which encode death receptors on the cell surface (Muller et al., 1998; Wu et al., 1997). Additional p53 target genes with roles in apoptosis are shown in Figure 1.3.

Together, these p53-dependent transcriptional programmes enable the cell to respond to many different cellular stresses. Transient growth arrest and repair of the damage, permanent growth arrest or apoptosis can be triggered by p53 depending on the extent of the damage and the duration and magnitude of the p53 induction (Bieging et al., 2014). The intricacy of the p53-mediated response is key to protecting the cells and is achieved primarily through its ability to regulate a multitude of different genes.

1.1.2. p53-mediated regulation of metabolism

I have provided an overview of what are considered the ‘classical’ p53 responses. However, more recent effector functions of p53 have been identified (Bieging et al., 2014). One of these is the p53-mediated regulation of cellular metabolism, achieved again mainly through regulation of target genes (Berkers et al., 2013; Frezza and Martins, 2012).

The metabolic gene expression mediated by p53 generally converges on the promotion of mitochondrial oxidative phosphorylation (OXPHOS) and catabolic pathways to provide adenosine triphosphate (ATP) and nutrients in conditions of stress, while inhibiting the glycolytic pathway and anabolic pathways which are conversely used to generate macromolecules (Figure 1.4; Berkers et al., 2013; Frezza and Martins, 2012). Transcriptional repression of the glucose transporters Glut1 (solute carrier family 2 member 1; Slc2a1) and Glut4 (Slc2a4), inhibits glucose uptake into the cell (Schwartzenberg-Bar-Yoseph et al., 2004), while additional steps in the glycolysis pathway can also be inhibited through the p53-mediated upregulation of TP53-induced glycolysis and apoptosis regulator (Tigar) and Parkin,

6 Glut1&4

Glucose G6PD Glutamine Gls2 Nucleotide PPP biosynthesis Glucose-6-P ROS management Tigar Sco2 p53 Aif Rrm2b

Pgm s Pten i

Lpin1 s

l s

i Tsc2

Crot o

s Parkin

n y

l Sestrin1 Cpt1a

m c

Pdk2 Srebf1c Mcd a t

y Sestrin2

u l

G mTORC1 G AMPK PDH Autophagy

Lipid synthesis Fatty acid oxidation

Oxidative Pyruvate Acetyl-CoA Phosphorylation

Lactate TCA cycle Glutamate

Figure 1.4. Metabolic regulation by p53.

p53 promotes fatty acid oxidation (FAO), oxidative phosphorylation (OXPHOS), glutaminolysis and autophagy while inhibiting glycolysis, the pentose phosphate pathway (PPP) and lipid synthesis. p53 promotes OXPHOS through transcriptional activation of Sco2, Aif and Rrm2b. p53-mediated transcriptional repression of Glut1 and Glut4 and inhibition of the glycolytic enzyme Pgm at the protein level inhibits glycolysis. Glycolysis is also repressed by transcriptional activation of Tigar and Parkin. Tigar dephosphorylates fructose-2,6-bisphosphate and therefore inhibits phosphofructose kinase 1 (Pfk1) activation whereas Parkin can promote activation of pyruvate dehydrogenase (PDH). p53 downregulates the PPP by interacting directly with glucose-6-phosphate dehydrogenase (G6PD) to inhibit its activity. Through inhibition of mTORC1 and Srebf1c, p53 negatively regulates lipid synthesis. p53-mediated inhibition of mTORC1 also promotes autophagy. FAO is promoted by p53 by the transcriptional activation of Lpin1, Crot, Cpt1a and Mcd and the activation of AMPK via increased expression of Sestrin 1/2. FAO and glutaminolysis generate acetyl coA and α-ketoglutarate which feed into the tricarboxylic acid (TCA) cycle. Adapted from (Liang, 2013).

7 and the downregulation of phosphoglycerate mutase (Pgm; at the protein level) and pyruvate dehydrogenase kinase 2 (Pdk2; Bensaad et al., 2006; Contractor and Harris, 2012; Kondoh et al., 2005; Zhang et al., 2013a). Furthermore, p53 can promote OXPHOS via induction of synthesis of cytochrome c oxidase 2 (Sco2) and apoptosis-inducing factor (Aif), which are required for the assembly of complex IV and the maintenance of complex I of the mitochondrial electron transport chain (ETC), respectively (Matoba et al., 2006; Stambolsky et al., 2006; Vahsen et al., 2004). Mitochondrial DNA integrity and mitochondrial function are also promoted by p53 via transcriptional activation of the ribonucleotide reductase subunit, p53R2 (Rrm2b) (Bourdon et al., 2007).

Other aspects of cellular metabolism can also be regulated by p53 such as glutaminolysis via the p53-mediated induction of glutaminase 2 (Gls2), and the pentose phosphate pathway (PPP), an important pathway in the biosynthesis of nucleotides (Hu et al., 2010). The PPP can be negatively regulated by p53 via the inhibition of glucose-6-phosphate dehydrogenase (G6PD) activity (Jiang et al., 2011). However, the role of p53 in the regulation of the PPP pathway is complex and the p53-dependent activation of Tigar can promote flux through the PPP (Bensaad et al., 2006; Cheung et al., 2012).

The management of reactive oxygen species (ROS) levels is another important role of p53 in the regulation of cellular metabolism. The PPP produces NADPH, which reduces oxidised (GSSG) to reduced glutathione (GSH, a key anti-oxidant) and so p53-mediated upregulation of the PPP is important for the anti-oxidant response (Cheung et al., 2012). A p53-mediated anti-oxidant response is also achieved via induction of Gls2 (Hu et al., 2010). In addition, p53 can elicit an anti-oxidant response in conditions of serine deprivation, channelling the reduced serine stores to the synthesis of GSH instead of nucleotide synthesis. p53 promotes the survival of cancer cells under conditions of serine depletion via this anti-oxidant response (Maddocks et al., 2013). While p53 promotes cell survival when the damage is low or transient, if damage within the cells is too high p53 can mediate the transcriptional upregulation of pro-oxidant genes, to promote ROS production and cell death (Berkers et al., 2013).

8 Catabolic pathways can also be promoted by the p53-mediated regulation of 5'- adenosine monophosphate-activated protein kinase (AMPK) and mTOR complex 1 (mTORC1). p53 positively regulates AMPK, via upregulation of Sestrin1 and Sestrin2, in conditions which cause ATP depletion such as hypoxia and nutrient deprivation (Budanov and Karin, 2008). AMPK can enhance the flux through energy producing pathways such as autophagy and fatty acid oxidation (FAO) while inhibiting biosynthetic pathways. Conversely, mTORC1 activity enhances cell growth and proliferation and is inhibited by p53. p53-mediated mTORC1 inhibition is achieved via AMPK activation and upregulation of various other p53 target genes which include Tsc2 (tuberous sclerosis complex 2) and Pten (phosphatase and tensin homologue) (Feng et al., 2007). The p53-mediated inhibition of mTORC1 also promotes autophagy, a pathway that can provide metabolic substrates for the cell via the breakdown of cellular components when the cell is under stress (Kim et al., 2011).

Finally, another branch of cellular metabolism that is regulated by p53 is lipid metabolism. p53 regulates both the synthesis and degradation of lipids. Srebf1c is transcriptionally repressed by p53. The sterol regulatory element binding transcription factor (Srebf2; otherwise known as Srebp) family of transcription factors mediate lipid biosynthesis and thus, p53-mediated inhibition of Srebf1c decreases the level of fatty acid synthesis (Yahagi et al., 2003). Conversely, p53 increases FAO levels through the transcriptional activation of lipin 1 (Lpin1) (Assaily et al., 2011). The transport of fatty acids into the mitochondria where they undergo β-oxidation is facilitated by p53- mediated induction of carnitine acyltransferases, carnitine-O-octanoyltransferase (Crot) and carnitine palmitoyltransferase 1A (Cpt1a) (Goldstein et al., 2012). In addition, p53 induces malonyl-CoA decarboxylase (Mcd), promoting CPT activity and thus FAO (Liu et al., 2014).

1.1.3. Tumour-suppressive functions of p53

As discussed above p53 plays important roles in protecting the cells from various cellular stresses and damage. Therefore, it is well-established that p53 is an essential tumour suppressor with inactivation of the p53 pathway occurring in the majority of cancers (Cancer Genome Atlas Research, 2014). The importance of the tumour suppressive role played by p53 was highlighted in early studies, which

9 compared mouse models of p53 deficiency to those with normal wild-type p53. Mice deficient for p53 displayed increased susceptibility to a range of different tumour types demonstrating the importance of the p53-mediated tumour suppressive functions in many different tissues (Donehower et al., 1992; Jacks et al., 1994). These observations were consistent with evidence from human disease where the familial Li-Fraumeni syndrome, which pre-disposes individuals to developing tumours such as sarcomas and breast tumours, is caused by germline mutations in TP53 (Li and Fraumeni, 1969a, b; Malkin et al., 1990; Srivastava et al., 1990).

Despite the importance of p53-mediated cell cycle arrest, DNA repair, senescence and apoptosis, mice engineered to carry mutations at three acetylation sites in the p53 gene, which render p53 incapable of mediating senescence, apoptosis or cell cycle arrest, are not prone to the early-onset tumour formation observed in the p53 null mice. This study indicates that p53 inhibits tumorigenesis by additional mechanisms, which include the p53-mediated regulation of cellular metabolism (Li et al., 2012). The role of p53 in regulating metabolic pathways, along with its more ‘classical’ functions such as cell cycle arrest and induction of apoptosis, likely explain why p53 function is lost in the majority of cancers.

1.2. TP53 mutations in cancer

The p53 pathway is disrupted in a large proportion of cancer and importantly, this is most commonly achieved by mutation in the p53 gene (Bullock and Fersht, 2001). The majority of these p53 mutations are missense mutations which cluster in the DNA-binding domain (Figure 1.5). By interfering with the ability of p53 to bind to its target genes, p53 mutations are considered to result in a loss of wild-type p53 function (Joerger and Fersht, 2007). Importantly, p53 mutations frequently result in the expression of a stable mutant p53 protein which accumulates in the tumour cells, instead of a loss of p53 expression (Iggo et al., 1990). The reason for the selection of missense mutations over p53 loss has been the subject of much investigation. In particular, it is not yet clear the extent to which p53 mutations exert gain-of-functions (GOF) or dominant-negative (DN) effects on remaining wild-type p53, which could explain why the mutations are selected for in many cancers (Figure 1.6).

10 R248 R273

R175

G245 R249 R282 Conformational Contact

p53 DBD

R273

R175

Figure 1.5. Mutations in the p53 gene in cancer.

Upper, examples of six hotspots for p53 missense mutations are shown (the labelled residues show human codons). P53 mutations fall into two classes: the conformational class (green) and the contact class (red). Adapted from (Bullock, 2001). Lower, structure of the human p53 core DNA-binding domain (DBD; blue) in contact with the DNA (grey) based on (Cho et al., 1994). Two frequently mutated sites are highlighted; conformational (green) and contact (red).

1.2.1. Different classes of p53 mutations

The p53 missense mutations can be divided into two classes based on the residues they affect and the overall impact on the structure of the p53 mutant protein (Figure 1.5). The first class is known as the ‘DNA-contact’ or ‘DNA-binding’ class and is formed of p53 mutations that affect residues that directly contact the DNA. The second class is known as the ‘conformational’ or ‘structural’ class. In the latter class, p53 mutations have significant effect on the overall conformation of the protein (Joerger et al., 2006; Joerger and Fersht, 2007). Examples of p53 mutations that frequently occur in human cancers are the contact mutations R248 and R273 and

Wild-type TP53 mutation Mutant

p53 p53

Accumulation due to oncogenic events

DN Wild-type Mutant p53 p53 LOF GOF

Growth suppression, Invasion/Metastasis, Apoptosis, DNA repair, Proliferation/Growth, Metabolic Regulation Drug Resistance

Adapted from Brosh and Rotter, 2009 Figure 1.6. Impact of p53 mutations in cancer.

The p53 gene is frequently mutated in cancer and the mutant p53 protein accumulates in the cancer cells. Mutations in p53 result in a loss of wild-type (WT) p53 function (LOF) and can potential exert dominant-negative (DN) effects on WT p53 and/or gain-of-functions. Examples of common functions WT p53 and GOFs of mutant p53 are shown (adapted from Brosh and Rotter, 2009).

the conformational mutations R175, G245, R249 and R282 (Figure 1.5).

Many studies in which the impact of p53 mutations has been explored did not contemplate that there may be functional differences between distinct p53 mutations. However, it is feasible that the mutant proteins function distinctly because the mutations have different effects on the protein (Joerger et al., 2006; Joerger and Fersht, 2007). In addition, conformational mutant proteins are reliant on the heat- shock protein 40 (Hsp40) family member Dnaja1 for protein stability whereas contact mutant proteins are not (Parrales et al., 2016). This implies that tumours with contact or conformational mutations could display different susceptibilities. Therefore, knowledge of the functional differences between p53 mutations is essential as it may

12 highlight differences in the therapeutic vulnerabilities of a group of tumours that is currently considered as one entity.

1.2.2. Mutant p53 gain-of-functions

The strong selective pressure for the retention of mutant p53 protein expression in tumours suggests that mutant p53 can acquire novel, oncogenic functions (gain-of- functions; GOF), that are advantageous to the cancer cells. Numerous mutant p53 GOFs have now been reported, which include the promotion of migration, invasion, proliferation and drug resistance (Figure 1.6; Muller and Vousden, 2014). Mutant p53 can increase signalling through growth factor receptors such as epidermal growth factor receptor (EGFR), platelet-derived growth factor receptor (PDGFR), hepatocyte growth factor receptor (MET) and transforming growth factor receptor β, which explains some of the tumour-promoting effects of mutant p53 (Adorno et al., 2009; Muller et al., 2009; Muller et al., 2013; Sauer et al., 2010; Weissmueller et al., 2014).

Importantly, p53 mutant tumours have been shown to be more aggressive and metastatic than their p53-null counterparts (Lang et al., 2004; Olive et al., 2004). Mutations in p53 are also associated with poor prognosis and increased resistance to standard cancer therapy (Cancer Genome Atlas Research, 2014; Masciarelli et al., 2014; Soussi and Beroud, 2001). Therefore, novel targeted therapies aimed at these tumours are urgently needed. Several of the oncogenic effects of mutant p53 are dependent on the p53 mutant binding and inhibiting p63 or p73, two transcription factors that belong to the p53 family (Adorno et al., 2009; Muller et al., 2009; Muller et al., 2013; Weissmueller et al., 2014). Therefore, mutant p53 likely exerts its oncogenic effects through interacting with other proteins and thereby regulating their function.

Interestingly, there is some in vivo evidence which hints that p53 mutations of different classes are not functionally equivalent; p53R172H/+ and p53R270H/+ (conformational and contact p53 mutations respectively) mice develop a distinct spectrum of tumours (Olive et al., 2004). Moreover, Li-Fraumeni patients with R248Q p53 mutations (contact) develop cancer earlier than those patients with p53 null or G245S mutations (conformational; Hanel et al., 2013). However, it is important to note that it may not be as simple as dividing the p53 mutations into classes as p53

13 mutations from within the same class, and even mutations at the same residue (depending on the amino acid incorporated), can have different effects on the structure or stability of the mutant protein (Joerger et al., 2006; Joerger and Fersht, 2007). Importantly, whether different p53 mutations can exert distinct GOFs in cancer cells and thus exhibit mutant-type specific susceptibilities still remains unclear.

1.2.3. Metabolic impact of p53 mutations

Interestingly, it is now emerging that mutant p53 can also exert its oncogenic effects through the manipulation of metabolic pathways. The mevalonate (MVA) pathway (Figure 1.7), the pathway of de novo cholesterol biosynthesis, is upregulated by mutant p53 in breast cancer and contributes to the invasive phenotype of the breast cancer cells (Freed-Pastor et al., 2012). The mutant p53-mediated upregulation of the mevalonate pathway and subsequent activation of Rho-GTPases activate the Yes Associated Protein 1 (YAP) and transcriptional coactivator with PDZ-binding motif (TAZ) oncogenes in breast cancer cells, promoting cell proliferation and self- renewal (Sorrentino et al., 2014). Lipid biosynthesis is required for the formation of new cellular membranes in proliferating cells and therefore upregulation of the cholesterol and fatty acid biosynthesis pathways is often observed in cancer cells (Santos and Schulze, 2012). As well as supporting membrane synthesis and cell proliferation via he synthesis of cholesterol, the MVA pathway generates intermediates such as geranyl geranyl pyrophosphate (GGPP) and farnesyl pyrophosphate (FPP) which are important for the lipid modification, membrane localisation and activation of the Ras superfamily of GTPases such as Rho and Ras (Konstantinopoulos et al., 2007). Therefore, the pathway has also been shown to be important for supporting not only the proliferation and growth of the cancer cells, but also their invasive potential via modulation of GTPases (Mullen et al., 2016).

The MVA pathway is regulated predominantly by the transcription factor Srebf2, although other Srebf family members (1a and 1c) may also play a role (Horton et al., 2003; Pai et al., 1998). Srebf2 is activated by insulin or when intracellular cholesterol levels are low or there is a high demand for cholesterol, for example to support high levels of proliferation. Inactive Srebf proteins are located in the endoplasmic reticulum (ER) membrane and they are activated by cleavage events which lead to

14 the translocation of the truncated active form to the nucleus where it can bind target genes (Ye and DeBose-Boyd, 2011).

Mutant p53 was also shown to drive glycolysis by increasing Glut1 translocation to the membrane, an effect mediated by RhoA activation (Zhang et al., 2013a), as well as promoting nucleotide biosynthesis to maintain proliferation and invasiveness (Kollareddy et al., 2015). An additional metabolic GOF of mutant p53 that has been reported is the inhibition of AMPK activation via the direct interaction of mutant p53 with the AMPKα subunit. Consequently, anabolic metabolism is not inhibited under conditions of energy stress, promoting tumour growth (Zhou et al., 2014).

I have previously described the role of wild-type p53 in the regulation of many aspects of cellular metabolism, which converge on an inhibition of glycolysis and anabolic, biosynthetic pathways and a promotion of catabolic pathways instead. These functions of p53 are in keeping with its tumour suppressive role as tumours cells frequently re-wire their metabolism to support proliferation and survival by increasing glucose uptake, glycolysis and flux through biosynthetic pathways (Berkers et al., 2013; Ward and Thompson, 2012). Studies have also shown that p53- /- cells display enhanced glycolysis and higher levels of synthesis of macromolecules such as lipids, relative to cells with wild-type p53 present (Jiang et al., 2011; Ma et al., 2007; Sinthupibulyakit et al., 2010). Interestingly, mutant p53 increases this metabolic rewiring further, with mutant cells displaying higher levels of glycolysis, lipid biosynthesis, MVA pathway upregulation and nucleotide biosynthesis than their p53 null counterparts (Freed-Pastor et al., 2012; Kollareddy et al., 2015; Zhang et al., 2013a; Zhou et al., 2014). It will be important to establish whether distinct p53 mutations differ in their ability to rewire metabolic pathways in tumours.

1.2.4. Mutant p53 dominant-negative effects

In addition to GOFs, it has been suggested that p53 mutant proteins can exhibit an inhibitory, dominant-negative (DN) effect on wild-type p53 function, thereby contributing to tumorigenesis (Muller and Vousden, 2014). As mutant p53 proteins accumulate in the cancer cells, partly due to a disruption of the Mdm2-mediated negative feedback loop, it is feasible that the majority of p53 mutants are able to

Acetyl-CoA Acetoacetyl-CoA Acat2

Hmgcs1 HMG-CoA Hmgcr Transcription factor:

Mevalonate Srebf2 Mvk Mevalonate-P Pmvk Mevalonate-PP Mvd

Isopentenyl-PP Squalene Idi2 Fdps Sqle Lss Geranyl-PP Cyp51a1 Dhcr24 Farnesyl-PP Tm7sf2 Msmol2 Geranylgeranyl-PP Dhcr7 Lipa Pggt1b Soat2 Farnesylated Rabggta Proliferation proteins e.g. Ras Cholesterol Geranylgeranylated Signalling

Invasive potential proteins e.g. Rho Ldlr Proliferation Cholesterol uptake

Figure 1.7. Mevalonate pathway in cancer.

The mevalonate pathway, depicted above, is the pathway of de novo cholesterol biosynthesis. Intermediates are also generated which are important for the farnesylation and geranylgeranlyation of GTPases such as Rho and Ras. Genes in this pathway are upregulated by the transcription factor, Srebf2 (orange). The pathway has been linked to cancer, supporting proliferation, cellular signalling and invasive potential of tumour cells. Acat2, Acetyl-coA acetyltransferase 2; Cyp51a1, Cytochrome P450 family 51 subfamily A member 1; Dhcr7, 7-dehydrocholesterol reductase; Dhcr24, 24-dehydrocholesterol reductase; Fdps, Farnesyl diphosphate synthase; Hmgcr, 3-hydroxy-3-methylglutaryl- CoA Reductase; Hmgcs1, 3-Hydroxy-3-Methylglutaryl-CoA Synthase 1; Idi2, Isopentenyl-diphosphate delta isomerase 2; Ldlr, Low density lipoprotein Receptor; Lipa, Lipase A, lysosomal acid type; Lss, Lanosterol synthase; Msmol, Methylsterol monooxygenase; Mvd, Mevalonate diphosphate decarboxylase; Mvk, Mevalonate kinase; Pggt1b, Protein geranylgeranyltransferase type I subunit beta; Pmvk, Phosphomevalonate kinase; Rabggta, Rab geranylgeranyltransferase alpha subunit; Soat2, Sterol O-acyltransferase 2; Sqle, Squalene epoxidase; Srebf, Sterol regulatory element binding transcription factor; Tm7sf2, Transmembrane 7 superfamily member 2.

16 exert DN activity, disrupting wild-type p53 function (Bartek et al., 1991). In agreement, p53 mutations have been reported to inhibit wild-type p53 function at least partially (Dong et al., 2007; Wang et al., 2011; Willis et al., 2004). As p53 functions as a tetramer, it is generally thought that mutant p53 proteins prevent the formation of functional wild-type tetramers by oligomerising with wild-type p53 monomers (Chan et al., 2004; Chene, 1998; Milner and Medcalf, 1991). Alternatively, the mutant p53 may form aggregates into which the wild-type proteins are sequestered and thus, inactivated (Xu et al., 2011).

However, the extent of the DN effects of mutant p53 in tumours is controversial. In human tumours, p53 mutations are often accompanied by loss of heterozygosity (LOH) at the p53 locus (Baker et al., 1990; Brosh and Rotter, 2009; Liu et al., 2016; Mitsudomi et al., 2000; Zienolddiny et al., 2001). This suggests that in many cancers the DN effects of p53 mutations on wild-type p53 are at best only partial as there is still selective pressure to lose wild-type p53 allele. Further investigation into the potential DN effect of p53 mutations is required in specific cancer types to ascertain the extent to which p53 mutations inhibit wild-type p53 function.

In summary, mutant p53 proteins may exert multiple effects in the cancer cells in addition to loss of wild-type function. Moreover, different p53 mutations may function distinctly in cancer cells, which needs to be explored further as any functional differences could uncover potential mutation-type specific dependencies and susceptibilities of the tumours.

1.3. Lung cancer

Lung cancer is the leading cause of cancer-related mortality. The disease has generally a very poor prognosis with only around 15% of patients surviving more than five years with the disease (Yang, 2009). There are two main subtypes of lung cancer based on the histology of the disease: small cell lung cancer (SCLC) and non- small cell lung cancer (NSCLC). NSCLC makes up more than 80% of lung cancers and can be divided further into different subclasses: adenocarcinomas and squamous cell, bronchioalveloar and large cell carcinomas (Chan and Hughes, 2015). My work will be focussed on the role of TP53 mutations in adenocarcinomas, the most common subtype of lung cancer (Parkin et al., 2005).

17 47%

Tp53

Deep deletion Mutation

100 no Tp53 mutation Tp53 mutation 80 P<0.05

a 60 v

v r

u s 40 %

0 0 200 400 600 800 1000

Survival (weeks)

Figure 1.8. TP53 mutations in human lung adenocarcinoma.

Upper: The frequency of TP53 mutations in lung adenocarcinoma Lower: overall survival of patients with p53 mutations versus patients with no p53 mutation. Data was taken from cBioPortal (Cancer Genome Atlas Research, 2014). Log-rank (Mantel-Cox) test.

1.3.1. Role of p53 mutations in lung cancer

The TP53 gene (Trp53, mouse; p53) is mutated in almost half of lung adenocarcinoma cases (Cancer Genome Atlas Research, 2014). Moreover, p53 mutations are associated with poor prognosis and reduced survival in NSCLC and so novel therapies to target these p53 mutant lung tumours are required (Figure 1.8;

18 Cancer Genome Atlas Research, 2014; Mao, 2001; Mitsudomi et al., 2000). Despite this, little is known about the functions of the mutant p53 mutations in this cancer type.

The extent of the GOFs that p53 mutations can exert in lung cancer is still unclear. Mouse models of lung adenocarcinoma have demonstrated that p53 mutations have no effect on lung tumour grade compared to p53 loss (Jackson et al., 2005). In contrast, overexpression of p53 mutants in lung tumour cells has been associated with oncogenic effects (Zhang et al. 2013; Muller et al., 2009; Muller et al., 2013).

DN effects of p53 mutations on wild-type p53 could also help explain this selection for p53 mutations observed in lung cancer. However, similar to mutant p53 GOFs, little is known regarding DN effects of p53 mutations in lung tumours. Common p53 mutations in lung cancer include R158, R175, R248 and R273 (Cancer Genome Atlas Research, 2014; Soussi et al., 2006). R175H and R273H mutant p53 proteins are among those which have demonstrated an inhibitory effect on wild-type p53 when expressed at high levels in vitro in other cancer types (Dong et al., 2007; Willis et al., 2004). However, loss-of-heterozygosity (LOH) of p53 has been observed in lung cancer suggesting that wild-type p53 is still able to function in the presence of the mutant as loss of wild-type p53 is still selected for even in this context (Mitsudomi et al., 2000; Zienolddiny et al., 2001).

1.3.2. Therapeutic strategies against p53 mutant lung tumours

Surgery is the standard therapeutic option for patients with early NSCLC (stages I-II), with or without adjuvant therapy (chemotherapy or radiotherapy; Hirsch et al., 2017). Treatment for advanced stage lung cancer continues to involve platinum-based doublet therapy or immunotherapy, unless the molecular alterations present allow the tumour to be treated with targeted therapy (for example the tyrosine kinase inhibitor erlotinib in EGFR mutant lung cancer or the ALK inhibitor crizotinib in lung cancer with ALK gene rearrangements (Hirsch et al., 2017). However, the prognosis is poor for those patients that progress after chemotherapy or develop resistance to targeted-therapy. Despite the development of drugs to treat some molecular subtypes of lung cancer (EGFR mutant positive or ALK gene rearrangement), p53 mutant lung tumours cannot be directly targeted in the clinic. Therefore, further work

19 is required to identify ways of specifically targeting these p53 mutant lung tumours, particularly since they represent nearly 50% of lung adenocarcinomas (Cancer Genome Atlas Research, 2014).

One approach to target p53 mutant tumours that is currently being investigated is the restoration of wild-type p53 functionality, as wild-type p53 can induce many tumour suppressive responses, as discussed earlier. It has been shown that the structural changes of some p53 mutations can be reversed by further point mutations, demonstrating that the structural changes induced by some p53 mutations are reversible (Joerger and Fersht, 2008). Following on from this, various compounds have demonstrated an ability to restore wild-type p53 function by stabilising the structure and promoting proper folding of the p53 protein (Boeckler et al., 2008; Bykov et al., 2002; Demma et al., 2010; Lehmann and Pietenpol, 2012; Soragni et al., 2016). One class of such molecules are the PRIMA-1 analogues which have reached clinical trials (Lehmann and Pietenpol, 2012). Importantly, one of the main strategies of interest for the treatment of p53 null tumours also involves the restoration of wild-type p53 activity (Cheok et al., 2011; Frezza and Martins, 2012). Therefore, it will be important to determine the extent to which different p53 mutations can exert DN effects on the wild-type p53. If wild-type p53 function prevails over the mutant, restoration of wild-type p53 may show efficacy in tumours even when the mutant p53 is still present.

Additional strategies to target p53 mutant tumours have centred around inducing the degradation of the mutant p53 proteins mediated by mdm2 and other ubiquitin ligases and targeting the heat-shock proteins which are required from stability of the mutant p53 proteins (Muller and Vousden, 2014). Moreover, inhibiting pathways downstream of mutant p53 in the tumour cells such as metabolic pathways (Freed- Pastor et al., 2012) and receptor-tyrosine kinase pathways (Weissmueller et al., 2014) may also be used to target p53 mutant tumours. In lung cancer less is known of the pathways downstream of the p53 mutations but if they exist their targeting could be of therapeutic benefit to patients with p53 mutant lung tumours.

20 1.3.3. Murine models of lung cancer

Mouse models have developed into an invaluable tool to model tumour evolution and response to therapy, in a physiologically relevant system. Mice are a useful species in which to model cancer due to their small size, their breeding ability and their genetic similarity to humans (Frese and Tuveson, 2007). We are now able to conditionally activate or inactivate genes in specific cell-types and also at any given time. Driver mutations can be introduced into target cells of interest, enabling the development of specific tumour types, for example lung tumours. This enables the investigation of how specific driver mutations contribute to the tumour evolution, phenotype and response to therapies. Moreover, multiple independent tumours form and evolve independently. Therefore, these tumours display distinct mutational profiles and so the role of the driver mutations can be examined in the context of disease heterogeneity (Frese and Tuveson, 2007).

One of the most well-studied mouse models of lung adenocarcinoma is based on the use of a Kras Lox-Stop-Lox(LSL)-G12D mutation (Jackson et al., 2005; Jackson et al., 2001). Kirsten Rat Sarcoma Viral Oncogene Homolog (Kras) is a GTPase, which is active in its guanosine triphosphate (GTP)-bound form, and can signal through multiple pathways such as the mitogen-activated protein kinase (MAPK) and phosphoinositide-3 kinase (PI3K) signalling pathways. Thus, Kras plays an important role in many cellular processes such as cell proliferation and survival. Mutations in KRAS occur in approximately 30% of human lung adenocarcinomas with the missense mutations resulting in a constitutively GTP-bound and active Kras protein (Cancer Genome Atlas Research, 2014; Gysin et al., 2011). Following Cre- recombinase administration, recombination and subsequent loss of the stop codon in these KrasLSL-G12D mouse models, the mutant KrasG12D is expressed from its endogenous promoter. This leads to the sporadic formation of Kras-driven lung tumours from the cells which have been in contact with the Cre-recombinase (Jackson et al., 2001).

Mice heterozygous for the KrasLSL-G12D allele can be crossed with other genetically engineered mice to allow the investigation of other genetic alterations in the context of KrasG12D-driven tumours, such as alterations in the p53 gene (e.g. missense mutations or p53 loss). Previous models of lung cancer driven by oncogenic Kras fail

21 RV TAM RV E8 ER p53ERTAM KI mice E5 E6 E7 E9 E10

- 4OHT + 4OHT

Activating p53ER protein p53ER protein signal Active Non-functional Functional WT p53 ~ p53 null ~ WT p53

Figure 1.9. p53ERTAM mouse model.

The conditional p53ERTAM allele allows wild-type p53 functionality to be toggled on and off in the animals or cell cultures depending on the presence or absence of 4-hydroxytamoxifen (4OHT), respectively. However, p53 activating signals are still required for wild-type p53 activity after 4OHT administration (Christophorou et al., 2005; Martins et al., 2006). KI: knock-in.

to recapitulate the later stages of human lung cancer, with lung tumours resembling early-stage human lung adenocarcinoma (Jackson et al., 2001). However, combining Kras-mutant with loss of p53 in the lung leads to the formation of lung tumours that better recapitulate the human disease, specifically demonstrating features of late- stage lung adenocarcinoma such as stromal desmoplasia and invasiveness (Jackson et al. 2005). Following administration of a viral titre of 106 PFU, Kras-mutant, p53 deficient mice (KrasLSL-G12D/+;p53Fx/Fx), develop tumours with 100 % penetrance and need to be sacrificed due to signs of lung tumour burden at 16 weeks after infection with AdenoCre. At this time the mice display multiple low grade, well-differentiated (grades 1-2) and high grade (grades 3-5) lung tumours, with the high-grade tumours representative of late-stage human lung adenocarcinoma.

The Kras-mutant, p53 deficient mouse model of lung adenocarcinoma (Jackson et al., 2005; Jonkers et al., 2001), is a useful way to model this disease type, firstly because it incorporates two key genetic alterations that often occur in lung

22 adenocarcinoma (p53 alterations occur in almost half of cases in addition to KRAS mutations occurring in almost a third of cases; Cancer Genome Atlas Research, 2014). This makes it a relevant model for investigating this disease. In addition, it is an endogenous tumour model with the Kras mutation and p53 mutations being expressed from their endogenous promoters rather than modelling the disease using supraphysiological levels of expression of the Kras and p53 mutant proteins. The Cre-recombination system utilised to activate the mutant genes also allows for spontaneous recombination events in the target cells in the lung after infection with Cre-expressing adenovirus through intranasal administration. This recapitulates the spontaneous nature of the mutations in human disease. Importantly, multiple tumours develop in each animal, each evolving independently, enabling the simultaneous analysis of tumours with distinct evolutionary ‘paths’ and thus, modelling the heterogeneity of human cancers (Jackson et al., 2005).

This Kras-mutant; p53-deficient mouse model can be utilised to investigate the impact of p53 loss or p53 mutations on tumour progression and response to therapy. In addition, incorporation of a conditional p53ERTAM allele into this model enables the restoration of wild-type p53 function in the mice in a time-controlled manner through administration of tamoxifen (hydroxytamoxifen, 4OHT; Christophorou et al., 2005). In the p53ERTAM mouse model the p53ER protein is always expressed but remains non- functional in the absence of its ligand 4OHT, due to the estrogen receptor (ER) domain which is fused to p53. This domain keeps p53 sequestered through interaction with heat-shock proteins until its 4OHT ligand binds. On ligand-binding p53ER dissociates from the heat-shock proteins, becoming functional (Christophorou et al., 2005). Therefore, this p53ERTAM allele enables the study of the effect of wild- type p53 in tumour models (Figure 1.9).

In summary, there is currently a lack of effective therapies to treat p53 mutant lung cancers, which represent a large proportion of lung adenocarcinoma cases. Therefore, further assessment of the roles of distinct p53 mutations in lung tumours is required to identify potential vulnerabilities in these tumours. Potential mutant p53 GOFs may highlight pathways downstream of mutant p53 which can be targeted to treat these tumours. We may be able to exploit these pathways to target all p53 mutant lung tumours or tumours with specific p53 mutations, depending on whether the GOFs identified are mutation-type specific. Moreover, analysis of the DN effects

23 of different p53 mutations would help us ascertain whether wild-type p53 restoration therapy would have efficacy in some or all p53 mutant lung tumours or whether the DN functions of the p53 mutations prevail over wild-type p53 function.

1.4. Aims of thesis

The overall objective of my thesis was to define the impact of p53 mutations in lung tumours. My research had the following aims:

1) To characterise the transcriptional impact of a conformational (R172H) and contact (R270H) p53 mutation in lung tumour cells, both in the presence and absence of wild-type p53.

2) To identify the gain-of-function effects of distinct p53 mutant classes in lung tumours.

3) To assess whether p53 mutants have an inhibitory effect on wild-type p53 function in lung tumours.

4) To define therapeutic strategies capable of targeting p53 mutant lung tumours.

24 2. Materials and Methods

Methods were adapted from (Turrell et al., 2017).

2.1. In vivo work

In vivo work was performed using mice with all the animal work for my project regulated under the Animals (Scientific Procedures) Act 1986 Amendment Regulations 2012 following ethical review by the University of Cambridge Animal Welfare and Ethical Review Body (AWERB). Animal work was carried out at Biological Resources Unit Cambridge Institute and Cambridge Biomedical Services. Mice were maintained under specific-pathogen-free (SPF) conditions, housed with no more than 5 animals/cage and were kept on an ad libitum diet. All animals were of from a mixed background (C57Bl/6/129/Sv). Compound mice were generated of the following genotypes: KrasLSL-G12D/+;p53Fx/Fx, KrasLSL-G12D/+;p53LSL-R172H/Fx, KrasLSL-

G12D/+;p53LSLR270H/Fx. KrasG12D/+;p53Fx/ER, KrasG12D/+;p53R172H/ER and KrasG12D/+;p53R270H/ER (Jonkers et al. 2001; Christophorou et al. 2005; Jackson et al. 2005).

2.1.1. Colony management

Breedings were set up with mice 6-8 weeks of age and the females were let to breed for typically 6 months. The health and the size of the litters were monitored. The presence of two KrasLSL-G12D alleles (absence of Kras) is embryonically lethal and so mice heterozygous for the KrasLSL-G12D allele were crossed with Kras+/+ animals. Animals from the p53R270H strain were not crossed with animals from p53R172H strain. P53R270H and p53R172H animals were not housed together. Examples of typical breeding pairs were: KrasG12D/+p53Fx/Fx and Kras+/+p53R270H/Fx; KrasG12D/+p53Fx/ER and Kras+/+p53R172H/Fx. Mice were earmarked for identification when weaned (3 weeks of age) and the ear biopsies were sent for genotyping. Genotypes (Kras and p53) of the mice were determined by a commercial vendor (Transnetyx, Cordova).

25 2.1.2. Induction of endogenous tumours

Mice were infected intransally with AdenoCre (AdCMV; Viral Vector Core, University of Iowa, UIowa-5) at 8 weeks of age and with 5x106 plaque-forming units (PFU) unless otherwise specified. The virus was thawed and diluted in Eagle’s Minimum Essential Media (EMEM; Sigma, M2279) to administer desired amount of

PFU/mouse in 64 μl. CaCl2 was added to a final concentration of 12 mM to activate the virus and the solution vortexed gently to mix and incubated for 20 minutes at room temperature. Meanwhile mice were anesthetised using isoflurane. After 20 minutes mice were administered with 32 μl of diluted virus to each nostril. After recovery from the anaesthesia, animals were kept in quarantine cages for 72 hrs to ensure any residual virus was inactive. After Adeno-Cre administration, multiple, endogenous lung tumours formed and animals were typically sacrificed 16-18 weeks post-infection.

2.1.3. Murine experimental cohorts

Animals were assigned to experimental cohorts (for example vehicle and drug treatments) to end up with similar numbers of males and females in each group. A minimum of 5 animals was used/cohort unless otherwise stated. We aimed to complete experiments using animals from the same infection round but where this was not possible we ensured we had comparable vehicle and treatment animals from the same infection group. Any animals which had been flagged as receiving the Adeno-Cre virus sub-optimally were not used in our experimental cohorts.

2.1.4. Genotyping and sequencing

Tails were collected for all mice in experimental cohorts when the animals were sacrificed to confirm the genotype of experimental animals. Genomic DNA (gDNA) was extracted by adding 500 μl of DNA lysis buffer (100 mM Tris.HCl pH 8.5, 5 mM EDTA, 0.2% sodium dodecyl sulphate, SDS, 200 mM NaCl) with 100 μg/ml Proteinase K (New England Biolabs, P8107S) added fresh. Tails were incubated in lysis buffer overnight at 55 C with gentle agitation. Samples were centrifuged to remove hair and gDNA was precipitated by adding 500 μl isopropanol to the tube and 26 inverting. Samples were centrifuged at maximum speed for ten minutes to pellet the gDNA which was then washed with 70 % ethanol and left to air dry before resuspending in ultra-pure water (18.2  filtered). The concentration of gDNA was determined using a Nanodrop spectrophotometer (Thermo Scientific) and input for the PCRs was 100 ng gDNA.

PCRs were performed as previously described for KrasG12D (https://jacks- lab.mit.edu/protocols) and p53Fx (Jonkers et al., 2001). p53ER PCR primers: (1) 5’ CCT CCA GCC TAG AGC CTT CCA AGC 3’; (2) 5’ GGT GAG ATT TCA TTG TAG GTG CC 3’; (3) 5’ GCA CAC AAA CTC TTC ACC CTG C 3’; WT band: 430bp; KI band: 700bp). P53ER PCRs were carried out using the following protocol: 1) 94 C for 2 minutes, 2) 94 C for 40 seconds, 3) 66 C for 40 seconds, 4) 72 C for 2 minutes, 5) go to 2) for 34 times, 6) 72 C for 10 minutes. Mutant p53 sequencing was carried our using the following primers: 172: (1) 5’ AGGTGTGTTGGCCATCTCTG 3’ (2) 5’ CTCAGGAGGGTGAGGCAAAC 3’; band: 743 bp; 270: 5’ TTTCTGTTCCACGAGTCCCG 3’, (2) 5’ AAAAGACCTGGCAACCTGCT 3’; band: 661 bp. Mutant p53 PCRs were carried out using the following protocol: 1) 95 C for 3 minutes, 2) 95 C for 40 seconds, 3) 66 C for 30 seconds, 4) 72 C for 1 minute, 5) go to 2) for 31 times, 6) 72 C for 10 minutes. TakaRa Ex Taq kit (RR001A) was used for all PCR reactions. Samples were run on a 1.5% agarose gel and visualised using Ethidium Bromide in a G:BOX gel imager (Syngene). Sequencing was performed by the Sanger Sequencing Service (Source BioScience) and analysed using 4Peaks software.

2.1.5. Tamoxifen and simvastatin treatments

For p53 restoration studies mice were treated at 15-17 weeks after Adeno-Cre- administration by intraperitoneal injection, (i.p.) with tamoxifen (1 mg/mouse once/day Sigma) or vehicle (Ctrl, peanut oil, Sigma) for 6 days. Simvastatin (200 mg/kg, EP, S0650000) or vehicle (Ctrl, 1% carboxymethylcellulose (Sigma) in dH2O) were administered 15 weeks after Cre, once daily for 6 days (unless otherwise stated) by oral gavage. For all in vivo treatments lungs were collected 24 hrs after the last treatment and a minimum of three mice/cohort were used (exact values are indicated).

27 2.1.6. Transplantation of lung tumour cell lines, imaging and survival studies

Transplant studies were carried out using 8- to 12-week-old wild-type mice, one cell line/genotype. Prior to transplantation cells were engineered to express luciferase (section 2.3.4). mice were irradiated (4 Gy, caesium source) 6 hrs before they received a 100 μl suspension of cells (in PBS) via tail-vein injection (performed by members of staff in the animal facility). Recipient mice received either 1x105 (metastatic potential cohorts) or 1x106 cells (simvastatin/Ctrl cohorts). Tumour growth was monitored weekly. Briefly animals were shaved over the chest area, injected (intraperitoneal) with D-luciferin (Caliper Life Sciences, 119222; 150 mg/kg), anesthetised using isoflurane and bioluminescence was measured ten minutes after injection by an IVIS Spectrum Xenogen machine (Caliper Life Sciences). Luciferase activity was measured using LivingImage software and normalized to control (luciferase-negative animal). Relative luciferase activity is the change from baseline (week 1 after transplantation) at indicated time point. All plotted mice developed lung lesions and additional thoracic tumors were observed in some mice. Simvastatin or control treatment (section 2.1.5) of transplanted models was carried out once/day, starting two weeks after transplantation and until endpoint. In all in vivo treatment groups lungs were collected 24 hrs after the last treatment and a minimum of three mice/cohort were used (exact values are indicated).

2.1.7. Tissue sampling and processing

Hearts were perfused with 5 ml phosphate-buffered saline (PBS) and the lungs were inflated with 10 % formalin. Tails were collected on dry ice and kept at -80 C until DNA extraction was performed. Lungs were kept in formalin overnight at room temperature. The next day the lungs were transferred over to 70 % ethanol and kept at 4 C until processing. Individual lobes were placed in tissue cassettes and submitted to the Histology core (Cambridge Institute, CRUK) or the Human Research Tissue Bank (Addenbrooke’s Hospital) for processing. Lungs were embedded in paraffin to expose as many lobes as possible and cut to display a minimum of 4 lobes/section. H&E staining was performed at the Histology Core or the Human Research Tissue Bank. Lung tumour burdern measurements were performed using 28 the H&Es and ImageJ to calculate the total area of the lungs and the area covered by lung tumours.

Single tumours that were collected from RNA analysis were isolated from the lungs and trimmed to remove any normal tissue. Tumours, or normal lung control tissue, were then immediately snap-frozen in liquid nitrogen and stored at -80 C until RNA extraction.

2.2. Immunohistochemistry

2.2.1. Ki67 staining and quantification

Histological analysis was carried out on the formalin-fixed, paraffin-embedded lung tissue. Sections (5 m) on glass slides were heated for 20 minutes at 55 C and then deparaffinised and hydrated by serial passaging through xylene (2x 5 minutes), 100 % ethanol ( 2x 2 minutes), 95 % ethanol (2x 2 minutes) , 70 % ethanol (2x 2 minutes) and water. Staining for Ki67 (Bethyl-Laboratories, IHC-00375) was performed on the Leica Bond-Max (Bond Polymer Refine Detection, DS9800). Slides were then dehydrated, mounted (DPX mounting media, Sigma) onto coverslips (VWR; 22 x 50 mm borosilicate glass) and scanned using the Zeiss Axio Scan Z1. The percentage of Ki67-positive cells/tumour was calculated by counting the number of cells (hematoxylin) and number of Ki67-positive cells by hand using the counter tool on ImageJ or by the automated Halo 2.0 software (Cytonuclear). The number of tumours counted are as indicated.

2.2.2. TUNEL staining and quantification

Terminal deoxynucleotidyl transferase dUTP Nick-End Labeling (TUNEL) staining was performed on formalin-fixed, paraffin-embedded (5 m) lung tissue. Slides were heated for 20 minutes at 55 C and then deparaffinised and hydrated as above. 20 μg/ml Proteinase K (New England Biolabs, P8107S) in PBS was added (150 μl/slide) for 15 minutes at room temperature. The staining was performed as stated by the manufacturer’s instructions with the following adaptions: 45 μl equilibration buffer was added/slide and then 5 seconds after 30 μl reaction buffer with 10 μl was added/slide. Slides with incubated for 1 hr at 37 C before the reaction was stopped 29 with Stop buffer. Slides with then washed with TBS-T (0.1 %) and mounted with Everbrite mounting media with DAPI (Biotium, 23004) onto coverslips (VWR; 22 x 50 mm borosilicate glass). Slides were then scanned using the Zeiss Axio Scan Z1. The percentage of TUNEL-positive cells/tumour was calculated by counting the number of cells (DAPI) and number of TUNEL-positive cells by hand using the counter tool on ImageJ or by the automated Halo 2.0 software (CytonuclearFL). The number of tumours counted are as indicated.

2.3. Isolation and maintenance of lung tumour cell lines

2.3.1. Isolation of lung tumour cell lines

KrasG12D/+;p53R270H/ER, KrasG12D/+;p53R172H/ER and KrasG12D/+;p53Fx/ER murine lung cancer cell lines were generated prior to the start of this research project. For the generation of the KrasG12D/+;p53R270H/Fx, KrasG12D/+;p53R172H/Fx and KrasG12D/+;p53Fx/Fx cell lines tumours were harvested from three KrasG12D/+;p53R270H/Fx and two KrasG12D/+;p53R172H/Fx mice at 14-15 weeks post-Adeno cre infection with 5x107 PFU and from two KrasG12D/+;p53Fx/Fx mice, 15-17 weeks post-Adeno cre infection (5x107 PFU). Tumours were collected in Hank’s Balanced Salt Solution (HBSS, with Ca2+ and Mg2+), mashed and collected again in fresh HBSS. After centrifugation at 1,000 rpm at 4C for 5 minutes the pellet was resuspended in HBSS (no Ca2+ and Mg2+) with 4 mg/ml collagenase dispase (Roche) and incubated at 37 C for 2.5 hours under agitation. HBSS (with Ca2+ and Mg2+) with 10 % fetal calf serum (FCS; Life Technologies) and 3.33 l/ml DNase I (Invitrogen; 1 U/μl; 18068-015) was added, incubating the cells on ice for 10 minutes. After centrifugation, cells were resuspended in Dulbecco’s Modified Eagle Medium (DMEM)/ F-12 (1:1) media (Life Technologies), supplemented with 10% FCS, 2mM L-Glutamine (Gibco), 100 U/ml penicillin (Sigma), and 0.1 mg/mL streptomycin (Sigma), filtered through a 70 m filter and plated. After 1.5 hours the media was re-plated to separate the epithelial population.

30 2.3.2. Genotyping of cell lines

Genomic DNA extraction and subsequent genotyping and sequencing of p53 for the cell lines was performed in the same way as for the genotyping for the mouse tails described above (section 2.1.4).

2.3.3. Maintenance of lung tumour cell lines

All lung tumour cell lines were maintained in DMEM/F-12 media supplemented with 10% FCS, 2mM L-Glutamine, 100 U/ml penicillin and 0.1 mg/mL streptomycin

(referred to as complete DMEM/F-12 media), at 37 °C and 5 % CO2. All cells tested negative for mycoplasma and were typically used at passages 10-20 after isolation.

2.3.4. Generation of luciferase-expressing lung tumour cell lines

Bacteria were grown up from bacterial glycerol stocks of the MSCV-luciferase-PGK hygromycin retroviral plasmid (Addgene, 18782) in LB broth (10 g Bacto Tryptone/5g

Bacto Yeast Extract/10 g NaCl/1L dH2O; adjusted to pH 7) with ampicillin (100 g/ml), overnight at 37 C. A Maxiprep (Qiagen, 12663) was performed to extract DNA and DNA was concentrated as required by performing ethanol precipitation (0.3 M sodium acetate, pH 5.2 and 100 % ethanol).

Phoenix Eco cells were plated at 2.5x106 cells/10 cm plate. The following day when the cells with 60-70% confluent they were transfected with 20 g plasmid/plate. For lung tumour cell line to be transduced, one plate of Phoenix Eco cells was transfected. For each plate to be transfected, 20 g plasmid (maximum volume 10 l) was added to 1.5ml Opti-MEM (Gibco, Life Technologies, 31985-062) and 60 l Lipofectamine 2000 (Invitrogen, 11668-027) was added to 1.5ml Opti-MEM in a second tube. The solutions were then incubated separately for 5 minutes before mixing and incubating for a further 20 minutes at room temperature. Complete DMEM media (3 ml) was added/plate of cells and then 3 ml of the DNA- Lipofectamine mix was added drop-wise on top. The next day the cells were rinsed with PBS and 5.5 ml complete DMEM/F-12 media was added. The murine lung

31 tumour cell lines were also plated at 1.0x106/10 cm plate. The following day the media was collected from the Phoenix Eco cells, which were re-fed with 5.5ml fresh media. The viral media was then filtered through a 0.45 l filter, 8 g /ml polybrene (Millipore, TR-1003-G) was added and the viral media was then added to the murine lung tumour receiver cells. A second round of retroviral transduction was performed the following day.

The next day, lung tumour cells (control and transduced plates) were then placed in antibiotic selection until the cells on the control plate had all detached (Hygromycin B selection, 250-450 g/ml, Invitrogen, 10687010). Luciferase-expressing cells were expanded and tested to confirm luciferase activity using a Dual-luciferase reporter kit (Promega, E1910), following the manufacturer’s instructions.

2.4. In vitro treatments

Treatments were performed in triplicate and, unless otherwise stated, treatments were refreshed daily. For p53 restoration studies, cells were treated with 100 nM 4OHT (Sigma, H7904) or Ctrl (ethanol) the following day after the cells were plated (2D and 3D cultures as indicated). For simvastatin experiments, cells were treated with vehicle or simvastatin (Sigma, S6196; 1 M) for 24 hrs (western blots) or 72 hrs (3D colony growth). Simvastatin was activated with addition of 125 l ethanol and 187.5 l NaOH (0.1M)/5 mg drug and incubation at 50 °C for 2 hours. Simvastatin was then diluted in water and aliquots of the stock solution were stored at -80 °C. Vehicle was made in the same way, without addition of the drug. Fatostatin (Sigma, F8932 was used at a working concentration of 10 M in DMSO; Sigma). All simvastatin and fatostatin experiments were performed in 3D cultures. Addback experiments were performed by addition of Mevalonate (0.3 mM in DMSO; Sigma),

GGPP (40 M; Sigma, G6025, in methanol:NH4OH, 7:3) or 1 x cholesterol (Sigma, S5442).

2.5. 2D proliferation assays

Cells were seeded at 15,000 or 5,000 cells/well in 6 well plates. The next day, where relevant, cells were treated with 4OHT/vehicle. Cells were trypinised and viable cells

32 were counted using trypan blue exclusion in triplicate every 24 hrs at indicated time points. Briefly, an equal volume of trypan blue stain (Gibco, Life technologies, 0.4 %) was added to the cell suspension and live cells were counted using a haemocytometer (blue cells were excluded).

2.6. 3D cultures

For all 3D cultures, cells were seeded around the edge of the wells in a 24-well plate in 3:2 Matrigel (Corning, 354234):DMEM/F-12 media unless otherwise stated. The plate was incubated at 37 C for 30 minutes to allow the Matrigel to set before the addition of media to the wells. 5,000 cells were seeded/well for the colony formation assays, with each condition seeded in triplicate. After 72 hours, 3-5 images were taken for each well at 20x magnification and the colony size measured using the Image J software. For all other 3D growth assays and for microarray sample preparation, 50,000 cells were seeded/well. For cell recovery, Matrigel was dissociated with Cell Recovery Solution (Corning, 354253) at 4 °C for 45 minutes before collection.

2.7. Microarray analysis

Microarray samples were collected by Meiling Gao, prior to the start of my PhD. Four independent lung tumour cells lines/genotype (KrasG12D/+p53FxER, KrasG12D/+p53R172H/ER and KrasG12D/+p53R270H/ER) were cultured in 3D Matrigel cultures for 48 hrs. Cultures were then treated with either vehicle (ethanol) or 4OHT for 2 hrs or 8 hrs. The media was then removed, the cultures washed with PBS and the Matrigel was dissociated and the cells collected as previously described (section 2.6). RNA was extracted using the RNeasy Qiagen kit (74106) with on-column DNase digestion.

Illumina MouseWG-6 version 2.0 Expression BeadChip (Department of Pathology, Cambridge University) was used for the microarray analysis. Standard Illumina protocols were used for the scanning of arrays. Data were normalised by Dr Shamith Samarajiwa using R (http://www.R-project.org) and Bioconductor (Gentleman et al., 2004) and spatial artefacts were also removed using BASH (Cairns et al., 2008) and

33 HULK algorithms from the beadarray package (Dunning et al., 2007). Data were then log2 transformed and quantile normalised and an average value from each sample was calculated from two runs.

Differentially expressed genes were defined as those where there was an average fold change between groups >1.4. A cut-off of a standard deviation<0.5 between log2 expression values within each cohort was also applied to account for intragenotype variation. Pathway analysis of differentially expressed genes was performed using Ingenuity Pathway Analysis (IPA, Qiagen) software and Fisher’s exact test was used to determine statistical significance of canonical pathways (P<0.05). The Multiple Experiment Viewer (MeV) software was used for the generation of heatmaps based on normalised log2 expression data values. Principal component analysis was performed using R (http://www.R-project.org) with the help of Sarah Davidson. Gene Set Enrichment Analysis (GSEA, Broad Institute) was used to generate enrichment plots for the Srebf2 target genes. Srebf2 target genes were based on target gene list from IPA and ChIPseq datasets (http://chip-atlas.org). p53 “canonical” targets were defined based on functional and expression studies (GSEA, Bieging et al., 2014; Lehar et al., 1996; Okamoto and Beach, 1994) or ChIP data (http://chip-atlas.org; Fischer, 2017).

2.8. qRT-PCR

Cells were pelleted from 10 cm plates (70-80% confluent) or 3D cultures as previously described (section 2.6) and RNA was extracted using the RNeasy kit (Qiagen, 74106), following the steps for on-column DNase (Qiagen) treatment. RNA concentration was measured using a Nanodrop spectrophotometer (Thermo Scientific). Complementary (c)DNA synthesis was performed using the iScript cDNA synthesis kit (Biorad, 1708891). For each reaction, 4 μl 5X iScript reaction mix, 1 μl iScript reverse transcriptase, 2 μl 300 ng/μl RNA and 13 μl nuclease-free water were combined and run on a MJ Research Tetrad PTC-225 Thermal Cycler using the standard kit protocol (5 min at 25 °C, 30 min at 42 °C, 5 min at 85 °C). The quantitative real-time polymerase chain reaction (qRT-PCR) was performed in 5 μl reactions using Taqman system (2.5 μl Taqman Universal PCR Master Mix (Applied

Biosystems, 4304437), 1.25 μl nuclease-free H2O, 0.25 μl probe and 1 μl (7.5 ng) cDNA) on 384 well plates. Plates were sealed with foil, spun at 2,000 rpm for 1 min

34 and loaded onto the LightCycler 480 II (Roche). The samples were analysed using standard Taqman hydrolysis programme: 1 pre-amplification step (95 °C, hold 10 min); 45 amplification cycles (95 °C, hold 10 sec, 60 °C, hold 30 sec, 72 °C, reading, hold 1 sec); 1 cooling step (40 °C, hold 30 sec). The following probes were used for Taqman analysis: Cdkn1a: Mm00432448_m1; Mdm2: Mm01233136_m1; Bbc3: Mm00519268_m1; Mvd: Mm00507014_m1; Pmvk: Mm01212763_m1; Sqle: Mm00436772_m1; Srebf2: Mm01306292_m1 Pilra: Mm04211819_m1; Nr2f2: Mm00772789_m1; Slc2a9: Mm00455122_m1; Phlda3: Mm00449846_m1; Sesn2: Mm00460679_m1; Bax: Mm00432051_m1; Gapdh: Mm99999915_g1; Euk 18S rRNA (4352930E) all from Applied Biosystems.

2.9. FACS

2.9.1. EpCAM/PDGFR

Cells were trypsinised and washed twice with cold PBS. Cells were then resuspended in 200 l staining buffer (2 % FCS in PBS) with either an allophycocyanin (APC)-conjugated antibody against epithelial cell adhesion molecule, EpCAM (1:100; Biolegend, 118213) or platelet-derived growth factor receptor, PDGFR (1:50; eBioscience, 17-1401-81) and incubated for 1 hr in the dark at 4 C. Cells were washed twice in staining buffer and resuspended in staining buffer for flow cytometry. Unstained controls for each sample were also collected. In addition, negative and positive controls for EpCAM (mouse embryonic fibroblasts and an epithelial tumour cell line respectively) and Pdgfra (an epithelial tumour cell line and mouse embryonic fibroblasts respectively) were also collected. Samples were run on an LSR II flow cytometer (BD Biosciences) and data was analysed using FlowJo software (Tree Star). Briefly, FSC-A x SSC-A and FSC-A x FSH-H gating was performed to exclude the debris and identify the single cell population and a gate for Pdgfra-positivity was generated on a sample which was a mix of the positive and negative controls. This gate was then applied to all samples to determine percentage Pdgfra-positivity (examples of the gating performed are shown in Appendix 4).

35 2.9.2. BrdU/PI

Cells were seeded in 3D Matrigel cultures at 50,000 cells/well in 24 well plates (3D cultures) or onto 6 well plates at 50,000/well (2D cultures). Cells were treated with ethanol (vehicle) or 4OHT for the indicated time points. Bromodeoxyuridine (BrdU; Thermo Fischer Scientific, 000103; 10 M) was then added 2 hours before collection and the treatment was refreshed when the media was changed. After the 2 hours incubation at 37 C, cells were washed with PBS and trypsinised. For 3D cultures, cells were dissociated and collected as previously described (section 2.6) and trypsinised. Cells were washed twice in cold PBS. After the last wash cells ice-cold ethanol was added drop-wise under agitation and cells were fixed overnight at 4 C. Cells were then washed with cold PBS twice more, spun down, resuspended in 200 l RNAse A (Qiagen, 0.5 mg/ml in PBS) and incubated at 37 C for 30 minutes. Cells were washed again and re-suspended in 200 l HCl/H2O/0.5% Triton X solution and incubated for 20 minutes at room temperature. Then 2 ml of 0.1M Na2B4O7 (pH 8.5) was added for neutralisation for 5 minutes. Cells were then washed in PBS-T (0.5 %). Anti-BrdU-FITC (Flow BrdU kit, BD Biosciences) was added to each sample (20 l antibody and 50 l PBS-T) and cells were incubated for 1 hr at room temperature in the dark. Cells were then washed with PBS-T and resuspended in PBS with 20 g/ml propidium iodide (PI, Sigma, P4170). Samples were run on an LSR II flow cytometer (BD Biosciences) with unstained and single-stained controls and data was analysed using FlowJo with compensation performed (Tree Star, version 10). Gating to exclude cellular debris and identify the single cell population was performed as above (section 2.9.1) and the BrdU-positive population was selected to determine the percentage of cells in S-phase (an example of the gating performed is shown in Appendix 5).

2.10. Cell death analysis

2.10.1. Ethidium homodimer/Calcein AM kit

Cells were seeded at 50,000 cells/100 μl Matrigel (Phenol red free, 356237)/Media mix. The following day the 3D cultures were treated with 4OHT/vehicle for 72 hrs in triplicate. The percentage dead cells was then measured and calculated using the

36 Ethidium homodimer/Calcein AM-based LIVE/DEAD Viability/Cytotoxicity Kit (Molecular Probes, MP03224), following the manufacturer’s instructions for the fluorescent microplate protocol.

2.10.2. Cleaved caspase-3 immunofluorescence

For immunofluorescence, cells were seeded at 50,000 cells/100 μl Matrigel (Phenol red free, 356237)/Media mix on glass chamber slides and were treated with 4OHT/Ctrl the next day for 72 hrs. Cultures were then at fixed in 2% paraformaldehyde (PFA; 1 hr) followed by 1% Glutaldehyde (1 hr), washed three times in wash buffer (PBS/0.15% Triton-X) and blocked in blocking buffer (PBS 0.15%/Triton-X/3% serum) for 30 mins at room temperature. Incubation with primary antibody (anti-Cleaved caspase-3, Cell Signaling, 9664S, 1:200) in blocking buffer was performed overnight at 4 C. Cultures were subsequently washed overnight at 4 C with wash buffer, with 3 changes, of buffer and incubated with secondary antibody (Thermo Scientific A11008; 1:200) in blocking buffer overnight at 4 C, in the dark. All subsequent steps after addition of the secondary antibody were also performed in the dark. Again the cultures were washed overnight at 4 C with wash buffer, with 3 changes in buffer. Following this, cells were incubated with DAPI (1 μg/ml in PBS) for 20 minutes at room temperature. Cultures were then washed in PBS for 3 minutes, fixed again with 4 % PFA for 15 minutes at room temperature and washed again in PBS for 10 minutes. PBS was refreshed and the 3D cultures were imaged using a Leica SP5 confocal microscope and analysed using Leica imaging software.

2.11. Senescence assay

The Senescence ß-galactosidase staining kit (Cell Signalling, 9860) was used following the treatment of cells in triplicate with 100 nM 4OHT or vehicle (ethanol) for 72 hours, refreshed daily. H460 and A549 lung cancer cells were treated with Palbociclib (Sigma, PD-0332991, 2 µM) for 72 hours, refreshed daily, as a positive control. As instructed by the manufacturers, cells were fixed with the fixative solution for 5 minutes and following washing, incubated with ß-galactosidase staining solution

37 (pH 6) overnight at 37˚C no CO2. Images were obtained from 5 random fields of view for each cell line in triplicate using an Olympus BX53 brightfield microscope.

2.12. Migration and invasion assays

2.12.1. Scratch assay

40,000 cells were seeded in a 24 well plate (2D). Cells were cultured for 48 hrs before each well was scratched twice using a yellow tip and media refreshed. Leica live cell microscopy was used to monitor scratch closure, with images acquired every hour for 24 hrs (Zeiss Live Cell microscope). Cells were cultured in complete DMEM/F-12 media while the closure of the scratch was measured over time. Velocity was calculated for each cell line from three measurements per scratch, measuring the horizontal distance moved by the cells/hour.

2.12.2. Invasive growth

Inverted invasive growth assays were performed to assay invasion with 40,000 cells seeded in each transwell (Costar, 3422; inverted) in 50 l of a growth factor-reduced Matrigel (Coring, 356231), and serum free basal DMEM (A14430-001) mix (4:1 Matrigel: media). The plate was incubated at 37 C for 30 minutes to allow Matrigel to set and then 200 l of the diluted Matrigel was added on top of the cells and the plate was incubated for a further 30 minutes at 37 C. 120 l of complete DMEM media with 20 % FCS was added on top of the Matrigel and 500 l of the serum free basal media was added below the transwell. The media were changed daily to maintain the gradient. After 8 days the transwells were washed in PBS and stained with 2 mM calcein AM (Life Technologies). Z-stack images with slices every 15 m were acquired on a Zeiss LSM 510 Meta confocal microscope (3-4 z-stacks for each well) and fluorescence intensities were calculated on Image J to assess the numbers of invaded cells.

38 2.13. Metabolomics

2.13.1. Metabolite extraction

Cells were seeded at 2.5-4.5 x 106 per well in 6 well plates to achieve 70-80 % confluency the next day. Each cell line was seeded in quadruplicate, with one well for each to allow determination of cell number. The following day the media was refreshed, with media also added to three empty wells to allow for metabolites to be collected from the media only as a control. After 8 h the cell number was determined using a haemocytometer and the metabolites were extracted. 50 μl of media from the remaining three replicates for each condition was collected in 750 μl pre-cooled metabolite extraction buffer (50 % methanol, 30 % acetonitrile, 100 ng/ml HEPES buffer in water). Samples were agitated on the thermomixer for 15 min at 4 °C and then spun at maximum speed for 10 min at 4 °C. The supernatant was transferred to mass spectrometry (MS) vials and the samples stored at – 80 °C. The cells were washed quickly in PBS, three times on ice. Pre-cooled metabolite extraction buffer was added (0.5 ml/ 106 cells) and then collected and processed as before.

2.13.2. Mass spectrometry

Liquid chromatography (LC)-MS analysis was performed by the Frezza lab on a Thermo QExactive Orbitrap mass spectrometer using a pHILIC column (150 mm × 2.1 mm, internal diameter (i.d.) 3.5 µm) with the guard column (20 mm × 2.1 mm i.d. 3.5 µm) (HiChrom, Reading, UK) and the following settings: flow rate: 180 µL/min; mobile phase: A: 20 mM ammonium carbonate plus 0.1% ammonia hydroxide in water, B: acetonitrile; gradient: 0 minutes 70% of B, 1 minute 70% of B, 16 minutes 38% of B, 16.5 minutes 70% of B, 25 minutes 70% of B.

2.13.3. Analysis

The peaks were analysed using the Spectra analysis software XCalibur Qual Browser and XCalibur Quan Browser (Thermo Scientific) and the values were normalised to the total metabolite levels (the sum of the peak areas) for each sample. MeV software was used for the generation of heatmaps. Principal

39 component analysis was performed using R (http://www.R-project.org; Gaude et al; 2013) with the help of Edoardo Gaude.

2.14. Extracellular flux assays

2.14.1. Plate preparation

The Seahorse Biosciences XFe24 system was used to assess extracellular flux in the previously mentioned murine lung tumour cell lines. Briefly, cells were plated in 24 well assay plates (20,000 cells per well) in triplicate and allowed to adhere overnight in full culture media, at 37 °C with 5 % CO2. Probe plates were prepared by adding 1 ml calibration buffer per well, and incubated overnight at 37 °C, without

CO2. The following day assay media was prepared by supplementing seahorse media (8.3g DMEM (Sigma D5030), 1.85 g NaCl, 0.015g Phenol Red in ultra-pure

H2O/1L, pH 7.4) with 0.45 % (w/v) glucose, 2 % FCS, 4 mM L-glutamine and sodium pyruvate (NaPyr; 0.5 mM) and corrected to pH 7.4. The 24 well assay plate media was then replaced with 675 μl per well of freshly prepared assay media, and the plate incubated at 37 °C, no CO2 for 30 min. The probe plate was prepared with the mitochondrial inhibitors: 75 μl 10 μM Oligomycin in injection port 1, 83 μl 5 μM CCCP in injection port 2, and 92 μl 10 μM Rotenone in injection port 3 in assay media.

2.14.2. Running the assay

The probe plate was inserted into the Seahorse XF analyser and then after 15 minutes of calibration the cell plate was inserted into the analyser. Oxygen consumption, extracellular acidification and proton pump rate at each step was calculated (basal, injection 1, injection 2 and injection 3), with a 2 minute mix and 1 minute read cycle repeated three times at each step. After the assay, the media was removed and the cells were lysed on ice for 15 minutes in 50 μl radioimmunoprecipitation assay (RIPA; 150 mM NaCl/0.1 % Triton X/0.5 % sodium deoxycholate/0.1 % SDS; 50 mM Tris-HCl pH 8.0) buffer/well. Protein quantification was performed by the bicinchoninic acid (BCA) assay (Thermo Scientific) with bovine serum albumin (BSA) standards, and results normalised accordingly.

40 2.15. Measurement of total protein and cell size

The size of trypsinised cells was measured using a Countess (Invitrogen) cell counter. To determine protein content, cells were trypsinised and counted using a haemocytometer. 10,000 cells were pelleted in triplicate/cell line. Cells were then lysed on ice in 30 μl RIPA buffer for 30 minutes. Protein concentration was measured using the BCA kit and BSA standards.

2.16. Intracellular lipid measurements

2.16.1. Oil red O staining

Cells were seeded (1x106 cells/well of a 6 well plate) in complete DMEM/F-12 media. In total 6 wells were seeded/cell line to perform the Oil Red O staining in triplicate and have wells to perform cell counts in order to normalise to cell number. The following day working Oil Red O solution was prepared by mixing 3 parts Oil Red O stock solution (Sigma, O1391) with 2 parts dH2O. The working solution was left to stand for 10 minutes and then filtered through filter paper. Three wells/ cell line were washed with PBS and fixed in 10 % formalin (Sigma) for 45 minutes and then washed twice in deionised (dH2O). 60 % isopropanol was added to the cells for 5 minutes. The additional three wells/cell line were trypsinised to determine the cell number/well. Working Oil Red O solution was then added to the cells for 10 minutes and then the cells were washed four times with dH2O. Fresh dH2O washed added to each well and images were taken using a Zeiss bright-field microscope. The wells were then dried, the stain was eluted with 100 % isopropanol and the absorbance at 500 nm was measured using a Tecan microplate reader. The absorbance was then normalised to cell number.

2.16.2. Cholesterol staining and quantification

Cells were cultured in glass chamber slides (Nunc, 177402) and then treated with Cell Mask Orange Plasma Membrane Stain (Molecular Probes, Life Technologies, C10045; 1:1000) for 10 minutes at 37 °C. Cells were then washed in PBS, fixed with 4 % PFA for 10 minutes and then washed with 0.1 % TBS-T. Cells were stained with

41 filipin (Sigma, 50 μg/ml) in PBS at room temperature for 1 hour in the dark. Cells were washed again with 0.1 % TBS-T and mounted onto coverslips using SlowFade Gold Antifade (Life Technologies, S36936). Random images were taken (10/cell line) using a Zeiss Live Cell microscope. Filipin staining was not assessed quantitatively because of its propensity to bleach extremely quickly and instead was used to assess cholesterol localisation.

Intracellular cholesterol was measured for cell lines plated in triplicate in 2D cultures. Cells were trypsinised and collected. The cholesterol was measured using the Amplex Red cholesterol kit (Molecular Probes, A12216), following the manufacturer’s instructions. Cholesterol values were normalised to total protein content, measured by the BCA kit and BSA standards.

2.17. SDS-PAGE and Western blotting

Cells were harvested from 2D or 3D cultures, as previous described. Pellets were lysed on ice for 30 min in lysis buffer (50 mM Tris.HCl pH 7.5/150 mM NaCl/20 mM EDTA/0.5 % NP40) with a protease inhibitor tablet added (1 tablet/50 ml lysis buffer; Roche, 19543200) and 1 mM phenylmethylsulfonyl fluoride (Thermo Scientific), 1 mM Dithiothreitol (Melford Laboratories), 1 mM sodium fluoride (Sigma), 1 mM - glycerophosphate, 1 mM sodium pyrophosphate (Sigma) and 0.2 mM sodium orthovanadate (Sigma) added fresh. The cells were spun at 14,000 rpm, 4 °C for 20 minutes. The protein lysate was retained and quantified using the BCA assay with BSA standards. 30 μg of total protein was then prepared with 4x sample buffer (8 % SDS/40 % glycerol/0.008% Bromophenol blue/20 % β-mercaptoethanol (added fresh)/250 mM Tris pH 6.8). The samples were denatured for 5 min at 95 °C and loaded to pre-cast gels 4-12 % Bis Tris (NUPAGE, NP0321BOX) gel, Novex Life Technologies). The gels were resolved at 100V in MOPS (NUPAGE; NP0001) running buffer until desired separation had been achieved.

The protein was then transferred from the gel to methanol-activated polyvinylidene fluoride membrane (Immobilon-P) using Mini Trans-Blot™ electrophoretic transfer cell system (BioRad Laboratories (UK) Ltd.) for 1 hr in Tris-Glycine buffer from Biorad, with 20 % methanol. Membranes were blocked for 1 h in 5 % milk in TBS-T (0.5 %). After blocking the membranes were incubated with primary antibodies (in 1 42 % milk in TBS-T for all antibodies other than Cell Signalling antibodies which were made up in 5 % BSA in TBS-T) overnight at 4°C with agitation. Membranes were then washed for 10 minutes in TBS-T three times and incubated secondary antibody (in 1 % milk in TBS-T) for 1 hr at room temperature. Three 15 minute washes in TBS- T were then performed, followed by 1 minute detection with enhanced chemiluminescence (ECL) western blot detection reagent (Amersham, RPN2106). Membranes were placed in acetate and exposed to X-ray film before developing using the Xograph Compact X4 automatic processor. The band intensities were measured using ImageJ software and normalised to the loading control.

Table 2.1. List of primary antibodies used for Western blotting.

Target Company Product Host Dilution Akt Cell Signalling 4691 Rabbit 1:1,000 Ampk Cell Signalling 2532 Rabbit 1:1,000 Beta-actin Sigma A5441 Mouse 1:10,000 Erk Cell Signalling 9102 Rabbit 1:1,000 LC3 Cell Signalling 2775 Rabbit 1:1,000 p53 Vector VP-P956 Rabbit 1:2,000 Phospho-Akt Cell Signalling 4090 Rabbit 1:1,000 Phospho-Ampk Cell Signalling 2535 Rabbit 1:1,000 Phospho-Erk Cell Signalling 4370 Rabbit 1:1,000 Phospho-S6k Cell Signalling 9205 Rabbit 1:1,000 S6k Cell Signalling 9202 Rabbit 1:1,000 Srebf2 Abcam ab30682 Rabbit 1:200 Santa Cruz sc-13552 Mouse 1:200 Santa Cruz sc-5603 Rabbit 1:200 Vinculin Santa Cruz sc-73614 Mouse 1:200

Secondary antibodies used were HRP-linked-anti-rabbit (1:2000; Cell Signaling, 7074S) or anti-mouse (1:3000; Invitrogen, 626520).

43 2.18. Immunoprecipitation and binding partner analysis

Cells were scraped and lysed in lysis buffer (20 mM Tris-HCl pH 7.4/0.2% NP-40/150 mM NaCl/2 mM sodium vanadate; 500 µl/ 15 cm plate), with other proteinase and phosphatase inhibitors added as before on ice for 30 minutes, and lysates were collected and protein concentration measured as in section 2.17. Samples were pre- cleared overnight by addition of 30 µl Dynabeads Protein A beads (10002D) to 5.5 mg (1.5 ml) protein and rotation at 4 ˚C. The supernatant was then transferred to a new tube and the lysate was incubated with the antibody against p53 (Cell Signalling, 2524; 1:500) or mouse IgG (IgG control, 7.5 µl/1.5 ml; Santa Crux, sc-2025) for 3 hours at 4 °C with rotation. 30 µl beads were then added to the lysates and incubated for 1 hour 4 °C with rotation. The supernatant was then removed and the beads were washed three times in lysis buffer. The supernatant was removed again and the samples were treated as before (section 2.17) to elute and denature protein precipitated complexes but with 10 minutes boiling at 95 °C. Samples were run on gels as in section 2.17.

The protein-binding partner analysis was performed by staining the gel with Coomassie blue R250 (Fischer) overnight at room temperature with shaking. The staining solution used was 45 % Methanol/10 % Acetic acid/45 % dH2O/3 g/L R250. The gel was then destained with destaining solution (5 % Methanol/7.5 % Acetic acid in dH2O) overnight. Bands were identified on the gel in the p53 mutant lanes and sent for mass spectrometry analysis and protein identification (Cambridge Centre for Proteomics). Protein binding partners of the p53 mutants were identified based on whether they were detected in the bands sent for Mass Spectrometry analysis (R172H or R270H sample) but not in the IgG control lane, which was analysed in parallel.

To confirm pull-down of p53, or to investigate whether Srebf2 was co-precipitated, western blotting of gels following immunoprecipitation was also. Western blotting was performed and antibodies were used were as in section 2.17.

44 2.19. Srebf2 knockdown

5 g of construct was reconstituted in 100 l ultra-pure H2O. DH5 competent cells (New England Biolabs, 50 l) were transformed with 1 l construct for plasmid expansion. Briefly, cells were mixed with DNA and incubated on ice for 15 minutes, heat-shocked at 42 °C for 2 minutes and then incubated on ice for a further 30 minutes. Bacteria were grown up in 1 ml LB broth and incubated at 37 °C with shaking (300 rpm) for 1 hour. Bacteria were then spun down and resuspended in 100 l LB broth and spread onto a LB agar + Kanamycin (30 µg/mL) plate. The plate was incubated overnight at 37 °C. A single colony was picked the following day and grown up in 5 ml LB broth at 37 °C with shaking for 8 hours and then the 5 ml was added to 200 ml LB broth with kanamycin and the bacteria were grown up overnight at 37 °C. A Maxiprep (Qiagen, 12663) was performed to extract DNA.

Cells were transiently transfected with 10 g green fluorescent protein (GFP)+ shRNA plasmids (TG510683A-D), or control non-targeting sequence (TR30013) from Origene; transfection as previously described (Section 2.3.4). 48hrs after transfection, cells were sorted for GFP positivity by the Flow Cytometry team (CIMR and MRC) and 50,000 GFP+ cells seeded in the Matrigel/DMEM-F12 mix. Colony size was measured 72 hrs after seeding and cells collected for RNA extraction as previously described (section 2.6).

2.20. Analysis of publically available datasets

2.20.1. ChIP datasets

ChIP datasets were analysed for the identification of potential Srebf and p53 targets from http://chip-atlas.org (Srebf2 and p53) and Fischer et al. 2017 (p53).

2.20.2. TCGA lung adenocarcinoma dataset

Lung adenocarcinoma patient TCGA (Cancer Genome Atlas Research, 2014) data was downloaded from cBioportal (http://www.cbioportal.org). Cases which had TP53

45 mutations were selected and these cases were subdivided further based on whether the TP53 mutation was a conformational, contact or non-missense mutation, or was unclassified. The classification of mutations into conformational or contact classes was performed using structural studies on the p53 DNA-binding domain. If the mutation occurred in a residue that directly contacted the DNA it was classified as a contact mutation, whereas if the mutation did not occur at a residue that directly contacted the DNA but instead had more effects on the overall structure of the protein it was classified as conformational (Cho et al., 1994; Joerger et al., 2006; Joerger and Fersht, 2007). The expression of MVA pathway genes (RNA Seq V2 RSEM) was analysed in these samples. Significance was calculated based on Significance Analysis for Microarrays (SAM) platform (Tusher et al., 2001), within Multiple Experiment Viewer (MeV).

2.21. Statistical analysis

All statistical analyses were performed using Prism 5.0 software (Graph Pad) unless otherwise indicated, with P < 0.05 considered statistically significant. P values for unpaired comparisons between two groups with comparable variance were calculated by two-tailed Student’s t-test. One-way ANOVA (between genotypes) was used for analysis between three groups with comparable variance, followed by Bonferroni post-test for individual comparisons. Two-way ANOVA (between treatments and genotypes) was used for analysis between data with two independent variables, followed by Bonferroni post-test for individual comparisons. Kaplan–Meier comparison with a Log-rank (Mantel-Cox) test was used for analysis of survival cohorts. Significance of altered canonical pathways and molecular functions was calculated by Fischer’s Exact test (within IPA) and SAM (Tusher et al., 2001), within MeV or a Student’s t-test was used to calculate significance of gene expression data as indicated.

46 3. Generation and transcriptional profiling of murine lung tumour cells

3.1. Introduction

Mouse models of cancer are frequently utilised to understand the mechanisms of tumour development and to assess how the tumours respond to therapy. They enable the accurate assessment of the effect of genetic alterations on tumour progression as well as the impact of drugs in an in vivo system with a functional immune system and a tumour microenvironment (Frese and Tuveson, 2007).

In order to investigate the impact of p53 mutations in lung tumours we used a well- established mouse model of lung adenocarcinoma (Kraslox-stop-lox(LSL)-G12D), in which the formation of the lung tumours is driven by mutant KrasG12D, a frequent driver mutation in lung adenocarcinoma (Cancer Genome Atlas Research, 2014; Jackson et al., 2001). KrasG12D expression can be induced in the lungs following the administration of a recombinant adenovirus expressing Cre recombinase (Adeno- Cre). Cre expression leads to the removal of the stop codon proceeding the KrasG12D gene, thus activating the expression of the mutant allele in the targeted tissue. This KrasG12D-driven model of lung adenocarcinoma recapitulates the human disease as tumorigenesis is driven by the sporadic activation of mutant Kras which is expressed from its endogenous promoter (Jackson et al., 2005; Jackson et al., 2001; Junttila et al., 2010; Kerr et al., 2016). In addition, the administration of the Adeno-Cre allows for the control of the number of tumours formed and the timing and tissue-specificity of tumour initiation (Jackson et al., 2001).

In the model used, the status of p53 can also be manipulated by generating compound mice with conditional mutations in Kras and p53 (Jackson et al., 2005). These compound mice develop more aggressive lung adenocarcinomas with shorter latency, and these lesions are considered to closely recapitulate those seen in humans (Jackson et al., 2005). In addition, incorporation of a conditional p53ERTAM allele into the compound mice enables the 4-hydroxytamoxifen (4OHT)-controlled restoration of wild-type p53 function at will (Christophorou et al., 2005).

47 The aim of this chapter is to utilise the KrasG12D-driven lung adenocarcinoma model with different p53 deficiencies to investigate the impact of p53 mutations on transcriptional programs. To achieve this, I will compare the transcriptional profiles of murine lung tumour cultures generated from independent lung tumours of distinct p53 genotypes and ascertain whether different p53 mutations (R172H and R270H) have gain-of-function (GOF) or dominant-negative (DN) transcriptional effects in lung tumour cells.

3.2. Results

3.2.1. Generation of p53 mutant tumours of different p53 deficiencies

Three distinct compound mice were generated: 1) KrasLSL/G12D/+;p53Fx/Fx, 2) KrasLSL-

G12D/+;p53LSL-R172H/Fx and 3) KrasLSL-G12D/+;p53LSLR270H/Fx. I administered Adeno-Cre intranasally to induce the expression of oncogenic Kras in the lungs, which drives the formation of the tumours. Adeno-Cre administration also induces recombination of the LoxP sites placed within the p53 locus: in 1) the exons 2-10 of the wild-type p53 allele flanked by two lox sites (p53LoxP) are deleted whereas in 2) and 3) the stop codon is removed, leading to the endogenous expression of the conformational p53 mutant R172H or the contact mutant R270H (corresponding to codons 175 and 273 in human cancers, Figure 3.1), respectively (Jackson et al., 2005; Jonkers et al., 2001). Therefore the final genotypes of the lung tumours formed in this Fx model are

KrasG12D/+;p53-/-, KrasG12D/+;p53R172H/- and KrasG12D/+;p53R270H/- which will be referred to as Null, R172H and R270H for simplicity. This model, referred to as the ‘Fx model’ in the text, was used to investigate the GOF effects of different p53 mutations in lung tumours: R172H and R270H compared to Null.

Furthermore, another variation of this murine lung adenocarcinoma model was also generated. In this second model, instead of a p53Fx allele, the mice also express a switchable p53 allele (p53ER) that enables the restoration of wild-type p53 function in a time-controlled manner, following the administration 4OHT. The p53ER protein is always expressed in the tumour cells but whether it is functionally active or not

A Fx Kras p53 model

KrasLSL-G12D/+;p53Fx/Fx KrasG12D/+;p53-/- (Null) KrasLSL-G12D/+;p53LSL-R172H/Fx + Cre KrasG12D/+;p53R172H/- (R172H) KrasLSL-G12D/+;p53LSL-R270H/Fx KrasG12D/+;p53R270H/- (R270H)

● mutant GOF effects ● impact of mutant p53-targeted therapy impact

B Kras p53ER model

KrasLSL-G12D/+;p53Fx/ER KrasG12D/+;p53-/ER (Null)

KrasLSL-G12D/+;p53LSL-R172H/ER + Cre KrasG12D/+;p53R172H/ER (R172H) KrasLSL-G12D/+;p53LSL-R270H/ER KrasG12D/+;p53R270H/ER (R270H)

+ 4OHT - 4OHT

WT p53 On WT p53 Off

● mutant DN effects ● mutant GOF effects ● impact of WT p53-targeted therapy impact

ER Fx Figure 3.1. p53 and p53 lung tumour models.

Mice of the above genotypes were administered Adeno-Cre (Cre) intranasally to drive G12D the Cre-induced recombination of the Kras and p53 alleles in the lungs. Kras ;p53- null, -R172H and -R270H tumours were generated in either the p53Fx (upper) or p53ER (lower) model. Lung tumours or cultures generated from lung tumours were analysed in vitro and in vivo. Wild-type p53 function can be restored in the ER model via administration of 4OHT (see Figure 1.9).

49 depends on the administration or removal of 4OHT respectively (Figure 1.9, 3.1; Christophorou et al., 2005). Therefore, the final genotypes of the lung tumours formed in these mice are KrasG12D/+;p53-/ER, KrasG12D/+;p53R172H/ER and KrasG12D/+;p53R270H/ER and these will also be referred to as Null, R172H and R270H. Whether these terms refer to the ‘ER model’ or the ‘Fx model’ will be indicated. This ER model will enable me to simultaneously investigate both the GOF (absence of 4OHT) and DN effects (presence of 4OHT) of different p53 mutations in lung tumours.

In my study lung tumours (from the Fx and ER models) were collected for analysis of the different p53 genotypes at endpoint, when the animals displayed visible signs of tumour burden (reduced activity, laboured breathing, hunched posture), with the latency period being 16-18 weeks post-Adenocre infection. Each animal develops multiple, independent lung tumours which can be analysed (Figure 3.2). All animals in our experimental cohorts were re-genotyped; PCRs were run on genomic DNA extracted from the tails of the mice for the p53mutant, p53ER and p53Fx alleles and the p53 gene was also sequenced. Figure 3.3A,B shows examples of the PCR and sequencing results obtained from the tails of the animals with different p53 genotypes.

3.2.2. Isolation and characterisation of p53 mutant and null lung tumour cell lines

In order to profile the transcriptional effects of p53 mutations we generated multiple lung tumour cell lines from independent, late-stage lung tumours. We hypothesised that the utilisation of multiple, independent cell lines that endogenously express mutant p53, or that are p53 null, would be a robust way to analyse the transcriptional and functional effects of p53 mutations in lung tumour cells.

KrasG12D/+;p53-/ER (Null), KrasG12D/+;p53R172H/ER (R172H) and KrasG12D/+;p53R270H/ER (R270H) lung tumour cell lines were generated prior to the start of my research project. In addition to these cell lines, I also generated KrasG12D/+;p53-/- (Null), KrasG12D/+;p53R172H/- (R172H) and KrasG12D/+;p53R270H/- (R270H) lung tumour cells lines from the ‘Fx model’ to further investigate and validate the GOF phenotypes identified.

Null

R172H

R270H

Figure 3.2. Lung tumour burden

Representative H&E images of p53 null, R172H and R270H lung sections. Multiple independent lung tumours are formed/animal. Scale bar: 2 mm.

Figure 3.3. Validation of genotypes of lung tumour mice.

A. PCRs were performed to test for the presence of the p53Fx (upper), p53ER (middle) and p53WT/Mut (lower) alleles in the experimental animals (unrecombined). B. Sequencing of the p53 locus confirms the presence of the corresponding p53 mutation in the R172H and R270H animals. Negative control: -ve; positive control: +ve; non-template control: NTC.

52 When generating the cell lines from the tumours, epithelial lung tumour cells were separated from tumour-associated fibroblasts by exploiting their differential attachment rates. I assessed expression of epithelial cell adhesion molecule (EpCAM; epithelial marker) and platelet-derived growth factor (Pdgfr; fibroblast marker) by flow cytometry to confirm epithelial populations based on a low percentage of Pdgfr-positive cells (<8%; Figure 3.4A,B). I also confirmed the genotypes of the murine lung tumour cell lines by PCR analysis and sequencing for presence of the p53 mutation Figure 3.4C,D). Immunoblot analysis also confirmed the expression of the mutant p53 protein in the R172H and R270H cell lines (Fx and ER). The expression of p53ER was at lower levels than that of the mutant p53 (Figure 3.4E).

Next, I measured the proliferation rates of the lung tumour cell lines generated to ensure the cell lines were all growing efficiently after isolation. The cell lines all grew well in 2D culture, with some cell lines growing faster than others but no overall genotype-specific difference in growth rates in either the ER or Fx model (Figure 3.5A). However, for transcriptional and functional analysis we decided to predominately utilise a Matrigel-based 3D culture system. 3D cultures are considered to better recapitulate the in vivo environment as cells are grown on a more physiologically relevant structure with cell-matrix interactions (Pampaloni et al., 2007). All lung tumour cells grew well in Matrigel when seeded at 50,000 cells in 100 l of a 3:2 Matrigel/media mix, forming spheroid colonies. Again I observed no genotype-specific difference in growth in these 3D cultures using cells from both Fx and ER models (Figure 3.5B).

3.2.3. Transcriptional impact of p53 mutations in lung tumour cultures

Transcriptome (microarray) analysis on four independent cell lines/genotype (p53 null, R172H and R270H; ER model), cultured in 3D Matrigel was performed by the lab prior to my arrival. Utilisation of the ER model enabled the investigation of the transcriptional effects induced by the R172H conformational and R270H contact p53 mutations in the presence (4OHT, p53 On; DN) and absence (vehicle, p53 Off; GOF) of wild-type p53.

53 A B Fibroblast 100 Epithelial

l l

e 80

v 60

a r

f 40 g

P 20 % 0 e e EpCAM Pdgfra -v v ll H H + u 2 0 N 7 7 1 2 R R

C D

R R /E /E H H R 2 0 E 7 7 x/ 1 2 F R R

172

270

x x E /F /F H H 2 0 7 7 1 2 R R -/ER R172H/ER R270H/ER p53 p53 p53 p53ER p53 (mutant) 172 Actb

p53 -/- p53 R172H/- p53 R270H/-

p53 270 (mutant) Codon WT Mutant 172 CGC CAC Actb 270 CGT CAT

Figure 3.4. Generation and characterisation of lung tumour cell lines.

A. Representative data showing EpCAM (epithelial marker; left) and Pdgfra (fibroblast marker; right) expression in murine lung tumour cell lines (Epithelial) and mouse embryonic fibroblasts (MEFs; Fibroblasts) based on FACS analysis. B. Percentage of Pdgrfa-positive cells for the lung tumour cell lines (Fx model). Epithelial cells were used as a negative (-ve) and MEFs as a positive (+ve) control. A,B. Two independent repeats were performed/cell line. C. PCRs were performed to test for the presence of the p53Fx (upper), p53ER (middle) and p53WT/Mut (lower) alleles in the Fx and ER cell lines (recombined). Negative control: -ve; positive control: +ve; non-template control: NTC. D. Sequencing of the p53 locus confirms the presence of the corresponding p53 mutation ER in the R172H and R270H ER and Fx cell lines. E. Expression of mutant p53 and p53 proteins in p53 null, R172H and R270H ER cell lines (upper) and the mutant p53 in Fx cell lines (lower), assessed by immunoblot analysis. Beta-actin (Actb) was used as loading control. Two independent repeats were performed/cell line.

-/- -/ER 8 p53 p53 1.5 p53R172H/- p53R172H/ER

R270H/- R270H/ER )

p53 ) p53 5

6 6

0 1

1 1.0

(

s l

l e

4 e

v i

i 0.5 L 2 L

0 0.0 0 24 48 72 96 0 24 48 72 96 Hours Hours

Null R172H R270H

ns l

o t

v i

r e

b 1

e C 0 Null R172H R270H

Figure 3.5. Characterisation of the growth of lung tumour cell lines.

A. Cells were seeded in 2D and the viable cell growth was assessed over time by trypan blue exclusion; left, ER cells; right, Fx cells. Representative data is shown for two independent cell lines of the indicated genotypes from the ER model (of n=4 cell lines assessed/genotype) and the Fx model (of n=3 cell lines analysed/genotype). Symbols depict average cell number for a cell line SD of triplicates. B. Upper, Representative images of p53 null and mutant cell lines grown in 3D Matrigel culture. Lower, viable cell number of p53 mutant and null cell lines seeded at 50,000 cells/well in Matrigel, assessed by trypan blue exclusion after 72 hrs. Cell number values relative to the null average are shown. Symbols represent independent cell lines of the indicated genotypes (Fx, squares; ER circles) and genotype average is shown SEM. P>0.05, One- way ANOVA. A minimum of two independent repeats were performed/cell line, with each cell line performed in triplicate in each experiment.

First, to establish if there are general mutant p53 transcriptional GOFs we compared the expression profiles of the p53 null (n=4) and p53 mutant (conformational and contact mutants combined, n=8) cell lines in the basal setting (vehicle) to identify whether there were transcriptional effects of the p53 mutants. The expression of anumber of genes was altered in p53 mutant cells compared to p53 null (Figure 3.6A). We applied a cut-off of a 1.4 average fold-change (FC) in expression and a standard deviation of <0.5 between expression values of cell lines of a single genotype to account for intragenotype variation observed for some genes. Using this analysis, we found 465 genes altered in p53 mutant cells compared to p53 null cells, indicative of p53 mutant GOF at a transcriptional level. Moreover, after further analysis of our transcriptome data, what struck me as particularly interesting and unexpected were the differences in the transcriptional profiles of lung tumour cells expressing the contact or the conformational mutant p53. We observed that the expression of many gene (n=972) was altered between lung tumour cell lines of distinct p53 mutational status (R172H and R270H; >1.4 average FC and <0.5 SD between samples in each group), indicating that individual p53 mutants have different functions in lung tumour cells (Figure 3.6B).

Finally, we analysed the transcriptional impact of the p53 mutations in the presence of wild-type p53 to establish whether wild-type p53 was still able to function in the presence of p53 mutations. We compared the transcriptional profiles of our p53 null cells and p53 R172H and R270H cells after wild-type p53 restoration (2 hrs and 8 hrs). Two time points of wild-type p53 restoration were used to help distinguish between direct (2 hrs) or indirect (8 hrs) wild-type p53 targets. As with our previous analysis we applied a 1.4FC cut-off between control and p53 restoration expression values with a standard deviation limit of 0.5 to exclude intra-genotype variability. Subsequently, we identified a subset of genes that were similarly induced or repressed in p53 null and p53 mutant lung tumour cells after wild-type p53 restoration. Eleven of these genes were altered 2 hrs after p53 restoration (Figure 3.7, upper) and 87 (including the 2 hrs subset) after 8 hrs restoration (Figure 3.7, lower). Importantly, many of these similarly regulated genes are known p53 targets (based on literature searches; (Bieging et al., 2014; Fischer, 2017a; Lehar et al., 1996; Lo et al., 2001; Okamoto and Beach, 1994; Quintens et al., 2015; Utrera et al., 1998); www.chip-atlas.org) including all the genes in the 2 hrs subset, which are likely to represent direct p53 targets.

A B

WT Null Mutant R172H R270H

-Min2.6 Max3.0 -Min2.7 Max2.8

Figure 3.6. p53 mutations display transcriptional GOF when compared to p53 null and also p53 mutant-specific effects.

A. Genes that are differentially expressed in p53 mutant (n=8) and p53 null (n=4) lung tumour cell lines are displayed in the heatmap. B. Genes differentially expressed between lung tumour cells with a R172H or a R270H mutation are shown. A,B. Each individual square in each row corresponds to an independent cell line. The microarray analysis was performed on one sample/cell line but n=4 cell lines were analysed/genotype. MeV was used to generate the heatmaps and the software normalised genes/row: Value=[(Value)- Mean(Row)]/SD(Row). The minimum and maximum normalised expression values and their corresponding colours are shown in the scale bars.

57 Null R172H R270H p53 On (hrs) p53 On (hrs) p53 On (hrs) Off 2 8 Off 2 8 Off 2 8 Bbc3 Ccng1 Cdkn1a Ddit4l Gtse1 Mdm2 Phlda3 Sesn2 Slc19a2 Tp53inp1 Znf365 Hist1h2aa Null R172HHist1h2ae R270H Hist1h2ab p53 On (hrs) p53 Mcm3On (hrs) p53 On (hrs) Aunip Off 2 8 Off Eprs 2 8 Off 2 8 Slc30a1 Chaf1b Bbc3 Ccng1 Clspn Cdkn1a Nrm Ddit4l Syncrip Gtse1 Mdm2 Birc5 Phlda3 Kiaa0101 Sesn2 Kif15 Slc19a2 Esco2 Tp53inp1 Mcm10 Znf365 Hist1h2ah Hist1h2aa Hist1h2ae Hist1h2aj Hist1h2ab Hist1h2am Mcm3 Hist1h2ap Aunip Pbk Eprs Slc30a1 Top2a Chaf1b Crip2 Clspn Evpl Nrm Gdpd5 Syncrip Birc5 Ddit4 Kiaa0101 Dyrk3 Kif15 Plau Esco2 Mcm10 Sh3bgrl2 Hist1h2ah Slc4a11 Hist1h2aj Mast4 Hist1h2am Upp1 Hist1h2ap Pbk Ggta1 Top2a Irf2bp1 Crip2 Mmp14 Evpl Nfe2l2 Gdpd5 Ddit4 Apaf1 Dyrk3 Aldh4a1 Plau Dcxr Sh3bgrl2 Gsto1 Slc4a11 Mast4 Prr15l Upp1 St3gal3 Ggta1 Dgka Irf2bp1 Mmp14 Kank3 Nfe2l2 Anxa8 Apaf1 Bscl2 Aldh4a1 Nek3 Dcxr Gsto1 Pgpep1 Prr15l Ephx1 St3gal3 Tmem150a Dgka Ccdc3 Kank3 Anxa8 Ivl Bscl2 Glipr1 Nek3 St3gal2 Pgpep1 Blcap Ephx1 Tmem150a Ckap2 Ccdc3 Areg Ivl Hes6 Glipr1 St3gal2 Ercc5 Blcap Klhl26 Ckap2 Fam212b Areg Rnf169 Hes6 Ercc5 Ccnd1 Klhl26 Itgb4 Fam212b Bax Rnf169 Inf2 Ccnd1 Itgb4 Pml Bax Hyal Inf2 Prkd2 Pml Ei24 Hyal Prkd2 Tmem19 Ei24 Trim11 Tmem19 B3galnt2 Trim11 B3galnt2 Cpped1 Cpped1 Commd3 Commd3 Zmat3 Zmat3

Min Max Min-2.5 Max2.5 Figure legend on next page

58 Figure 3.7. Wild-type p53 restoration similarly induces or represses a subset of genes in p53 mutant and null cells.

Microarray analysis was performed on p53 mutant and null cell lines after Ctrl treatment (p53 Off) or 2 hrs or 8 hrs wild-type p53 restoration (p53 On). Genes that are similarly induced or repressed by WT p53 in mutant p53 and null cells at 2 hrs (n=11, upper) or 8 hrs (n=87, lower) are depicted in the heatmap. Genes identified as known p53 targets are highlighted in blue (see sections 2.7.4 and 2.19.1). Each individual square in each row corresponds to an independent cell line. MeV was used to generate the heatmaps and the software normalised genes/row: Value=[(Value)-Mean(Row)]/SD(Row). The relative minimum and maximum expression values are the limits of the scale bars which are shown.

The microarray analysis was performed on one sample/cell line/condition but n=4 cell lines were analysed/genotype for each condition. Work on the analysis of commonly regulated genes was performed with the help of Emma Kerr and David Shorthouse.

However, despite a large subset of genes being similarly regulated by wild-type p53 in p53 null and mutant cells, upon further analysis of our transcriptome data some genotype-specific transcriptional effects induced by WT p53 in lung tumours cultures were identified. By performing comparative analysis of the genes altered by wild-type p53 across the different genotypes we concluded that after 2 hrs wild-type p53 restoration there are 27, 1 and 11 genes uniquely altered in p53 null, R172H and R270H cells respectively. At the 8 hrs time point this is increased to 213, 46 and 257 genes for the respective genotypes (Figure 3.8A). This indicates that there are genotype-specific effects following WT p53 restoration which include potential DN effects of p53 mutations on wild-type p53.

Taking the subset of genes that are not altered after 8 hrs of wild-type p53 restoration in R172H cultures (213+90) and R270H cultures (213+27) I assessed the DN effect of the p53 mutants. I classified these genes further based on whether a similar trend the (>1.2FC) was observed for the genes in the p53 mutant cells compared to the p53 null cells after restoration (decreased ability) or whether there was complete failure (<1.2FC) of the induction or repression of these genes. In some cases, the reason for the decreased ability or complete failure for the wild-type p53 to induce or repress the genes in the mutant cells was because the basal levels of expression in the p53 mutant cells were similar to those in the p53 null cells after restoration (<1.4FC between mutant p53 Off and Null p53 On; wild-type-like). In all cases a standard deviation limit of 0.5 was imposed as before and a small number of genes were excluded based on this. As a result, we observed DN effects for both mutants,

59 A B

p53 On 2 hrs p53 On 8 hrs 100 12 24 R R 80 H 2 H 2 2 7 2 7 s 7 0 7 0 e 60 1 0 H 1 19 H n

R 1 11 R 46 257 e 47 69 g

11 87 40 1 4 27 90 % WT-like 20 29 Decreased ability 27 213 19 0 Complete failure R172H R270H Null Null n=302 n=227

C D

Mutant: DN effects Mutant: “WT-like” Null R172H Null R270H Null R172H Null R270H p53 p53 p53 p53 p53 p53 p53 p53 Off On Off On Off On Off On Off On Off On Off On Off On Pilrb Sdf2l1 Itpripl1 Nedd9 Pilra Cav2 Igf1r Nr2f2 Apoa1 Eva1c Smg1 Cd3eap Glyat Bloc1s2 Fam199x Dnajc10 Tnfrsf18 Rps27L Nr2f2 Sptbn1 Mppe1 Pidd1 Sptbn1 Tsc22d2 Slc2a9 Apobec1 Zmynd11 Zeb2 Sgsm2 Ccp110 Ep300 Ints7 Trim6 Fas Golga1 Irf3 Mapkapk2 Ddx6 Dnajb2 Ranbp3l Pomt2 Otud7b Cd151 Cenpf Ddx49 Ubtd2 Hmgb2 Ncapg2 H2afx Rad18 Cenpn Mt2 Mier3 Tubb2b

-2.0 2.5 -1.9 2.4 -2.2 2.7 -2.5 2.4

Figure 3.8. p53 mutants display DN effects on wild-type p53 transcriptional function.

A. Venn diagrams depict the number of genes altered after 2 hrs (left) or 8 hrs (right) of

wild-type p53 restoration for the cell lines of different p53 status. Shaded regions represent genes changed in the null but not in R172H (green), R270H (red) or both mutants (yellow) after 8 hrs p53 restoration. B. Classification of genes altered in the p53 null cells but not in the R172H or R270H cells (A) following 8 hrs wild-type p53 restoration to assess DN functions. C. Heatmaps depict examples of DN (left) and wild- type-like (right) effects of mutant p53 for the R172H mutant and the R270H mutant. Previously identified targets are shown in blue (see sections 2.7.4 and 2.19.1). MeV was used to generate the heatmaps and the software normalised genes/row: Value=[(Value)-Mean(Row)]/SD(Row). The minimum and maximum normalised expression values are the limits of the scale bars which are shown.

some common and some mutant-type specific, as well as interestingly some wild- type-like function of p53 mutants (Figure 3.8B-D).

In conclusion, we have utilised genetically modified mice to model KrasG12D-driven lung tumorigenesis in the background of various p53 deficiencies. Multiple lung

60 tumour cell lines were generated from independent, advanced lung tumours of different p53 genotypes which enabled us to characterise the transcriptional impact of p53 mutations in vitro, both in the presence and absence of wild-type p53.

Based on our initial transcriptome analysis we devised a model of the impact of p53 mutations in lung cancer to test (Figure 3.9). We hypothesised that:

1) R172H and R270H mutations exhibit common GOFs in lung tumour development culminating in different phenotypes of p53 mutant and null tumours. 2) p53 mutations may exert inhibitory effects on wild-type p53 (DN), which may also influence the mutant p53 lung tumour phenotype. 3) R172H and R270H mutants may function distinctly in lung tumours.

We further hypothesised that by characterising the impact of p53 mutants in lung tumour cell cultures and ultimately advanced lung tumours, we may uncover novel therapeutic vulnerabilities of p53 mutant lung tumours.

3.3. Discussion

Previous studies have identified oncogenic effects of p53 mutants, covering a range of cancer types (Adorno et al., 2009; Freed-Pastor et al., 2012; Muller et al., 2009; Muller et al., 2013; Sauer et al., 2010; Sorrentino et al., 2014; Weissmueller et al., 2014; Zhang et al., 2013a). However, the majority of these studies were based on investigation into the effect of mutant p53 in cell lines by either over-expressing mutant p53 in p53 null cell lines or knocking down mutant p53 in p53 mutant cell lines. In contrast, we wanted to avoid using non-physiological conditions such as over-expression. In addition, mutant p53 knock-down in cells may have detrimental effects on the growth of the cells if they evolved in tumours which were addicted to the mutant p53. Therefore, we focussed our investigations on comparing multiple, independently evolved cell lines which endogenously expressed mutant p53 or were p53 null. By using multiple independent cell lines/genotype that exhibited an array of other different mutations and alterations we hypothesised that any p53 genotype- dependent phenotypes that were identified would be robust and more likely to

Accelerated lung tumour progression p53null

p53R172H GOF? p53mut =? p53R270H DN? p53WT

Therapeutic vulnerabilities?

Figure 3.9. Model predicting the effects of p53 mutations in lung tumours based on transcriptional data.

The schematic depicts the hypothesised effects of p53 mutations in lung tumours based on transcriptional data. p53 mutants may exert gain-of-function (GOF) or dominant-negative (DN) effects to promote tumorigenesis. Mutant-specific effects may also exist.

translate into similar phenotypes in vivo. Moreover, our model enabled us to investigate the effects of p53 mutations in the presence and absence of wild-type p53, enabling us to extend our investigations to include the DN effects as well as GOFs of mutant p53. Through use of these models we aimed to provide the first comprehensive analysis of the impact of distinct p53 mutations in this cancer type.

Importantly, our transcriptome data also highlighted potential differences in the function of distinct p53 mutations. Such differences have been largely overlooked so far, with both types of p53 mutation considered to result in a loss-of-wild-type function (Joerger and Fersht, 2007) and with the majority of GOFs described as general functions of mutant p53 (Adorno et al., 2009; Freed-Pastor et al., 2012; Muller et al., 2009; Muller et al., 2013; Sorrentino et al., 2014; Weissmueller et al., 2014; Zhang et

62 al., 2013). Characterisation of the potential functional distinctions between p53 mutations is required as they may highlight mutant-specific therapeutic vulnerabilities in a class of tumours which are currently considered as one entity in the clinic.

63 4. Defining mutant p53 gain-of-functions in lung tumours

4.1. Introduction

As discussed in Chapter 1.2.2, gain-of-functions (GOFs) of mutant p53 that have been previously described often involve mutant p53 manipulating growth factor signalling pathways (Adorno et al., 2009; Muller et al., 2009; Muller et al., 2013; Sauer et al., 2010; Weissmueller et al., 2014) and generally result in a more aggressive phenotype relative to p53 loss. Further work into characterising the oncogenic effects of p53 mutations in lung cancer is required as it may provide insight into how to target these p53 mutant lung tumours.

Previously, it was shown that KrasG12D-p53 mutant lung tumours were a similar grade to their p53 null counterparts, arguing against additional oncogenic functions of mutant p53 in lung tumours (Jackson et al., 2005). However, the authors report that KrasLSL-G12D-p53LSL-mutant/Fx animals had an increased mortality following Adeno-Cre infection compared to KrasLSL-G12D-p53Fx/Fx animals. This was not due to differences in the lung tumours formed but instead was explained by an enhanced susceptibility to forming adenocarcinomas in the upper respiratory tract. These findings suggest that there are additional, oncogenic functions of p53 mutants. Our initial transcriptome analysis of multiple p53 mutant and null lung tumour cultures indicates that p53 mutants can exert additional effects in lung tumour cells (Figure 3.5A).

Therefore, the aim of this chapter was to compare our p53 null, R172H and R270H murine lung tumour cell lines and endogenous lung tumours to investigate the potential GOF effects of p53 mutations in lung cancer.

64 4.2. Results

4.2.1. Functional profiling of p53 mutant and p53 null lung tumour cells

As described in Chapter 3, our transcriptome analysis lead to the identification of 465 genes that were differentially expressed between p53 mutant (conformational and contact, n=8 cell lines) and p53 null (n=4 cell lines) 3D cultures (Figure 3.5A). Next, I sought to analyse which pathways may be altered by the presence of p53 mutations by performing pathway analysis on the differentially expressed genes. Through ingenuity pathway Analysis (IPA), the top significant molecular functions altered were cellular survival, movement and proliferation and cell cycle regulation. More specifically, when I analysed the canonical pathways it was various growth factor signalling pathways that were the top significantly altered pathways in p53 mutant cells, such as phosphoinositide-3-kinase/Akt, 14-3-3 and nerve growth factor (NGF) signalling (Figure 4.1A).

Next, I proceeded to assess whether there was a functional difference between the p53 mutant and null lung tumour cells. I assessed levels of phospho(p)-Akt, a read- out of PI3K/Akt pathway activity, but observed no difference in p-Akt levels between cells of the different p53 genotypes (Figure 4.1B). This result suggested that activity of this pathway is not altered between p53 null and mutant cells despite genes involved in these pathways being differently expressed. At the same time, I assessed p-Erk levels, which is downstream of many growth factor signalling pathways but similarly observed no genotype-specific difference (Figure 4.1B). As previously shown, we found no difference in the growth of p53 mutant and null cells in 2D or 3D cultures (Figure 3.4).

An oncogenic function of mutant p53 that has been frequently reported is an enhanced invasive potential of mutant p53 tumour cells, relative to p53 null cells (Adorno et al., 2009; Freed-Pastor et al., 2012; Muller et al., 2009; Muller et al., 2013; Weissmueller et al., 2014). Therefore, next I assessed the ability of the lung tumour cells to form colonies in the 3D Matrigel culture when seeded at a low density (5,000 cells/100l Matrigel/Media mix). I hypothesised that the presence of the mutant may

65 A

Molecular Functions Canonical Pathways

- log10(p-value) 0 1 2 3 4 5 Cell Death and Survival NSCLC Signalling Gene Expression PI3K/Akt Signalling Cellular Movement Myc-mediated Cell Cycle Apoptosis Signalling Cellular Growth and Proliferation 14-3-3-mediated Signalling Cellular Development NGF Signalling

2 3 4 5 6 0 10 20 30 40 50

- log10(p-value) % impact on pathway

Null R172H R270H

p-Akt

Akt

p-Erk

Erk

Vinculin

Actb

Figure 4.1. Analysis of growth-factor signalling in p53 mutant cells (ER model).

A. Pathway analysis was performed on genes that were differentially regulated (>1.4FC and <0.5 SD cut-off between samples in each group) between the p53 mutant and p53 null lung tumour cell lines. Top molecular functions (left) and canonical pathways (right) significantly altered in p53 mutant cells compared to p53 null cells. Statistical significance was determined by Fisher’s exact test. B. Expression of p-Akt and p-Erk, and their corresponding total proteins, was analysed in lung tumour cell lines of indicated p53 genotypes (ER cells shown). -actin (Actb) and vinculin were used as a loading controls.

66 increase the cells’ ability to survive and form colonies, indicative of their colonisation potential in vivo. Using multiple independent cell lines from both ER and Fx models we did notobserve a significant increase in the size or number of p53 mutant colonies compared to p53 null, demonstrating that the mutant p53 was providing no extra growth advantage in 3D cultures, nor did we observe a difference in the morphology of the colonies (Figure 4.2A).

In addition, we performed scratch assays to assess the migratory capacity of the cell lines (done in collaboration with Emma Kerr). We observed a trend towards enhanced movement of the p53 mutant lung tumour cells across a 2D surface compared to p53 null cells but no significant difference (Figure 4.2B). I proceeded to assess the invasive potential of the lung tumour cell lines in 3D cultures by performing invasive growth assays, using a similar Matrigel-based culture to that used for the transcriptome analysis. However, p53 null and mutant lung tumour cells were similarly invasive through Matrigel (Figure 4.2C).

Finally, we transplanted p53 null, R172H and R270H cell lines that I engineered to express luciferase into wild-type mice to assess their ability to survive, colonise the lung and form lung tumours. I measured the impact of mutant p53 on the location and burden of the tumours, and subsequent survival of the mice that formed lung tumours. I observed no difference in any of these parameters when comparing animals that were transplanted with lung tumours cells expressing a conformational or contact mutant p53 or p53 null cells (Figure 4.3).

4.2.2. p53 null, R172H and R270H lung tumours are similarly aggressive

Data from our cell lines indicate that p53 mutant and null lung tumour cells are similarly aggressive. To assess whether there were any differences in the endogenous p53 mutant and null lung tumours, I infected ‘Fx’ mice with AdenoCre, sacrificed the animals 16 weeks post-infection and compared the murine p53Mutant/Fx and p53Fx/Fx lung tumours. The time point of 16 weeks was chosen as the end point as the animals were displaying moderate signs of lung tumour burden. Animals

67 A ns ns 15 1.5

ns ns

n V

o 1.0 10

a l

u e

0.5 N

R 5

l ll H l H H u 2 H u 2 0 N 7 0 N 7 7 1 7 1 2 2 R R R R B 0 hrs 12 hrs 24 hrs

ns Null ns

)

r h / 40

m µ

R172H (

t i

c o

l 20 e V R270H 0 Null R172H R270H

C Invaded distance (μm) ns

m 50 ns

0 60 120 180 240 300 µ

0 2

1 40

Null g n

i 30

n i

R172H s

i n

o 10 l

o c R270H

% 0 Null R172H R270H

Figure 4.2. p53 mutant cells do not exhibit increased invasiveness compared to p53 null cells.

A. Average size of colonies (left) and number of colonies (right) formed when cells were seeded at low density (5,000/ well). Values are relative to a randomly picked reference. Colony size was measured after 72 hrs using ImageJ software. and genotype average is shown SEM. Ns: non-significant, P>0.05, One-way ANOVA. B. Migrational potential of p53 mutant and null lung tumour cell lines was assessed by measuring the distance travelled in one direction/time over 24 hrs after an initial scratch was made in the cell monolayer. Left, representative images of p53 null and mutant cell lines migrating over time to close the scratch. Scale bar: 300 m. Right, velocity of p53 mutant and null cell lines with the average value shown/genotype SEM. A,B. Symbols represent independent cell lines of the indicated genotypes (Fx, squares; ER circles). C. Invasive growth assays were performed on p53 mutant and null colonies in Matrigel (ER cells shown). Left, representative images of z- stack slices used to assess the ability of the colonies to invade through Matrigel. Scale bar: 200 m. Right, percentage of colonies that invaded >120m through the Matrigel was calculated. Representative data shows one cell line/genotype with the average of replicates SD shown. Ns: non-significant, P>0.05, One-way ANOVA.

Figure 4.3. Transplantation of p53 mutant cells, compared to p53 null, did not reduce host survival

Mice were transplanted with luciferase-expressing p53 null and mutant lung tumour cells (a minimum of 5 animals/cohort). A. Tumour growth was monitored by in vivo bioluminescence imaging with example images of lung tumour burden shown. B. Average luciferase activity/genotype 7 (left) and 14 days (right) after transplantation SEM. C. Percentage lung tumour burden at endpoint with the average value shown/genotype SEM D. Survival (upper) and tumour location (lower) after transplantation with null or mutant lung tumour cells. All animals shown had lung tumours at endpoint (one cell line/genotype). Additional tumours were occasionally found on the rib cage and in the chest cavity (thoracic tumours). Log-rank (Mantel-Cox) test. B,C. Ns: non-significant, P>0.05, One-way ANOVA.

69 showed similar signs of disease at this stage regardless of the p53 genotype, indicating a similar tumour latency of p53 mutant and null lung tumours. Consistent with this I observed comparable lung tumour burden in p53 mutant and null lung tumours, measured by percentage tumour coverage and number of lung tumours (Figure 4.4A,B). In addition, p53 mutant and null lung tumours displayed similar levels of proliferation as measured by Ki67 staining (Figure 4.4C).

In summary, I performed extensive analysis of the impact of different p53 mutations (conformational and contact) on cancer-associated phenotypes in lung tumour cells but observed no significant genotype-specific difference across multiple independent p53 mutant and null cells. Moreover, endogenous R172H or R270H p53 mutant expression does not affect the aggressiveness of the lung tumours, as p53 null, R172H and R270H show comparable lung tumour burden at endpoint as well as latency and proliferation levels in the tumours.

4.3. Discussion

Our data on the comparison of p53 null and mutant lung tumours and corresponding cell lines is inconsistent with the effect of p53 mutations in other cancers where the mutant p53 was shown to increase the growth and invasiveness of tumour cells (Freed-Pastor et al., 2012; Muller et al., 2009; Muller et al., 2013; Weissmueller et al., 2014), and tumours (Lang et al., 2004; Olive et al., 2004). As mentioned earlier, the discrepancy between the phenotypes of the tumour cell lines is probably explained by the use of endogenous p53 mutant and null cell lines compared to using mutant p53 over-expression or knock-down. Previous studies which have demonstrated a GOF effect of p53 mutations in vivo have compared p53 null mice with mice that express mutant p53 throughout the body (Lang et al., 2004; Olive et al., 2004). The mutant p53 animals have a predisposition to develop a different spectrum of tumours relative to p53 null. Therefore, it is likely that the p53 mutations have tissue specific effects. Instead our data, along with previously published data, indicates that there is no difference in the aggressiveness of p53 mutant and lung null lung tumours (Jackson et al. 2005). However, Jackson et al. report an increased mortality of KrasLSL-G12D- p53LSL-mutant/Fx animals compared to KrasLSL-G12D-p53Fx/Fx due to the

70 A

Null R172H R270H

ns l ns a ns

60 m

ns i 100

g n

a 80

e c

v 40 e

o 60 s

u u o 40 o

m 20

t 20

g n

u 0 0 L l ll H H l H H u 2 0 u 2 0 N 7 7 N 7 7 1 2 1 2 R R R R

C ns

r u

o 40

/ s

l 30

v i

t 20

i s

7 10 6

% 0 Null R172H R270H

Figure 4.4. Endogenous p53 mutant and null lung tumours display comparable tumour burden and proliferation.

A. Representative H&E images of the lung for p53 null, R172H and R270H tumour-bearing mice. B. Assessment of lung tumour burden in Null, R172H and R270H tumour-bearing animals. Left, The percentage tumour coverage of the lung shown as an average percentage coverage/genotype SEM. Right, average number of lung tumours/animal. Symbols represent independent animals. A minimum of 5 animals/genotype was analysed. C. Quantification of % Ki67-positive cells/tumour. The average percentage/genotype is shown with symbols representing individual tumours. A minimum of 4 animals and 25 tumours were analysed/genotype, B,C. ns: non-significant, P>0.05, One-way ANOVA.

71 development of sinonasal adenocarcinomas, a difference in mortality that we do not observe. This is likely explained by the lower AdenoCre titre used in the other study (5x105 plaque-forming units; PFU). Our mice, which receive a higher titre (5x106 PFU), are culled due to primary lung tumour burden across all genotypes and likely before other respiratory-tract tumours present.

72 5. Distinct transcriptional and metabolic profiles in p53 R172H and p53 R270H mutant lung tumour cells

5.1. Introduction

There are many different p53 mutations found in lung adenocarcinoma cases (Cancer Genome Atlas Research, 2014). As described in detail in Chapter 1.2.1, p53 mutations fall into two classes, conformational (R172H) and contact (R270H), based on the residues they affect. It is feasible that, because of the different effects they have on the structure of the protein, the p53 mutants may exert distinct functions within the cancer cells. However, the distinct of effects of different p53 mutations in tumorigenesis have not been well-characterised.

Surprisingly, the number of genes differentially expressed between p53 R172H and R270H lung tumour cells was considerably larger than the number whose expression was differentially altered between p53 null and mutant (R172H and R270H) cells (Figure 3.5B). Therefore, my initial analysis of the transcriptome data indicated that the different p53 mutations may indeed function distinctly in lung tumour cells. Despite these differences, comparison of KrasG12D-driven lung tumours, and corresponding cell lines, by myself and others (Jackson et al., 2005), has demonstrated that there is no difference in the aggressiveness of the p53 R172H and R270H mutant lung tumours, at least from a gain-of-function (GOF) perspective. However, I hypothesise that p53 null, R172H and R270H tumours may be dependent on different cellular pathways in tumorigenesis to reach a comparable, aggressive end-point. Delineation of these pathways has the potential to highlight unique therapeutic vulnerabilities of p53 mutant lung tumours, as has been described in other cancer types (Freed-Pastor et al., 2012; Weissmueller et al., 2014)

Consequently, the aim of this chapter is to further explore the transcriptional differences between R172H and R270H lung tumour cultures and to establish whether p53 mutants of different classes can exhibit distinct phenotypes in the cells.

73 5.2. Results

5.2.1. Mutant-type specific transcriptional and metabolic profiles

R172H and R270H p53 mutant lung tumour cells displayed distinct transcriptional profiles, as reported in Chapter 3. In agreement, when the expression of all gene was analysed by principal component analysis, the cell lines separated based on the p53 mutation present (Figure 5.1A). To explore these transcriptional differences further I performed pathway analysis (IPA). This analysis indicated that it was the cholesterol biosynthesis pathway and related pathways that were the most significantly altered in the R270H lung tumour cells compared to those that express the R172H mutation (Figure 5.1B, left). Furthermore, multiple areas of metabolism were significantly altered between the two types of p53 mutant lung tumour cells, notably lipid and vitamin and mineral metabolism, and protein synthesis when I performed molecular function analysis on IPA (Figure 5.1B, right).

Next, I looked specifically at the expression of the genes involved in these three metabolic functions (Figure 5.2A). There was some overlap between genes involved in lipid metabolism and vitamin and mineral metabolism and these included many genes involved in cholesterol biosynthesis. For the genes involved in these two metabolic areas there were GOFs that were specific to R172H, that is where the R172H cells were different from the null and R270H cells, and some that appeared to be R270H-specific GOFs. In both cases these were true GOF effects of the specific p53 mutant as expression in the mutant cells was altered compared to both p53 null and p53 ‘wild-type’ cells (null cells with wild-type p53 restored; with 8 hrs 4OHT). On the other hand, the genes involved in protein synthesis that were differently expressed in lung tumour cells expressing distinct mutations, were predominantly R172H-specific GOF effects. Finally, using a compiled list of all known metabolic genes (compiled by Edoardo Gaude), IPA comparison analysis confirmed that the most significant difference in expression of metabolic genes existed between the cells expressing different p53 mutations (Figure 5.2B). As before areas of lipid and protein metabolism were significantly altered but also mitochondrial metabolism was highlighted as significantly different in the R172H cells.

74 Subsequently, to further investigate the metabolic differences between R172H and R270H mutant cells we performed metabolomic analysis on the independent lung tumour cell lines to confirm whether conformational and contact p53 mutations regulate metabolic pathways distinctly. Cellular metabolite levels were measured by mass spectrometry in collaboration with the Frezza lab. Principal component analysis on the metabolites measured for four independent cell lines of each genotype was performed with the help of Edoardo Gaude. The principal component analysis revealed that the R172H cells separated away from the R270H cells apart from one R172H cell line which was an outlier and displayed very different cellular metabolite levels to all other cell lines (Figure 5.3A). Nonetheless the metabolomics data supports the idea that lung tumour cells expressing distinct p53 mutations have different metabolic profiles with three independent R172H cell lines displaying altered levels of metabolites compared to four R270H cell lines.

I further analysed the metabolites that were altered between the lung tumour cells expressing a R172H mutation or a R270H mutation. To do this I excluded the R172H outlier from the analysis and focussed on the metabolic differences between the remaining p53 mutant cell lines. Encouragingly, the metabolites that were different between the R172H and R270H cell lines were fatty acids, amino acids and TCA cycle intermediates (Figure 5.3B), which supports the conclusion from our transcriptome analysis that lipid metabolism, protein synthesis and mitochondrial metabolism were altered between the cells expressing the different p53 mutations. However, it is important to note that the metabolites involved in lipid metabolism that were identified for the mass spectrometry analysis were fatty acids and their derivatives. We were unable to detect any metabolites in the MVA pathway, likely due to the hydrophilic metabolite extraction used which prevents the detection of sterols.

Both the transcriptome and metabolomics analysis indicated that R172H and R270H cells may have different levels of mitochondrial respiration; genes involved in mitochondrial dysfunction and the TCA cycle were highlighted by the transcriptome pathway analysis and TCA cycle intermediates were present at different levels in the metabolomics analysis. Therefore, I performed mitochondrial stress test extracellular flux assays to assess whether there were any differences in the levels of mitochondrial respiration in the R172H and R270H cells (Figure 5.4).

Figure 5.1. Metabolic genes are differentially expressed in p53 R172H and R270H cells.

A. Principal component analysis on total gene expression data separated p53 R172H and R270H mutant cells. 61.1% of the total variance is covered. Each symbol represents an individual cell line. B. Pathway analysis was performed on genes that were differentially regulated (>1.4FC and <0.5 SD cut-off between samples in each group) between the R172H mutant and R270H mutant lung tumour cell lines. Left, top canonical pathways significantly altered in R270H cells compared to R172H cells. Right, metabolic functions that are significantly altered in cells expressing distinct p53 mutations. Statistical significance was determined by Fisher’s exact test.

H H H H 2 0 2 0 ll 7 7 ll 7 7 T u 1 2 T u 1 2 W N R R W N R R

Lipid metabolism Protein Synthesis

Vitamin and Mineral metabolism

-2.6 2.8 -2.4 3.0

Molecular Functions Canonical Pathways

H H 2 2 ll ll 7 ll ll 7 u u 1 u u 1 -N -N -R -N -N -R H H H H H H 2 0 0 2 0 0 7 7 7 7 7 7 1 2 2 1 2 2 R R R R R R Synthesis of Lipid Superpathway of Cholesterol Biosynthesis Metabolism of Membrane Lipid Mitochondrial Dysfunction Synthesis of Terpenoid Superpathway of Geranylgeranyldiphosphate Biosynthesis Metabolism of Terpenoid Oxidative Phosphorylation Concentration of lipid Mevalonate Pathway I Fatty Acid Metabolism TCA Cycle II (Eukaryotic) Metabolism of Cholesterol Ketogenesis Metabolism of Carbohydrate Cholesterol Biosynthesis I Synthesis of Cholesterol Cholesterol Biosynthesis II (via 24,25-dihydrolanosterol) Metabolism of Amino Acids Cholesterol Biosynthesis III (via Desmosterol)

0 7.8 0 22.2 -log (p-value) -log (p-value) 10 10

Figure 5.2. Altered expression of genes involved in lipid, protein and mitochondrial respiration in p53 R172H and R270H cells.

A. Expression of the genes involved in the most significantly altered metabolic functions between R172H and R270H cells are shown in heatmaps. Expression values were normalised/row and the scale bar shows the minimum and maximum values. Each individual square in each row represents an independent cell line of the indicated genotype. Expression in the p53 R172H and R270H cells is compared to p53 null and p53 WT (p53 null with p53 On (+4OHT)) cells. B. Ingenuity comparison analysis of a list of total metabolic genes highlighted the extent of the differences in metabolic gene expression between p53 null, R172H and R270H cells. Top significantly altered functions (left) and canonical pathways (right) are shown. The likelihood of the overlap between a set of differentially expressed genes and a canonical pathway being significant was determined by Fisher’s exact test with Benjamini-Hochberg multiple testing correction within the IPA software, P<0.05 was considered significant.

A B

. R172H R270H 1 Null Homocysteine Creatinine R172H Proline

5 GABA . R270H

Gluconate 0 ) Glycerate

% Nacetylaspartate

0 Hexose

. Acetoacetate

8 0 .

( Aconitate

0 Carnitine 4

Cystathionine

C GSSG

Valine P 5 . Urea

0 Agmatine - Serotonin AMP Phosphoglycerate

0 ADP

. Adenine 1

- Phosphoserine Aginosuccinate -1.0 -0.5 -0.0 0.5 1.0 ATP Phosphogluconolactone PC 1 (39.5%) Gluconate6P Oleic acid Stearic acid Glycerophosphocholine Ribose ribulose5P SAM SAH Uridine Epinephrine Glucose Glucosamine HexosePhosphate Hydroxybutyrate Fumerate Malate Kynurenic acid Glyceraldehyde3P Kynurenine Dihydrothymine Glutamine Pyroglutamate Cinnamate Tryptophan Phenylalanine Leucine Methionine Acetylserine Histidine Tyrosine Thymine

-2.0Min Max2.0

Figure 5.3. p53 R172H and R270H cells display altered metabolic profiles.

Metabolite levels in p53 null, R172H and R270H were measured by mass spectrometry analysis for four independent cell lines/genotype (ER model). A. Principal component analysis was performed in collaboration with the Frezza group using the R. A total of 47.5% of the variance is covered by the analysis. Symbols show triplicates values for each cell line for the four cell lines/genotype. B. Heatmap displaying cellular metabolite levels in R172H and R270H lung tumour cell lines. Each square represents a triplicate average for an independent cell line (n=3 R172H cell lines, n=4 R270H cell lines). Heatmap was generated by MeV in which metabolite values were normalised/row. The scale bar limits depict the colour corresponding to the minimum and maximum normalised values. Metabolites shown have a greater than 1.3 average fold change between the two genotypes.

78 Using a Seahorse Extracellular Flux analyser, I measured the oxygen consumption rate (OCR) and the extracellular acidification rate (ECAR) of the media, which provide a read-out of mitochondrial respiration (OXPHOS) and glycolysis, respectively. Levels of OCR and ECAR were measured in basal conditions as well as upom administration of various mitochondrial stressors: oligomycin, which inhibits ATP synthase, was used to obtain levels of mitochondrial respiration associated with ATP production; CCCP, which is an electron transport chain (ETC) uncoupler and results in uninhibited electron flow through the ETC, was used to measure the maximal respiration levels; and Rotenone was used to inhibit all mitochondrial respiration. There was no difference in the basal levels of mitochondrial respiration between the R172H and R270H lung tumour cells but a trend was observed for a higher spare respiratory capacity (maximal respiration achieved after CCCP injection minus basal respiratory level) in the R270H cells compared to the R172H cells with three out of four R270H cell lines displaying higher levels (Figure 5.4B, left and centre) This suggests altered mitochondrial metabolism and ability to compensate under stress conditions in the R270H p53 mutant cells relative to R172H cells. In addition, higher levels of glycolysis were observed in the R270H cells compared to the R172H cells (Figure 5.4B, right).

Genes involved in protein metabolism and amino acids were significantly altered between R172H and R270H mutant cells, as shown by our transcriptome and metabolomics analysis respectively. Therefore, next I sought to investigate the potential differences in protein metabolism between cells expressing the contact or conformational p53 mutant. First, I measured protein levels in our p53 null, R172H and R270H cell lines to establish whether there were any differences in the overall protein levels. However, we observed no genotype-specific differences with only one R172H cell line displaying higher protein levels which could be explained by its larger cell size (Figure 5.5A). We also assessed ribosomal S6-kinase (S6K) phosphorylation which is downstream of the mTOR pathway and promotes protein synthesis. Once S6K is phosphorylated and activated it phosphorylates the ribosomal protein S6 which upregulates protein synthesis (Hay and Sonenberg, 2004). However, we observe no genotype-specific difference in P-S6K levels (Figure 5.5B). Finally, the different levels of amino acids we observed in R172H and R270H cells may be a result of altered levels of autophagy, the breakdown of macromolecules.

Figure 5.4. p53 R270H cells display higher levels of glycolysis and a trend for increased mitochondrial respiratory capacity.

Mitochondrial stress test extracellular flux assays were performed on p53 null, R172H and R270H cell lines (ER model) by measuring OCR and ECAR in response to mitochondrial inhibitors. A. Representative mitochondrial stress test profile showing two cell lines/genotype. Measurement over time (1-12) refers to the measurements the machine takes as the assay is running. B. Average basal respiration levels (OCR) for p53 mutant relative to p53 null cells (left). Average spare respiratory capacity (Maximal OCR-basal OCR) for p53 mutant relative to p53 null cells (middle). Average basal glycolysis levels (ECAR) p53 mutant relative to p53 null cells (right), n=4 cell lines assessed/genotype with symbols depicting average values for independent cell lines. The average level/genotype is shown SEM. *P<0.05, t-test.

80 A

s l

5 4

/ n

e t

o 2 r

g μ 0 Diameter (μm) 9.8 9.3 9.3 9.9 12.2 9.5 9.5 8.6 8.9 9.4 8.4 9.4

Null R172H R270H

p-S6k

S6k

p-Ampk

Ampk

LC3I LC3II

Vinculin

Actb

Figure 5.5. Analysis of protein levels and expression of metabolic regulators in p53 mutant and null cells.

A. Cellular protein content was assessed by BCA protein quantification and normalised to cell number, with n=3 independent repeats performed/ cell line, n=4 independent cell lines analysed (ER). The average amount of protein/10,000 cells is shown for each cell line SD of the triplicates. Cell diameter was also measured using a Countess cell counter. B. Expression of p- S6K and p-AMPK and their corresponding total proteins, as well as LC3 levels were analysed in lung tumour cell lines of indicated p53 genotypes (ER cells shown). -actin (Actb) and vinculin were used as loading controls.

81 However, after assessment of cleaved LC3 levels (LC3II) we again find no genotype- specific difference in autophagy (Figure 5.5B).

Finally, lipid metabolism was also highlighted as an area of metabolism differently regulated by R172H and R270H cells in our transcriptome (cholesterol biosynthesis) and metabolomics (fatty acids levels) analysis. AMP-activated protein kinase (AMPK) is one of the key metabolic sensors involved in the regulation of lipid metabolism, promoting fatty acid oxidation (the breakdown of lipids) in conditions of low energy, while opposing the synthesis of lipids (Long and Zierath, 2006). Also, a previous metabolic GOF that has been attributed to mutant p53 is an inhibition of AMPK signalling (Zhou et al., 2014). Therefore, we assessed levels of AMPK activation measured by its phosphorylation and observed a trend for decreased activity in the p53 R270H mutant cells although one R172H cell line had lower AMPK activation too (Figure 5.5B). Next, I assessed lipid levels in the lung tumour cell lines by oil red o staining and although I observed a trend for higher lipid levels in the R270H cells compared to the R172H when I solubilised and quantified the stain, the difference was not significant (Figure 5.6A).

A type of lipid that is predicted to be present at different levels between cells expressing the two types of p53 mutation based on our transcriptome analysis is cholesterol, a sterol and an essential component of cellular membranes. Therefore, I assessed both the localisation and total amount of intracellular cholesterol in the R172H and R270H lung tumour cell lines and found that cholesterol was present at significantly higher levels in the R270H cells compared to the cells expressing the R172H mutation (Figure 5.6B).

Overall, we observed metabolic differences between the lung tumour cells expressing different classes of p53 mutation, particularly in levels of glycolysis and intracellular cholesterol as well as a considerable number of other metabolites.

82 A

Null R172H R270H ns d 3

e ns

e t

m 2

O μ

l i 0 O l l H H u 2 0 N 7 7 1 2 R R

B Cholesterol PM Merge

Null

***

l l

l 1.5

R172H e

e e

v 1.0

R270H C 0.5 l l H H u 2 0 N 7 7 1 2 R R

Figure 5.6. R270H mutant cells display higher levels of intracellular cholesterol than R172H cells.

A. Oil red O was used to stain lipids in the p53 null and mutant cells. Left, representative images of staining of lipid droplets in the lung tumour cells. Scale bar: 20 m. Right, quantification of the Oil red O stain was performed by measuring absorbance after the stain was resuspended and values were normalised to protein content. Representative data (from n=2 cell lines/genotype, ER model) is shown. Average values of staining/genotype are shown relative to p53 null average SEM. Each symbol represents an independent cell line (n=4 analysed/genotype, ER cells). B. Left, filipin (cholesterol) and cell mask orange (plasma, membrane, PM) staining were performed to assess cholesterol levels and localisation in p53 null, R172H and R270H cells. Scale bar: 60 m. Right, cellular cholesterol levels were quantified using a fluorescence-based Amplex red kit, and fluorescence values were normalised to total protein content and are shown relative to average null. Representative data shows two cell lines/genotype with experimental repeats of n=4 cell lines tested (Fx and ER model). Symbols represent individual cell lines with the average/genotype shown SD. ns: non-significant, P>0.05, ***P<0.001, One-way ANOVA P>0.05, One- way Anova.

83 5.2.2. Mevalonate pathway gene upregulation is a gain-of- function specific to p53R270H lung tumours

After observing the difference in cholesterol levels between the R172H and R270H lung tumour cells I proceeded to analyse the cholesterol biosynthesis genes that were differentially expressed.

The pathway responsible for the de novo synthesis of cholesterol is known as the mevalonate (MVA) pathway and it is a pathway which has been implicated in tumorigenesis by supporting cancer cell growth and invasion (Mullen et al 2016). We observed that upregulation of MVA pathway gene expression is a GOF of the p53 R270H mutant in lung tumour cells as MVA pathway genes are upregulated compared to p53 null cells. In addition, it is a true GOF phenotype rather than a retention of wild-type p53 activity by the mutant protein as the genes are upregulated compared to p53WT cells (p53 null + 4OHT) as well. However, what is most interesting is MVA pathway gene upregulation appears to be a mutant-type specific GOF, as it is not observed in the R172H p53 mutant cells. (Figure 5.7A,B). I confirmed that these MVA pathway genes were upregulated in the R270H mutant cells by qRT-PCR (Figure 5.7C).

To establish whether we observe this R270H mutant-specific function in vivo I assessed the expression of MVA pathway genes in lung tumours. To do this, I used tumours from animals from our Fx lung adenocarcinoma model. I collected independent late-stage lung tumours from multiple animals/genotype and snap-froze the tissue in liquid nitrogen immediately after isolation. I also collected normal lung from non-tumour animals as a control. I lysed the tissue and isolated the RNA for qRT-PCR analysis. Encouragingly, I observed higher expression of MVA pathway genes in R270H tumours compared to p53 null and R172H tumours and the normal lung (Figure 5.7D).

The Srebf (also known as Srebp) family of transcription factors controls the expression of genes involved in lipid and cholesterol biosynthesis and Srebf2 in particular is the main regulator of the cholesterol biosynthesis genes (Pai et al. 1998;

84 A B

H H 2 0 ll 7 7 T u 1 2 W N R R Acat2 Acetyl-CoA Acetoacetyl-CoA Hmgcs1 Acat2 Mvk Pmvk Mvd Hmgcs1 Idi2 Fdps HMG-CoA Sqle Hmgcr Transcription factor: Lss Srebf2 Dhcr24 Mevalonate Cyp51a1 Mvk Tm7sf2 Msmo1 Mevalonate-P Dhcr7 Lipa Pmvk Soat2 Pggt1b Mevalonate-PP Rabggta Mvd Srebf2 Ldlr Isopentenyl-PP Squalene Idi2 -1.75 1.75 Fdps Sqle Lss Geranyl-PP Cyp51a1 C Dhcr24 Null Farnesyl-PP Tm7sf2 Msmol2 R172H *** Geranylgeranyl-PP Dhcr7 12 *** R270H Lipa

e 8 Pggt1b Soat2

v i

t *** Rabggta

l Farnesylated

a 3 *** l

l *** u

e *** Cholesterol

r proteins

N Geranylgeranylated

n ***

3 2

o i

5 *** proteins

s Ldlr

o e

r 1 p

x Cholesterol uptake E 0 k d le f2 v v q b m M S re P S D

* s

r ** u

o 30 *

m 20 *

l 10

t 8

v i

t 6

l e

i s

s 2

r p

x 0 E k e k d le k d le k d le v vd l v v q v v q v v q m q m M S m M S m M S P M S P P P Normal Null R172H R270H

Figure 5.7. Higher MVA pathway gene expression in p53 R270H cells and lung tumours.

A. Expression of mevalonate (MVA) pathway genes in p53 null, R172H and R270H cells are displayed in the heatmap (based on microarray data). MeV was used to generate the heatmaps and the software normalised genes/row: Value=[(Value)- Mean(Row)]/SD(Row). The minimum and maximum normalised expression values and their corresponding colours are shown in the scale bars. Null cells are shown in the presence (WT) and absence (Null) of WT p53 restoration (4OHT treatment; 8 hrs) to compare gene expression in mutant cells relative to p53 WT and null conditions. B. Schematic representation of the MVA pathway with genes significantly altered in R172H and R270H shown in red (t.test, P<0.05). C. Expression levels of indicated MVA pathway genes in the p53 null, R172H and R270H lung tumour cells grown in 3D cultures (qPCR). Expression values were normalised to Gapdh. Representative data (of n=3 cell lines/ genotype) show average expression of one cell line/genotype SD of triplicates. ***P<0.001, One-way ANOVA. D. Expression of indicated MVA pathway genes (qPCR) in normal lung tissue and lung tumours of the indicated p53 genotypes (Fx model). Values were normalised to 18S expression. Average expression values for each group are shown relative to the average expression of p53 null lung tumours SEM/genotype (n=10 normal tissue, n=15 tumours/genotype; a minimum of 4 animals analysed/genotype). **P<0.01, * P<0.05, t. test. 85

Horton et al. 2003). Interestingly, the expression of Srebf2 is also upregulated in the R270H lung tumour cells compared to the cells of other p53 deficiencies (Figure 5.7A-C). Therefore, I. hypothesised that Srebf2 was responsible for the upregulation of these genes in the R270H mutant lung tumour cells.

To investigate this hypothesis, I compiled a list of genes considered to be transcriptional targets of Srebf2 based on IPA data of known Srebf2 targets and genes that Srebf2 had been shown to bind in ChIP-seq datasets (www.chip- atlas.org). A number of these targets were significantly upregulated in R270H cells compared to R172H cells (Figure 5.8A, left), which includes many other genes in addition to the MVA genes that I identified earlier. Furthermore, based on the target genes of Srebf2, GSEA identifies Srebf2 function as enriched in the R270H cells compared to the R172H cells (Figure 5.8A, right). This strengthens the hypothesis that it is Srebf2 activity that is altered in the R270H cells, resulting in upregulation of its target genes which includes genes in the MVA pathway.

To establish whether the increased MVA pathway gene expression in the R270H cells was Srebf2-mediated I used an inhibitor of Srebf activation, fatostatin (Kamisuki et al., 2009), on the p53 null, R172H and R270H 3D cultures and examined the effect on the expression of the MVA pathway genes. Fatostatin reduced the expression of the MVA pathway genes to similar levels across all cell lines demonstrating that Srebf transcription factors regulate the expression of these genes in the lung tumour cell lines (Figure 5.8B). To confirm it was the Srebf2 isoform, and not other Srebf proteins, that was regulating the expression of these genes, Srebf2 knockdown was independently performed (by Emma Kerr) in R270H tumour cell cultures. Knockdown of Srebf2 was confirmed by analysing expression of Srebf2 by qRT-PCR (Figure 5.8, upper left). Srebp2 protein levels following knockdown could not be assessed due to a lack of available Srebp2 antibody that reliably detects inactive and active forms. Srebf2 knockdown also resulted in a decrease of MVA pathway gene expression demonstrating that Srebf2 is largely responsible for regulating the expression of MVA pathway genes in R270h lung tumour cells (Figure 5.8C). However, even though Srebf2 knockdown results in a decrease in target gene expression, the correlation between Srebf2 and target gene RNA levels is not always strong. This likely reflects the variability of the qRT-PCR analysis. Targets that are expressed at lower levels following knockdown of Srebf2 may be subject to more variability in the analysis. In

86 A B

Null R172H R270H Ctrl Srebf inhibitor Upregulated in R270H

60 *** 40 20

k 2.0

m o

i P

s 1.5

e ns

p a

x 1.0

R 0.5 ns 0.0 l ul H H N 72 0 1 27 -2.7 3.0 R R

*** C 20 ** ** 10 *** * ** 2.0

* d v ** n

* o

i M

1.5

** * s ns

r t

k p

a 1.0

2 x

1.0 n f

1.0 e

P e

s ns

i s

e 0.5 s

e p

r 0.5 v

a 0.5

t p e

e 0.0

l R

e e

R l 0.0 ul H H 0.0 N 2 0 l l 7 7 C 1 2 3 4 o C 1 2 3 4 o 1 2 T # # # # o T # # # # o R R N P N P shSrebf2 shSrebf2 *** * ** 40 ** * 20 ns ** ** ** ** 2.0

* e

o i

S 1.5

d 1.0

e s

l ns

e q

1.0 i

a 1.0

e e

r t

0.5 R

p a

l 0.5 l x 0.5

x ns

e R R 0.0 0.0 0.0 l l C 1 2 3 4 o C 1 2 3 4 o ll T # # # # o T # # # # o u H H N P N P N 2 0 17 27 shSrebf2 shSrebf2 R R

Figure 5.8. Srebf2 regulates MVA pathway gene expression in the lung tumour cells.

A. Left, heatmap depicting Srebf2 targets which are differently expressed in the R172H and

R270H lung tumour cultures (genes are significantly altered by SAM statistical package within MeV). Expression values were normalised/row and the scale bar shows the minimum and maximum values. Right, enrichment plot for Srebf2 target genes in R172H and R270H cultures (GSEA). B. Expression of indicated genes analysed by qRT-PCR in p53 null, R172H and R270H cultures in the presence or absence (Ctrl) of the Srebf inhibitor, fatostatin. Representative data of n=3 independent runs shows one cell line/genotype SD of triplicates. Two-way ANOVA, ns: non-significant, *** P<0.001. C. Expression of genes analysed by qRT-PCR in p53 null, R172H and R270H cultures with or without short-hairpin

(sh) Srebf2 knockdown. Four constructs (1-4) and a pool of all constructs were analysed. Representative data of n=2 independent runs shows one cell line/genotype SD of triplicates, relative to non-targeting control (NTC). t-test, ns: non-significant, * P<0.05, **P<0.01, *** P<0.001.

addition, levels of active Srebf2 protein remaining in the samples following knockdown may be subject to some deviation because of slight variations in the cell sorting of different shRNA cultures or the separate 3D cultures influencing the stability of the protein. On finding a suitable antibody, it will be important to analyse remaining active/nuclear Srebf2 protein levels following knockdown.

Both p53 mutants were expressed at high levels compared to the p53ER (wild-type) protein, which is in keeping with the accumulation of mutant p53 that is frequently observed in cancer cells (Figure 3.3E) (Bartek et al., 1991). We observed a tendency for higher expression of the R270H mutant over the R172H mutant but this was not significant across all lung tumour cell lines. Moreover, the R172H mutant cells did not display an intermediate phenotype between the null and R270H cell lines for R270H- specific GOFs, such as the higher MVA pathway expression. This indicates that these mutant-type specific GOFs are not due to a difference in the expression of the p53 mutants. In addition, previous data from our lab indicates that it is not a difference in localisation of the p53 mutant that explains the distinct mutant p53- specific transcriptional profiles as both displayed a similar cytosolic and nuclear localisation.

Instead, the mutant-specific functions are likely to involve differences in their ability to bind other proteins. In line this this, mutant p53 proteins have been shown to exert their GOFs via binding to other proteins, such as the transcription factors p63 and p73 (Adorno et al., 2009; Muller et al., 2009; Muller et al., 2013; Weissmueller et al., 2014), in the cancer cells. Moreover, mutant p53 was previously shown to bind Srebf2 and the mutant p53 that the authors analysed in this study was the R273H contact mutant (Freed-Pastor et al., 2012). Therefore, I hypothesised that the difference in the Srebf2-mediated genes expression observed between R172H and R270H lung tumour cells may not only be due to higher Srebf2 expression in the R270H cells, and that there may also be differential binding of the distinct p53 mutants to Srebf2, influencing its activity. However, despite trying different Srebf2 antibodies, I could not find one that reliably detected the active form of Srebf2 to assess binding of the p53 mutants to Srebf2.

Together, these data demonstrate that Srebf2-dependent MVA pathway gene upregulation is a GOF of mutant p53 that is relevant in lung tumours although

88 importantly it is specific to the p53 contact mutant, highlighting that distinct p53 mutations have different functions and may exploit different pathways during tumorigenesis.

5.3. Discussion

Mutant p53 proteins have been shown to exert GOFs through manipulation of metabolic pathways, in addition to growth factor signalling pathways, as was discussed in detail in Chapter 1.3.4. Loss of wild-type p53 has been shown to alter metabolism due to the loss of the p53-mediated metabolic regulation. For instance, p53 null cells are more reliant on glycolysis than p53 wild-type cells (Ma et al., 2007; Sinthupibulyakit et al., 2010). However, the mutant p53 appears to alter metabolism further, as p53 mutant cells have been shown to be more glycolytic than p53 null, enhancing the glycolytic changes further to support tumorigenesis (Zhang et al, 2013). In contrast, our data imply that metabolic differences between the cells expressing distinct p53 mutations are more pronounced than between p53 mutant cells as a single entity compared to p53 null cells. Therefore, it is important to assess the metabolic impact of different p53 mutations.

Importantly, the upregulation of MVA pathway gene expression (via Srebf2) has been described as a function associated with mutant p53 in breast cancer. Interestingly, the authors analysed p53 R273H and R280K mutant breast cancer cell lines, two p53 mutations that belong to the contact mutant subgroup (Freed-Pastor et al., 2012). Our data demonstrate that while it is a GOF of the contact R270H mutation relevant to lung cancer, it is not the case for all p53 mutations. Under physiological conditions, MVA pathway activity is inhibited by high intracellular sterol concentrations or when there is less demand for cholesterol such as when cell growth ceases. However, the MVA pathway is frequently deregulated in cancer cells to provide increased cholesterol synthesis and production of other MVA pathway intermediates to support tumorigenesis (Mullen et al., 2016). Therefore, we propose the higher cholesterol levels we observe in R270H mutant lung tumour cells are indicative of aberrant regulation of Srebf2 and a higher requirement for MVA pathway activity in these cells.

89 6. Characterisation of the dominant-negative effects of p53 mutants in lung tumour cells

6.1 Introduction

Mutations in the p53 gene are selected for in many human cancers, including lung adenocarcinoma (Cancer Genome Atlas Research, 2014) and are more common than p53 loss. Therefore, it is considered that the mutant p53 proteins exert effects in the tumour cells that are additional to just loss of wild-type p53 function. I have analysed and discussed the GOF effects of two p53 mutations in lung tumours, with the GOFs representing one potential reason for the selection for p53 mutations. This chapter will instead focus on the potential dominant negative (DN) effects of mutant p53.

As discussed in detail in chapter 3, there is evidence for p53 mutations interfering with the activity of the wild-type p53 protein. However, the extent of this DN activity remains unclear with loss-of-heterozygosity (LOH) occurring in many cancers including lung cancer, suggesting that p53 mutants do not completely inhibit wild-type p53 (Baker et al., 1990; Brosh and Rotter, 2009; Liu et al., 2016; Mitsudomi et al., 2000; Zienolddiny et al., 2001). It will be important to characterise the extent of the DN effects of the p53 mutants on WT p53 to establish whether p53-restoration therapy could have potential efficacy in p53 mutant lung tumours. Moreover, as I reported earlier, different p53 mutations are not functionally equivalent and therefore the extent of wild-type p53 inhibition may vary depending on the type of mutation present.

In this chapter, I aim to identify and compare the DN activity of the R172H and R270H p53 mutations with the help of the p53ER model.

90 6.2 Results

6.2.1 Transcriptional analysis of wild-type p53 activity in p53 mutant lung tumour cells

As reported in Chapter 3, a substantial subset of genes was similarly induced or repressed by wild-type p53 restoration (4OHT treatment) in p53 null, R172H and R270H lung tumour cultures. Many of these genes were typical p53 targets, identified based on evidence of a functional effect of wild-type p53 on target gene expression or on evidence of p53 binding the target gene from the literature (Figure 3.6; (Bieging et al., 2014; Fischer, 2017a; Lehar et al., 1996; Lo et al., 2001; Okamoto and Beach, 1994; Quintens et al., 2015; Utrera et al., 1998); www.chip-atlas.org). In support of this, the most altered canonical pathway following wild-type p53 restoration was p53 signalling (Figure 6.1A). The cellular functions that some of these identified p53 target genes are involved in regulating was identified by performing further literature searches (Aksoy et al., 2012; Barak et al., 1993; Bersani et al., 2014; Bieging et al., 2014; Collavin et al., 2000; Jerga et al., 2007; Kannan et al., 2000; Kawase et al., 2009; Lehar et al., 1996; Lo et al., 2001; Nithipatikom et al., 2014; Okamoto and Beach, 1994; Tanikawa et al., 2017; Watanabe et al., 1995; Zhang et al., 2013b; Zhao et al., 2012; Zhou et al., 2002), http://chip-atlas.org). Interestingly, this subset of commonly-regulated genes includes p53 targets involved in a range of different p53 functions (Figure 6.1B). This suggests that various tumour suppressive responses may still be induced by wild-type p53, at least to some extent, in p53 mutant cells.

Next, I validated the expression of the well-established, ‘canonical’ p53 target genes Mdm2, p21 (Cdkn1a), Puma (Bbc3), Sesn2, Phlda3 and Bax in the lung tumour cell lines. Firstly, I analysed the expression of three of these genes in the basal setting and I observed no genotype-specific expression difference for these genes (Figure 6.2A). To confirm that these genes were similarly regulated by wild-type p53 across all genotypes, I performed qRT-PCR on four cell lines/genotype following 2hrs and 8 hrs 4OHT treatment (p53 On). All genes were induced by p53 restoration which supports our transcriptome data; no genotype-specific differences in the levels of

Figure 6.1. Wild-type p53 restoration altered the expression of genes involved in various tumour suppressive functions of p53.

A. Ingenuity pathway analysis was performed on the genes similarly altered by p53 restoration across p53 mutant and null cells (8 hrs cohort). The top six canonical pathways associated with these gene expression changes are displayed. Significance was determined by Fischer’s exact test. B. Diagram showing examples of p53 targets genes that are commonly regulated in all genotypes, grouped according to their cellular functions. Ccng1, cyclin G1; Ei24, autophagy associated transmembrane protein; Ddit4, DNA damage inducible transcript 4; Dgka, Diacylglycerol kinase alpha; Ephx1, Epoxide hydrolase 1; Slc19a2, solute carrier family 19 member 2; Trp53inp1, transformation related protein 53 inducible nuclear protein 1; Upp1, uridine phosphorylase 1. Other genes as described previously.

a 4

m n

6 2

u o

p 2

s e

s s

i e

e 4

v r

i 2

p p t l

1 p

e l

e e

e 2

R R 0 0 0 Null R172H R270H Null R172H R270H Null R172H R270H

p53 Off p53 On 2 hrs p53 On 8 hrs 30

40 20 3

20 a

1 10

s 8 s

15 P

r t

v 6

a t

p 10

e l

e 4

e R

5 R

2 0 0 Null R172H R270H Null R172H R270H

300 30

200 20 2

m 100

n 10

n i

30 a

s o

M 8

e e

r i

v 6

e i

t 20

a l

e 4

x e

e e

R 10

2 0 Null R172H R270H Null R172H R270H

30 30

2 20

20 n

a 10

s n

e o n m 10

i 3

s 8

v 2

i e

r i

6 a

x e

l 4 1

e R

R 2 0 0

Null R172H R270H Null R172H R270H

Figure 6.2. Typical p53 targets are similarly induced by wild-type p53 in p53 mutant and null cells.

A,B. Expression of p53 indicated target genes was assessed by qPCR analysis in p53 mutant and lung tumour cells (2D). A. Expression values shown were normalised to 18S expression and are relative to a randomly chosen p53 null cell line. Bars depict expression levels for independent cell lines as indicated SD of triplicates. B. Expression of p53 indicated target genes was assessed following 2 hrs or 8 hrs WT p53 restoration (p53 On). Expression values shown were normalised to 18S expression and are relative to Ctrl (p53 Off) values for a null cell line. Representative data of two independent runs is shown. Bars depict average fold-change in expression for an independent cell line SD of triplicates.

induction of these p53 target genes were observed in the lung tumour cells (Figure 6.2B). The transcriptome and qRT-PCR data were intriguing as they indicated that wild-type p53 was able to induce expression of several of its canonical genes when a p53 mutantprotein was present and to levels similar to those achieved by p53 restoration in p53 null cells.

I also performed time-course experiments to assess the impact of wild-type p53 restoration across a longer period of time and thus determine whether these expression changes were sustained in the cells. Encouragingly, the p53 targets were still upregulated 24-72 hrs after p53 restoration was initiated, although there was a trend observed for higher levels of induction in the p53 null cells at the 72 hrs time point (Figure 6.3).

Together, these data argue towards limited or lack of DN effects of the p53 mutants on wild-type p53 function because multiple canonical p53 target genes are similarly induced or repressed in p53 mutant (both R172H and R270H) and p53 null cells. However, as reported in Chapter 3, we also observed genotype-specific transcriptional effects following wild-type p53 restoration (Figure 3.7). These transcriptional differences included genes that were altered in the p53 null cells but not in one or both of the p53 mutant cells (DN or ‘wild-type-like’ effects). Examples were also validated by qRT-PCR (Figure 6.4A). To identify which wild-type p53 functions the DN effects may be impacting in the p53 mutant cells I investigated the pathways that these genes are involved in. The most significant molecular functions highlighted by the DN effects of both p53 mutants were proliferation and the cell cycle (Figure 6.4B). Therefore, we predicted that p53-mediated cell cycle arrest may be hindered in the p53 mutant cells.

6.2.2. Effect of p53 mutant on wild-type p53 tumour- suppressive functions

Next we aimed to assess the functional impact of these wild-type p53-induced gene expression changes in the p53 mutant lung tumour cells. The investigation of these functional effects was a collaborative effort between myself and other members of the Martins’ group.

30 80

1 n

2 20

e 20

v i

l x

e 10

e R

0 0 p53 On 0 24 48 72 0 24 48 72 0 24 48 72 hrs 0 24 48 72 0 24 48 72 0 24 48 72 hrs Null 1 R172H 1 R270H 1 Null 2 R172H 2 R270H 2

40 80

40 30 30

M s 20

s 20

l e

e 10 10

0 0 p53 On 0 24 48 72 0 24 48 72 0 24 48 72 hrs 0 24 48 72 0 24 48 72 0 24 48 72 hrs Null 1 R172H 1 R270H 1 Null 2 R172H 2 R270H 2

5 8

a n

m 3

s 4

r v

i 2

l e e 2 R 1

0 0 p53 On 0 24 48 72 0 24 48 72 0 24 48 72 hrs 0 24 48 72 0 24 48 72 0 24 48 72 hrs Null 1 R172H 1 R270H 1 Null 2 R172H 2 R270H 2

Figure 6.3. p53 target genes are upregulated at later time points in p53 null and mutant cells.

Expression of the indicated target genes was assessed by qPCR analysis in p53 mutant and null lung tumour cells (2D, 2 cell lines/genotype) following 24, 48 or 72 hrs of wild-type p53 restoration (p53 On). Expression values shown were normalised to 18S expression and are shown relative to Ctrl (p53 Off) values. Representative data of two independent runs is shown. Bars depict average fold- change in expression for each cell line SD of triplicates.

Figure 6.4. Validation of DN and wild-type-like transcriptional effects.

A. Expression of the indicated p53 target genes was assessed by qPCR analysis in p53 mutant and null lung tumour cells (2D; 2 cell lines/genotype) following 8 hrs of wild-type p53 restoration (p53 On). Expression values shown were normalised to 18S expression and are relative to average Ctrl (p53 Off) values for null. Representative data of two independent runs is shown. Bars depict average fold-change in expression for an independent cell line SD of triplicates. B. Top two molecular functions following pathway analysis of the DN transcriptional effects of the R172H and R270H mutants. The significance was determined by Fischer’s exact test.

First, we measured the impact of p53 restoration on the live cell number in p53 null, R172H and R270H cells. Cells were treated with vehicle or 4OHT and viability was determined by trypan blue exclusion at specific time points post-restoration. Importantly, we observed no significant difference in the response to p53 restoration across the p53 genotypes, demonstrating that wild-type p53 results in a similar reduction in viable cells in p53 mutant and null cultures (Figure 6.5A,B). We also assessed the effect of 4OHT treatment on the number of live cells in p53-/- cells (Fx model) of found no effect of 4OHT on cell viabiity. Thus, we were able to confirm that the impact of 4OHT treatment on the p53 null and mutant cells (ER model) was due to wild-type p53 restoration and not another effect of the 4OHT treatment (Figure 6.5C).

To determine what was responsible for this reduction in live cell number following wild-type p53 restoration in these cells we performed BrdU and PI staining and subsequent FACS analysis, to assess whether wild-type p53 induced cell cycle arrest. We observed a similar induction of cell-cycle arrest (lower percentage of cells in S-phase) following wild-type p53 restoration in p53 mutant and null cells (Figure 6.5D). Importantly, this result was recapitulated in 3D cultures which were the model used for transcriptome analysis (Figure 6.5E).

Wild-type p53 can also induce apoptosis and senescence (Bieging et al., 2014), which may also be partly responsible for the p53-mediated reduction in live cell number observed in the lung tumour cells. We observed a small induction of apoptosis following p53 restoration in the p53 mutant and null 3D cultures (Figure 6.6A,B) but interestingly not in 2D cultures (data not shown). In addition, we observed no induction of senescence in the p53 null and mutant lung tumour cells (Figure 6.6C).

To summarise, despite some evidence of transcriptional DN effects of the p53 mutants on WT p53 function, the tumour suppressive activity of the wild-type protein (cell cycle arrest and apoptosis) prevailed over mutant p53, and independently of the type of p53 mutation present. Therefore, the DN effects of the p53 mutants in these lung tumour cells are limited and the subset of similarly-regulated p53 target genes are sufficient to induce significant and long-lasting tumour-suppressive responses in p53 mutant cells.

A B Null R172H R270H 72 hrs (2D) 100

T ns 0.5 0.4 0.4

p53 Off H

) ) ) p53 On 80

0.4

6 6 6

0.3 0.3 r

0 0 0 e

1 1 1

( ( (

0.3 a

s s

s 60

l l l

0.2 s

l l l

0.2 l

e e e

c c

c 0.2

e e

e 0.1 40

v v

v 0.1

i i i

l l l

0.1 i l

0.0 0.0 0.0 % 20 0 12 24 48 72 hrs l 0 12 24 48 72 hrs 0 12 24 48 72 hrs l H H u 2 0 N 7 7 1 2 R R C D Fx/Fx Null R172H R270H 2.0

ns )

6 1.8 ^ 41.2% 36.2%

0 41.0%

1 (

1.6

c U

1.4 U

r e

p53 Off r

B B l 1.2

1.0 Ctrl 4OHT PI PI PI E 72 hrs (3D) ns 26.4% 24.7%

s 26.4%

l 100 ns

t 80

i d

t 60

s r

p53 On r

o e

p i

U a

d e r

r 20

% 0 ll H H PI PI PI u 2 0 N 7 7 1 2 R R

Figure 6.5. Similar reduction in proliferation observed in p53 null and mutant cell lines after wild-type p53 restoration.

A. Growth of p53 null and mutant cell lines (1 cell line/genotype shown of 4 analysed) following wild-type p53 restoration (p53 On) or vehicle treatment (p53 Off). The number of viable cells, measured by trypan blue exclusion, at indicated time points is shown. Symbols show triplicate average SD B. Percentage viability of p53 null and mutant cell lines after 72 hrs of wild-type p53 restoration (4OHT treatment), measured by trypan blue exclusion. Four independent cell lines/genotype are shown SEM. C. Number of live cells following 72 hrs of Ctrl or 4OHT treatment in a p53Fx/Fx cell line. Triplicate mean is shown SD. Representative data of n=3 cell lines analysed. D.

Representative images of FACS plots showing the change in the percentage S-phase cells (BrdU-positive) following p53 restoration (24 hrs, 2D). E. Percentage of BrdU- positive cells in p53 null, R172H and R270H 3D cultures following 72 hrs p53 restoration (4OHT treatment), relative to Ctrl analysed by FACS. Symbols represent independent cell lines and SEM of average percentage/genotype is shown. C,E, ns: non-significant P>0.05, One-way ANOVA. BrdU analysis was performed by Meiling Gao and myself.

98 A C 72 hrs (3D) Palbociclib-treated 6 ns * ns

WT/WT

p53

s 5

H c

d 4

a p53 Off p53 On

d f

% 3 Null D 2 p53 Off On Off On Off On R172H Null R172H R270H

R270H

Null R172H R270H

p53 Off

p53 On

DAPI Cleaved caspase 3

Figure 6.6. p53 restoration induces apoptosis, but not senescence, in p53 mutant and null cell cultures.

A. Percentage of dead cells 72 hrs after vehicle or p53 restoration (4OHT treatment). Representative data is shown, displaying one cell line/genotype of three cell lines analysed. The triplicate mean is shown SD. B. Representative immunofluorescence images of cleaved caspase 3 staining 72 hrs after p53 restoration in p53 null and mutant cell lines (3D cultures). Examples of positive staining are shown by arrows. Scale bar: 100 m. C. Images show representative -galactosidase staining in p53 null, R172H and R270H lung tumour cells following 72 hrs p53 restoration (p53 On). One cell line/genotype is shown of two cell lines analysed. Palbociclib treatment of wild-type p53 human cancer cells was used as a positive control. Scale bar: 40 m. Cell death analysis was performed by Emma Kerr and the senescence assay was performed by Hannah Thorpe. The image analysis of the senescence assay was performed by myself.

99 6.3 Discussion

Our data argue against there being significant DN effects of the p53 mutations which can impact on the well-established tumour-suppressive functions of p53 (cell cycle arrest and apoptosis). This is despite previous evidence for DN effects of mutant p53 (including R175H and R273H) on wild-type p53 (Aurelio et al., 2000; Dong et al., 2007; Wang et al., 2011; Willis et al., 2004). Instead our data may explain the LOH phenotype observed in many tumours (Baker et al; Brosh and Rotter et al 2009, Mitsudomi et al. 2000; Zienolddiny et al. 2001; Liu et al. 2016). In other words, even if the p53 mutation occurs first tumour cells still profit significantly from counter- selection of the wild-type p53.

The functional impact of the DN transcriptional effects observed for the p53 mutants is unclear. It may be that these DN effects manifest at other levels not covered by our analysis. Moreover, although we did not observe any difference in the DN effects of R172H and R270H mutants in vitro, it has been demonstrated that these different mutants can affect wild-type p53 distinctly in vivo (Jackson et al., 2005; Olive et al., 2004, Wang et al., 2011). Next, it will be important to assess the impact of p53 restoration in the lung tumours to establish whether the presence of R172H and R270H mutants can affect the p53-mediated tumour suppressive responses in vivo.

The wild-type-like transcriptional effects of the two mutants that we observed are also intriguing, as they suggest that p53 mutants can retain at least some wild-type functions. According to published ChIP datasets (www.chip-atlas.org;(Fischer, 2017a)), there is direct-binding evidence for p53 binding to promoter regions of the majority (~70%) of these ‘wild-type-like’ genes. These data indicate that p53 mutants retain some transcriptional functions of wild-type p53, contrary to the view that these p53 mutations lose all wild-type DNA-binding ability (Joerger and Fersht, 2007). It will be important to investigate further these ‘wild-type-like’ transcriptional effects will be important to investigate further as it is possible that the mutants may retain parts of wild-type p53 activity that they can exploit in tumorigenesis.

100 7. Targeting p53 mutant lung tumours

7.1. Introduction

Lung tumours with p53 mutations are associated with poor overall survival and comprise almost half of lung adenocarcinoma cases, as described in chapter 3 (Cancer Genome Atlas Research, 2014). Importantly, there is currently a lack of targeted therapies against mutant p53 and so novel therapies which target p53 mutant lung tumours are urgently needed.

The overall aim of this chapter is to investigate whether we can uncover any therapeutic susceptibility of p53 mutant tumours based on our transcriptional and functional profiling of p53 null, R172H and R270H lung tumour cells. I will establish whether wild-type p53 restoration can be beneficial in p53 mutant lung tumours. This will allow us to determine the potential impact of p53-targeted therapies (i.e. restoring wild-type function) in p53 mutant lung tumours. I will also determine whether R270H p53 mutant tumours are dependent on the MVA pathway and whether by inhibiting this pathway downstream of the R270H mutant we can uncover a way of targeting these R270H mutant lung tumours.

7.2. Results

7.2.1. Similar tumour-suppressive responses are induced by wild-type p53 restoration in p53 mutant and null tumours

Previously, it was shown that wild-type p53 restoration successfully targeted p53 null lung tumours (Feldser et al., 2010; Junttila et al., 2010), and we aimed to determine to what extent wild-type p53-mediated tumour-suppressive could be effective in p53 mutant lung tumours. To address this, we restored wild-type p53 in advanced p53 mutant and null lung tumours and assessed the resulting levels of proliferation and apoptosis in these tumours. To restore wild-type p53, tamoxifen (which is metabolised into 4OHT in the liver (Crewe et al., 1997)) was administered daily to the

101 animals. The in vivo work involving p53 restoration was initiated by Meiling Gao and completed by myself and Emma Kerr.

After 6 days of wild-type p53 restoration or control treatment we harvested the lungs and stained the sections with Ki67, a proliferation marker. Importantly, we observed a statistically significant and comparable decrease in proliferation in p53 null and p53 mutant lung tumours (Figure 7.1A). To confirm that this anti-proliferative effect was specific to wild-type p53, and not another effect of the 4OHT treatment, I treated KrasG12Dp53Fx/Fx animals with 4OHT. As was the case in vitro, no cytostatic response was observed in this model, demonstrating that the effect was specific to the p53ER model and thus, to p53 restoration (Figure 7.1B).

A different cohort of ER mice was treated with tamoxifen or vehicle for 24 hours to identify immediate effects of p53 restoration. We observed an induction of apoptosis in p53 null, R172H and R270H tumours, following wild-type p53 restoration, which we measured by TUNEL staining (Figure 7.2A). Quantification of the staining revealed that apoptosis was significantly induced across all genotypes (Figure 7.2B). In addition, as with the cytostatic response, the 4OHT-induced cytotoxic response was not observed in KrasG12Dp53Fx/Fx animals and so was again due to p53 restoration (Figure 7.2C).

Taken together these results demonstrate that wild-type p53 tumour suppressive functions prevail even in the presence of the mutant p53 protein. Thus, any DN effects of the mutant p53 are limited in vivo, as responses to wild-type p53 restoration are comparable in p53 mutant and null lung tumours. This highlights the potential of using p53-targeted therapies for the treatment of p53 mutant and null lung tumours.

7.2.2 p53R270H lung tumours display enhanced sensitivity to mevalonate pathway inhibition

Next, I sought to determine whether we could use our knowledge of the transcriptional differences between p53 null, R172H and R270H lung tumour cells to identify genotype-specific susceptibilities. To achieve this, I explored the sensitivity of

102 A ER model

25 ** *** ***

r u

o 20

l l

e 15

t i

s 10

i K

5 %

0 p53 Off On Off On Off On n=26 n=21 n=31 n=28 n=37 n=27 Null R172H R270H B Fx model

50 ns

r u

m 40

l e

c 30

i s

o 20 p

i K 10 % 0 Ctrl Tam n=22 n=31 Fx/Fx Figure 7.1. Similar anti-proliferative effects in p53 null and mutant lung tumours following wild-type p53 restoration.

A. Percentage Ki67-positive cells/tumour in p53 null, R172H or R270H animals treated with Ctrl (p53 Off) or Tamoxifen (p53 On) for 6 days. A minimum of 3 animals were analysed/cohort and the number of tumours analysed is indicated. Average positivity/cohort SEM is shown. Treatment of this cohort of mice and subsequent analysis was a collaborative effort between Emma Kerr, Meiling Gao and myself. B. Percentage Ki67-positive cells/tumour in p53Fx/Fx animals treated with Ctrl or Tamoxifen (Tam) for 6 days. A minimum of 3 animals were analysed/cohort and the number of tumours analysed is indicated. Average positivity/cohort SEM is shown. ns: non-significant, P>0.05 ** P<0.01 ***P<0.001, t.test.

103 A

Null R172H R270H

p53 Off

p53 On

B C

ER model Fx model r

r ns

u 2.5

** ** ** u 5.0

o m

m 4.0

u u

t /

2.0 /

s s

l 3.0

e c

c 2.0

1.5

e v

t i

i 1.5

s o

1.0 o

1.0

L E

E N

0.5 N 0.5

% 0.0 0.0 % p53 Off On Off On Off On Ctrl Tam n=28 n=34 n=81 n=69 n=43 n=32 n=18 n=32 Null R172H R270H

Figure 7.2. Wild-type p53 restoration induces similar apoptotic responses in p53 null and mutant lung tumours

A. Representative images of TUNEL staining (arrows) in p53 null, R172H and R270H tumours. Scale bar: 60 m. B. Percentage TUNEL cells/tumour in p53 null, R172H and R270H animals following p53 restoration (p53 On, 24 hrs) or vehicle treatment (p53 Off). Average positivity/cohort is shown SEM. A minimum of 3 animals were used/cohort with the number of tumours analysed indicated. A,B. Treatment of this cohort of mice and subsequent analysis was a collaborative effort between Emma Kerr, Meiling Gao and myself. C. Percentage TUNEL-positive cells/tumour in p53Fx/Fx animals treated with vehicle or Tamoxifen (Tam) for 6 days. A minimum of 3 animals were analysed/cohort and the number of tumours analysed is indicated. Average positivity/cohort SEM is shown. ns: non-significant, P>0.05, **P<0.01, t.test.

104 each genotype to MVA pathway inhibition. Inhibition of the MVA pathway can be achieved via statins which inhibit Hmgcr, an enzyme that catalyses an early step in the MVA pathway cascade (Figure 7.3A; (Istvan and Deisenhofer, 2001)). Therefore, I treated p53 null, R172H and R270H 3D cultures (using both Fx and ER models) with simvastatin for 72 hours to assess whether statins impacted colony growth of cells expressing a particular subclass of p53 mutation. I observed an inhibition of colony growth after statin treatment in all cell lines but importantly, R270H lung tumour cells displayed higher sensitivity to the treatment (Figure 7.3B). Importantly, this statin-induced inhibition of cell growth was rescued by addition of mevalonate (MVA), the metabolite downstream of Hmgcr, demonstrating the growth inhibitory effect of statins is due to inhibition of MVA pathway activity (Figure 7.3C).

Subsequently, I aimed to investigate the mechanism for the increased sensitivity to statin treatment that we observed in the R270H contact mutant cell lines. It was reported that statins could induce a reduction in mutant p53 stability, which rendered the cells sensitive to statin treatment (Parrales et al., 2016). Therefore, I investigated the impact of statin treatment on mutant p53 expression in our p53 R172H conformational mutant and R270H contact mutant lung tumour cultures. I observed statin-induced mutant p53 degradation in our lung tumour cells that ranges from a 10 to 55% reduction in mutant p53 levels but this reduction does not correlate with sensitivity to statin treatment (Figure 7.4A)

Furthermore, we previously concluded that MVA pathway gene expression was regulated by the transcription factor Srebf2. Therefore, to confirm that R270H lung tumour cells were dependent on the upregulation of the genes in this pathway, we assessed the impact of the Srebf inhibitor (Fatostatin) on 3D cultures. While fatostatin treatment reduced the expression of the MVA genes to similar levels across all genotypes (Figure 5.8), R270H cells were more sensitive to this inhibitor. (Figure 7.4B). We also confirmed that this growth-inhibitory phenotype was specific to Srebf2 inhibition by assessing the impact on 3D growth after Srebf2 knockdown (Figure 7.4C), knock-down performed by Emma Kerr). These findings demonstrated that R270H mutant cells are more dependent on MVA pathway gene expression, as well as pathway activity shown by their sensitivity to statin treatment Figure 7.3B).

105

A C

Transcription factor Ctrl Stn Stn+Mva Srebf2

Acat2 Acetyl-CoA Acetoacetyl-CoA ns ns ns Hmgcs1 120

HMG-CoA e v

Hmgcr i Statins t

a 100

Mevalonate l

Mvk t 80

C l

Mevalonate-P

t c

Pmvk 60

Mevalonate-PP v i l 40 Mvd Isopentenyl-PP % Idi2 Squalene Null R172H R270H Fdps Sqle Geranyl-PP Lss Cyp51a1 ns ns Farnesyl-PP Dhcr24 * Tm7sf2 100 Msmol2 e Geranylgeranyl-PP v

Farnesylated i

Dhcr7 t

proteins Pggt1b a 80 Lipa l

Rabggta e

l r

Soat2 r

Geranylgeranylated s

l 60

C l proteins Cholesterol

Cholesterol uptake Ldlr

e 40 v

i l

20 % B Null R172H R270H * 100 *

ns ns n

t *

80 e

i 120

t l

f e

60 l a

r 100

e 80 o

c 40

v i

v 60

20 % % 40 0 Null R172H R270H Null R172H R270H

Figure 7.3. p53 R270H cells display increased sensitivity to treatment with simvastatin compared to p53 null and R172H cells.

A. Schematic representation of the MVA pathway shows the step at which statins inhibit the pathway. B. Percentage live cells relative to vehicle after 72 hrs of simvastatin (Stn) treatment (ER and Fx cells, 3D culture). Average percentage live cells/genotype is shown SEM, with symbols representing independent cell lines. Data is representative of n=2 independent repeats. * P<0.05, One-way ANOVA. C. Mevalonate (MVA) addition to Stn-treated 3D cultures on live cell number for three cell lines/genotype performed at independent times (Fx model, one cell line/genotype shown in each graph). Bars shown triplicate average for a cell line SD. ns, * P<0.05, t.test.

106 Following this, I sought to investigate why the R270H cells were dependent on this higher MVA pathway gene expression. The MVA supports multiple cellular functions. On one hand, it provides cholesterol to the cell, which supports the synthesis of new membranes and therefore cell proliferation. In addition, the MVA pathway produces many intermediates that are important for the modification and activity of the Ras- superfamily of GTPases. These GTPases require prenylation in the form of farnesylation or geranylgeranylation for their localisation to the plasma membrane and subsequent activation (Mullen et al., 2016)

Therefore, I assessed the impact of the addition of downstream intermediates to establish which branch of the pathway the R270H cells depended on. Addition of the intermediate GGPP to the statin-treated cells partially rescued the growth inhibitory effect of the statins in the R270H tumour cells (Figure 7.4, left). These data suggest that the R270H cells may rely more on the activity of a Rho-family GTPase member which requires geranylgeranylation and that its activity is supported through increased MVA pathway expression. However, addition of exogenous cholesterol also partially rescued the growth inhibitory effect of statin treatment, implying that the cells are at least to some extent dependent on cholesterol generated by the MVA pathway (Figure 7.4, right). Thus, it appears that the R270H cells are not solely dependent on one branch of the MVA pathway. Further analysis of the effects of these downstream intermediated in p53 null and R172H cells will provide more information as to which branch of the pathway R270H cells are more reliant on, compared to the other genotypes.

After demonstrating that p53 R270H mutant lung tumour cells displayed higher sensitivity to MVA pathway inhibition than p53 R172H or null cells, the main question we sought to ask next was whether we could use statins to target R270H p53 mutant lung tumours. For that purpose, I treated our endogenous KrasG12D/+;p53Fx/Fx, KrasG12D/+;p53R172H/Fx and KrasG12D/+;p53R270H/Fx (Fx model) advanced lung tumours for 6 days with simvastatin. We observed a significant decrease in the percentage of proliferating cells measured by Ki67-positivity in the R270H tumours treated with simvastatin but not the R172H or p53 null tumours (Figure 7.5A). In addition, the statin-treated R270H tumours displayed an induction of apoptosis as assessed by TUNEL staining (Figure 7.5B). Following statin treatment there was a significant increase in the percentage TUNEL-positive cells/tumour but no induction of apoptosis

107 A

R172H R270H R172H R270H 100 R270H R172H n

Stn - + - + - + - + - + - + l

o 80

e 60

o r

p53 t

e x

(mutant) v e

i 40

3 a

5 2 e

p R =0.1173 r

Actb % P=0.5063 0 0 20 40 60 80 100 % live cells relative to vehicle

B C

Ctrl R270H Fatostatin *** ns ns ** ***

*** l

r ***

t C

100 ***

i s

a 1.0

l o

l 50

c e

v i

e 0.5

R %

0 0.0 l Null R172H R270H C 1 2 3 4 o T # # # # o N P shSrebf2 D

1.2 1.2

r 1.0 1.0

t 0.8

e 0.8

t a l e 0.6

r 0.6

l e

0.4 0.4

i l

% 0.2 0.2

0.0 0.0 l l l tr n a tr tn a o t v P C S v h C S M P M C + G + + tn G tn tn S + S S tn S

Figure 7.4. Determining why p53 R270H cells are more sensitive to simvastatin.

A. Left, expression levels of mutant p53 and Actb (loading control), assessed by immunoblotting, in p53 null, R172H and R270H lung tumour cell lines (Fx model, 3D) 24 hrs after simvastatin (Stn) treatment. Right, correlation between % mutant p53 protein levels following Stn treatment (relative to Ctrl), measured by band densitometry plotted against the % live cells after Stn treatment (relative to Ctrl). B. Live cell number following 72 hrs fatostatin (Srebf inhibitor) or Ctrl treatment of 3D cultures of p53 null, R172H and R270H cells. Representative data of n=3 independent runs shows one cell line/genotype SD of triplicates. Two-way ANOVA, ns: non-significant, ** P<0.01. C. Average colony size of R270H cells grown in 3D cultures with and without shSrebf2 knockdown. Four constructs and a pool of the constructs were analysed. Representative data of n=2 independent runs shows one cell line/genotype SD of triplicates, relative to non-targeting control (NTC). t-test, ns: non-significant, *** P<0.001. D. Percentage live cells relative to Ctrl shown for indicated treatments (72 hrs). Representative data showing one cell line of two analysed (Fx model). Triplicate mean is shown SD.

108

in p53 null or R172H tumours (Figure 7.5C). Therefore, our data demonstrate that simvastatin treatment can selectively target p53 R270H mutant, but not p53 R172H or null, lung tumours.

To better examine the impact of simvastatin treatment on tumour growth over time, I used an orthotopic model of p53 R270H mutant lung cancer by transplanting luciferase-expressing R270H cells into recipient mice. I imaged the animals weekly to assess tumour burden and found that after 2 weeks of daily statin treatment the tumour burden, measured by luciferase activity, was significantly reduced compared to the control-treated animals (Figure 7.6A). Furthermore, simvastatin treatment resulted in improved host survival but these results were not significantly significant (Figure 7.6B).

Finally, I asked the relevance of these findings to human cancer. To assess this question I utilised publically available data for human lung adenocarcinoma cases (Cancer Genome Atlas Research, 2014). First, I classified p53 mutations into the two classes of p53 mutant, conformational and contact, based on evidence of the location of the mutations in the DNA-binding domain and overall impact on the structure of the protein (Table 6.1; (Cho et al., 1994; Joerger et al., 2006; Joerger and Fersht, 2007). Then I analysed the expression data for these tumours and excitingly I found a similar upregulation of MVA pathway genes in the R273 and other contact mutant tumours compared to the R175 and other conformational mutant tumours (Figure 7.7). Hence, the transcriptional signatures that underlie the sensitivity to simvastatin were also observed in these human samples, suggesting that use of statins to treat these tumours may be clinically relevant.

To summarise, I have shown that lung tumours expressing a R270H p53 mutation can be successfully targeted by treatment with statins and that the anti-proliferative and apoptotic effects of statin treatment are only observed in the R270H lung tumours and not in p53 null or R172H tumours.

109 A C

ns ns *** ns ns ***

r 20 u

40 o

r 10

m o

u 8

l t

/ 30

l l

e s

v 20 i

t 4

10 N 2

% % 0 0 Ctrl Stn Ctrl Stn Ctrl Stn Ctrl Stn Ctrl Stn Ctrl Stn Null R172H R270H Null R172H R270H

B Null R172H R270H

Ctrl

Stn

50 µm 50 µm

Figure 7.5. Statin treatment induces a cytostatic and cytotoxic response in p53 R270H lung tumours but not in R172H and null tumours.

A. Percentage Ki67-positive cells/tumour in p53 null, R172H or R270H animals (Fx model) treated with Ctrl or simvastatin (Stn) for 6 days. A minimum of 47 tumours from 5 animals were analysed/cohort. Average positivity/cohort SEM is shown. B. Representative TUNEL staining (arrows) in p53 null, R172H and R270H (Fx model) lung tumour sections from Ctrl or statin (Stn)-treated mice (6 days). Scale bar: 50 m. C. Percentage TUNEL cells/tumour in p53 null, R172H or R270H animals (Fx model) treated with Ctrl or simvastatin (Stn) for 6 days. Average positivity/cohort is shown SEM. A minimum of 93 tumours (n=5 mice) was analysed/cohort. A,C ***P<0.001, t-test.

110

Figure 7.6. Simvastatin treatment reduces lung tumour burden and may improve host survival in p53 R270H lung tumour model.

A. Lung tumour burden (luciferase activity) in animals transplanted with luciferase-expressing R270H lung tumours cells (i.v.) following vehicle (ctrl) or simvastatin (Stn) treatment. The luciferase activity shown is relative to the start of treatment (week 0). Symbols represent animals (n≥5 mice/cohort) and the average relative luciferase activity is shown/cohort SEM ***P<0.001, t-test. B. Survival of recipient mice transplanted with R270H lung tumor cells and treated with Ctrl or Stn. Log- rank (Mantel-Cox) Test. A,B. Daily treatments were started 14 days after transplantation (week 0).

111 7.3 Discussion

We have explored two avenues of targeting p53 mutant lung tumours in this chapter. We have demonstrated that restoration of wild-type p53 induces similar tumour- suppressive responses in p53 mutant (conformational and contact) and p53 null lung tumours. A number of therapies based on restoring wild-type p53 function in p53 null tumours have been investigated (Cheok et al., 2011; Frezza and Martins, 2012). Importantly, our findings imply that wild-type p53 restoration therapies may show efficacy in the clinic in p53 mutant as well as p53 null tumours. Our data also confirm a lack of DN function of the p53 mutants in vivo, at least when considering the cytostatic and cytotoxic functions of p53, similar to what we observed previously in vitro. As discussion in Chapter 6.3, these results may help explain the LOH of p53 that is observed frequently in tumours (Baker et al., 1990; Jackson et al., 2005; Liu et al., 2016; Mitsudomi et al., 2000; Zienolddiny et al., 2001).

Furthermore, in this chapter we showed that p53 R270H mutant lung tumours are sensitive to treatment with statins, a class of clinically available drugs. This finding importantly highlights that tumours expressing different p53 mutations can display distinct responses to therapy, a concept that has been under-explored previously, as p53 mutant tumours are generally considered as a single entity. In contrast to our findings, it was reported by one group that p53 conformational mutations were more sensitive to statin treatment (Parrales et al., 2016). Our findings are instead in line with what has been observed in breast cancer, where breast cancer cells expressing R273H and R280K p53 mutations were reported to be sensitive to simvastatin treatment (Freed-Pastor et al., 2012). Therefore, it may be that the effect of statin treatment is tissue-specific. The response to statin treatment may also differ depending on the in vitro system used: 2D (Parrales et al., 2016) or 3D (Freed-Pastor et al., 2012) cultures. Importantly, our in vitro data translates into comparable responses following simvastatin treatment in endogenous lung tumours.

112

Table 7.1. Classification of TP53 mutations in lung adenocarcinoma cases (TCGA, Nature 2014).

113

-1.5 1.5 -2.0 4.5

Figure 7.7. Increased MVA gene expression occurs in human lung tumours that express a contact p53 mutation compared to a conformational p53 mutation.

MVA pathway gene expression in lung adenocarcinoma patient

samples (TCGA, Nature 2014) with a R175 or R273 mutation (left) or conformational or contact mutation (right) shown in heatmaps. Expression values were normalised/row and the scale bar shows the minimum and maximum values. Genes significantly upregulated in R273/contact groups following SAM (Significance Analysis for Microarrays) analysis are displayed. * denotes a sample with a R175H and a C277F (contact) mutation.

The anti-cancer impact of statins in the clinic has so far been somewhat controversial (Mullen et al., 2016), but the use of statins has been correlated with improved survival of lung cancer patients (Cardwell et al., 2015; Lin et al., 2016). Treating a subset of lung tumours that have specific molecular alterations (such as contact p53 mutations that result in increased dependence on the MVA pathway may improve the efficacy of statin-based anti-cancer therapy in the clinic.

114 8. Discussion

8.1. Summary

In this dissertation I have provided the first comprehensive transcriptional analysis of the effect of two common p53 mutations (R172H and R270H) in lung tumour cells, both in the presence and absence of wild-type p53. The transcriptional and functional profiling of the lung tumour cells led us to identify a p53 genotype-independent response of Kras mutant lung tumours to wild-type p53 restoration but a p53 mutation-type specific sensitivity of lung tumours to simvastatin treatment. Importantly, p53 mutant tumours can no longer be considered as a single entity as they do not always function similarly, particularly in their regulation of cellular metabolism, and can therefore display distinct responses to therapeutic agents.

8.2. Impact of p53 mutations in lung cancer

Although p53 mutations were established as an important oncogenic event around 30 years ago (Weisz et al., 2007), the impact of p53 mutations in tumorigenesis has not been fully delineated. Mutations in p53 are considerably more common than p53 loss in human lung adenocarcinoma cases, with p53 mutations occurring in almost 50% of cases and only three cases of p53 loss reported in TCGA (Cancer Genome Atlas Research, 2014). The presence of a p53 mutation is associated with poor survival in human lung adenocarcinoma when compared to cases without a p53 mutations, and the same is true in mice (Cancer Genome Atlas Research, 2014; Jackson et al., 2005; Jackson et al., 2001).

Our analysis has defined the impact of R172H (conformational) and R270H (contact) p53 mutations in KrasG12D-driven lung tumours. When expressed throughout the mouse, p53 mutations cause a predisposition to a different spectrum of tumours, including more invasive carcinomas, compared to p53 null animals (Lang et al., 2004; Olive et al., 2004). However, as we observed no difference in aggressiveness between p53 null and p53 mutant (R172H and R270H) lung tumours, it is likely that mutant p53 has different effects in the lung tumours compared to other tissues. We have established how distinct p53 mutations differ from each other functionally, an area of investigation which has been so far largely overlooked. Mutant p53 has 115 previously been associated with enhanced glycolysis, lipid synthesis, and mevalonate (MVA) pathway upregulation as well as AMPK inhibition in cancer cells (Freed-Pastor et al., 2012; Zhang et al., 2013a; Zhou et al., 2014). Interestingly, in our lung tumour cells these metabolic gain-of-function phenotypes were associated with the R270H contact and not the R172H conformational mutant protein. These discrepancies are likely due to tissue-specific effects of the p53 mutants or the method used to investigate the impact of mutant p53: p53 knockdown or overexpression compared to our analysis which investigated mutant p53 in a physiologically relevant setting.

However, the metabolic impact of the R172H mutant still remains to be elucidated. The majority of the R172H-specific metabolic effects centred around protein and amino acid metabolism. One possibility is that specific amino acid biosynthesis or degradation pathways are altered in the R172H cells. For instance, the serine biosynthesis pathway, a pathway emerging with important roles in tumorigenesis (Amelio et al., 2014), may be differently regulated in the R172H cells. Amino acid depletion experiments would help ascertain whether the R172H cells are more reliant on the exogenous supply of particular amino acids.

The mutant-specific functions likely occur due to the p53 mutants binding different proteins. Mutant p53 has been reported to bind Srebf2 in R273H breast cancer cells (Freed-Pastor et al., 2012). It is possible that R172H and R270H differ in their ability to bind to Srebf2 or that both mutant proteins bind but only the Srebf2-R273H p53 complex is recruited to the Srebf2-response elements. This can be addressed by performing ChIP-seq analysis of the R172H and R270H proteins, to identify if only the R270H mutant binds to Srebf2 target genes. Interestingly, a preliminary, unbiased investigation into the binding partners of the R172H and R270H mutants by mass spectrometry analysis indicated that these mutants bind different proteins in lung tumour cells. Of the binding partners identified 15% were common to both mutants, 32% were bound to the R172H mutant and 53% were bound by the R270H mutant protein (data not shown). Further analysis into the binding partners will be key to improving our understanding of the distinct functions of p53 mutant.

We show that wild-type p53 function prevails over mutant p53 in lung tumours, providing an explanation for the frequent loss-of-heterozygosity (LOH) observed in

116 tumours, including lung tumours (Baker et al., 1990; Jackson et al., 2005; Liu et al., 2016; Mitsudomi et al., 2000; Zienolddiny et al., 2001). Mutant p53 is more highly expressed in our lung tumour cells that the p53ER (wild-type) protein but quantification of the western blots indicates that the mutant is on average only 2.1 fold more highly expressed than the wild-type protein. Interestingly, it has been shown that for p53 mutants at least three mutant subunits are required to inhibit the activity of the tetramer (Chan et al. 2004). Therefore, it is likely that the mutants, which are expressed at endogenous levels, are not expressed at a sufficient level to form these predominately mutant tetramers. This provides a potential explanation for the discrepancy with previous studies, which demonstrate that the mutant p53 inhibits wild-type p53 function (Chan et al., 2004; Dong et al., 2007; Milner and Medcalf, 1991; Willis et al., 2004; Xu et al., 2011). In vivo, the R172H mutation was previously shown to have DN effects on a hypomorphic wild-type p53. Importantly, the mutant could not fully inhibit the wild-type p53 when it was expressed at normal levels, again indicating that the mutant p53 needs to be present at much higher levels than the wild-type to exert an inhibitory effect on wild-type p53 function (Wang et al., 2011).

It has previously been reported that lung tumours from KrasG12D;p53R270H/+ mice displayed a higher tumour grade than those from KrasG12D;p53R172H/+ and KrasG12D;p53-/+ mice. KrasG12D;p53R270H/+ mice also exhibited a higher tumour burden/animal (Jackson 2005). These differences were not due to GOF effects of the R270H mutant as KrasG12D;p53-/-, KrasG12D;p53R172H/- and KrasG12D;p53R270H/- lung tumours were similar. Importantly, LOH occurred at similar frequencies in KrasG12D;p53R270H/+ and KrasG12D;p53R172H/+ lung tumours demonstrating that wild-type p53 still retained some function in the presence of either mutant as its loss was still selected for in all cases (Jackson, 2005). Instead, these data suggest that the R270H mutant displayed DN functions but that this effect was partial. Of note, in our model wild-type p53 was only restored for a short duration in advanced tumours, potentially explaining the lack of an obvious DN phenotype and indicating that the DN effect of the R270H mutant observed previously occurred earlier in tumorigenesis (Jackson et al., 2005). Indeed, we did not observe any decrease in tumour burden after either 1 or 6 days of p53 restoration in all cases (data not shown). Importantly, we utilised this duration of p53 restoration as we aimed to model the DN effects of the p53 mutants as well as the therapeutic impact of wild-type p53 restoration in these advanced p53 mutant lung tumours, using a similar approach to what was used previously in p53

117 null lung tumours (Feldser et al., 2010; Junttila et al., 2010). These studies demonstrated that wild-type p53 only becomes activated in advanced p53 null lung tumours following p53 restoration when there are sufficient levels of oncogenic stress and activating signals (p19ARF) present. However, it could be that the presence of a p53 mutation provides activating signals earlier on in tumorigenesis to explain the DN effect of the R270H mutant when it is present from tumour initiation (Jackson et al. 2005). Alternatively, it may be explained by transient activation of p53 earlier on in tumorigenesis or if p53 is functioning downstream of stress mediators other than oncogenic stress and p19ARF activation.

It is important to note that we observed some DN effects of the R172H and R270H mutants on the transcriptional function of wild-type p53 that need to be considered. Interesting, the R172H mutant appeared to exert stronger, ‘complete-failure’ DN effects than the R270H mutant, even though it is the R270H mutant cells that appear more resistant to a p53-mediated proliferative arrest in vitro. We established that R172H and R270H potentially bind multiple different proteins. Therefore, it is feasible that the p53 mutants bind the promoters of genes distinctly due to the mutations causing different structural effects in the mutant proteins. It may be that the R172H forms tetramers with wild-type p53 that prevent the complex binding to some promoter regions whereas the R270H-wild-type complexes may retain transcriptional activity at these sites. On the other hand, it is interesting that it is also the R172H that retains more ‘wild-type-like’ transcriptional functions and it could alternatively be that the wild-type genes expressed in the R172H create redundancy for certain genes or negative-feedback regulation on wild-type p53 function, thus limiting the transcription of some p53 targets when wild-type p53 function is restored in these cells. It is important to note that the majority of these ‘wild-type-like’ genes were found in ChIP- seq datasets (http://chip-atlas.org) and so they likely represent direct wild-type p53 target genes. This finding indicates that the p53 mutants, and particularly the R172H mutant, retain some aspect of wild-type p53 transcriptional activity.

The functional impact of these DN and ‘wild-type-like’ effects remains to be elucidated and they likely manifest at other levels not covered by our analysis. It will be important to investigate these further, particularly on whether they translate into any difference in phenotype in vivo. The ‘wild-type-like’ transcriptional effects are particularly intriguing. It may be that p53 mutants are able to utilise some activity of

118 wild-type p53 in tumorigenesis. For instance, the regulation of metabolism by wild- type p53 could potentially be exploited in tumour cells. Tumours cells that reside in hypoxic or nutrient-deprived regions, or are metastasising to other sites may be more dependent on catabolic pathways (such as FAO and autophagy which can be promoted by wild-type p53) for cell survival in low-nutrient or matrix-detached conditions, respectively. Therefore, there may be some benefit in certain contexts for p53 mutant tumour cells retaining some aspects of wild-type p53 activity.

As discussed it is likely that the distinct transcriptional effects observed by the R172H and R270H mutants are explained by the formation of heterotetramers with wild-type p53 that exhibit differential cooperativity and distinct promoter affinities (Schlereth et al., 2013). As well as providing insight into the mechanism underlying the GOF effects of the different p53 mutants, ChIP-seq analysis could increase our understanding of the DN effects of the mutants. Performing ChIP-seq experiments to analyse the DNA bound by p53 in p53Fx/ER (p53 off; control), p53Fx/ER (p53 On), p53R172H/ER (p53 On) and p53R270H/ER (p53 On) samples would provide information regarding which genes are bound by mutant p53-wild-type p53 complexes.

8.3. Clinical Implications

By exploiting the most altered pathway between the R172H and R270H mutant lung tumours, the MVA pathway, we were able to therapeutically target R270H mutant tumours but not R172H tumours using simvastatin. Therefore, we argue that in the future, when considering therapeutic strategies against these mutant tumours, the responses of tumours with of distinct p53 mutations are explored. Moreover, the expression signature that underlies this sensitivity to statins was also found in p53 contact mutant human lung adenocarcinomas, suggesting this R270H-specific susceptibility is clinically relevant. Statin-use has been associated with improved survival of lung cancer patients (Cardwell et al., 2015; Lin et al., 2016). Identification of molecular subsets of lung tumours that display enhanced sensitivity to statin treatment, such as the p53R273H-positive subset proposed in this study, will likely improve the anti-tumorigenic impact of statins.

The MVA pathway expression signature observed in humans suggests the statin- sensitivity phenotype may be applicable to other contact mutant proteins. However, it

119 will be important to establish whether this is true for lungs tumours expressing other p53 contact mutants as it is important to determine the percentage of p53 mutant- lung adenocarcinoma this susceptibility could be applicable to. Indeed, the most common p53 mutation is the R248 contact mutation (Soussi et al., 2006). However, it is not clear whether this p53 mutation would have the same functional effects as the R273H mutation. Even though both mutations affect a residue that directly contacts the DNA, these mutations still exert different effects on the structure of the DNA- binding face and thus, its potential DNA-binding ability (Joerger and Fersht, 2007). Importantly, our study has highlighted numerous transcriptional differences unique to the R172H lung tumour cells. Future identification of the pathways that the R172H mutant lung tumour cells depend on may uncover therapeutic susceptibilities unique to lung tumours of this p53 mutation type.

Our data has also highlighted the effectiveness of using 3D cultures as an in vitro system for the investigation of therapeutic responses. We found that the phenotypes we observed in 3D Matrigel culture translated better in vivo than 2D responses. Indeed, wild-type p53 induced apoptosis in 3D cultures and in tumours but not in 2D cultures and the MVA pathway phenotype (expression and sensitivity to statin) was more pronounced in 3D cultures than 2D cultures. The differences between 2D and 3D cultures may also explain why we and others found contact p53 mutant tumour cells are sensitive to MVA pathway inhibition by statins (Freed-Pastor et al., 2012; Turrell et al., 2017) but others have not (Parrales et al., 2016). A potential reason for the p53-mediated induction of an apoptotic response in 3D cultures but not in 2D cultures could be explained by differences in the stress signals in the cells cultured in these environments. If the stress signals (p53-activating signals) are higher in cells in 3D cultures it could support the induction of cell death in a proportion of the cells following restoration of wild-type p53. In addition, it has been shown that 3D cultures represent an intermediate transcriptional and metabolic phenotype between that of 2D cells and tumours (Smith et al., 2012). Metabolic phenotypes may differ between 2D and 3D cultures because culturing cells in 3D results in oxygen and nutrient gradients, with the cells at the centre of the spheroid having a more limited nutrient or oxygen supply. Specifically, the MVA pathway phenotype may differ in 3D and 2D cultures because of these gradients but also potentially because the interactions to cells form between themselves (2D and 3D) and the matrix (3D) differ. The MVA pathway is important for cellular membrane integrity as it generates cholesterol but

120 also is important for the activity of GTPases of the Ras superfamily, such as Rho and Ras. Signalling downstream of these GTPases is directly regulated by integrin-matrix interactions and so the dependency on these pathways is likely altered in 3D cultures (Legate et al., 2009).

Restoration of wild-type p53 function has previously been shown to therapeutically target mutant Kras, p53 null lung tumours (Feldser et al., 2010; Junttila et al., 2010). We have now demonstrated that wild-type p53 restoration similarly targets p53 mutant lung tumours, a cohort much more relevant to human lung adenocarcinoma (Cancer Genome Atlas Research, 2014). However, the efficacy of p53 restoration is limited to advanced lung tumours, which display high levels of oncogenic stress and p19ARF, thereby providing the activating signals required for wild-type p53 function (Feldser et al., 2010; Junttila et al., 2010). We have also shown that this is the case for at least the p53-mediated proliferative arrest response in p53 mutant lung tumours (Turrell et al., 2017). However, as the majority of lung cancer patient typically present at late-stages (Scheff and Schneider, 2013), p53-targeted therapies may still be of benefit against a large proportion of human lung cancers.

Future analysis of the different binding partners of the p53 mutants may uncover additional ways to target a group of lung tumours with a specific mutation. However, we have identified almost 500 genes that are similarly regulated by both p53 mutants but have distinct expression in both p53 null and wild-type cells. Therefore, it may be more useful from a therapeutic perspective to explore these differences further. Even though we did not identify a cancer-associated phenotype common to both mutants (enhanced proliferation or invasiveness), further analysis of these genes may uncover a pathway similarly exploited by both mutants in tumorigenesis and subsequently, common vulnerabilities. Ultimately, these insights may lead to the identification of other generic susceptibilities of p53 mutant lung tumours which, if clinically relevant in humans, would be applicable to a larger proportion of lung adenocarcinoma cases.

121 Bibliography

Adorno, M., Cordenonsi, M., Montagner, M., Dupont, S., Wong, C., Hann, B., Solari, A., Bobisse, S., Rondina, M.B., Guzzardo, V., et al. (2009). A Mutant-p53/Smad complex opposes p63 to empower TGFbeta-induced metastasis. Cell 137, 87-98.

Aksoy, O., Chicas, A., Zeng, T., Zhao, Z., McCurrach, M., Wang, X., and Lowe, S.W. (2012). The atypical E2F family member E2F7 couples the p53 and RB pathways during cellular senescence. Genes Dev 26, 1546-1557.

Amelio, I., Cutruzzola, F., Antonov, A., Agostini, M., and Melino, G. (2014). Serine and glycine metabolism in cancer. Trends Biochem Sci 39, 191-198.

Assaily, W., Rubinger, D.A., Wheaton, K., Lin, Y., Ma, W., Xuan, W., Brown-Endres, L., Tsuchihara, K., Mak, T.W., and Benchimol, S. (2011). ROS-mediated p53 induction of Lpin1 regulates fatty acid oxidation in response to nutritional stress. Mol Cell 44, 491-501.

Aurelio, O.N., Kong, X.T., Gupta, S., and Stanbridge, E.J. (2000). p53 mutants have selective dominant-negative effects on apoptosis but not growth arrest in human cancer cell lines. Mol Cell Biol 20, 770-778.

Baker, S.J., Preisinger, A.C., Jessup, J.M., Paraskeva, C., Markowitz, S., Willson, J.K., Hamilton, S., and Vogelstein, B. (1990). p53 gene mutations occur in combination with 17p allelic deletions as late events in colorectal tumorigenesis. Cancer Res 50, 7717-7722.

Barak, Y., Juven, T., Haffner, R., and Oren, M. (1993). mdm2 expression is induced by wild type p53 activity. EMBO J 12, 461-468.

Bartek, J., Bartkova, J., Vojtesek, B., Staskova, Z., Lukas, J., Rejthar, A., Kovarik, J., Midgley, C.A., Gannon, J.V., and Lane, D.P. (1991). Aberrant expression of the p53 oncoprotein is a common feature of a wide spectrum of human malignancies. Oncogene 6, 1699-1703.

Bensaad, K., Tsuruta, A., Selak, M.A., Vidal, M.N., Nakano, K., Bartrons, R., Gottlieb, E., and Vousden, K.H. (2006). TIGAR, a p53-inducible regulator of glycolysis and apoptosis. Cell 126, 107-120.

Berkers, C.R., Maddocks, O.D., Cheung, E.C., Mor, I., and Vousden, K.H. (2013). Metabolic regulation by p53 family members. Cell Metab 18, 617-633.

Bersani, C., Xu, L.D., Vilborg, A., Lui, W.O., and Wiman, K.G. (2014). Wig-1 regulates cell cycle arrest and cell death through the p53 targets FAS and 14-3- 3sigma. Oncogene 33, 4407-4417.

Bieging, K.T., and Attardi, L.D. (2012). Deconstructing p53 transcriptional networks in tumor suppression. Trends Cell Biol 22, 97-106.

122 Bieging, K.T., Mello, S.S., and Attardi, L.D. (2014). Unravelling mechanisms of p53- mediated tumour suppression. Nat Rev Cancer 14, 359-370.

Boeckler, F.M., Joerger, A.C., Jaggi, G., Rutherford, T.J., Veprintsev, D.B., and Fersht, A.R. (2008). Targeted rescue of a destabilized mutant of p53 by an in silico screened drug. Proc Natl Acad Sci U S A 105, 10360-10365.

Bourdon, A., Minai, L., Serre, V., Jais, J.P., Sarzi, E., Aubert, S., Chretien, D., de Lonlay, P., Paquis-Flucklinger, V., Arakawa, H., et al. (2007). Mutation of RRM2B, encoding p53-controlled ribonucleotide reductase (p53R2), causes severe mitochondrial DNA depletion. Nat Genet 39, 776-780.

Brosh, R., and Rotter, V. (2009). When mutants gain new powers: news from the mutant p53 field. Nat Rev Cancer 9, 701-713.

Brugarolas, J., Chandrasekaran, C., Gordon, J.I., Beach, D., Jacks, T., and Hannon, G.J. (1995). Radiation-induced cell cycle arrest compromised by p21 deficiency. Nature 377, 552-557.

Budanov, A.V., and Karin, M. (2008). p53 target genes sestrin1 and sestrin2 connect genotoxic stress and mTOR signaling. Cell 134, 451-460.

Bullock, A.N., and Fersht, A.R. (2001). Rescuing the function of mutant p53. Nat Rev Cancer 1, 68-76.

Bullock, A.N., Fersht, A. R. (2001). Rescuing the function of mutant p53. Nature Reviews Cancer 1.

Bykov, V.J., Issaeva, N., Shilov, A., Hultcrantz, M., Pugacheva, E., Chumakov, P., Bergman, J., Wiman, K.G., and Selivanova, G. (2002). Restoration of the tumor suppressor function to mutant p53 by a low-molecular-weight compound. Nat Med 8, 282-288.

Cairns, J.M., Dunning, M.J., Ritchie, M.E., Russell, R., and Lynch, A.G. (2008). BASH: a tool for managing BeadArray spatial artefacts. Bioinformatics 24, 2921- 2922.

Cancer Genome Atlas Research, N. (2014). Comprehensive molecular profiling of lung adenocarcinoma. Nature 511, 543-550.

Cardwell, C.R., Mc Menamin, U., Hughes, C.M., and Murray, L.J. (2015). Statin use and survival from lung cancer: a population-based cohort study. Cancer Epidemiol Biomarkers Prev 24, 833-841.

Carvajal, L.A., Hamard, P.J., Tonnessen, C., and Manfredi, J.J. (2012). E2F7, a novel target, is up-regulated by p53 and mediates DNA damage-dependent transcriptional repression. Genes Dev 26, 1533-1545.

Chan, B.A., and Hughes, B.G. (2015). Targeted therapy for non-small cell lung cancer: current standards and the promise of the future. Transl Lung Cancer Res 4, 36-54.

123 Chan, W.M., Siu, W.Y., Lau, A., and Poon, R.Y. (2004). How many mutant p53 molecules are needed to inactivate a tetramer? Mol Cell Biol 24, 3536-3551.

Chene, P. (1998). In vitro analysis of the dominant negative effect of p53 mutants. J Mol Biol 281, 205-209.

Cheok, C.F., Verma, C.S., Baselga, J., and Lane, D.P. (2011). Translating p53 into the clinic. Nat Rev Clin Oncol 8, 25-37.

Cheung, E.C., Ludwig, R.L., and Vousden, K.H. (2012). Mitochondrial localization of TIGAR under hypoxia stimulates HK2 and lowers ROS and cell death. Proc Natl Acad Sci U S A 109, 20491-20496.

Cho, Y., Gorina, S., Jeffrey, P.D., and Pavletich, N.P. (1994). Crystal structure of a p53 tumor suppressor-DNA complex: understanding tumorigenic mutations. Science 265, 346-355.

Christophorou, M.A., Martin-Zanca, D., Soucek, L., Lawlor, E.R., Brown-Swigart, L., Verschuren, E.W., and Evan, G.I. (2005). Temporal dissection of p53 function in vitro and in vivo. Nat Genet 37, 718-726.

Cole, A.M., Ridgway, R.A., Derkits, S.E., Parry, L., Barker, N., Clevers, H., Clarke, A.R., and Sansom, O.J. (2010). p21 loss blocks senescence following Apc loss and provokes tumourigenesis in the renal but not the intestinal epithelium. EMBO Mol Med 2, 472-486.

Collavin, L., Monte, M., Verardo, R., Pfleger, C., and Schneider, C. (2000). Cell-cycle regulation of the p53-inducible gene B99. FEBS Lett 481, 57-62.

Contractor, T., and Harris, C.R. (2012). p53 negatively regulates transcription of the pyruvate dehydrogenase kinase Pdk2. Cancer Res 72, 560-567.

Crewe, H.K., Ellis, S.W., Lennard, M.S., and Tucker, G.T. (1997). Variable contribution of cytochromes P450 2D6, 2C9 and 3A4 to the 4-hydroxylation of tamoxifen by human liver microsomes. Biochem Pharmacol 53, 171-178. de Stanchina, E., Querido, E., Narita, M., Davuluri, R.V., Pandolfi, P.P., Ferbeyre, G., and Lowe, S.W. (2004). PML is a direct p53 target that modulates p53 effector functions. Mol Cell 13, 523-535.

Demidenko, Z.N., and Blagosklonny, M.V. (2008). Growth stimulation leads to cellular senescence when the cell cycle is blocked. Cell Cycle 7, 3355-3361.

Demidenko, Z.N., Zubova, S.G., Bukreeva, E.I., Pospelov, V.A., Pospelova, T.V., and Blagosklonny, M.V. (2009). Rapamycin decelerates cellular senescence. Cell Cycle 8, 1888-1895.

Demma, M., Maxwell, E., Ramos, R., Liang, L., Li, C., Hesk, D., Rossman, R., Mallams, A., Doll, R., Liu, M., et al. (2010). SCH529074, a small molecule activator of mutant p53, which binds p53 DNA binding domain (DBD), restores growth- suppressive function to mutant p53 and interrupts HDM2-mediated ubiquitination of wild type p53. J Biol Chem 285, 10198-10212.

124 Deng, C., Zhang, P., Harper, J.W., Elledge, S.J., and Leder, P. (1995). Mice lacking p21CIP1/WAF1 undergo normal development, but are defective in G1 checkpoint control. Cell 82, 675-684.

Donehower, L.A., Harvey, M., Slagle, B.L., McArthur, M.J., Montgomery, C.A., Jr., Butel, J.S., and Bradley, A. (1992). Mice deficient for p53 are developmentally normal but susceptible to spontaneous tumours. Nature 356, 215-221.

Dong, P., Tada, M., Hamada, J., Nakamura, A., Moriuchi, T., and Sakuragi, N. (2007). p53 dominant-negative mutant R273H promotes invasion and migration of human endometrial cancer HHUA cells. Clin Exp Metastasis 24, 471-483.

Dunning, M.J., Smith, M.L., Ritchie, M.E., and Tavare, S. (2007). beadarray: R classes and methods for Illumina bead-based data. Bioinformatics 23, 2183-2184. el-Deiry, W.S., Tokino, T., Velculescu, V.E., Levy, D.B., Parsons, R., Trent, J.M., Lin,

D., Mercer, W.E., Kinzler, K.W., and Vogelstein, B. (1993). WAF1, a potential mediator of p53 tumor suppression. Cell 75, 817-825.

Feldser, D.M., Kostova, K.K., Winslow, M.M., Taylor, S.E., Cashman, C., Whittaker, C.A., Sanchez-Rivera, F.J., Resnick, R., Bronson, R., Hemann, M.T., et al. (2010). Stage-specific sensitivity to p53 restoration during lung cancer progression. Nature 468, 572-575.

Feng, Z., Hu, W., de Stanchina, E., Teresky, A.K., Jin, S., Lowe, S., and Levine, A.J. (2007). The regulation of AMPK beta1, TSC2, and PTEN expression by p53: stress, cell and tissue specificity, and the role of these gene products in modulating the IGF- 1-AKT-mTOR pathways. Cancer Res 67, 3043-3053.

Ferbeyre, G., de Stanchina, E., Querido, E., Baptiste, N., Prives, C., and Lowe, S.W. (2000). PML is induced by oncogenic ras and promotes premature senescence. Genes Dev 14, 2015-2027.

Fischer, M. (2017). Census and evaluation of p53 target genes. Oncogene 36, 3943- 3956.

Freed-Pastor, W.A., Mizuno, H., Zhao, X., Langerod, A., Moon, S.H., Rodriguez- Barrueco, R., Barsotti, A., Chicas, A., Li, W., Polotskaia, A., et al. (2012). Mutant p53 disrupts mammary tissue architecture via the mevalonate pathway. Cell 148, 244- 258.

Frese, K.K., and Tuveson, D.A. (2007). Maximizing mouse cancer models. Nat Rev Cancer 7, 645-658.

Frezza, C., and Martins, C.P. (2012). From tumor prevention to therapy: empowering p53 to fight back. Drug Resist Updat 15, 258-267.

Gaude, E., Chignola, F., Spiliotopoulos, D., Spitaleri, A., Ghitti, M., Garcia-Manteiga, J.M., Mari, S., Musco, G. (2013). muma, An R package for metabolomics univariate and multivariate statistical analysis. Curr Metabolomics 1, 180-189.

125 Gentleman, R.C., Carey, V.J., Bates, D.M., Bolstad, B., Dettling, M., Dudoit, S., Ellis, B., Gautier, L., Ge, Y., Gentry, J., et al. (2004). Bioconductor: open software development for computational biology and bioinformatics. Genome Biol 5, R80.

Goldstein, I., Ezra, O., Rivlin, N., Molchadsky, A., Madar, S., Goldfinger, N., and Rotter, V. (2012). p53, a novel regulator of lipid metabolism pathways. J Hepatol 56, 656-662.

Gysin, S., Salt, M., Young, A., and McCormick, F. (2011). Therapeutic strategies for targeting ras proteins. Genes Cancer 2, 359-372.

Hanel, W., Marchenko, N., Xu, S., Yu, S.X., Weng, W., and Moll, U. (2013). Two hot spot mutant p53 mouse models display differential gain of function in tumorigenesis. Cell Death Differ 20, 898-909.

Harper, J.W., Adami, G.R., Wei, N., Keyomarsi, K., and Elledge, S.J. (1993). The p21 Cdk-interacting protein Cip1 is a potent inhibitor of G1 cyclin-dependent kinases. Cell 75, 805-816.

Hay, N., and Sonenberg, N. (2004). Upstream and downstream of mTOR. Genes Dev 18, 1926-1945.

Hirsch, F.R., Scagliotti, G.V., Mulshine, J.L., Kwon, R., Curran, W.J., Jr., Wu, Y.L., and Paz-Ares, L. (2017). Lung cancer: current therapies and new targeted treatments. Lancet 389, 299-311.

Ho, J., and Benchimol, S. (2003). Transcriptional repression mediated by the p53 tumour suppressor. Cell Death Differ 10, 404-408.

Hollander, M.C., Sheikh, M.S., Bulavin, D.V., Lundgren, K., Augeri-Henmueller, L., Shehee, R., Molinaro, T.A., Kim, K.E., Tolosa, E., Ashwell, J.D., et al. (1999). Genomic instability in Gadd45a-deficient mice. Nat Genet 23, 176-184.

Horn, H.F., and Vousden, K.H. (2007). Coping with stress: multiple ways to activate p53. Oncogene 26, 1306-1316.

Horton, J.D., Shah, N.A., Warrington, J.A., Anderson, N.N., Park, S.W., Brown, M.S., and Goldstein, J.L. (2003). Combined analysis of oligonucleotide microarray data from transgenic and knockout mice identifies direct SREBP target genes. Proc Natl Acad Sci U S A 100, 12027-12032.

Hu, W., Zhang, C., Wu, R., Sun, Y., Levine, A., and Feng, Z. (2010). Glutaminase 2, a novel p53 target gene regulating energy metabolism and antioxidant function. Proc Natl Acad Sci U S A 107, 7455-7460.

Iggo, R., Gatter, K., Bartek, J., Lane, D., and Harris, A.L. (1990). Increased expression of mutant forms of p53 oncogene in primary lung cancer. Lancet 335, 675-679.

Istvan, E.S., and Deisenhofer, J. (2001). Structural mechanism for statin inhibition of HMG-CoA reductase. Science 292, 1160-1164.

126 Jacks, T., Remington, L., Williams, B.O., Schmitt, E.M., Halachmi, S., Bronson, R.T., and Weinberg, R.A. (1994). Tumor spectrum analysis in p53-mutant mice. Curr Biol 4, 1-7.

Jackson, E.L., Olive, K.P., Tuveson, D.A., Bronson, R., Crowley, D., Brown, M., and Jacks, T. (2005). The differential effects of mutant p53 alleles on advanced murine lung cancer. Cancer Res 65, 10280-10288.

Jackson, E.L., Willis, N., Mercer, K., Bronson, R.T., Crowley, D., Montoya, R., Jacks, T., and Tuveson, D.A. (2001). Analysis of lung tumor initiation and progression using conditional expression of oncogenic K-ras. Genes Dev 15, 3243-3248.

Jackson, R.J., Engelman, R.W., Coppola, D., Cantor, A.B., Wharton, W., and Pledger, W.J. (2003). p21Cip1 nullizygosity increases tumor metastasis in irradiated mice. Cancer Res 63, 3021-3025.

Jeffers, J.R., Parganas, E., Lee, Y., Yang, C., Wang, J., Brennan, J., MacLean, K.H., Han, J., Chittenden, T., Ihle, J.N., et al. (2003). Puma is an essential mediator of p53- dependent and -independent apoptotic pathways. Cancer Cell 4, 321-328.

Jerga, A., Lu, Y.J., Schujman, G.E., de Mendoza, D., and Rock, C.O. (2007). Identification of a soluble diacylglycerol kinase required for lipoteichoic acid production in Bacillus subtilis. J Biol Chem 282, 21738-21745.

Jiang, P., Du, W., Wang, X., Mancuso, A., Gao, X., Wu, M., and Yang, X. (2011). p53 regulates biosynthesis through direct inactivation of glucose-6-phosphate dehydrogenase. Nat Cell Biol 13, 310-316.

Joerger, A.C., Ang, H.C., and Fersht, A.R. (2006). Structural basis for understanding oncogenic p53 mutations and designing rescue drugs. Proc Natl Acad Sci U S A 103, 15056-15061.

Joerger, A.C., and Fersht, A.R. (2007). Structure-function-rescue: the diverse nature of common p53 cancer mutants. Oncogene 26, 2226-2242.

Joerger, A.C., and Fersht, A.R. (2008). Structural biology of the tumor suppressor p53. Annu Rev Biochem 77, 557-582.

Jones, J.M., Cui, X.S., Medina, D., and Donehower, L.A. (1999). Heterozygosity of p21WAF1/CIP1 enhances tumor cell proliferation and cyclin D1-associated kinase activity in a murine mammary cancer model. Cell Growth Differ 10, 213-222.

Jonkers, J., Meuwissen, R., van der Gulden, H., Peterse, H., van der Valk, M., and Berns, A. (2001). Synergistic tumor suppressor activity of BRCA2 and p53 in a conditional mouse model for breast cancer. Nat Genet 29, 418-425.

Junttila, M.R., Karnezis, A.N., Garcia, D., Madriles, F., Kortlever, R.M., Rostker, F., Brown Swigart, L., Pham, D.M., Seo, Y., Evan, G.I., et al. (2010). Selective activation of p53-mediated tumour suppression in high-grade tumours. Nature 468, 567-571.

127 Kamisuki, S., Mao, Q., Abu-Elheiga, L., Gu, Z., Kugimiya, A., Kwon, Y., Shinohara, T., Kawazoe, Y., Sato, S., Asakura, K., et al. (2009). A small molecule that blocks fat synthesis by inhibiting the activation of SREBP. Chem Biol 16, 882-892.

Kannan, K., Amariglio, N., Rechavi, G., and Givol, D. (2000). Profile of gene expression regulated by induced p53: connection to the TGF-beta family. FEBS Lett 470, 77-82.

Kastan, M.B., Onyekwere, O., Sidransky, D., Vogelstein, B., and Craig, R.W. (1991). Participation of p53 protein in the cellular response to DNA damage. Cancer Res 51, 6304-6311.

Kastan, M.B., Zhan, Q., el-Deiry, W.S., Carrier, F., Jacks, T., Walsh, W.V., Plunkett, B.S., Vogelstein, B., and Fornace, A.J., Jr. (1992). A mammalian cell cycle checkpoint pathway utilizing p53 and GADD45 is defective in ataxia-telangiectasia. Cell 71, 587-597.

Kawase, T., Ohki, R., Shibata, T., Tsutsumi, S., Kamimura, N., Inazawa, J., Ohta, T., Ichikawa, H., Aburatani, H., Tashiro, F., et al. (2009). PH domain-only protein PHLDA3 is a p53-regulated repressor of Akt. Cell 136, 535-550.

Kerr, E.M., Gaude, E., Turrell, F.K., Frezza, C., and Martins, C.P. (2016). Mutant Kras copy number defines metabolic reprogramming and therapeutic susceptibilities. Nature 531, 110-113.

Kim, J., Kundu, M., Viollet, B., and Guan, K.L. (2011). AMPK and mTOR regulate autophagy through direct phosphorylation of Ulk1. Nat Cell Biol 13, 132-141.

Kitayner, M., Rozenberg, H., Kessler, N., Rabinovich, D., Shaulov, L., Haran, T.E., and Shakked, Z. (2006). Structural basis of DNA recognition by p53 tetramers. Mol Cell 22, 741-753.

Kollareddy, M., Dimitrova, E., Vallabhaneni, K.C., Chan, A., Le, T., Chauhan, K.M., Carrero, Z.I., Ramakrishnan, G., Watabe, K., Haupt, Y., et al. (2015). Regulation of nucleotide metabolism by mutant p53 contributes to its gain-of-function activities. Nat Commun 6, 7389.

Kondoh, H., Lleonart, M.E., Gil, J., Wang, J., Degan, P., Peters, G., Martinez, D., Carnero, A., and Beach, D. (2005). Glycolytic enzymes can modulate cellular life span. Cancer Res 65, 177-185.

Konstantinopoulos, P.A., Karamouzis, M.V., and Papavassiliou, A.G. (2007). Post- translational modifications and regulation of the RAS superfamily of GTPases as anticancer targets. Nat Rev Drug Discov 6, 541-555.

Kubbutat, M.H., Jones, S.N., and Vousden, K.H. (1997). Regulation of p53 stability by Mdm2. Nature 387, 299-303.

Lane, D.P. (1992). Cancer. p53, guardian of the genome. Nature 358, 15-16.

128 Lang, G.A., Iwakuma, T., Suh, Y.A., Liu, G., Rao, V.A., Parant, J.M., Valentin-Vega, Y.A., Terzian, T., Caldwell, L.C., Strong, L.C., et al. (2004). Gain of function of a p53 hot spot mutation in a mouse model of Li-Fraumeni syndrome. Cell 119, 861-872.

Legate, K.R., Wickstrom, S.A., and Fassler, R. (2009). Genetic and cell biological analysis of integrin outside-in signaling. Genes Dev 23, 397-418.

Lehar, S.M., Nacht, M., Jacks, T., Vater, C.A., Chittenden, T., and Guild, B.C. (1996). Identification and cloning of EI24, a gene induced by p53 in etoposide-treated cells. Oncogene 12, 1181-1187.

Lehmann, B.D., and Pietenpol, J.A. (2012). Targeting mutant p53 in human tumors. J Clin Oncol 30, 3648-3650.

Li, F.P., and Fraumeni, J.F., Jr. (1969a). Rhabdomyosarcoma in children: epidemiologic study and identification of a familial cancer syndrome. J Natl Cancer Inst 43, 1365-1373.

Li, F.P., and Fraumeni, J.F., Jr. (1969b). Soft-tissue sarcomas, breast cancer, and other neoplasms. A familial syndrome? Ann Intern Med 71, 747-752.

Li, T., Kon, N., Jiang, L., Tan, M., Ludwig, T., Zhao, Y., Baer, R., and Gu, W. (2012). Tumor suppression in the absence of p53-mediated cell-cycle arrest, apoptosis, and senescence. Cell 149, 1269-1283.

Liang, Y., Liu, J., Feng, Z. (2013). The regulation of cellular metabolism by tumour suppressor p53 Cell and Bioscience 3, 1-10.

Lin, J.J., Ezer, N., Sigel, K., Mhango, G., and Wisnivesky, J.P. (2016). The effect of statins on survival in patients with stage IV lung cancer. Lung Cancer 99, 137-142.

Liu, Y., Chen, C., Xu, Z., Scuoppo, C., Rillahan, C.D., Gao, J., Spitzer, B., Bosbach, B., Kastenhuber, E.R., Baslan, T., et al. (2016). Deletions linked to TP53 loss drive cancer through p53-independent mechanisms. Nature 531, 471-475.

Liu, Y., He, Y., Jin, A., Tikunov, A.P., Zhou, L., Tollini, L.A., Leslie, P., Kim, T.H., Li, L.O., Coleman, R.A., et al. (2014). Ribosomal protein-Mdm2-p53 pathway coordinates nutrient stress with lipid metabolism by regulating MCD and promoting fatty acid oxidation. Proc Natl Acad Sci U S A 111, E2414-2422.

Lo, P.K., Chen, J.Y., Tang, P.P., Lin, J., Lin, C.H., Su, L.T., Wu, C.H., Chen, T.L., Yang, Y., and Wang, F.F. (2001). Identification of a mouse thiamine transporter gene as a direct transcriptional target for p53. J Biol Chem 276, 37186-37193.

Long, Y.C., and Zierath, J.R. (2006). AMP-activated protein kinase signaling in metabolic regulation. J Clin Invest 116, 1776-1783.

Ma, W., Sung, H.J., Park, J.Y., Matoba, S., and Hwang, P.M. (2007). A pivotal role for p53: balancing aerobic respiration and glycolysis. J Bioenerg Biomembr 39, 243- 246.

129 Maddocks, O.D., Berkers, C.R., Mason, S.M., Zheng, L., Blyth, K., Gottlieb, E., and Vousden, K.H. (2013). Serine starvation induces stress and p53-dependent metabolic remodelling in cancer cells. Nature 493, 542-546.

Malkin, D., Li, F.P., Strong, L.C., Fraumeni, J.F., Jr., Nelson, C.E., Kim, D.H., Kassel, J., Gryka, M.A., Bischoff, F.Z., Tainsky, M.A., et al. (1990). Germ line p53 mutations in a familial syndrome of breast cancer, sarcomas, and other neoplasms. Science 250, 1233-1238.

Mao, L. (2001). Molecular abnormalities in lung carcinogenesis and their potential clinical implications. Lung Cancer 34 Suppl 2, S27-34.

Martins, C.P., Brown-Swigart, L., and Evan, G.I. (2006). Modeling the therapeutic efficacy of p53 restoration in tumors. Cell 127, 1323-1334.

Masciarelli, S., Fontemaggi, G., Di Agostino, S., Donzelli, S., Carcarino, E., Strano, S., and Blandino, G. (2014). Gain-of-function mutant p53 downregulates miR-223 contributing to chemoresistance of cultured tumor cells. Oncogene 33, 1601-1608.

Matoba, S., Kang, J.G., Patino, W.D., Wragg, A., Boehm, M., Gavrilova, O., Hurley, P.J., Bunz, F., and Hwang, P.M. (2006). p53 regulates mitochondrial respiration. Science 312, 1650-1653.

McLure, K.G., and Lee, P.W. (1998). How p53 binds DNA as a tetramer. EMBO J 17, 3342-3350.

Milner, J., and Medcalf, E.A. (1991). Cotranslation of activated mutant p53 with wild type drives the wild-type p53 protein into the mutant conformation. Cell 65, 765-774.

Mitsudomi, T., Hamajima, N., Ogawa, M., and Takahashi, T. (2000). Prognostic significance of p53 alterations in patients with non-small cell lung cancer: a meta- analysis. Clin Cancer Res 6, 4055-4063.

Mullen, P.J., Yu, R., Longo, J., Archer, M.C., and Penn, L.Z. (2016). The interplay between cell signalling and the mevalonate pathway in cancer. Nat Rev Cancer 16, 718-731.

Muller, M., Wilder, S., Bannasch, D., Israeli, D., Lehlbach, K., Li-Weber, M., Friedman, S.L., Galle, P.R., Stremmel, W., Oren, M., et al. (1998). p53 activates the CD95 (APO-1/Fas) gene in response to DNA damage by anticancer drugs. J Exp Med 188, 2033-2045.

Muller, P.A., Caswell, P.T., Doyle, B., Iwanicki, M.P., Tan, E.H., Karim, S., Lukashchuk, N., Gillespie, D.A., Ludwig, R.L., Gosselin, P., et al. (2009). Mutant p53 drives invasion by promoting integrin recycling. Cell 139, 1327-1341.

Muller, P.A., Trinidad, A.G., Timpson, P., Morton, J.P., Zanivan, S., van den Berghe, P.V., Nixon, C., Karim, S.A., Caswell, P.T., Noll, J.E., et al. (2013). Mutant p53 enhances MET trafficking and signalling to drive cell scattering and invasion. Oncogene 32, 1252-1265.

130 Muller, P.A., and Vousden, K.H. (2014). Mutant p53 in cancer: new functions and therapeutic opportunities. Cancer Cell 25, 304-317.

Munoz-Espin, D., and Serrano, M. (2014). Cellular senescence: from physiology to pathology. Nat Rev Mol Cell Biol 15, 482-496.

Nithipatikom, K., Endsley, M.P., Pfeiffer, A.W., Falck, J.R., and Campbell, W.B. (2014). A novel activity of microsomal epoxide hydrolase: metabolism of the endocannabinoid 2-arachidonoylglycerol. J Lipid Res 55, 2093-2102.

Okamoto, K., and Beach, D. (1994). Cyclin G is a transcriptional target of the p53 tumor suppressor protein. EMBO J 13, 4816-4822.

Olive, K.P., Tuveson, D.A., Ruhe, Z.C., Yin, B., Willis, N.A., Bronson, R.T., Crowley, D., and Jacks, T. (2004). Mutant p53 gain of function in two mouse models of Li- Fraumeni syndrome. Cell 119, 847-860.

Pai, J.T., Guryev, O., Brown, M.S., and Goldstein, J.L. (1998). Differential stimulation of cholesterol and unsaturated fatty acid biosynthesis in cells expressing individual nuclear sterol regulatory element-binding proteins. J Biol Chem 273, 26138-26148.

Pampaloni, F., Reynaud, E.G., and Stelzer, E.H. (2007). The third dimension bridges the gap between cell culture and live tissue. Nat Rev Mol Cell Biol 8, 839-845.

Parkin, D.M., Bray, F., Ferlay, J., and Pisani, P. (2005). Global cancer statistics, 2002. CA Cancer J Clin 55, 74-108.

Parrales, A., Ranjan, A., Iyer, S.V., Padhye, S., Weir, S.J., Roy, A., and Iwakuma, T. (2016). DNAJA1 controls the fate of misfolded mutant p53 through the mevalonate pathway. Nat Cell Biol 18, 1233-1243.

Pearson, M., Carbone, R., Sebastiani, C., Cioce, M., Fagioli, M., Saito, S., Higashimoto, Y., Appella, E., Minucci, S., Pandolfi, P.P., et al. (2000). PML regulates p53 acetylation and premature senescence induced by oncogenic Ras. Nature 406, 207-210.

Quintens, R., Verreet, T., Janssen, A., Neefs, M., Leysen, L., Michaux, A., Verslegers, M., Samari, N., Pani, G., Verheyde, J., et al. (2015). Identification of novel radiation-induced p53-dependent transcripts extensively regulated during mouse brain development. Biol Open 4, 331-344.

Santos, C.R., and Schulze, A. (2012). Lipid metabolism in cancer. FEBS J 279, 2610-2623.

Sauer, L., Gitenay, D., Vo, C., and Baron, V.T. (2010). Mutant p53 initiates a feedback loop that involves Egr-1/EGF receptor/ERK in prostate cancer cells. Oncogene 29, 2628-2637.

Scheff, R.J., and Schneider, B.J. (2013). Non-small-cell lung cancer: treatment of late stage disease: chemotherapeutics and new frontiers. Semin Intervent Radiol 30, 191- 198.

131 Schlereth, K., Heyl, C., Krampitz, A.M., Mernberger, M., Finkernagel, F., Scharfe, M., Jarek, M., Leich, E., Rosenwald, A., and Stiewe, T. (2013). Characterization of the p53 cistrome--DNA binding cooperativity dissects p53's tumor suppressor functions. PLoS Genet 9, e1003726.

Schwartzenberg-Bar-Yoseph, F., Armoni, M., and Karnieli, E. (2004). The tumor suppressor p53 down-regulates glucose transporters GLUT1 and GLUT4 gene expression. Cancer Res 64, 2627-2633.

Seo, Y.R., Fishel, M.L., Amundson, S., Kelley, M.R., and Smith, M.L. (2002). Implication of p53 in base excision DNA repair: in vivo evidence. Oncogene 21, 731- 737.

Sinthupibulyakit, C., Ittarat, W., St Clair, W.H., and St Clair, D.K. (2010). p53 Protects lung cancer cells against metabolic stress. Int J Oncol 37, 1575-1581.

Smith, S.J., Wilson, M., Ward, J.H., Rahman, C.V., Peet, A.C., Macarthur, D.C., Rose, F.R., Grundy, R.G., and Rahman, R. (2012). Recapitulation of tumor heterogeneity and molecular signatures in a 3D brain cancer model with decreased sensitivity to histone deacetylase inhibition. PLoS One 7, e52335.

Soragni, A., Janzen, D.M., Johnson, L.M., Lindgren, A.G., Thai-Quynh Nguyen, A., Tiourin, E., Soriaga, A.B., Lu, J., Jiang, L., Faull, K.F., et al. (2016). A Designed Inhibitor of p53 Aggregation Rescues p53 Tumor Suppression in Ovarian Carcinomas. Cancer Cell 29, 90-103.

Sorrentino, G., Ruggeri, N., Specchia, V., Cordenonsi, M., Mano, M., Dupont, S., Manfrin, A., Ingallina, E., Sommaggio, R., Piazza, S., et al. (2014). Metabolic control of YAP and TAZ by the mevalonate pathway. Nat Cell Biol 16, 357-366.

Soussi, T., Asselain, B., Hamroun, D., Kato, S., Ishioka, C., Claustres, M., and Beroud, C. (2006). Meta-analysis of the p53 mutation database for mutant p53 biological activity reveals a methodologic bias in mutation detection. Clin Cancer Res 12, 62-69.

Soussi, T., and Beroud, C. (2001). Assessing TP53 status in human tumours to evaluate clinical outcome. Nat Rev Cancer 1, 233-240.

Srivastava, S., Zou, Z.Q., Pirollo, K., Blattner, W., and Chang, E.H. (1990). Germ-line transmission of a mutated p53 gene in a cancer-prone family with Li-Fraumeni syndrome. Nature 348, 747-749.

Stambolsky, P., Weisz, L., Shats, I., Klein, Y., Goldfinger, N., Oren, M., and Rotter, V. (2006). Regulation of AIF expression by p53. Cell Death Differ 13, 2140-2149.

Symonds, H., Krall, L., Remington, L., Saenz-Robles, M., Lowe, S., Jacks, T., and Van Dyke, T. (1994). p53-dependent apoptosis suppresses tumor growth and progression in vivo. Cell 78, 703-711.

Tang, W., Willers, H., and Powell, S.N. (1999). p53 directly enhances rejoining of DNA double-strand breaks with cohesive ends in gamma-irradiated mouse fibroblasts. Cancer Res 59, 2562-2565.

132

Tanikawa, C., Zhang, Y.Z., Yamamoto, R., Tsuda, Y., Tanaka, M., Funauchi, Y., Mori, J., Imoto, S., Yamaguchi, R., Nakamura, Y., et al. (2017). The Transcriptional Landscape of p53 Signalling Pathway. EBioMedicine 20, 109-119.

Tao, W., and Levine, A.J. (1999). P19(ARF) stabilizes p53 by blocking nucleo- cytoplasmic shuttling of Mdm2. Proc Natl Acad Sci U S A 96, 6937-6941.

Tron, V.A., Trotter, M.J., Ishikawa, T., Ho, V.C., and Li, G. (1998). p53-dependent regulation of nucleotide excision repair in murine epidermis in vivo. J Cutan Med Surg 3, 16-20.

Turrell, F.K., Kerr, E.M., Gao, M., Thorpe, H., Doherty, G.J., Cridge, J., Shorthouse, D., Speed, A., Samarajiwa, S., Hall, B.A., et al. (2017). Lung tumors with distinct p53 mutations respond similarly to p53 targeted therapy but exhibit genotype-specific statin sensitivity. Genes Dev.

Tusher, V.G., Tibshirani, R., and Chu, G. (2001). Significance analysis of microarrays applied to the ionizing radiation response. Proc Natl Acad Sci U S A 98, 5116-5121.

Utrera, R., Collavin, L., Lazarevic, D., Delia, D., and Schneider, C. (1998). A novel p53-inducible gene coding for a microtubule-localized protein with G2-phase-specific expression. EMBO J 17, 5015-5025.

Vahsen, N., Cande, C., Briere, J.J., Benit, P., Joza, N., Larochette, N., Mastroberardino, P.G., Pequignot, M.O., Casares, N., Lazar, V., et al. (2004). AIF deficiency compromises oxidative phosphorylation. EMBO J 23, 4679-4689.

Villunger, A., Michalak, E.M., Coultas, L., Mullauer, F., Bock, G., Ausserlechner, M.J., Adams, J.M., and Strasser, A. (2003). p53- and drug-induced apoptotic responses mediated by BH3-only proteins puma and noxa. Science 302, 1036-1038.

Wang, Y., Suh, Y.A., Fuller, M.Y., Jackson, J.G., Xiong, S., Terzian, T., Quintas- Cardama, A., Bankson, J.A., El-Naggar, A.K., and Lozano, G. (2011). Restoring expression of wild-type p53 suppresses tumor growth but does not cause tumor regression in mice with a p53 missense mutation. J Clin Invest 121, 893-904.

Ward, P.S., and Thompson, C.B. (2012). Metabolic reprogramming: a cancer hallmark even warburg did not anticipate. Cancer Cell 21, 297-308.

Watanabe, S., Hino, A., Wada, K., Eliason, J.F., and Uchida, T. (1995). Purification, cloning, and expression of murine uridine phosphorylase. J Biol Chem 270, 12191- 12196.

Weissmueller, S., Manchado, E., Saborowski, M., Morris, J.P.t., Wagenblast, E., Davis, C.A., Moon, S.H., Pfister, N.T., Tschaharganeh, D.F., Kitzing, T., et al. (2014). Mutant p53 drives pancreatic cancer metastasis through cell-autonomous PDGF receptor beta signaling. Cell 157, 382-394.

Weisz, L., Oren, M., and Rotter, V. (2007). Transcription regulation by mutant p53. Oncogene 26, 2202-2211.

133 Williams, A.B., and Schumacher, B. (2016). p53 in the DNA-Damage-Repair Process. Cold Spring Harb Perspect Med 6.

Willis, A., Jung, E.J., Wakefield, T., and Chen, X. (2004). Mutant p53 exerts a dominant negative effect by preventing wild-type p53 from binding to the promoter of its target genes. Oncogene 23, 2330-2338.

Wu, G.S., Burns, T.F., McDonald, E.R., 3rd, Jiang, W., Meng, R., Krantz, I.D., Kao, G., Gan, D.D., Zhou, J.Y., Muschel, R., et al. (1997). KILLER/DR5 is a DNA damage- inducible p53-regulated death receptor gene. Nat Genet 17, 141-143.

Wu, X., Bayle, J.H., Olson, D., and Levine, A.J. (1993). The p53-mdm-2 autoregulatory feedback loop. Genes Dev 7, 1126-1132.

Xu, J., Reumers, J., Couceiro, J.R., De Smet, F., Gallardo, R., Rudyak, S., Cornelis, A., Rozenski, J., Zwolinska, A., Marine, J.C., et al. (2011). Gain of function of mutant p53 by coaggregation with multiple tumor suppressors. Nat Chem Biol 7, 285-295.

Xu, Y. (2003). Regulation of p53 responses by post-translational modifications. Cell Death Differ 10, 400-403.

Xue, W., Zender, L., Miething, C., Dickins, R.A., Hernando, E., Krizhanovsky, V., Cordon-Cardo, C., and Lowe, S.W. (2007). Senescence and tumour clearance is triggered by p53 restoration in murine liver carcinomas. Nature 445, 656-660.

Yahagi, N., Shimano, H., Matsuzaka, T., Najima, Y., Sekiya, M., Nakagawa, Y., Ide, T., Tomita, S., Okazaki, H., Tamura, Y., et al. (2003). p53 Activation in adipocytes of obese mice. J Biol Chem 278, 25395-25400.

Yang, P. (2009). Epidemiology of lung cancer prognosis: quantity and quality of life. Methods Mol Biol 471, 469-486.

Ye, J., and DeBose-Boyd, R.A. (2011). Regulation of cholesterol and fatty acid synthesis. Cold Spring Harb Perspect Biol 3.

Yin, C., Knudson, C.M., Korsmeyer, S.J., and Van Dyke, T. (1997). Bax suppresses tumorigenesis and stimulates apoptosis in vivo. Nature 385, 637-640.

Zhang, C., Liu, J., Liang, Y., Wu, R., Zhao, Y., Hong, X., Lin, M., Yu, H., Liu, L., Levine, A.J., et al. (2013a). Tumour-associated mutant p53 drives the Warburg effect. Nat Commun 4, 2935.

Zhang, Y., Shin, S.J., Liu, D., Ivanova, E., Foerster, F., Ying, H., Zheng, H., Xiao, Y., Chen, Z., Protopopov, A., et al. (2013b). ZNF365 promotes stability of fragile sites and telomeres. Cancer Discov 3, 798-811.

Zhao, Y.G., Zhao, H., Miao, L., Wang, L., Sun, F., and Zhang, H. (2012). The p53- induced gene Ei24 is an essential component of the basal autophagy pathway. J Biol Chem 287, 42053-42063.

Zhou, G., Wang, J., Zhao, M., Xie, T.X., Tanaka, N., Sano, D., Patel, A.A., Ward, A.M., Sandulache, V.C., Jasser, S.A., et al. (2014). Gain-of-function mutant p53

134 promotes cell growth and cancer cell metabolism via inhibition of AMPK activation. Mol Cell 54, 960-974.

Zhou, M., Gu, L., Li, F., Zhu, Y., Woods, W.G., and Findley, H.W. (2002). DNA damage induces a novel p53-survivin signaling pathway regulating cell cycle and apoptosis in acute lymphoblastic leukemia cells. J Pharmacol Exp Ther 303, 124- 131.

Zienolddiny, S., Ryberg, D., Arab, M.O., Skaug, V., and Haugen, A. (2001). Loss of heterozygosity is related to p53 mutations and smoking in lung cancer. Br J Cancer 84, 226-231.

135 Appendices

Appendix 1: Published article

Appendix 2: Microarray data accession number

Microarray data were deposited in Gene Expression Omnibus (accession number: GSE94758).

Appendix 3: R scripts

1.Microarray principal component analysis

#load in data data <- read.delim("~/Desktop/FT data.txt", row.names=1, stringsAsFactors=FALSE) #extracting groups groups <- as.data.frame(data[1,]) #remove group names and transpose data.t <- as.matrix(t(data[-1,])) #perform PCA pc=prcomp(data.t, center = T, scale =T) #Variances summary(pc)

#defining groups: group1 = RT172H (green), group2 = R270H (brown) groupColours_all <- as.factor(as.character(groups[1,])) levels(groupColours_all) <- c("forestgreen", "brown3") # These must be in the correct order cols <- groupColours_all

#plotting graph install.packages("scatterplot3d") library(scatterplot3d) pdf("PCA.plot.pdf", width=6, height = 7)

136 scatterplot3d(pc$x[, 1], pc$x[, 2], pc$x[,3], main = "PCA", xlab = "PC1", ylab = "PC2", zlab = "PC3", color=cols, xlim = c(-100, 150), zlim = c(-150, 150), ylim = c(-150,150), scale.y=0.5, angle = 110, pch = 19, grid=F, box=T ) dev.off()

2. Metabolomics data principal component analysis

Principal component analysis was performed using the muma R package (Gaude et al, 2013).

Appendix 4: Gating for Pdgfa FACS analysis

Examples of gating performed for Pdgfra FACS analysis.

Upper, gating performed on a sample that was a mix of the negative (epithelial cell line) and positive (mouse embryonic fibroblasts) controls for Pdgfra expression. The gate for Pdgfra positivity was set for this sample and applied to subsequent samples. Lower, gating performed on a lung tumour cell line to identify the % Pdgfra-positivity. See Figure 3.4.

137 Appendix 5: Gating for BrdU/PI FACS analysis

Example of gating performed for BrdU/PI FACS analysis.

Gating performed on a lung tumour cell line following vehicle treatment to assess percentage BrdU-positivity. See Figure 6.5.

138