Investigation of A

Investigation of deoxythymidylate kinase (dTYMK) as an imaging and therapeutic target

Alice Beckley

A dissertation submitted for the degree of Doctor of Philosophy

IMPERIAL COLLEGE LONDON Department of Surgery and Cancer

Declaration of originality

I declare that the contents of this dissertation are my original work and conducted by myself, except where otherwise stated and appropriately acknowledged. This thesis was conducted between March 2015 and March 2019 under the supervision of Prof. Eric Aboagye at Imperial College London UK.

The copyright of this thesis rests with the author and is made available under a

Creative Commons Attribution Non-Commercial No Derivatives licence. Researchers are free to copy, distribute and transmit the thesis on the condition that they do not change, transform nor build upon the document. It must not be used for commercial purposes.

For any reuse or redistribution, researchers must make clear to others the licence terms of this work.

Abstract

The uncontrolled proliferative capacity of tumour cells, a hallmark of cancer, has been the main focal point for imaging modalities that provide non-invasive and quantitative estimates of tumour growth. Over the past few decades, several tracers have been developed for use with positron emission tomography (PET) to assess cell proliferation. More commonly, the exploitation of thymidine kinase-1 (TK1) substrate 18F-labeled-3-deoxy-3-fluorothymidine (18F-FLT) uptake for imaging of proliferation has been broadly accepted but, its limitation in accurately depicting the

S-phase fraction has become more apparent over the past 10 years. This study explores the use of deoxythymidylate kinase (dTYMK) as a plausible imaging and therapeutic target since dTYMK participates in the only known pathway to synthesise deoxythymidine diphosphate and ultimately deoxythymidine triphosphate

(dTTP). We introduce the first use of a novel squaramide-nucleotide radiotracer combined with the sensitivity of PET imaging to trace the proliferative tumour fraction with respect to the convergent enzyme, dTYMK. Initial in vitro 18F-SqFLT uptake in salvage proficient (HCT-116), de novo proficient (OST TK1) and CRISPR/Cas9 edited dTYMK knockdowns (B1 and B5) was found to be significantly low (~0.2 %

ID/mg protein) when compared to 18F-FLT (~20 % ID/mg protein) suggesting that, 18F-

SqFLT is not a substrate for dTYMK and, its rate-limiting step may be due to a low passive diffusion. As a pilot study, our observations were extended into an in vivo setting, which revealed non-significant tumour uptake in both wild-type and dTYMK knockdown models when compared to muscle. The highest accumulation of 18F-

SqFLT occurred in the kidney, liver and bladder. A high uptake was also observed in

the gall bladder indicating partial excretion via the biliary pathway. While 18F-SqFLT was unsuccessful in tracing the tumour proliferative fraction, the study still provided pharmacodynamic information into the increasing interest of nucleoside analogues, presenting squaramide phosphate mimics, as potential biologically active cancer agents. Moreover, the CRISPR/Cas9 edited dTYMK knockdown models served as a good platform for understanding some of the mechanisms that may account for dTYMK targeted radiotracer accumulation and retention in cells. A key finding in this study was the disparity between in vitro and in vivo growth rate of dTYMK knockdown models. It was concluded that a dTYMK bypass mechanism that becomes more apparent in vitro than in vivo, may exist to sustain DNA synthesis and maintain genomic integrity. Given the increasing interest in targeting dTYMK as part of an adjunct therapy, these models present as a good system for future pharmaceutical application. To conclude, the exploitation of dTYMK from an imaging endpoint remains challenging; however, success will allow detailed evaluation of the cellular metrics of proliferation and overcome the key limitations associated with 18F-FLT imaging.

Acknowledgements

I want to start by expressing my most profound appreciation to Professor Eric

Aboagye for providing me with the opportunity to embark on the journey of research science. My personal experience left me with a passion for pursuing a career in cancer research, with the hope of having an impact on people’s lives, since my teenage years. I am forever grateful to you for giving me this opportunity and providing the platform I needed to fulfil my passion. You are an exceptional supervisor who has inspired me to be better and challenged my critical thinking. This project, our discussions along with the other collaborative work you have involved me in, has dramatically broadened my research skills and stimulated my drive for research.

I am grateful to my siblings (Lola, Peter, Phillip and Paul), to Yeshua and my parents

(Janet and Elijah Beckley), who have provided moral, emotional and practical support all through my life. I extend further thanks to Peter and Abi for their care and love during my write up and, to Paul for his outstanding presence in my life. Your patience, love and constant care have seen me through all aspects (from the joy of teaching me to ride a bike at 7 years old to soothing my pain during the most turbulent moments of my life). You are nothing short of a blessing to me. Thank you also to my all my nieces and nephews. You may all be to young to know right now but, your laughter, love and carefree nature fuelled my heart and gave me strength.

I want to also express my gratitude to my best friends Marta, Lorraine and Willis for constantly pushing at me and getting the best out of me in your own ways.

Together we have laughed, cried our way from childhood to adulthood. I will always cherish this. I would like to extend a special thanks to Marta. Going through this

PhD journey with you was possibly the best experience I could have. Your support for me both mentally and practically is invaluable. We grew together not just as scientists but as best friends. A very special thanks goes out to all the funders of

Cancer Research UK for providing the funding for this and numerous research projects. It would not be possible without you all.

Lastly, but certainly not the least, I would like to thank the whole of Aboagye lab for the warm atmosphere, the ability to bounce ideas and the fun we have had over the years.

Table of Contents

TABLE OF CONTENTS ...... 6

CHAPTER 1 ...... 18

INTRODUCTION ...... 18

1.0. BACKGROUND ...... 19

1.1. NUCLEOTIDE SYNTHESIS AND DTYMK FUNCTIONALITY ...... 21

1.1.2. STRUCTURE AND CATALYTIC MACHINERY OF DTYMK ...... 24

1.1.3. CYTOPLASMIC AND MITOCHONDRIAL DTYMK ...... 25

1.1.4. CURRENT RESEARCH ON TARGETING DTYMK ...... 28

1.2. PET IN ONCOLOGY ...... 31

1.2.1. PET IMAGING OF CELLULAR PROLIFERATION ...... 34 i) - 2-[11C]Thymidine ...... 35 ii) – 18F-FMAU ...... 36 iii) - 76Br-BrdU and 76Br-BFU ...... 37

1.2.2. 18F-FLT ...... 39

1.2.3. SUMMARY OF 18F-FLT UPTAKE CHARACTERISTICS – ITS ADVANTAGES AND LIMITATIONS ... 40

1.3. SELECTION OF A SUITABLE RADIOTRACER ...... 44

1.3.1. BIOISOSTERES FOR PHOSPHATE MIMICRY ...... 45

1.4. THESIS OBJECTIVES ...... 47

CHAPTER 2 ...... 48

MATERIALS AND METHODS ...... 48

2.1. CELL CULTURE ...... 49

2.2 WESTERN BLOTTING ...... 49

2.3 DNA CELL CYCLE ANALYSIS USING ﬂOW CYTOMETRY ...... 50

2.4. SYNTHESIS OF 18F-SQFLT, 18F-FLT, 18F-D4-FCH AND 18F-FDG ...... 51

2.5. IN VITRO UPTAKE OF RADIOTRACERS ...... 52

2.6. HIGH-PERFORMANCE LIQUID CHROMATOGRAPHY (HPLC) ...... 53

2.7. LIPOSOME ENCAPSULATION ...... 53

2.8. IN VITRO UPTAKE TO DETERMINE THE ACID INSOLUBLE FRACTION OF LABELLED NUCLEOTIDES 54

2.9. ENZYMATIC ASSAY OF DTYMK ...... 55

2.10. SULFORHODAMINE B PROLIFERATION ASSAY ...... 55

2.11. GENERATING HYPOXIA CONDITIONS ...... 56

2.12.1. GENERATION OF KNOCKDOWN CELL LINES WITH CRISPR/CAS9 ...... 56 2.12.2. Polymerase Chain Reaction (PCR) ...... 57 2.12.3 Agarose gel electrophoresis ...... 57 2.12.4 DNA sequencing ...... 58

2.13. SIRNA TRANSFECTIONS ...... 58

2.14. IMMUNOFLUORESCENCE ...... 58

2.15. TUMOUR MODELS ...... 59

2.16. MOLECULAR DOCKING CALCULATIONS ...... 60

2.17. IN VIVO PET IMAGING OF 18F-SQFLT AND 18F-FLT ...... 61

2.18. BIODISTRIBUTION OF 18F-SQFLT ...... 62

2.19. ENZYME ASSAYS FOR PHOSPHORYLATION POTENTIAL ...... 62

2.20. LOGD7.4 DETERMINATION ...... 63

2.21. METABOLITE ANALYSIS USING HYDROPHILIC INTERACTION LIQUID CHROMATOGRAPHY HIGH-

RESOLUTION MASS SPECTROMETRY ...... 63

2.22. CACO-2 AND PAMP ASSAY ...... 65

2.23. DATA ANALYSIS ...... 65

CHAPTER 3 ...... 66

EVALUATING THE ESSENTIALITY OF DTYMK DURING CELL PROLIFERATION ...... 66

3.1. INTRODUCTION TO DTYMK EXPRESSION AND FUNCTIONALITY ...... 67

3.2. BIOINFORMATICS ANALYSIS OF GENETIC DEPENDENCY TO DTYMK ...... 69

3.3.1 CELL LINE CHARACTERISATION ...... 73

3.3.2 MODULATION OF DTYMK DURING CELL CYCLE INHIBITION ...... 75

3.3.3 18F-FLT UPTAKE IN CELL CYCLE ARRESTED CELLS ...... 81

3.4. DTYMK INHIBITION WITH YMU1 ...... 84

3.4.1. MODULATION OF THE SALVAGE PATHWAY WITH YMU1 ...... 87

3.4.1. INVESTIGATING THE EFFECT OF EFFLUX TRANSPORTERS ON YMU1 ...... 89

3.4.2. LIPOSOMAL YMU1 DRUG DELIVERY ...... 91

3.5. VALIDATION OF THE HCT-116 MODEL BY RNA INTERFERENCE ...... 96

3.6. GENERATION OF A HCT-116 CELL LINE ENCODING MUTANT DTYMK ...... 101

3.7. CRISPR-CAS9 CELL LINE CHARACTERISATION ...... 108

3.8. PROFILING OF 18F-FLT PHOSPHATE SPECIES IN HCT-116 DTYMK MUTANT CELL LINES .... 111

3.9. IN VIVO ASSESSMENT ...... 117

3.10. LOCALISATION OF DTYMK BY FLUORESCENT MICROSCOPY...... 121

3.11. DISCUSSION ...... 124

CHAPTER 4 ...... 133

EVALUATION SQUARAMIDE THYMIDINE MONOPHOSPHATE DERIVATIVE FOR PET IMAGING OF CELL PROLIFERATION ...... 133

4.1. IMPETUS FOR USE OF TDR MONOPHOSPHATE ...... 134

4.2. IN SILICO MODELLING OF 18F-SQFLT FOR TARGET VALIDATION ...... 136

4.3. IN VITRO EVALUATION ...... 140

4.3.2. EVALUATION OF COMPETITION AND TRANSPORTER MODULATION ...... 142

4.3. THYMIDINE SALVAGE AND GLOBAL METABOLISM ...... 145

4.4. IN VIVO PET IMAGING OF 18F-SQFLT AND 18F-FLT ...... 149

4.5. BIODISTRIBUTION OF 18F-SQFLT IN BALB/C TUMOUR BEARING MICE ...... 153

18 4.6. INVESTIGATION OF F-SQFLT LOGD7.4 ...... 155

4.7. DISCUSSION ...... 156

5.0. SUMMARY AND CONCLUDING REMARKS ...... 164

5.1. FUTURE DIRECTIONS ...... 168

APPENDIX ...... 171

APPENDIX B – SPECTRA ANALYSIS ...... 180

REFERENCES ...... 184

List of Figures

Figure 1. The ten underlying principles of cancer as reported by Hanahan and Weinberg (2008)...... 21 Figure 2. Schematic representation of the intra and extracellular interactions of thymidine...... 23 Figure 3 (Ostermann et al. (2000)). A ball-and-stick stereoview of the open P-loop of dTYMK in complex with TMP and ADP...... 25 Figure 4. Schema of the mitochondrial-cytosolic exchange of thymidine nucleotides for DNA synthesis...... 27 Figure 5. (Chen et al. 2016). Chemical Structures of 1a, Arene-Fused Isothiazolone Analogues 1b−3b synthesised as dTYMK inhibitors...... 31 Figure 6 (Phelps, 2000). Use of radiotracers like 18F-FDG in PET imaging...... 33 Figure 7. Chemical structures of 2-[11C]-thymidine and its mimic 18F-FMAU for PET imaging...... 37 Figure 8. Structural representation of thymidine and its mimic 76Br-BrdU and 76Br- BFU utilised in PET imaging...... 39 Figure 9 (McKinley et al. (2013)). Schematic representation of the intracellular uptake of 18F-FLT and thymidine...... 40 Figure 10 (Sato et al. (2002)). Evaluation of the chemical structures of squaric acid (1), squarates (2), squaryldiamide (3); and corresponding mesomeric effects of squaric acid derivatives (2 and 3)...... 46 Figure 11. Synthetic pathway of 18F-SqFLT. Synthesis was conducted by Dr Diana Brickute with the help of Chris Barnes and Dr Louis Allot of Imperial College London...... 52 Figure 12. The Cell Line Encyclopaedia (CCLE) gene expression data of dTYMK...... 71 Figure 13. Gene expression boxplots of copy number vs CRISPR/shRNA enrichment scores where neutral refers to ≥ 2 copy numbers and loss = ≤ 1 copy number...... 72 Figure 14. Analysis of TK1, TS and dTYMK expression along with the determination of S phase fraction in the indicated cancer cell lines...... 74

Figure 15. Cell cycle distribution of HCT-116 and BT474 cells after treatment with the indicated cell cycle inhibitors for 24h...... 76 Figure 16. Expression profile of essential enzymes of the thymidine nucleoside pathway following drug-induced cell cycle arrest of HCT-116 and BT474 cells...... 77 Figure 17. Expression profile of the essential enzymes in the thymidine nucleoside pathway...... 80 Figure 18. 18F-FLT cell uptake with cell cycle arresting agents...... 83 Figure 19. Correlation graph of dTYMK expression over 18F-FLT cell uptake following HCT-116 cell cycle arrest...... 84 Figure 20. Cytotoxic effects YMU1 on specified cell lines...... 86 Figure 21. Metabolite assessment of 18F-FLT uptake in YMU1 treated HCT-116 cells...... 88 Figure 22. Cell viability assay in HCT-116 cells after treatment with verapamil hydrochloride in a range of concentrations...... 90 Figure 23. The effect of YMU1 liposomal drug delivery on nucleoside flux in HCT-116 cells...... 93 Figure 24. Caco-2 and PAMPA permeability assay of YMU1...... 95 Figure 25. siRNA mediated dTYMK knockdown in HCT-116 cells...... 97 Figure 26. Effect of transcription and translation inhibition on dTYMK expression. A...... 99 Figure 27. Effect of dTYMK knockdown in HCT-116 cells. Cells were exposed to nontargeting siRNA luciferase or dTYMK 6 and 14 for 48 h...... 100 Figure 28. DTYMK gene organisation and Location of coding exon1 CRISPR guide RNAs...... 103 Figure 29. DTYMK sequencing in wild type and dTYMK knockdown HCT116 (B1 and B5)...... 105 Figure 30. Allele sequencing results for CRISPR/CAS9 generated mutant dTYMK cells...... 107 Figure 31. Expression profile of the essential enzymes in the thymidine nucleoside pathway...... 108 Figure 32. Comparison of doubling times in parental HCT-116 and mutant cells.. . 109 Figure 33. Cell cycle analysis of mutant vs wild type cells...... 110

Figure 34. Assessment of macromolecule incorporation of 18F-FLT with or without 100 μg/ml 5FU treatment in HCT-116, B1 and B5 cells...... 113 Figure 35. Metabolite assessment of 18F-FLT uptake of HCT-116, B1 and B5 cells treated with either 100 µg/ml or 0.33 % DMSO...... 115 Figure 36. Tumour growth curves of HCT-116 and B1...... 118 Figure 37. Expression profile of the essential enzymes of the thymidine nucleoside pathway in HCT-116 and B1 xenografts...... 118 Figure 38. Effect of hypoxia on proliferation following CRISPR/CAS9 mediated dTYMK genomic knockdown...... 120 Figure 39. dTYMK mutants B1 and B5 show unaltered subcellular localization.. ... 123 Figure 40. Schema of alternative contribution to dTTP biosynthesis involving the possible import of dUMP into the mitochondria...... 132 Figure 41. Structure of 18F-FLTMP and species of 18F-SqFLT during charge redistribution...... 136 Figure 42. Stick representation of human ADP-bound dTYMK residues of the P-loop (13−17) coloured in white and enclosing either TMP or 18F-SqFLT...... 138 Figure 43. Structural overlay of active site residues for dTYMK and 18F-SqFLT...... 139 Figure 44. The effect of 18F-5’Squaryl-FLT and 18F-FLT cell uptake in specified cell lines...... 141 Figure 45. Assessment of thymidine or 19F- 5’squaryl-FLT competition with 18F-FLT...... 143 Figure 46. Metabolite assessment of thymidine or 19F- SqFLT competition with 18F- FLT by rHPLC...... 144 Figure 47. Uptake of 18F-FDG and 18F-D4-Choline in HCT-116 and mutant cells. .... 146 Figure 48. Variation patterns of metabolites involved in purine and pyrimidine metabolism, glycolysis and TCA cycle in B1 and B5 cells...... 148 Figure 49. Representative PET images selected from a 60-min dynamic PET scan of viable tumour-bearing HCT-116 or B1 mice imaged with 18F-SqFLT and then re- imaged with 18F-FLT 48 h later...... 151 Figure 50. Time activity curves of HCT116 and B1 tumour bearing mice scanned with either 18F-FLT or 18F-SqFLT over a 60 min period. r...... 152

Figure 51. Biodistribution study of 18F-SqFLT in HCT-116 and 3B1 tumour-bearing BALB/c nude mice following PET imaging...... 154 Figure 52. Schematic diagram of the interactions of the methionine and folate cycle for dTMP synthesis...... 159 Figure 53. Suspected bypass mechanism for dTDP synthesis following dTYMK inhibition...... 166

Table of Tables

Table 1. Liposome dilution protocol for control and YMU1 treatment of HCT -116 cells ...... 54 Table 2. Cell cycle distribution of HCT-116 cells after induction of cell cycle arrest with indicated drug concentrations for 24 h ...... 76 Table 3. Percentages of 18F-FLT cell uptake in HCT-116 after the indicated treatments for 24h...... 83 Table 4. integration of rHPLC 18F-FLT chromatograms under specified conditions .. 93 Table 5. Transcript (Splice variants) of DTYMK as reported by ENSEMBL (protein codable variants have been highlighted in orange) ...... 102 Table 6. Integration of 18F-FLT following 5FU treatment ...... 115 Table 7. B3LYP/6-31G(d): UFF ...... 138 Table 8. Integration of rHPLC 18F-FLT chromatograms under specified conditions 144

18 18 Table 9. Calculated LogP (cLogP) values and measured LogD7.4 for F-FLT, F- FLTMP and 18F-SqFLT...... 155

Abbreviations

β+: Positron 5-FU: 5-fluorouracil

11C: Carbon-11

18F: Fluorine-18

76Br: Bromide-76

µg: microgram

µl: microliter

µM: micromolar

APH: Aphidicolin

ADP: Adenosine Diphosphate

ATP: Adenosine Triphosphate

AZT: 3’-azido-2’,3’-dideoxythymidine

BSA: Bovine Serum Albumin

BrdU: Bromodeoxyuridine

2 D4-FCH: fluoromethyl-[1,2- H4]choline

DAPI: 4’, 6-diamidino-2-phenylindole

DMEM: Dulbecco’s Modified Eagle’s Medium

DMSO: Dimethyl Sulfoxide

DNA: Deoxyribonucleic Acid dNTP: Deoxyribonucleotide Triphosphate dTDP: Deoxythymidine Diphosphate

dTMP: Deoxythymidine Monophosphate dTTP: Deoxythymidine Triphosphate dTYMK: Deoxythymidylate Kinase dUMP: Deoxyuridine monophosphate

E. coli: Escherichia coli

EDTA: Ethylenediaminetetraacetic acid

FACS: Fluorescence-Activated Cell Sorting

FBS: Fetal Bovine Serum

FDG: Fluorodeoxyglucose

FLT: Fluorothymidine

FMAU: 1-(2’-deoxy-2’-fluoro-β-D-arabinofuranosyl)-5-methyluracil

G1: Gap 1 phase

G2: Gap 2 phase

HIF-1α: Hypoxia-inducible factor 1α

HIV: Human immunodeficiency virus

HPLC: High-performance liquid chromatography

HRP: Horse Radish Peroxidase

ID: Injected Dose

KDa: Kilodalton

M: Mitosis mg: Milligram ml: millilitre mM: millimolar

MRI: Magnetic Resonance Imaging mRNA: messenger RNA mtdTYMK: Mitochondrial dTYMK

NDPK: Nucleotide Diphosphate Kinase nM: nanomolar

NOC: Nocodazole

OMP: Orotidine 5′-monophosphate

PAGE: Polyacrylamide Gel Electrophoresis

PBS: Phosphate Buffered Saline

PBS-T: Phosphate Buffered Saline-0.1% Tween 20

PET: Positron Emission Tomography

PI: Propidium Iodide pM: picomolar

PRPP: phosphoribosyl pyrophosphate

RIPA: Radioimmunoprecipitation assay

RNA: Ribonucleic Acid

RNase: Ribonuclease

RNR: Ribonucleotide Reductase

ROI: Region Of Interest

Rpm: rounds per minute

RPMI: Roswell Park Memorial Institute Medium 1640

S-phase: Synthetic phase

SDS: Sodium Dodecyl Sulphate

Ser: Serine shRNA: short hairpin RNA siRNA: small interfering RNA

TAC: Time-activity curves

TDR: Thymidine

TK1: Thymidine Kinase 1

TK2: Thymidine kinase 2

TE: Tris EDTA buffer

TP: Thymidine phosphorylase

TS: Thymidylate Synthase

Chapter 1

Introduction

1.0. Background

Cancer develops as a result of a clonal expansion of abnormal somatic cells that kill by invading and eroding healthy tissues1. The fundamental processes that propagate tumour pathogenesis and metastasis are evolutionary and thus, occur through

Darwinian natural selection of inherited or acquired mutations. Over the last 6 decades, advances in research have enabled the identification of multistep processes that fuel the propagation of aberrant cells that confer an increased survival of mutant clones2. Hanahan and Weinberg summarised these acquired capabilities into ten underlying principles known as, the hallmarks of cancer1 (figure 1). Although neoplasia often requires a combination of the ten principles, in almost all instances, deregulated cell proliferation and suppressed cell death collectively provide the central platform for neoplastic progression. Coherently, the primary focus of the research community is to identify and understand the critical events of DNA synthesis that drive aberrant cell division and evaluate their potential as targets for cancer therapy/imaging applications. Uncontrolled cell proliferation (an integral hallmark in the cancer phenotype1) occurs as a consequence of accumulated aberrations in multiple cell regulatory systems thereby changing their expression, conformation and stability3. Tumour growth is commonly used to describe the relationship between cell division and cell death which are intricately regulated processes through a network of signal transduction pathways3,4. In most cases, cell regulators like receptor tyrosine kinases (RTK) and p53 are highly and continuously activated or deregulated (for the latter) in human cancers. The result is an exacerbation of proliferation accompanied by defects of the apoptotic mechanisms3.

The assessment of cell proliferation is thus essential for the identification of not only tumour masses but, for the differentiation of benign and cancerous lesions. During

DNA synthesis, a steady supply of deoxyribonucleotides is maintained and tightly modulated by a complex of enzymes that initiate a cascade of events to aid successful

DNA synthesis. Over the last 60 years, attention has been on the functionality of these enzymes as targets for anti-cancer therapy5,6. One such enzyme thymidylate synthase (TS) is a target of the chemotherapy 5-fluorouracil (5-FU). 5-FU is a cytotoxic agent that becomes intracellularly converted into the active metabolites, 5- fluoroxyuridine monophosphate (F-UMP) and 5-5-fluoro-2'-deoxyuridine-5'-O- monophosphate (F-dUMP). F-UMP replaces uracil monophosphate and is incorporated into RNA thereby inhibiting RNA processing and preventing cell growth.

F-dUMP inhibits the action of thymidylate synthase causing a depletion of thymidine triphosphate (dTTP) pool for DNA synthesis/repair6. 5-FU has been mainly used for advanced colorectal cancer therapy with a response rate of approximately 10–15%7.

This low response rate and clinical resistance led to the modulation of treatment programs that employed the synergistic effects of drugs like methotrexate to intensify the effect of 5FU8. An increasing focus has been on, amongst others, targeting deoxythymidylate kinase (dTYMK) as a more desirable pharmacological endpoint for sensitising cancer cells to DNA damage9. The pharmacodynamic response of tumours to these therapies would, therefore, need evaluation on a molecular scale, and one possibility is to use positron emission tomography (PET)10–

12. PET is a non-invasive imaging tool that utilises short-lived radioactive isotopes to provide functional information on the pathophysiology of disease12,13. Although other imaging modalities, like MRI14, can also be used, PET is preferred due to its high

sensitivity and ability to target most cell surface and intracellular receptors or enzyme systems in vivo during the characterisation and quantification of disease progression10,12.

Figure 1. The ten underlying principles of cancer as reported by Hanahan and Weinberg (2008). These principles outline the key cancer phenotypes that form the basis of malignancy.

1.1. Nucleotide synthesis and dTYMK functionality

Two pathways produce precursors for DNA synthesis; the de novo pathway and the salvage pathway (figure 2). In the energy expensive de novo pathway, the synthesis of nucleotides occurs as a series of enzymatic reactions to convert small molecules into nucleotides (i.e. dUMP). The initial step involves carbamoyl phosphate synthetase II which catalyses the formation of carbamoyl phosphate by combining

available CO2 to glutamine. The formed carbamoyl phosphate is combined with water and aspartate before being dehydrogenated in a series of enzymatic reactions to form orotic acid. The resulting product is then merged with a ribose-5-phosphate ring by the action of Orotate phosphoribosyl transferase to form, Orotidine 5′- monophosphate (OMP). OMP is later decarboxylated to produce uridine monophosphate (UMP) which becomes deoxygenated to form the deoxyribonucleotide, deoxyuridine monophosphate (dUMP). Thymidylate synthase then methylates dUMP to produce deoxythymidine monophosphate (dTMP) for the deoxythymidine triphosphate (dTTP) biosynthesis pathway. The de novo pathway plays an essential role in regulating a balanced supply of dTMP to prevent death due to lack of thymine that may occur from dTMP shortage15. In the salvage pathway, free nucleosides are actively transported into cells to aid the synthesis of ribonucleotides and deoxyribonucleotides. Thymidine is recycled from the extracellular space and rapidly imported into the cell by nucleoside transporters16,17.

Upon entry, thymidine is intracellularly trapped by the action of thymidine kinase 1

(TK1) which catalyses the transfer of a phosphate group (the terminal phosphate of

ATP) to the 5’-hydroxyl end of thymidine to produce dTMP 13,17. Although cancer cells are heavily reliant on the de novo synthesis of nucleotides, the energy inexpensive salvage pathway also plays a vital role in aiding de novo synthesis for the balance dTMP required for the DNA synthetic pathway. Deoxythymidylate kinase (dTYMK)

(EC 2.7.4.9) is an essential enzyme for pyrimidine synthesis that catalyses the phosphorylation of dTMP to deoxythymidine diphosphate (dTDP) in the presence of

ATP and magnesium18,19. The product, dTDP, is phosphorylated further into dTTP by the action of nucleoside diphosphate kinase (NDPK). DTTP is then incorporated into

the DNA by DNA polymerase alpha which adds dTTP to the three prime (3')-end of a

DNA strand (figure 2)18. Therefore, dTYMK represents the first enzymatic step by which the de novo and salvage pathway converge for dTTP biosynthesis. Unlike other nucleoside monophosphate kinases (NMPKs), dTYMK is cell-cycle regulated with its expression peaking during the S phase fraction during DNA synthesis18,19. Since dTYMK participates in the only known pathway to synthesis dTDP, the development of a potent and selective inhibitor of the enzyme could be advantageous for anticancer therapy and improving patient prognosis.

Figure 2. Schematic representation of the intra and extracellular interactions of thymidine. For TTP biosynthesis, thymidine is transported across the cell membrane and intracellularly trapped (phosphorylated) by TK1 to produce thymidine monophosphate (TMP). TMP is then phosphorylated into thymidine diphosphate (TDP) and later thymidine triphosphate (TTP) for incorporation into DNA. In the de novo pathway, deoxyuridine monophosphate (dUMP) is methylated by TS to produce TMP and converges with salvage to produce TTP in the same manner. Key: dUDP (deoxyuridine diphosphate), UMP (uridine monophosphate, dNT (deoxynucleotidase), TP (Thymidine phosphorylase) and TdR (thymidine)

1.1.2. Structure and catalytic machinery of dTYMK

The structure of dTYMK and its functionality has been widely studied in a number models including yeast, Escherichia coli and human18,20. The enzyme presents as a globular homodimer with each subunit consisting of a central five-stranded parallel

β-sheet surrounded by 7-11 α-helices structure that folds similarly to other NMPKs21.

From the sequence homology and biochemical data of yeast and E. coli dTYMK, the enzyme displays a three-loop system that is essential for its function in producing dTDP during dTTP biosynthesis20,22. The first is a highly conserved P-loop motif, presenting a consensus sequence GX1X2X3X4GKS/T in single-letter amino acid code21, whose primary function is to bind and position the α- and β-phosphoryl groups of the phosphoryl donor (i.e. ATP) through an intricate network system between the amide backbone hydrogens and phosphate oxygen atoms20,21 (figure 3).

The dTYMK phosphoryl acceptor (TMP) interacts with the 13–17 residue of the human dTYMK P-loop via a hydrogen bond between the 3′-hydroxyl group of the TMP

20,21 along with a carboxylic acid residue at position X2 of the P-loop motif . Although the exact function of this residue remains unclear, it is reported to be an integral part of the catalytic mechanism which can halt dTYMK functionality if the residue becomes deleted21. The second critical loop encompasses a DR(Y/H) motif that is a characteristic of all isoforms of dTYMK20,21. Unique however to the human enzyme, is the aspartic acid residue (Asp96) to which the magnesium ion and ATP complex are bound and positioned. Adjacent is the arginine residue (Arg97) which functions as a clamp drawing together, the phosphate group of TMP and the γ-phosphate of phosphoryl donor (ATP) for phosphoryl donation. The last loop is known as the flexible LID region which spans residues 135–150 of the enzyme. As the name

suggests, it covers phosphoryl donor-ATP complex preventing water from accepting the transferred phosphoryl group20,21.

Figure 3 (Ostermann et al. (2003)). A ball-and-stick stereoview of the open P-loop

structure of human thymidylate kinase with NH2TMP and AppNHp. The phosphates of AppNHp are kept in place through intricate interactions between the oxygen atoms and the P-loop main chain amides along with the catalytically essential magnesium ion as previously described. Stabilisation within the P-loop is mediated by the seven water

molecules that form hydrogen bonds between the P-loop and the NH2TMP substrate.

1.1.3. Cytoplasmic and mitochondrial dTYMK

Genomic analysis suggests that dTYMK exists in two forms. The first is a human cytoplasmic dTYMK which is often referred to as TMPK1 and the second is a human

23 mitochondrial dTYMK (mtdTYMK) that is so termed TMPK2 . Since mitochondrial

DNA synthesis requires a steady supply of thymidine triphosphate (dTTP) that occurs

23 independently of nuclear DNA replication , mtdTYMK is speculated to exist and contain a mitochondria localisation sequence at the N-terminusl24.

Unlike nuclear DNA synthesis, mitochondrial DNA replication is not cell cycle regulated and continues during the S-phase and quiescence stages of the cell23,25–27.

In proliferating cells, dTDP is synthesised in the cytoplasm and imported into the mitochondrial matrix via deoxynucleotide transporters23. Mitochondrial NDP kinases catalyse the conversion of dTDP to dTTP for DNA incorporation28. In quiescent and differentiating cells, the expression levels of ribonucleotide reductase, cytoplasmic

TK1 and TS (amongst others) are significantly reduced due to cell cycle arrest24,25. As a result, thymidine is imported from the cytoplasm (from degradation products) into the mitochondrial (via TOM20 transporter) and phosphorylated by TK227,29. The presence of TK2, mitochondrial NDP kinase and observations of rapid exchange between the cytosol and mitochondrial levels of dTMP23,30, suggests that a mitochondrial dTYMK that phosphorylates dTMP to form dTDP in its matrix must exist. Chen et al. (2008) published a report that seemingly fitted the criterion for mitochondrial dTYMK however, their attempt to define mtdTYMK specificity and kinetic properties were not successful nor were they able to use their purified recombinant mtdTYMK protein to detect sufficient levels of the enzyme activity in vitro24. The group speculated that a critical co-factor was amiss from the reaction that proved to be the rate-limiting step of their in vitro activity assay. Nonetheless, they were first group to report the discovery of a mtdTYMK to bridge the gap for the second phosphorylation step that forms dTDP in the mitochondria.

Figure 4. Schema of the mitochondrial-cytosolic exchange of thymidine nucleotides for DNA synthesis. Schematic representation of the metabolic pathways that supply dNTPs for mtDNA replication. dTdR: thymidine; TP: thymidine phosphorylase; TK1 (cytosolic) and TK2 (mitochondrial) thymidine kinases; dCK: deoxycytidine kinase; cdN and mdN: cytosolic and mitochondrial deoxynucleotidases; TS: thymidylate synthase; RNR: ribonucleotide reductase. ENT1/2: equilibrative nucleoside transporter 1/2, which has been found in human mitochondria. Evidence, as described by Chen et al. (2008), of a highly concentrative dTMP transport by mitochondrial import receptor TOM20. The arrow and question mark indicate the possibility of an unidentified TTP/CTP carrier. A dCTP transport activity has been reported by Bridges et al. (1999).

1.1.4. Current research on targeting dTYMK

As a critical enzyme in the dTTP biosynthesis pathway, dTYMK is tightly controlled/ modulated during different phases of the cell cycle. During DNA repair, dTYMK and ribonucleotide reductase (RNR) are activated to orchestrate the production of the four dNTPs required for double-strand break repair. A decline in dTYMK functionality would, in turn, increase RNR supplied dUTP misincorporation into DNA during DNA double-strand break (DSB) repair. Initial inhibitor studies of dTYMK involved co- crystallisation of 3’-azido-2’,3’-dideoxythymidine monophosphate (AZTMP) with dTYMK to aid the understanding of the rate-limiting step in 3’-azido-2’,3’- dideoxythymidine (AZT)activation. Information from this study initiated a series of inhibitor research on dTYMK as an alternative to targeting TS or TK1. One such study is that of Hu and Chang 2008 who evaluated the potential for dTYMK as an anticancer target by silencing, using lentiviral-based small hairpin RNA, the expression of dTYMK in HCT-116 p53 proficient and null cells. A dTYMK knockdown was hypothesised to induce cell cytotoxicity due to an increase in deoxyuridine triphosphate (dUTP) misincorporation into DNA following dTTP shortage. Surprisingly, under the genotoxic insults of short hairpin RNA, a knockdown appeared not to impair cell viability. Interestingly, the same group noted that a knockdown of dTYMK led to an increase of DNA lesions following treatment with doxorubicin (an inhibitor of DNA topoisomerase II). The group hypothesised that these effects were due to an increase in DNA uracil content that could be countered by overexpression of dUTPase to clear dUTP from the nucleotide pool. In all, it was concluded that a knockdown of dTYMK in colon cancer cells predisposes cells to lower doses of doxorubicin irrespective of p53 status. On this basis, Hu et al. (2012) proceeded to search for inhibitors of dTYMK

that might be useful as an approach for the retardation of DSB repair and, sensitisation of tumour cells to doxorubicin without genetic manipulation. The group screened a library of 21,120 small molecules through a luciferase-coupled dTYMK in vitro assay optimised by Hu and Chang (2008). Of the 21,120 potential inhibitors,

YMU1 was selected as a potent inhibitor of dTYMK with an IC50 of 0.6 µM). Of importance was the selectivity of the compound which was reported to be avid to dTYMK and inactive to human thymidine kinase (TK1) at physiological concentrations

≤ 10 µM. The authors proceeded to model YMU1 in complex with dTYMK in attempt to understand YMU1’s interaction with the enzyme active site and, determine the binding free energy. Results from in silico modelling suggests that YMU1 inhibits the interaction of Mg2+ with Asp15 residue in the P-loop, thereby, disrupting phosphoryl donation to TMP through its competition with ATP for the Asp15 residue. Their initial cell-based assay revealed that YMU1 sensitised cancer cells to low dose doxorubicin but treatment with the lead compound alone did not affect cell viability. Although dTYMK is understood to be essential in the dTTP biosynthesis pathway, its mechanism of tolerance in cells, following its knockdown, remains to be determined.

The authors propose the possibility of cells producing variant isoforms of dTYMK as a compensatory mechanism. The study, however, did not proceed to validate this hypothesis with additional genetic and biochemical analysis. Nevertheless, the findings of Hu et al. (2012) provided an insight into the potential of targeting dTYMK as part of an adjunct therapy31. In another study, 3’-C-branched chain-substituted substrate analogues of dTMP in the ribo and 2’-deoxyribose series were explored as novel anti-tuberculosis agents targeted to dTYMK. Compounds K1, K2, K7 and K10 were found to exhibit cytotoxic effects on mycobacterial growth (inhibition of DNA

synthesis) at varying intensities19. However, these were less active than YMU1 and in turn rejected for translation into human cancer cells. A recent report by Chen et al.

(2016)9 provided promising evidence of arene-fused isothiazolone inhibitors, which disrupt connections within dTYMK active site (LID and P-loop) that are crucial for phosphotransfer, as plausible inhibitors of dTYMK worthy of biological evaluation.

With similarities to Hu et al. ’s (2012) methodology, the authors employed the luciferase-coupled TMPK assay to evaluate the inhibitory activities of their library of benzene-fused isothiazolones. Of the various inhibitors investigated, compound 3a

(figure 5) was found to be an eﬀective hdTYMK inhibitor (70% inhibition at 2 μM) and would be worth future validations in vitro 9. Regarding cell type sensitivity, Lui et al.

(2013) reported a deficit in the nucleotide metabolism of LKB1-mutant lung cancer cells that increased the sensitivity of the cancer cells to dTYMK inhibition. The authors, in turn, suggested dTYMK as a potential therapeutic target in LKB1 -mutant human cancer32. With the increasing interest in targeting dTYMK as part of adjunct therapy31, the development of a radiotracer which elucidates drug-induced changes of dTYMK activity or expression would be indispensable in understanding the pharmacodynamics of these emerging therapies for pharmaceutical application. A prerequisite for the development of such a radiotracer requires a detailed investigation into the relationship between dTYMK gene expression, its enzymatic activities and correlation with the growth rate of proliferating cells.

Figure 5. (Chen et al. 2016). Chemical Structures of 1a, Arene-Fused Isothiazolone Analogues 1b−3b synthesised as dTYMK inhibitors.

1.2. PET in Oncology

PET utilises short-lived positron-emitting isotopes to provide quantitative and qualitative information on tissue biochemistry, disease aetiology or drug kinetics13,33.

Positron-emitting isotopes like 11C, 13N, 76Br, 124I, and 18F can be incorporated into many compounds of biological interests to produce radiotracers33,34. These radionuclides undergo β+ decay and emit a positron that travels a short distance before annihilating with an electron. Annihilation produces two γ-rays (two 511KeV photons) which propagate at nearly 180⁰ in opposing directions, and the resulting photons are detected by scintillation crystals present within the scanner to generate a coincidence event (within ~10 ns window)35. Many of these events are summed to provide the distribution of the radiotracer35,36. The coincidence window permits signal localisation and calculation of annihilation locus, thereby giving rise to a PET image35. For our application, 18F is the radioisotope of choice due to its high branching fraction, overall stability, 109 min half-life and in turn good feasibility in clinical practice13. Moreover, the radioisotope emits low kinetic energy positrons, with short-range travel, that result in a high PET image resolution. Other radioisotopes,

depending on required characteristics37, may be employed for pharmacokinetic studies (i.e. microdosing of human subjects before phase I trials38 and antibody labelling) or assessment of gene delivery. PET imaging is an attractive molecular imaging tool for clinical decision making since it is surgically non-invasive, permits repeat analysis in the same patient (before and after therapy) and detects heterogeneity within tumours33,34,39. 2-deoxy-2-[18F]fluoro-D-glucose (18F-FDG)

(glucose analogue) is currently the most clinically used method for diagnosing, staging and evaluating (interim) response to therapy (figure 6)39. Its uptake into tissues is not a direct measure of cell proliferation but instead, a surrogate for cell viability that is determined mainly by high activities of glucose transporter and hexokinase along with low glucose-6-phosphatase activity in tumours39,40. Although

18F-FDG PET has been shown to detect small tumour foci and possess great diagnostic accuracy in distinguishing scar tissue from malignant tumours, it does present with limitations including its inability to delineate proliferating tumour from inﬂammation including those stemming from radiotherapy, its consideration of high glucose-6- phosphatase activity in liver and the high uptake exhibited in the brain41,42. As a result, its use in pancreatic, prostate (high excretion through the kidney obscuring prostrate), liver and brain tumours/metastasis is restricted. Other PET tracers have been employed for specific pathologies (i.e. 18F-T808, a biomarker of tau aggregates in Alzheimer)43 or other biological endpoints of therapy. Exploratory radiotracers like

ICMT-11 (an apoptotic marker via assessment of cleaved caspase-3/7 activation44), and 3-fluoro-2,2-dimethylpropionic acid 18F-FPIA (an analogue of [11C]acetate for tracing fatty acid synthesis45) are in variable stages of clinical approval46. However,

18F-ICMT-11 uptake is known to be less sensitive in depicting the changes in apoptosis

in the presence of extensive pre-existing necrosis44. With these emerging limitations increasing focus has been on imaging cell proliferation as a more robust method of understanding focal pathophysiology and monitoring drug response.

Figure 6 (Phelps, 2000). Use of radiotracers like 18F-FDG in PET imaging. Following the injection of positron-emitting 18F-FDG, the radionuclide starts to lose its kinetic energy in the tissue before combining with its antiparticle, the electron. This annihilation process emits 2 gamma photons at opposite directions which is collected by a pair of coincident detectors. The signals are then reconstructed to determine the loci of radiotracer accumulation.

1.2.1. PET imaging of cellular proliferation

As previously mentioned, 18F-FDG-PET depicts the glucose metabolic pathway and permits the identification of neoplastic lesions since tumours require high glucose uptake owing to the Warburg effect. However, its limitations (as mentioned in section 1.2.) limits its use as an overall decision-making marker for tumour staging/grading and pharmacodynamic response. This limitation provided an impetus for the research community to develop a proliferation radiotracer that compliments and overcomes the limitations of 18F-FDG. The replication of DNA is essential for proliferating cells, and thymidine (TdR) is known to be actively incorporated into Deoxyribose Nucleic Acid (DNA) and not Ribose Nucleic Acid (RNA) during S-phase47. The ability to measure TdR uptake provides a means to characterise tumour behaviour and allow direct assessment of changes in tumour growth when validating biological response13,47. Over the years, a number of pyrimidine nucleosides (i.e. 2-[11C]thymidine and its analogues, 3’-deoxy-3’[18F]fluorothymidine

(18F-FLT), 2-fluoro-5-[11C]-methyldeoxyuracil-β-D-arabinofuranoside (18F-FMAU) and

[76Br]bromodeoxyuridine (76Br-BrDU)) have been synthesised for pre-clinical and clinical assessment of cell proliferation13,48,49. These radiotracers are actively transported into cells via nucleoside transporters before being intracellularly trapped by the action of TK1 to form their corresponding TMPs50. However, the rate of phosphorylation between these nucleoside tracers differs significantly due to their avidity to TK1. Of clinical relevance is the high in vivo stability of nucleoside analogues with a halogen substitution in the sugar-ring (e.g. 18F-FLT, 18F-FMAU) since the metabolism of radiotracers with less stability can complicate the interpretation of data13. The following bullet points expands on some well-reviewed radiotracers i.e.

2-C-11-Thymidine, methyl-11C-dThd, 1-(2’-deoxy-2’-18F-fluoro-β-D- arabinofuranosyl) thymidine (18F-FMAU), 1-(2’-deoxy-2’-18F-fluoro-βD- arabinofuranosyl) uracil (18F-FAU) and [76Br]-bromodeoxyuridine (76Br-BrdU) as thymidine analogues for real-time imaging of cell proliferation. 3’-deoxy-3’-

[18F]fluorothymidine (18F-FLT) will be examined in greater detail in section 1.2.3. i) - 2-11C-Thymidine

The concept of using labelled dThd was first introduced by Wantzin and Killmann

(1977) who provided quantitative estimates of proliferation rate in human leukemic blasts using tritiated thymidine (3H-thymidine)51. Over the years, more interest has been on the investigation of 2-11C-thymidine and 11C-methyl-thymidine in place of tritiated thymidine for clinical use. A report by Lonneux et al. (1996)52 showed successful incorporation of 11C-methyl-thymidine into DNA however, the metabolism of this radiotracer resulted in labelled thymine and other labelled degradation products that made it difficult to prove specific labelling during the interpretation of

PET images53. 2-11C-thymidine was later developed with the hope of improving tumour accumulation while reducing degradative products. Unfortunately, this bioisostere was metabolised at a similar rate in vivo as its predecessor 11C-methyl- thymidine. Compartmental models52 and mathematical analysis were subsequently

11 11 developed to separate the contributions of 2- C-thymidine and C-CO2, a labelled degradation product of labelled thymidine cleared into the blood, improving the interpretation of results54. On this basis, groups like Eary et al. (1999) continued to explore the diagnostic potential of 2-11C-thymidine with particular interest on PET imaging of brain tumours55. Because 18F-FDG reflects energy metabolism, the naturally high uptake of glucose (in healthy segments of the brain) can obscure the

subtle increases during neoplasia. Furthermore, significant differences in tumour-to- tumour metabolism of 18F-FDG vs glucose can also impede the interpretation of tumour energetics since, the transport and phosphorylation of the 2 hexos substrates occur at different rates with hexokinase (enzyme responsible for glucose or FDG phosphorylation) which lays preference to glucose over FDG. In turn, the lumped constant (a correction factor) is required to infer the glucose metabolic rate from

FDG metabolic rate 39,56,57. For these reasons, labelled thymidine has been suggested as a better measure of tumour proliferation in brain imaging. This suggestion was validated by Eary et al. (1999) who reported a superior differentiation ability of 11C- thymidine over 18F-FDG in distinguishing between, tumour and normal brain/white matter. It can be concluded that tumour viability derived from imaging of cell proliferation may be more sensitive for monitoring disease activity and progression than 18F-FDG uptake, especially in tissues like the brain. However, the short half-life of 11C (20.4 min), its rapid metabolism following incorporation into cells needed to be addressed. Accordingly, second-generation probes using 18F were developed to avert these limitations along with complex modelling analysis that is associated with

11C uptake 48,54,58. ii) 18F-FMAU

FMAU was originally discovered as an anti-viral and anti-neoplastic compound but, its severe cytotoxicity made it an unusable therapeutic agent for clinical practice17,59,60. Its rejuvenation as a proliferation marker, used at trace doses, provided a new prospect for its return into the clinic, and unlike the aforementioned labelled thymidine analogues, its 3’ hydroxyl group of deoxyribose was not metabolised. The extensive catabolism associated with both 11C-thymidine and of

11C- methyl-thymidine led to the interest of labelling FMAU with either 11C or 18F61.

FMAU later proved to be an inadequate marker presenting significantly low uptake values in highly proliferating tumours such as triple-negative breast cancer62. Tehrani et al. (2008) attributed this observation to the radiotracer’s preference for mitochondrial TK2. Unlike TK1, TK2 functions without the regulations of the cell cycle and is continuously expressed at low activity in both proliferative and quiescent cells23,62. Although 18F-FMAU has been disregarded as an adequate marker of proliferation, its low physiologic uptake in normal bone marrow may confer an advantage in detecting and monitoring bone marrow metastases13,48,62. Since its uptake in tumours is still higher than most healthy tissues, it can also be used to evaluate the pharmacodynamic response of conditions that present an increase in mitochondrial mass due to oxidative, reductive and energy stress63,64.

Figure 7. Chemical structures of 2-[11C]-thymidine and its mimic 18F-FMAU for PET imaging.

iii) 76Br-BrdU and 76Br-BFU

76Br-BrdU, a well-reviewed analogue of thymidine for the immunohistochemical analysis of cell proliferation, was adapted for PET imaging by introducing a 16 h half-

life 76Br positron-emitting radionuclide into thymidine. In a study by Bergström et al.

(1998), the radiotracer showed promising results in rats and pigs and, offered insight to its potential as a biomarker of proliferation in vivo65. However, within moments of injection, it was noted that a large fraction 76Br-BrdU is rapidly metabolised into 76Br- bromide which dominates the radioactivity in the plasma and non- proliferative/slowly proliferating tissues. Due to this slow plasma clearance

(elimination half-life of 9 to 12 days), PET-76Br-bromide images require mathematical corrections for 76Br-Br to boost contrast. Second generation (2′-deoxy-2′-fluoro-β-d- arabinofuranosyl)-5-[76Br]bromouracil (76Br-BFU) was synthesised as a modified version of 76Br-BrdU substituting the hydrogen at the 2′ position of the sugar moiety with a fluorine atom. This modification was hypothesised to reduce the 76Br-bromide formation and prolong tracer biological half-life66. Preclinical evaluations of 76Br-BFU showcased encouraging results that suggested the radiotracer as a valid proliferation marker in vivo. Though 76Br-BrdU is incorporated into DNA at high levels, it is still dependent on TK1 activity and in turn reports on the salvage arm of DNA synthesis.

Moreover, labelling with 76Br is not as direct as 18F, and the radionuclide generates a high radiation dose which can only be counteracted by reducing patient injection dose. The result is a noisy image that makes interpretation difficult. This, together with slow plasma clearance, makes 76Br-BFU an unsuitable tracer for clinical evaluation. Taken together, the adverse effects accompanying these radiotracers meant that their use as a robust biological and pharmacological endpoint was restricted.

Figure 8. Structural representation of thymidine and its mimic 76Br-BrdU and 76Br-BFU utilised in PET imaging.

1.2.2. 18F-FLT

FLT was initially discovered as a chemotherapeutic agent against leukaemia (Blau et al., 1989)67. A later study by Kong et al. (1992)68 revealed the anti-HIV activity potential of FLT. However, high systemic toxicity precluded its use as a treatment for

HIV17. In that time, the potential of thymidine as a tracer of cellular proliferation was being evaluated. Shields et al. (1998) later proposed fluorinated thymidine analogue,

18F-FLT (figure 9), presenting an adequate half-life (109.8 min) and resistance to phosphorylase induced degradation for diagnostic use with PET. This resistance to degradation conferred an advantage over 11C labelled thymidine since it minimised non-specific activity and increased the tumour to background ratio16.

1.2.3. Summary of 18F-FLT uptake characteristics – its advantages and limitations

The most promising of all proliferation markers is 18F-FLT; an 18F labelled thymidine at the 3’OH position of the deoxyribose sugar. 18F-FLT gains access into cells through passive diffusion and active transport by pyrimidine transporters ENT-1 and ENT-2

(alongside concentrative nucleoside transporters 1 and 3). Upon entry, the salvage pathway regulator TK1 catalyses the intracellular trapping of thymidine and 18F-FLT into their highly charged nucleotide monophosphates. During the stages of the cell cycle, TK1 is highly expressed in S-phase and rapidly degraded when cells enter mitosis/G1 arrest. The rate of intracellular trapping of 18F-FLT correlates, therefore, with the S-phase fraction in relation to TK1 activity. Nucleotide monophosphates are subsequently phosphorylated into their respective nucleotide diphosphate and triphosphates by the action dTYMK (for TDP), and nucleoside diphosphate kinase

(NDPK), a ubiquitous and S phase modulated kinase, for synthesising dTTP. The energy inexpensive anabolic arm of the DNA salvage pathway is used by cancer cells to maintain the equilibrium of intracellular nucleotide pools required for DNA synthesis16. Since the contributions of the salvage nucleotide pool are rapid16, excess deoxynucleotides produced are degraded by nucleotidases and thymidine phosphorylase (TP) to be exported as nucleosides to avert nucleotide imbalance16.

18F-FLT however, is inert to TP due to the halogen modification that stabilises the glycosidic bond, and in turn, dismisses the contributions of TP in regulating thymidine accumulation. This is of importance when interpreting PET images of tumour types that present higher levels of TP. Grierson et al. (2005) suggested the possibility of a putative deoxynucleotidase (dNT) capable of dephosphorylating 18F-FLTMP into FLT and in turn, offer an answer to the primary mechanism for tracer excretion from the cell16. However, to our knowledge, the role of this enzyme in the metabolism of 18F-

FLT is yet to be validated. The authors also noted a higher accumulation of 18F-FLTMP over 18F-FLTDP and 18F-FLTTP. It is believed that 18F-FLTMP has reduced substrate avidity towards dTYMK16 resulting in a decrease in the accumulation of its diphosphate species. This theory is coherent with the findings of Ostermann et al.

2000, who reported a limitation in the conversions of 3’-azido-2’,3’- dideoxythymidine (AZT), an analogue of thymidine with an azido group in place of the 3’ hydroxyl group, from its mono to diphosphate form21. In silico modelling of the nucleotide complexes, performed by the same group, revealed a steric hindrance by the azido group preventing the formation of the catalytically active closed conformation of the dTYMK P-loop. Owing to these results, it is generally accepted that a loss of 3’OH group in thymidine confers a reduced dTYMK avidity and, inhibits

nucleotide incorporation into DNA. Unlike thymidine, the phosphorylated form of

18F-FLT cannot be incorporated into DNA (<1%) and therefore acts as a chain terminator during DNA replication13,69,70. Nevertheless, 18F-FLT has been highly reviewed as an adequate biomarker of cell proliferation, due to its incorporation into cells and its high specificity to TK1 which, permits the formation of radiolabelled 18F-

FLT phosphates inside cells thus increasing the target to noise ratio; a prerequisite for PET-imaging71,72.

Several tumor models such as gastric cancer (NCI-N82), leukemia (K562)73, head and neck squamous cell carcinoma (HNSCC FaDu,)74 and neuroblastoma (K-N-SH)75 have been reported to accumulate a more considerable amount of 18F-FLT than 18F-FDG while, comparable levels of uptake has been reported in gastrointestinal stromal tumours (GIST882)76. A review by Schelhaas et al. (2017) noted that 83 % of preclinical studies monitoring the modulation of 18F-FLT uptake after tumour treatment, report a decrease of radiotracer accumulation as a pharmacodynamic response. Although 18F-FLT has shown great promise, it has rarely been used for tumour therapy follow-up in clinical trials. This is due to the presence of confounding factors which regulate 18F-FLT uptake making it difficult to interpret the outcome of specific therapeutics. The default mechanism of action of 18F-FLT results in its principal limitation as proliferation marker 18F-FLT is unable to delineate moderately proliferative tumours that rely on the salvage pathway from highly proliferating tumours that are dependent on de novo and therefore TS77. The results are a generation of an inaccurate tumour proliferative index/ false negatives that impedes on effective therapeutic mediation77. Tumour-specific characteristics such as thymidine phosphorylase (TP) and serum/tissue thymidine levels also impact the

level of 18F-FLT uptake. Thymidine competes with 18F-FLT for nucleoside transporters at the cell surface and TK1 activity for its retention within cells. Because TK1 exhibits a natural preference to thymidine relative to 18F-FLT16, it impedes on the interpretation of the tumour proliferative fraction since the uptake of the radiotracer is attenuated due to thymidine competition; a display that becomes more evident in tumour models with an elevated thymidine pool43. An inverse correlation of 18F-FLT uptake to plasma thymidine concentration in HCT-116 xenografts was observed by

Zhang et al. 2005 further offering support to this notion. However, the group also noted that mouse models have a 100-fold higher serum thymidine level than humans, so this effect may not be so pronounced in a clinical setting. Other confounding factors such as species dependent glucuronidation, possible efflux of 18F-

FLT from cells, the role of a putative deoxynucleotidase and differences in expression or modulation of ENT-1/2 further impact the rate of 18F-FLT uptake. Furthermore, the biological effect of upstream proteins (i.e. the effect of deregulated p53 signalling on maintaining a rational relationship between TK1 activity and S-phase) and the effect of drugs that induce G2/M-phase block (causing hyperphosphorylation and inactivation of TK1) all impact the level of 18F-FLT uptake.

Furthermore, 18F-FLT is reported to be a poor substrate for TK213,78 and in turn, does not offer information on the mitochondrial-cytoplasmic exchange of nucleotides.

During the clinical imaging and staging of several tumour types, 18F-FLT has been seen to exhibit high uptake in the liver, marrow and renal system thus restricting use in these organs. While it is acknowledged that study designs of preclinical research are less standardized than clinical approaches with factors like (i) type of PET imaging protocol, (ii) method of reconstruction/ PET quantification and (iii) number of

imaging time points, increasing variability during data interpretation, pre-clinical imaging still aids the assessment of pharmacological objectives during the development of targeted therapy (see review by Schelhaas et al. (2017)) 79. As the magnitude of 18F-FLT uptake is context-dependent, its inability (at times) to delineate a true-negative from a false-negative result limits77 its use as an adequate decision- making biomarker during drug development.

1.3. Selection of a suitable radiotracer

To overcome the critical limitation associated with 18F-FLT, its dependence on TK1 activity firstly needs to be eliminated. Imaging downstream of the TMP biosynthetic pathway (convergent point of the de novo and salvage pathways) provides an attractive option for overcoming this dependence by monitoring the tumour proliferative fraction with respect to dTYMK activity. The ideal radiotracer would mimic 18F-FLTMP, be recognised as a monophosphate and, be phosphorylated by dTYMK and subsequently, NDPK. Like 18F-FLT, this radiotracer would not be incorporated into DNA due to the 18F- positioned in place of OH at the 3’ deoxyribose to maintain metabolic stability. However, the phosphate group which gives the radioligand a high overall charge would be the rate-limiting factor in its ability to permeate the cell membrane. Since membrane permeability and avidity to dTYMK are of paramount significance, an investigation into the use of phosphate bioisosteres as adequate phosphate mimics was subsequently reviewed.

1.3.1. Bioisosteres for phosphate mimicry

An aspect of medicinal chemistry has focused on the use of chemical moieties for developing nucleoside/nucleotide analogues whose properties exert antiviral, anticancer, antibacterial and antiparasitic activity. As briefly reviewed in section

1.2.2., nucleoside analogues have had a high success in inhibitor research owing to their ability to permeate the cell membrane. Similar biological achievements for the use of nucleotide analogues, however, are yet to be reported due to the highly charged nature of the phosphate moiety that renders them cell-impermeant. There has been increasing interest in developing biosoteric groups of phospho and pyrophosphate, with similar physicochemical and biological properties, to the lead compound with improved pharmacokinetics. One such mimic is squaric acid (3,4- dihydroxy-3-cyclobutene-1,2-dione) which is formulated as a cyclobutenyl ring with a 2π electrons system that can be stabilised through the delocalisation of bonding electrons in its pi orbital (figure 10). Initial squaramide research showed successful incorporation of the moiety into an array of oligonucleotides by replacing the phosphodiester linkage. This adaptation amplified both the binding and permeability properties of the oligonucleotide thereby, increasing its incorporation into DNA duplexes. Though the modified oligo displayed an increase in its resistance to nuclease-mediated degradation, the authors (Sekine et al.) noted some destabilisation of the overall structure following binding. In later studies, Sato et al.

(2002) and Seio et al. (2005) explored the potential of squaramides as plausible phosphate mimics for thymidine and cytosine deoxynucleoside 5’-phosphate analogues80,81. However, no biological evaluation has been reported despite the authors acclaiming its potential as antiviral and anticancer agents. While other

bioisosteres such as methyl phosphonates, phosphorothioates and boranophosphates have been reported as adequate phosphate mimics, the use of squaramide is of particular interest since its use with thymidine, by Sato et al. (2002), showed great potential and is worth exploring in the context of PET imaging.

Figure 10 (Sato et al. (2002)). Evaluation of the chemical structures of squaric acid (1),

squarates (2), squaryldiamide (3); and corresponding mesomeric effects of squaric acid derivatives (2 and 3).

1.4. Thesis objectives

With the increasing interest on targeting dTYMK as part of adjunct therapy and, progressing knowledge of nucleoside phosphate mimics like squaramide deoxynucleosides and deoxynucleotides as potential anticancer agents, there is a need for the development of an adequate decision-making proliferation biomarker that supersedes 18F-FLT for clinical and pharmaceutical application. A prerequisite for the rational design of such a compound warrant a detailed investigation and understanding of the catalytic nature of dTYMK.

Accordingly, this thesis sets out to explore using a multipronged approach;

 The gene essentiality of dTYMK for DNA synthesis by assessing functionality

following therapeutic and genomic manipulations; both in vitro and in vivo.

 Utilise 18F-FLT and 18F-FDG as well-established tool compounds to evaluate

the overall flux of the nucleoside salvage pathway and changes in glucose

metabolism

 Assess the potential of a novel bioisostere of 18F-FLTMP combined with the

unparallel contrast/sensitivity PET imaging to provide the first in vitro and in

vivo semiquantitative and visual analysis of squaramide-nucleoside

phosphate mimics for reporting the rate of cell proliferation with respect to

dTYMK activity.

Chapter 2

Materials and Methods

2.1. Cell culture

HCT116 (human colon cancer; LGC Standards, Teddington, Middlesex, UK), Caco-2

(human epithelial colorectal adenocarcinoma) HOS and OST TK1- (human osteosarcoma; a gift from Prof. Vera Bianchi, Department of Biology, University of

Padua, Italy) cells were maintained in 10% CO2 and in Dulbecco’s Modified Eagle’s

Medium (DMEM; Sigma, Cat no: DN5546); BT474 (breast carcinoma) and IGROV-1

(ovarian cancer; a gift from Dr Euan Stronach, Department of Surgery and Cancer,

Imperial College London) cells were cultured in Roswell Park Memorial Institute

Medium 1640 (RPMI-1640; Sigma, Cat no: R5886). A549 (Non-small-cell lung carcinoma), H23 (Non-small-cell lung carcinoma) isogenically matched PC9 (erlotinib- sensitive; Non-small-cell lung carcinoma) and PC9ER (erlotinib-resistant; Non-small- cell lung carcinoma), H1975 (Non-small-cell lung carcinoma) presented in Appendix figure 1 and 2, were a kind gift of Dr Olivier Pardo’s laboratory, Imperial College

London. These cells were also maintained in RPMI-1640 (Sigma, Cat no: R5886). Both culture media were supplemented with 10% foetal bovine serum (FBS; First Link, Cat no: 02-00-850), 2% Penicillin-Streptomycin (5,000 U/mL) (Life Technologies, UK) and

1% L-Glutamine (Life Technologies, UK). The cells were maintained in a 5% CO2 humidified incubator at 37˚C. For most assays, cells were grown in 6-well plates at

300,000 cells/ well for the stated cell lines. All cells were grown for 24 h in a volume of 2ml/well unless stated otherwise.

2.2 Western blotting

Cells were lysed in radioimmunoprecipitation assay (RIPA) buffer containing protease and phosphatase inhibitors (all from Sigma-Aldrich). Snap-frozen tumour samples

were grounded in using a mortar and pestle with liquid nitrogen to maintain protein integrity. These were subsequently homogenised in Radioimmunoprecipitation assay

(RIPA) buffer again containing protease and phosphatase inhibitors. Equal amounts of protein (20 µg) were resolved on 4–15% mini-protean TGX gels (Biorad) and transferred to PVDF membranes (Trans-Blot Turbo Transfer Packs, Biorad).

Membranes were subsequently blocked for 1 h in 5% non-fat dry milk in phosphate buffered saline containing 0.1% v/v tween-20 (PBST) and incubated with β- Actin

(Abcam, Cat. Nr.: ab6276), dTYMKAbcam (Abcam, ab15486), dTYMKtAtlas (Atlas antibodies, HPA042719) Thymidine Kinase-1 (Cell signalling, 8960), Thymidylate

Synthase (Cell signalling, 3766) and Nucleoside Diphosphate Kinase (NME1/NDKA

(D98) (New England Biolabs, 3345S), COX IV (New England Biolabs, 4844S) or GAPDH

(Cell Signalling, 5174S) overnight at 4°C. Following incubation, membranes were washed 3 x 15 min in PBST. Secondary HRP-conjugated mouse (Santa Cruz

Biotechnology, sc-2004) and rabbit antibodies (Santa Cruz Biotechnology, sc-2005) were added to membranes and incubated for 1 h at room temperature. Signals were visualised using Amersham ECL Western Blotting Detection Reagent (GE Healthcare) and Amersham Hyperﬁlm (GE Healthcare).

2.3 DNA cell cycle analysis using ﬂow cytometry

Cells were grown in complete media conditions (RPMI supplemented with 10% foetal bovine serum (FBS; First Link, Cat no: 02-00-850), 2% Penicillin-Streptomycin (5,000

U/mL) (Life Technologies, UK) and 1% L-Glutamine (Life Technologies, UK)) and allowed to settle for 24 h. For cell cycle inhibition studies, cells were serum starved or treated with aphidicolin, Nocodazole or 5FU (all from Sigma) for another 24 h.

Supernatant and cells were collected, washed and centrifuged. The formed cell pellet was washed twice in ice cold PBS, centrifuged and resuspended dropwise with ice- cold 70 % ethanol under mild vortex to prevent clumping. Cells were ﬁxed for at least

2 hours at –20°C. Cells were centrifuged and rehydrated in PBS for 15 minutes.

Samples were stained in 500 µL buffer containing 100 mM Tris pH 7.4, 150 mM NaCl,

1 mM CaCl2, 0.5 mM MgCl2, 0.1% Triton X-100, 0.1 mg/ml RNase A, 50 µg/ml propidium iodide (all items from Sigma-Aldrich) at room temperature for 1 h protected from light. Data from 10000 cells per treatment were acquired on a BD

FACSCanto ﬂow cytometer (BD Bioscience) and analysed using FlowJo 7.6.4 software

(Tree Star).

2.4. Synthesis of 18F-SqFLT, 18F-FLT, 18F-D4-FCH and 18F-FDG

18F-SqFLT (figure 1) and cold 19F-SqFLT were synthesised primarily by Dr Diana

Brickute with the help of Chris Barnes and Dr Louis Allot of Imperial College London.

18 18 2 18 F-FLT and F-fluoromethyl-[1,2- H4]choline ( F-D4-FCH) were produced by Chris

Barnes on a FASTlab synthesiser (GE Healthcare, UK). Semipreparative HPLC purified tracers, and radiochemical purity was confirmed by co-elution with reference compounds 18F-FLT, 18F-SqFLT and 18F-D4-FCH. 18F-FDG was supplied by PETNET UK.

Figure 11. Synthetic pathway of 18F-SqFLT. Synthesis was conducted by Dr Diana Brickute with the help of Chris Barnes and Dr Louis Allot of Imperial College

London.

2.5. In vitro uptake of radiotracers

Cells for in vitro uptake were maintained in complete media conditions and plated as previously described in 2.1. 18F-FLT, 18F-FDG, 18F-D4-FCH or 18F-SqFLT was added to each well yielding a final concentration of 0.74 mBq in a volume of 1ml. Cells were subsequently incubated at 37˚C and 5% CO2 for the times required in a New

Brunswick Galaxy 14S incubator (Eppendorf, North America). Following incubation, cells were washed three times with 1X PBS (Oxoid, UK) and lysed on ice for 15 min using 1ml RIPA buffer/well (Sigma, UK). Cell lysates were homogenised and transferred to respective radioactivity counting tubes. The associated radioactivity in each sample was counted using a Cobra II Auto-Gamma Counter (Perkin Elmer,

London, UK). For radio-high performance liquid chromatography (rHPLC), cells were scraped into 2 ml PBS and lysed with 8 ml 100 % acetonitrile. Cells were homogenised with ultraturax before being centrifuged and the lysates collected and prepared for injection into HPLC. Protein concentration in each gamma counted sample was quantified using the Pierce™ BCA protein assay Kit (Thermo Fisher Scientific, UK)

method. Decay corrected counts were normalised to protein concentration in each sample and is presented as percentage Incubated Dose per milligram of cellular protein (%ID/mg protein).

2.6. High-performance liquid chromatography (HPLC)

Radio-high-performance liquid chromatography (HPLC) metabolite analysis was carried out on an Agilent 1100 series HPLC system (Agilent Technologies, Stockport,

UK) equipped with a γ-RAM Model 3 gamma-detector (IN/US Systems Inc., FL, USA) and Laura 3 software (LabLogic, Sheffield, UK). HPLC separations used a Partisil-SAX column (250 X 4.6 mm; 5 µm particles; Hichrom, Reading, UK) or a Phenomenex

Gemini 5 µm particles C18 column (150 x 4.6 mm) for separation of nucleoside or phosphonucleosides. The Partisil-SAX column (an ion exchange column) was eluted with mobile phase-A (10% acetonitrile-water) and a buffer in line B (0.25 M

(NH4)2HPO4, pH 7.5) at a gradient of 100 % mobile phase-A from 0 – 10mins, 5→ 95% from 10 – 16 min, 95% → 5% organic for 2 min with a flow rate of 1mL/min and 1mL injection loop. Phenomenex Gemini 5µ C18 150 x 4.6 mm column was also used with a mobile phase comprising of water (0.1%TFA)/MeCN with a gradient of 5% organic for 1 min, 5→95% in 16 min, 95% organic for 2 min, 95→5% organic in 2 min, delivered at a flow rate of 1 ml/min. Thymidine, TMP, TDP and TTP standards were used to calibrate the ion exchange column. Reagents were purchased from Sigma-

Aldrich or otherwise as indicated.

2.7. Liposome encapsulation

YMU1 was solubilised into 2.954 mM in 100 % DMSO; 10 µl of 2.954 mM YMU1 (max volume of 100 % DMSO permitted per vial) was transferred to liposome lyophilised

powder, which contains: cholesterol, L-α-phosphatidylcholine, and stearylamine. 200

µl of aqueous medium was subsequently added before being vortexed at room temperature. Afterwards, 800 µl of aqueous medium was added before the total 1 ml mixture was subjected to agitation for 30 min. The final concentration of YMU1 was 29.54 µM. The 29.54 µM (liposome infused YMU1) LP-YMU1 was diluted into media for a final concentration of 190 nM or 1.9 µM YMU1. Cells were incubated for

12 h ensuring concentrations of liposome constituent was the same in both control and YMU1 conditions. Refer to table 5 for full detail.

Table 1. Liposome dilution protocol for control and YMU1 treatment of HCT -116 cells

2.8. In vitro uptake to determine the acid insoluble fraction of labelled nucleotides

HCT-116 cells were maintained in complete culture medium and seeded into 150 x

15 mm petri dishes (n=4 plates per condition) at 5 million cells per dish (total volume:

10 ml). Cells were allowed to adhere and grow for 48 h prior to radiotracer uptake.

18F-SqFLT or 18F-FLT was added to each well yielding a final concentration of 3.7 mBq in a volume of 10 ml. Following an hour incubation, cells were washed three times in

ice-cold PBS. Cells were then washed twice with 10ml ice cold 0.5 M perchloric acid.

These washings were kept as the ‘acid soluble’ fraction as was the subsequent wash with 10 ml PBS (total volume of acid soluble fraction = 30 ml). Cells were then lysed with 10 ml of 1M NaOH and collected into as ‘acid insoluble fraction’. The lysates were vortexed, to ensure homogeneity, and a 1ml sample removed and placed into counting tubes for analysis on a Cobra II Auto Gamma Counter (Perkin Elmer, London,

UK). The normalisation of fractions were to total protein.

2.9. Enzymatic assay of dTYMK

Nucleoside conversions of 18F-FLT were performed as per Sigma-Aldrich protocol 31.

A 3.10 ml reaction mix was made with 73 mM triethanolamine, and 0.87 mM TMP,

100 µCi of 18F-FLT, 1.7 mM adenosine 5'-triphosphate, 16 mM magnesium sulphate and 65 mM potassium chloride, and heated to 37˚C for 5 min before 1 mg nucleoside monophosphate kinase was added. The full reaction mixture was incubated at 37˚C for 30 min in an open vial to allow oxidative metabolism. The product was diluted 15

X in 10% acetonitrile-water for HPLC analysis.

2.10. Sulforhodamine B Proliferation Assay

Cells were cultured in completed media conditions and plated at a seeding density of

5000 cells/well in transparent 96 well plates. After 24 h incubation, cells were treated with siRNA or transferred into hypoxia chambers for hypoxia studies or kept under normoxia for 24, 48 or 72 hr. Trichloroacetic acid (TCA) fixing solution, prepared by dissolving 50g TCA (Sigma, UK) in 100ml Milli-Q dH20, was added to produce a final concentration of 10% per well. TCA fixed cells were then incubated at 4˚C for 30 min.

Following incubation, plates were washed four times using tap water and air dried at room temperature. 100μl of 0.4% Sulforhodamine B solution, prepared by dissolving

0.4g Sulforhodamine B powder (Sigma, UK) in 100ml 1% acetic acid glacial (VWR, UK), was added to each well and incubated at room temperature for 30 min. Following incubation, Sulforhodamine B solution was washed out with 1% acetic acid and air- dried. Once dry, samples were solubilised with Tris base, and absorbance was read using a Tecan Sunrise plate reader at 492nm. Absorbance was used as a substitute method of cell density quantification in each well.

2.11. Generating hypoxia conditions

HCT-116, B1 and B5 cells were seeded into 6-well plates in full media conditions for

24 h. Cells were subsequently transferred into Oxoid AnaeroJa 2.5L (Thermo

Scientific, AG0025A) with Oxoid AnaeroGen 2.5L Sachet (Thermo Scientific,

AN0025A) and closed within 60 seconds to prevent sachet saturation. Cells were again incubated for either 24 or 72 h and analysed protein expression determined by western blot or proliferation determined by SRB assay.

2.12.1. Generation of knockdown cell lines with CRISPR/Cas9

Guide RNA (gRNA) sequences for CRISPR/Cas9 were generated from CRISPR design web site (http://crispr.mit.edu/), provided by the Feng Zhang Lab82. HCT-116 cells were co-nucleofected with dTYMK CRISPR gRNAs 25926363

(GCGCGGGGCTCTCATAGTGC) or 25926373 (GCCACCGCGCCGAACTGCTC), and

Streptococcus pyogenes Engen Cas9NLS (SpCas9) programmed with tracrRNA to produce a Cas9 Ribonucleoprotein mix82,83. The two dTYMK gRNAs target the exon 1 of the dTYMK gene. Complementary oligonucleotides to gRNAs of interest were

annealed and cloned into SpC CRISPR/Cas9-Puro vector (Addgene, Cambridge, MA).

After two days, following transfection, cells were treated with 1 μg/ml of puromycin for three days. Two weeks later, colonies were isolated with the cloning cylinders.

Clones were then prepared for DNA sequencing along with western blotting.

2.12.2. Polymerase Chain Reaction (PCR)

PCR reactions were performed using Reddymix PCR Master mix (Thermo Scientific;

UK) set up in the following reaction mix: 12.5μl Reddymix (2xpre-aliquoted

Reddymix™;Thermo ), 50- 100ng of DNA template, 2.5μl of 10μM forward primer

(SA4760DTYMKex1f gaggccgggaaatactagct) 2.5μl of 10μM reverse primer

(SA4761DTYMKex1r cagttcacagcacgccag) and ddH2O to make a final volume of 25μl.

Thermal cycling conditions of 95°C for 2 minutes, 30-35 cycles of 95°C for 1 min, 55°C for 1 min and 72°C for 1 min followed by a final elongation cycle of 72°C for 2 min were selected for PCR. Reactions were placed in 4°C before recovery from the programmable heating block (Applied Biosystems, Cheshire).

A NanoDrop 1000 spectrophotometer (Thermo Scientific) was used to give a direct readout of DNA concentration. Agarose gel electrophoresis permitted the assessment of products by size separation.

2.12.3 Agarose gel electrophoresis

1-2% agarose (Fisher Scientific, UK) gels were produced by boiling in 150ml 1x TBE buffer. 15μl SYBR® Safe DNA gel stain (Invitrogen) was added before allowing the gel to set. Once the gel had hardened, samples along with ladders were loaded onto the gel and gel electrophoresis performed with TBE buffer at 150Mv, on a horizontal gel electrophoresis apparatus (24x 24cm gel HU13 (Thistle Scientific, UK)), for up to 1 h.

MassRuler™ Low Range DNA Ladder (Thermo Scientific) and MassRuler™ High Range

DNA Ladder (Thermo Scientific) were also included during loading to give a size range that allows sample detection. The UVIpro Platinum Gel Doc System ultraviolet transilluminator (UVItec, UK) was employed to identify DNA products qualitatively.

The full-length PCR product size was expected at 336 nucleotides (nt), while CRISPR

25926363 - 25926373 deleted PCR products was expected at 247 nt.

2.12.4 DNA sequencing

Genomic DNA was prepared from resuspended cells using the QIAprep Spin Miniprep

Kit (Qiagen, 27104) as per the manufacturer’s instructions. A total of 10 μl sample mix containing 500 ng plasmid DNA or 10ng per 100 bp PCR product and 3.2 pmole sequencing primer were sent to Beckman Genomics Sequencing service (Beckman

Coulter, High Wycombe, UK), to be externally sequenced.

2.13. siRNA transfections

HCT116 cells were transfected with siRNAs targeting DTYMK (FlexiTube Gene

Solution for 5 nM dTYMK- 4 siRNA; QIAGEN, GS1841) or nontargeting scramble control (SCR; 12.5 nM; Dharmacon, D-001810-10-2). Plated cells were transfected

(reverse transcription) with RNAiMAX (Invitrogen) according to the manufacturer’s instructions.

2.14. Immunofluorescence

HCT-116, B1 and B5 cells at 50,000 cells/well were plated in slide chambers (Nunc

Lab-Tek, Sigma-Aldrich) and incubated for 24 h. Cells were then fixed with 4% formaldehyde in PBS for 15 min at room temperature. Cells were subsequently washed three times (3 x 10 min) with PBS and incubated with blocking and

permeabilisation buffer (PBS + 1% bovine serum albumin (BSA) + 0.1% Triton-X100

(Sigma-Aldrich)) for 1 h at room temperature. Next, samples were incubated overnight at 4°C with rabbit anti-dTYMK Atlas or Abcam antibody (1:100) in blocking buffer (PBS + 1% BSA). Following primary antibody incubation, cells were again washed three times with PBS before being incubated with Alexa Fluor® 488 goat anti- rabbit IgG secondary antibody (1:400; Molecular Probes™, Thermo Fisher Scientific) in blocking buffer for 1 h at room temperature. Due to light sensitivity of fluorophore, chambers were incubated in the dark. Following incubation, cells were washed with

PBS and incubated further with Alexa Fluor® 594 Phalloidin (1:100; Molecular

Probes™) for 20 min at room temperature without light. Cells were washed for the final time and mounted using ProLong® gold antifade mounting reagent with 4’-6- diamidino-2-phenylindole (DAPI) (Life Technologies Ltd., Thermo Fisher Scientific).

Immunofluorescence imaging was performed using a 40X UPlanAPO objective lens on an Olympus BX-51 wide-field microscopy UIS2 optical system (Olympus Life

Science Europa GMBH, Hamburg, Germany) equipped with a DP70 digital camera.

Qualitative images were attained by an Olympus U-RFL-T epifluorescence source and

DPController 1.2.1.108 imaging software (Olympus Optical Co. Ltd, Tokyo, Japan) in the red, blue and green channels.

2.15. Tumour models

Licensed investigators conducted all in vivo experiments to the guides of United

Kingdom Home Office Guidance on the Operation of The Animals (Scientific

Procedures) Act 1986 (HMSO, London, UK, 1990) and within the published guiding principle for the Welfare of Use of Animals in Cancer Research Institute Committee

on Welfare of Animals in Cancer Research 84. Xenografts were established by injecting

HCT-116 or B1 (5x106 cells in phosphate-buffered-saline (PBS)) subcutaneously on the back of the neck of BALB/C athymic nude mice (Charles River UK Ltd.). Tumour dimensions were determined by calliper measurements and volumes calculated by

휋 the ellipsoid formula for estimating tumour mass (푉표푙푢푚푒 (푚푚3) = × 푎 × 푏 × 푐 6 where a, b and c represent 3 orthogonal axes of the tumour). Mice where imaged when tumour volume reached approximately 120 – 140 mm3 (at around 4-6 weeks post-induction for all tumour models).

2.16. Molecular docking calculations

All calculations were performed at Imperial College High-Performance Computing cluster with the help of Dr Karl Thorley who conducted the gaussian optimisation and defined the binding free energy. Ligands were drawn in ChemDraw, imported into

Maestro (version 9.2.112, Schrodinger) and prepared using gaussian 09 software.

The 3D crystallographic structure of the DTYMK was retrieved from the RCSB Protein

Data Bank (http://www.rcsb.org/pdb/), under the accession code 1841. Before the molecular docking, the geometry of the initial structure of 18F-SqFLT was built and optimized by using the Gaussian 09 software85. Geometry optimisation was performed using density functional theory (DFT) at the B3LYP (Becke, three- parameter, Lee-Yang-Parr)/6-31G (d, p) level86,87. For the docking studies, MOE

(Molecular Operating Environment) software was used as an estimator of the binding free energy associated with the enzyme-ligand complex. A Protein Preparation

Wizard was used through the study to remove water molecules, add hydrogens, hydrogen bonding network optimisation and ﬁnally for restrained energy

minimisation. The enzyme-ligand complex was minimized to a gradient of 0.01 kcal/(mol·Å), and hydrogens were added and charges on the enzyme and ligand assigned using the UFF force field. An oniom method (using micro-iterations and quadratic coupled algorithm) was employed to geometrically optimise the system and produce poses of 18F-SqFLT. Docking was evaluated both in the presence and in the absence of ADP and crystal water molecules near the binding site.

2.17. In vivo PET imaging of 18F-SqFLT and 18F-FLT

HCT-116 and B1-tumour bearing mice were imaged when tumours reached approximately 120 - 140 mm3. Dynamic PET imaging was conducted in mice anaesthetised with 2.5% isoflurane/O2 and placed in a thermostatically controlled dedicated small animal Genisys4 PET scanner (SOFIE Biosciences, Culver City, USA).

Mice were subsequently injected with 0.74 MBq of either 18F-SqFLT or 18F-FLT and dynamic PET images acquired in list mode format over 0-60 min to give decay- corrected values of radioactivity accumulation in tissues. Images were sorted into 19- time frames for reconstruction by maximum-likelihood expectation maximisation

(4×15 seconds, 4×60 seconds and 11×300 seconds). The regions of interest (ROIs) for tumours and selected organs were defined using Siemens Inveon Research

Workplace software (Siemens Molecular Imaging Inc., Knoxville, USA). Count densities were averaged obtained for all ROIs at each time point were used to generate time-activity curves (TACs). Tissue associated radioactivity values were subsequently normalised to average whole-body radioactivity at 40-60 min (NUV40-

60 min). The area under the curve (AUC), calculated as the integral of the NUV from 0

to 60 min, were used for comparison. The values generated for each mouse were statistically analysed with GraphPad Prism 5 for Windows, Version 5.01.

2.18. Biodistribution of 18F-SqFLT

Biodistribution studies were conducted in the same cohort of animals soon after dynamic PET imaging (60 min after radioactivity injection). Tissue samples were quickly excised, and its associated radioactivity determined by γ-counting. These values were normalised to sample tissue weight.

2.19. Enzyme Assays for phosphorylation potential

HCT-116 cells were maintained in RPMI full medium and allowed to reach 60 – 70 % confluency. Cells were subsequently washed with ice-cold PBS, collected by scraping and centrifuged at 1500 rpm at 4 ᵒC. Cell pellets were resuspended in 0.5 ml homogenisation buffer containing; 10 mM Tris-HCL pH 7.5, 1 mM EDTA pH 8.0, 5 mM

2-mercaptoethanol and sonicated for 3 x 10 seconds on low speed. Homogenates were centrifuged at 15,000 rpm for 5 min and the supernatant stored on ice. The protein concentration of supernatants were determined by NanoDrop spectrophotometer. Following this, a reaction mixture containing 0.1 mg of cell homogenate, 50 mM Tris-HCL pH 7.5, 5 mM ATP, 5 mM MgCl2 and 3.7 mBq of either

18F-FLT or 18F-SqFLT were incubated for 30 min at 37 ᵒC. Samples were quenched with ice-cold 10 % acetonitrile-water and passed through a Milex 0.2 µm filter (Millipore,

Billerica USA) and injected onto the rHPLC through a 1 ml loop.

2.20. LogD7.4 determination

The partition coefficient (LogD) was determined by Dr Diana Brickute and Dr Louis

Allot. To a 1.5 mL microcentrifuge tube was added PBS pH 7.4 (500 µL) and octanol

(500 µL). The tube was shaken for 30 min, followed by centrifugation (13,000g, 15 min). 18F-SqFLT (1 MBq) in PBS (<5% EtOH) was added and the samples were shaken for 30 min, followed by centrifugation (13,000 g, 15 min). Aliquots from the octanol layer and PBS layer were removed (100 µL) and placed in counting tubes.

Radioactivity was measured using a γ-counter, and the partition coefficient was calculated using LogD7.4 = log10[Oct/PBS]. The LogD7.4 value was reported as Mean ±

SD (n = 3, triplicate analysis).

2.21. Metabolite analysis using Hydrophilic Interaction Liquid Chromatography

High-Resolution Mass Spectrometry

HCT-116, B1 and B5 cells were seeded into 15 cm plates to a 60 – 70 % confluency for 24 h. Media were transferred into 50 ml falcon tubes, centrifuged and placed into

– 80 ᵒC for future analysis of spent metabolites. Cells were immediately washed twice with Ringer’s buffer (¼ strength tablets (96724, Sigma) in 500 ml dH20) before being quenched with ice-cold 80 % methanol. 10 µl of internal standard mix per 1x106 cells were added to plates and incubated at – 20 ᵒC for 20 min. Supernatants were subsequently scraped into corresponding Eppendorf tubes. Ice-cold, 100 % methanol, was further added to wash the plates and washes were pooled into corresponding Eppendorf tubes. Pooled supernatants were vortexed and centrifuged to remove debris and produce clean/clear samples. Supernatants were transferred

into corresponding LC-MS vials and evaporated under nitrogen gas. Dried samples were processed by Dr Alexandros Siskos and Dr Eirini Kouloura of Imperial College

London for Hydrophilic Interaction Liquid Chromatography High-Resolution Mass

Spectrometry (HRMS) analysis. Cell pellets were reconstituted in 75 µl of an aqueous internal standard (IS) mix. The IS mix consists of 10 µg/mL of citric-2,2,4,4,-d4 acid, L-

13 13 13 aspartic acid- C4, L-glutamic acid- C5, L-malic acid- C4 and 25 µg/mL of lactic acid-

13 C3. The sample was then split into two aliquots of 37.5 µl. To the first aliquot 337.5

µl of AcCN were added so that the final reconstitution solvent was AcCN/H2O 9/1. To the second aliquot, 150 µl of H2O was added resulting in a fully aqueous reconstitution solvent. The two aliquots of the samples were then analysed with

UPLC-MS/MS.

Hydrophilic interaction liquid chromatography (HILIC) chromatographic separation of the cell extracts was conducted using an Agilent 1290 UPLC system on an ACQUITY

UPLC BEH Amide Column, 1.7 µm, 3 mm X 150 mm (Waters). Binary gradient elution was used, comprising of mobile phase A (acetonitrile + 10mM NH4OH) and mobile phase B (water + 20 mM ammonium acetate, 10 mM NH4OH). Mass spectrometric analysis of the chromatographic eluent was performed using a triple quadrupole

SCIEX QTRAP 4000 spectrometer in the negative electrospray ionisation mode. The column temperature was kept at 50 ᵒC, and injection volume was 5 µL.

Peak picking and peak integration were carried out using the mass spectrometer manufacturer’s software (SCIEX, Analyst 1.6.2). The metabolites were quantified as

13 15 AUC of the analyte over AUC of internal standard, whereas the IS was TMP C10, N2

13 for the metabolites detected using the aqueous sample aliquot and lactic acid C3 for the AcCN/H2O 9/1 aliquot.

2.22. Caco-2 and PAMP assay

Caco-2 and PAMPA permeability assay were conducted by the Institute of Cancer

Research. Briefly, in the PAMPA studies, drug solutions at different pH’s were dosed in each well and incubated. The artificial membrane was prepared and experiments conducted as per manufacturers instructions.

Caco-2 cells were maintained in full culture medium, seeded on polycarbonate filter inserts and cultivated in DMEM fully supplemented media containing; 10% FCS, 2 mM l-glutamine, 1% non-essential amino acids, penicillin (100 units/mL), and streptomycin (100 μg/mL). The effect of verapamil at pH 4.8–8.0 were also measured in this study.

2.23. Data Analysis

All data analyses were carried out using GraphPad Prism 6.0, Microsoft Windows 8.1 and Microsoft Excel 2014. Data are stated as Mean ± SEM (standard error of the mean) unless otherwise stated. Unpaired t-test, one- and two-ANOVA analysis were carried out where appropriate with a post-hoc analysis using Bonferroni multiple comparison test.

Chapter 3

Evaluating the essentiality of dTYMK during cell proliferation

3.1. Introduction to dTYMK expression and functionality

It is known that an aberration in the functional integrity of the cell cycle regulatory systems can induce uncontrolled tumour proliferation1. Over the last 60 years, nucleoside inhibitor research for anticancer therapy has focused on early S-phase and in particular, thymidine synthesis88. More recently, interests have been on, amongst others, targeting dTYMK as part of an adjunct therapy since a disruption in its functionality suppresses the supply of dTTP and promotes deoxyuridine triphosphate (dUTP) misincorporation during DNA repair31,32. Hu et al. (2012) provided evidence for YMU1, an ATP-competitive inhibitor of dTYMK, as a promising agent for chemosensitising tumour cells to sub-lethal doses of doxorubicin. YMU1 is reported to hinder the functionality of the Mg2+ ion pointing toward the catalytic

Asp15 residue in the ATP pocket of dTYMK and turn, inhibit dTYMK catalytic efficiency9. Intriguingly, their study presented observations of YMU1 having little or no impact on cell viability without the synergistic effects of doxorubicin. Given its integral role in dTDP production, it remains challenging to elucidate why cells can endure dTYMK inhibition in the absence of extrinsic stresses. The authors suggested multiple dTYMK isoforms as a plausible cause of YMU1’s lack of potency; however, further genetic analysis was not performed. A high dTYMK expression has been associated with the aggressive progression, and poor prognosis of NSCLC in males and in turn highlights the potential of the enzyme as a prognostic marker and a novel molecular target for not only NSCLC treatment but other cancers as well. From an imaging prospective, the development of a dTYMK targeted radiotracer as a better reflection of the tumour proliferative fraction would be advantageous in understanding the pharmacodynamics of these emerging dTYMK targeted therapies.

Thus, it became of interest to explore the essentiality of dTYMK for DNA synthesis and monitor its oscillation under therapeutic and genetic manipulation and, evaluate its potential as an imaging marker of proliferation. Furthermore, we aimed to develop a suitable cell line as a negative control for assessing dTYMK targeted tracer avidity. While 18F-FLT is a well-established compound for investigating the S-phase fraction; it is important to emphasise that 18F-FLT only traces cell proliferation with respect to TK1 activity (section 1.2.4). This needs to be taken into consideration when interpreting results pertaining to TS proficient tumour types. To date, 18F-FLT is yet to be succeeded as the universal marker of proliferation which opens up the need for the development of TK1 independent bioisosteres of TdR for the apprehension of 18F-

FLT limitations. In the meantime, 18F-FLT will be employed as a compound for nucleoside assessment.

3.2. Bioinformatics analysis of genetic dependency to dTYMK

To determine whether a change of dTYMK gene copy number gives rise to genetic vulnerability and identify an appropriate cell panel for in vitro examination, we undertook a detailed investigation of gene expression, copy-number changes, mutations and mRNA expression data available through the Cancer Cell Line

Encyclopaedia (CCLE). In figure 12A, the R Pearson value of 0.49 suggests that dTYMK

DNA copy number variations were not responsible for the increased mRNA expression seen in cancer cells (figure 12A). It is possible that intrinsic regulatory mechanisms may be the driving factor of enzyme expression as opposed to genetic disparity since a negative correlation between enzyme expression and copy number was not observed. Through analysis of the CCLE database, no dTYMK mutations could be identified although, common single nucleotide polymorphisms were excluded from the analysis. Cancers of the bone marrow and glioma showed less sensitivity/dependency to dTYMK knockdown (figure 12B) while cancers of the colon, breast, prostate and pancreas appear to show sensitivity. Of the twenty-three CCLE box plotted cancer types (figure 12B), we chose cancer cell lineages from ovarian

(IGROV-1), breast (BT474), colon (HCT-116) and osteosarcoma (HOS and OST TK1-) with mean essentiality scores of μ = - 0.22, -0.14, 0.05 and -0.14 as appropriate cell models to derive a quantitative understanding of the role of dTYMK in nucleoside metabolism. Moreover, HCT-116, BT474 and IGROV-1 cells have been well characterised and known to produce tumour xenografts in mice89–91. For a more systematic evaluation of gene function and candidate gene enrichments, we utilised the CERES dataset (https://depmap.org/ceres/) together with Project Achilles

(https://portals.broadinstitute.org/achilles). CERES uses a computational method of

estimating gene-dependency through the CRISPR–Cas9 system and measures depletion as a sum of unknown gene knockout and copy-number effects. Project

Achilles sets out to identify genetic vulnerabilities that may have profound effects on cell viability in 216 cancer cell lines by interrogating 11 000 genes, using a lentivirally delivered pool (~54 000) of short hairpin RNA (shRNA). Cells were transduced with

ﬁve shRNAs per gene, followed by next-generation sequencing to allow for the quantification of gene enrichment/depletion of each targeted shRNA by, comparing the shRNA population at the end of the screen to the initial reference pool. Coupling the CERES dataset to the Project Achilles shRNA dataset, potentially increases precision in detecting dTYMK gene essentiality and candidate gene enrichment.

Figure 13A and 13B evaluated copy number loss vs neutral in both shRNA and CRISPR datasets. The plots suggest a genetic dependency of tumour cells to dTYMK with the one-tailed t-test flanking the screening scores as significant (p= 0.00610 CERES and

0.0397 shRNA). We speculated that tolerance of dTYMK knockdown shown in the

CCLE dataset was a consequence of compensatory candidate genes because of its low impact on reducing cell survival. The two screens were overlapped, and p-value adjusted to multiple t-tests (FDR of 0.05) for the identification of the plausible genes compensating for dTYMK knockdown. Twenty-three negatively correlated genes were identified as statistically significant (P<0.05). A basic local alignment search tool

(BLAST) of the identified genes showed no obvious correlation to the DNA thymidine synthesis pathway, but highlighted lipid formation and transcription factors amongst others. Nonetheless, these candidate genes are likely to be biologically meaningful

(figure 13C) and would be worthy of future validations in vitro.

8 R2 = 0.49 7

2 mRNA mRNA expression (RNAseq): DTYMK

1 -1 -0.5 0 0.5 1 1.5 DNA copy number

Figure 12. The Cell Line Encyclopaedia (CCLE) gene expression data of dTYMK. A. dTYMK mRNA expression versus DNA copy number generated from CCLE database. B. CCLE data detailing essentiality score of dTYMK shRNA knockdown across specified cancer cell lines. The datasets to the left (more positive scores) are less sensitive/dependant whilst the right (more negative scores) appear to be dependent.

A B

C D

Figure 13. Gene expression boxplots of copy number vs CRISPR/shRNA enrichment scores where neutral refers to ≥ 2 copy numbers and loss = ≤ 1 copy number. A. CERES score of gene dependency across copy number levels. B. shRNA score of gene dependency as per Project Achilles. C. A volcano plot showing FDR in -log10 ratio of the whole genome database with respect to dTYMK. Each gene across cell lines is correlated with its copy number measurements, before and after CERES correction. The distribution of Pearson correlation coefficients is shown for all genes with the negative correlation highlighted in blue and positive in red. Significance where p < 0.05 is indicated by values above the dotted line. D. Corresponding table to volcano graph detailing significant genes that are negatively correlated to dTYMK. Expression levels for 20532 genes in 433 code read cases (RNA Seq V2 RSEM)

3.3.1 Cell line characterisation Our first experiments aimed to characterise the expression of dTYMK and compare this to TK1 and TS, the latter being the enzyme responsible for the methylation of uracil monophosphate (dUMP) to TMP. We investigated HCT116, BT474, IGROV-1,

HOS and OST TK1- (TK1-deficient human osteosarcoma cell line) as a cell panel for studying the mechanisms of DNA synthesis in both the de novo and salvage pathway.

From the literature, HCT-116, BT474 and IGROV-1 have been shown to express high levels of TK114,92. Correspondingly, our HCT116, BT474 and IGROV-1 also showed high expressions of TK1 with values of 1.296 ± 0.074, 1.220 ± 0.086 and 0.745 ± 0.022, respectively (figure 14B). OST TK1- and HOS were selected as differential cell models to assess the contributions of de novo vs salvage pathway in DNA synthesis. As expected, OST TK1- cells expressed at least a 3-fold higher level of TS compared to

HCT-116, BT474 and IGROV-1, with HOS cells expressing higher levels of both TK1 and

TS. However, dTYMK was highly and uniformly expressed across all cell lines; further suggesting its high demand during DNA synthesis.

Flow cytometric analysis permitted quantification of the S phase fraction by staining with DNA intercalating agent 7-aminoactinomycin D (7-AAD). Under the culture conditions used, the S phase fractions were defined as; 24 %, 21 %, 23 %, 26 % and

21% in OST TK1-, HOS, HCT -116, BT474 and IGROV-1 respectively. We then attempted to correlate these fractions with the PET marker of cell proliferation, 18F-

FLT. As anticipated, OST TK1- cells presented little or no 18F-FLT uptake while the other four TK1 high/salvage proficient cell lines exhibited uptake of 20 to 40-fold difference (figure 14D).

A OST TK1- HOS HCT-116 BT474 IGROV-1

TK1 26 kDa

TS 30 kDa dTYMK 24 kDa β-actin 42 kDa

B C

D E

5 0

) n

i 4 0

e e

t p

3 0

L F

m 2 0

I 1

% 1 0

(

0 - S 6 4 1 1 1 - O 7 V K H -1 4 T T T O T C B R S G O H I

Figure 14. Analysis of TK1, TS and dTYMK expression along with the determination of S phase fraction in the indicated cancer cell lines. For all experiments, cells were maintained at a 60-70% confluence level. A Representative western blot showing TK1, TS and dTYMK expression in the five cell lines tested. B. Quantification of TK1 expression normalized to β actin (n=3). C. Cell cycle distribution graph of the five analysed cell lines 18 visualised at an Excitation⁄Emission (nm): 546⁄647 wavelength on FACsCanto. D. F-FLT uptake following 1 h incubation with 0. 74 MBq. The graph represents the average values

(n=6) and the standard error of the mean. E. Table showing S phase distribution of cells stained with 7-AAD (total DNA stain)

3.3.2 Modulation of dTYMK during cell cycle inhibition

Given the lack of variation of dTYMK expression between cell lines (figure 14A), we investigated whether modulation occurs with well-established cell cycle arresting agents. For synchronisation in specific phases of the cell cycle, Sala et al (2014)92 suggested treatment concentrations of 3.38 µg/ml (10 µM) Aphidicolin (APH), 100

µg/ml 5- fluorouracil (5FU), 0.5 µg/ml (1.7 µM) nocodazole or serum-free medium

(0% FBS) for 24 hours92,93. Synchronisation to G0/G1 was achieved by replacing full growth media with growth media containing 0 % FBS. Serum-free conditions cause a reduction in the basal activity of cells due to the exemption of serum growth factors and cytokines that are essential for the activation of the signal pathways related to survival and proliferation94. APH permitted an arrest at the S-phase border by reversibly inhibiting DNA polymerase α and δ23,95 whereas 5FU exerted its anticancer effects by inhibiting de novo thymidylate synthase (TS), amongst others, and encouraging mis-incorporation of its metabolites (FdUTP) into RNA and DNA6.

Treatment with nocodazole permitted G2/M phase arrest through the inhibition of spindle formation causing a cessation of cell division92. Of the five cell lines previously mentioned, HCT-116 and BT474 cells were selected for their reported dose response92,93 to APH, 5FU and NOC following 24 h treatment. Their DNA content was then assessed by FACs to confirm effective phase synchronisation. Indeed, 5FU and APH successfully induced late G1/early S-phase accumulation in both salvage proficient cell lines. In the representative set of data (Table 2), the proportions of

HCT-116 and BT474 cells in the G1/S phases increased from 70 and 68 % (control) to

97, and 96 % under 100 µg/ml 5FU. A similar increase from 70 and 68 % (control) to

96, and 95 % with 10 µM APH treatment can also be seen in both cell lines. Lastly, 1.7

µM NOC caused an increased in control G2/M phase from 30 and 32 % to 93 and 70

% in HCT-116 and BT474 respectively.

Table 2. Cell cycle distribution of HCT-116 cells after induction of cell cycle arrest with indicated drug concentrations for 24 h Cell line TREATMENT G0/G1 S phase G2/M

HCT-116 Control 48, 52, 51 (%) 22, 21, 18 (%) 30, 27, 31 (%) DMSO (0.33 %) 44, 46, 49(%) 30, 28, 27, (%) 26, 26, 24 (%)

Serum starvation (S.S) 68, 71, 74 (%) 13, 11, 11 (%) 19, 18, 15 (%) 5FU 33, 28, 26 (%) 61, 63, 66(%) 06, 09, 08 (%)

APH 34, 33, 30 (%) 63, 61, 61 (%) 03, 06, 09 (%) NOC 02, 05, 03(%) 05, 06, 05 (%) 93, 89, 92 (%)

BT474 Control 44, 43, 41 (%) 24, 22, 25(%) 32, 35, 34 (%) DMSO (0.33 %) 41, 41, 45 (%) 22, 23, 20 (%) 37, 36, 35 (%)

Serum starvation (S.S) 76, 73, 72 (%) 17, 16, 19 (%) 07, 11, 09 (%) 5FU 35, 35, 38 (%) 61, 59, 55 (%) 04, 06, 07 (%)

APH 31, 29, 29 (%) 64, 65, 69 (%) 05, 06, 02 (%) NOC 22, 18, 19 (%) 08, 09, 13 (%) 70, 73, 68 (%)

Figure 15. Cell cycle distribution of HCT-116 and BT474 cells after treatment with the indicated cell cycle inhibitors for 24h. A. Representative graph of cell cycle distribution of HCT-116 and BT474 DNA content. PI, propidium iodide. Table 2. representing the percentages of cells in each phase of the cell cycle. Replicates represent n=3 independent repeats.

Following the confirmation of phase synchrony, lysates were prepared from treated

HCT-116 and BT474 cell lines to assess TK1, TS and dTYMK expression by western blot. Unexpectedly, the target enzyme dTYMK (anticipated to oscillate) remained evenly expressed in both cell lines (figure 16). However, it was noted that the dTYMK antibody produced by ATLAS (dTYMKATLAS) presented high background and multiple non-specific bands that made it difficult to truly interpret the expression profile of dTYMK. These additional bands did not correspond with molecular weights associated with multimers or post-translational modifications that are usually associated with multiple bands. Due to this unresolvable issue, the dTYMKATLAS antibody was abandoned.

Figure 16. Expression profile of essential enzymes of the thymidine nucleoside pathway following drug-induced cell cycle arrest of HCT-116 and BT474 cells. A. Representative western blot showing expression of dTYMK, by ATLAS antibody, along with TK1 and TS.

Experiments were repeated three times and the best of the three blots is presented in the above.

A new antibody (Abcam, ab15486) raised against recombinant full-length dTYMK was retested in cell cycle arrested cells. As expected, dTYMK expression oscillated during periodic density fluctuation (figure 17). Under serum starvation (G1 enriched),

dTYMK expression reduced to 28% and 32% in HCT-116 and BT474 cells with respect to control. A similar expression profile was seen in TK1 which, according to literature, is reduced in G1 since DNA replication is no longer required92,96. Thus, TK1 enzymatic activity is down-regulated to decrease the production of thymidine monophosphates and prevent thymidine induced cytotoxicity92. On this notion, dTYMK expression is speculated to have also been reduced as a consequence of the low requirement of dNTPs for the DNA pool.

Aphidicolin and 5FU-treated cells (S-phase enriched) presented a reduction in dTYMK expression (APH - 67% and - 57%; 5FU - 60% and -36 %) in both HCT-116 and BT474 cells. Our observations were dissimilar to that demonstrated for human dTYMK97, and to yeast dTYMK98 which shows a peak expression of dTYMK during S-phase.

While these studies bring to light the requirement of dTYMK for progression through the S-phase of the cell cycle; they do not consider the feedback mechanisms that may regulate dTYMK expression in response to genotoxic insults. Centrifugal elutriation was the preferred method of synchrony for many of these studies (i.e. Huang et al.

(1994)22 over pharmacological manipulation which can perturb biological systems, disrupt the metabolic state and induce cell death94. For our application, it was important to understand the clinical relevance of these genotoxic insults on the cellular enzymes responsible for nucleotide biosynthesis.

Assuming that the gene function of dTYMK is required to sustain progression through the S-phase of the cell cycle, the level and rate of dTYMK catalytic activity should be intimately tied to that of cell proliferation. We postulate the reduction of dTYMK expression in the presence of APH and 5FU to be a consequence of decreased cell

proliferation due to replication collapse and double-strand break formation99,100. The salvage enzyme TK1 increased in aphidicolin-treated HCT-116 (+ 44 %) and BT474 cells (+ 36 %). This is consistent with literature reports of TK1 hyper-phosphorylation and reduced degradation92. In nocodazole-treated HCT-116 and BT474 cells (G2/M phase enriched), a reduction of dTYMK expression was also detected; albeit to a lesser extent. This effect could be explained by the studies of Coppock and Pardee

1987101, amongst others102,103, who reported a drop in cellular dTTP, during the G2/M phase, to a low level until the mid-G1 phase of the next cell cycle. In contrast, an increase of TK1 expression could be seen in NOC -treated HCT-116 cells. Because the microtubule-depolymerising drug only arrests cells in prometaphase and prevent mitotic exit, the anaphase-promoting complex/cyclostome (APC/C) mediated proteolysis of Ser13 phosphorylated TK1 is inihibited104. TK1 proteolysis is not instigated leading to an accumulation of the TK1 protein without any catalytic functionality104.

An increase of TS expression was observed in 5FU-treated HCT-116 and BT474 cells while a G1 arrest resulted in a decrease. This is in line with Marvetti et al. 2009105 and

Ligabue et al. 2012106 who have shown that 5FU elevates the steady-state expression level of TS in tissues and cells following its administration. In addition, the group noted a doubling of TS half-life to almost 15 h in 5FU-treated cells and concluded the increase to be due to the translation of TS mRNA synthesised before the addition of

5FU and/or increased stability of the ternary complex105,106. Unsurprisingly, nucleoside diphosphate kinase (NDPK) oscillates in the same manner as dTYMK across all conditions perhaps due to similar pressures resulting from the need for or

otherwise of nucleotides for DNA synthesis. NDPK is thus dependent on dTYMK for the TDP pool required for the exchange of terminal phosphate between TDP and TTP.

Figure 17. Expression profile of the essential enzymes in the thymidine nucleoside

pathway. Representative western blot showing expression of stated proteins in HCT116 and BT474 cells.

Altogether, dTYMK expression fluctuates under serum starvation (G1 enriched), 5FU and APH (S phase enriched) and nocodazole (G2/M enriched) treatment. The oscillation appears to be in accordance with a decrease in cell proliferation owing to, the genotoxic insults exerted by the chosen cell cycle arresting agents. However, knowledge from literature and our previous data suggest that a differential expression of protein does not always correlate with activity. To sustain our conclusion, we initially attempted to correlate these modulations with nucleoside uptake using 18F-FLT.

3.3.3 18F-FLT uptake in cell cycle arrested cells

Having provided evidence of nucleoside salvage pathway modulation across the phases of cell cycle, we conducted incorporation experiments with 18F-FLT to assess the salvage pathway activity with respect to TK1. G0/G1 arrest in HCT-116 and BT474 resulted in an 85% (p<0.0001) and 74 % decrease (p<0.0001), respectively, of 18F-FLT uptake, when compared to DMSO control (Table 3). This corresponds with the reduced TK1 expression seen in figure 17 since dTMP production is no longer required. S phase arrest (while inducing a high TK1 expression) resulted in a decrease of 18F-FLT uptake by - 58 % (p<0.0001) and - 53 % (p<0.0001) in APH treated HCT-116 and BT474. This supports the findings of Sala et al. (2014) on reduced TK1 activity due to the presence of both non-phosphorylated and phosphorylated forms of the protein. The accumulation of dTTP, due to DNA polymerase inhibition, may have resulted in a feedback mechanism that modulates TK1 activity thereby preventing genetic instability through nucleotide imbalance. An introduction of 100 µg/ml 5FU treatment also presented a - 53% (p<0.0001) and - 51 % (p<0.0001) decrease in HCT-

116 and BT474. However, a 2.5-fold increase of 18F-FLT uptake (flare response) was noted using the same concentration after 2 h treatment (appendix, figure 7). The decreased 18F-FLT accumulation under 24 h 5FU conflicts with the majority of literature reports on 5FU administration. It is possible that the high 100 µg/ml 5FU concentration, at least 5-fold higher than that reported in the literature107,108, may have enhanced in vitro cytotoxicity by metabolite 5-fluoro-dUTP (dFdUTP) incorporation into DNA. A higher cell kill results in reduced dTNPs flux and therefore reduced TK1 activity (conversion of thymidine to its phosphate derivative)109.

Furthermore, our results concur with Barthel et al. (2005) gamma counter

measurements ex vivo where the group noted a parallel reduction in both ATP and

18F-FLT following a 24 h and 48 h 5FU administration69. Nocodazole treatment resulted in a decrease of 18F-FLT trapping since TK1 activity is no longer required post-

DNA synthesis (Table 3).

Next, glucose metabolism was assessed by 18F-FDG (a gold standard radiotracer) to briefly determine whether a change in nucleotide metabolism is associated with loss of cell viability and/or glucose metabolism often determined in radiotracer studies.

Exposure to 0 % FBS, 5FU and NOC resulted in a decrease of 18F-FDG uptake, with the greatest reduction (- 76%: p<0.001) occurring under serum starvation in HCT-116 cells. A similar reduction in 18F-FDG uptake was also noted in BT474 cells under 0 %

FBS (-61 %: p < 0.001) and (-51 %: p< 0.001). However, there was little difference between control and NOC. Exposure to APH for 24 h, had little effect on 18F-FDG uptake (p=0.0933) which is in contrast to that noted under 5FU (S-phased enriched) treatment. Although the experiment was repeated 3 x independently, it was later suspected that APH activity was found to have been lost in 18F-FDG experiments and so data have been excluded.

Table 3. Percentages of 18F-FLT cell uptake in HCT-116 after the indicated treatments for 24h.

Treatment % Uptake % Uptake HCT-116 BT474 Control 96 103 DMSO 100 100 S.S (0%FBS) 15 26 5-FU 47 49 APH 42 47 NOC 26 49 % Relative to DMSO

Figure 18. 18F-FLT cell uptake with cell cycle arresting agents. A. HCT116 and BT474 cells were seeded to 60-70% confluency and incubated with 0% FBS medium, 100 μg/ml 5- fluorouracil (5-FU), 10 μM aphidicolin (APH), 0.5 μg/ml nocodazole (NOC) for 24 h 0.74 MBq of either 18F-FLT or 18F-FDG was co-incubated with cells for 1h before being washed

thrice and gamma counted. Table 3. % uptake of 18F-FLT normalised to DMSO control of respective cell lines.

The differential expression dTYMK, exerted by specified cell cycle arresting agents, was subsequently correlated to the accumulation of 18F-FLT in salvage proficient HCT-

116 cells (figure 19). Under the therapeutic stress of 0% FBS, 5FU and APH, a high correlation (p = 0.0016) between 18F-FLT uptake and dTYMK protein levels were

evident, although the transient effect of nocodazole was excluded from the statistical analysis.

To conclude, we have shown, in advance of exploiting dTYMK as an imaging target, that an alteration in the nucleotide biosynthesis pathway modulates dTYMK expression, however, this phenotypic response was indirectly established by targeting several enzymes, i.e. TS by 5FU, involved in the cross-talk interactions of the salvage and de novo nucleotide synthesis. To avert this context-dependent regulation of dTYMK, we explored the use of YMU1 as a direct inhibitor of dTYMK activity.

Figure 19. Correlation graph of dTYMK expression over 18F-FLT cell uptake following HCT-116 cell cycle arrest. Cells were treated with indicated conditions for 24 h before

being subjected to either western blot analysis or radiotracer uptake. A graph of expression against activity (with respect to TK1 substrate 18F-FLT) was derived to see if a 3.4. dTYMK inhibition with YMU1 correlation could be observed.

Hu et al. 2012 proposed the use of YMU1 (a non-toxic, potent, reversible, and ATP- competitive inhibitor of dTYMK) as a mild chemosensitisation regimen that sensitises tumour cells to sublethal doses of doxorubicin110. YMU1 was anticipated to adversely increase the dUTP:dTTP ratio causing genome toxicity or cytotoxicity via increased

dUTP mis-incorporation into DNA. Unexpectedly, the authors noted that YMU1 when administered singularly, had little or no effect on cell proliferation110. It then became of interest to assess the degree to which YMU1 could inhibit dTYMK and whether the resulting inhibition of the enzyme’s catalytic activity could modulate the nucleotide flux with respect to 18F-FLT uptake.

Initially, the effect of YMU1 on the proliferation of cells employed in this thesis was analysed, by incubating HCT-116, BT474 and IGROV-1 cells with 0.0001 to 16.5 µm

YMU1 for 72h (figure 20A). The cells were fixed, SRB stained and cell density determined at 510 nm using a microplate reader. Consistent with Hu et al. ’s. (2012) findings, YMU1 exerted very little or no effect on cell growth (figure 20A). Introducing

0.1 µM doxorubicin in combination with YMU1 indeed reduced cell viability (relative to untreated control) by 70 %, 77 % and 64 % in HCT-116, BT474 and IGROV-1 respectively (figure 20C). A similar phenotypic response, albeit to a lesser extent, was also observed under the influence of doxorubicin alone (figure 20C). We can infer that cytotoxicity was predominantly induced by doxorubicin.

B C

Figure 20. Cytotoxic effects YMU1 on specified cell lines. A. HCT-116, BT474 and IGROV-

3 1 cells were plated onto 96-well plates at 5x10 cells/well. After overnight culture, the cells were treated with vehicle (0.625 % DMSO), 1 nM, 10 nM, 100 nM, 1µM, 10 µM and 16.5 µM of YMU1. After 72 h, cell viability was determined by Sulforhodamine B colorimetric assay for cytotoxicity screening. Error bars represent SD (n= 6 x 3 independent repeats). B. Structure of YMU1. C. % Cell viability of YMU1 + dox combination treatment. Error bars represent SEM (n= 6 x 3 independent repeats).

3.4.1. Modulation of the Salvage pathway with YMU1

Following the lack of significant cytotoxicity above, we assessed whether an inhibition of dTYMK by YMU1 changes the intracellular conversion of 18F-FLTMP to

18F-FLTDP. For this, cells were treated for 24 h with 0.625 % DMSO, 1 µM or 10 µM and then co-incubated with 18F-FLT for 1h. Cell extracts were then prepared (as described in methods and materials) and injected into a SAX ion exchange column.

Chromatograms of metabolites revealed very little/non-significant differences in the

18F- FLT-TMP: 18F-FLT-TDP/FLT-TTP ratio under control and YMU1 (figure 21) treated conditions.

Given the essential role of dTYMK in dTDP production, it was puzzling as to why cells can tolerate YMU1-induced dTYMK inhibition in the absence of extrinsic stressors. As an IC50 of 0.61 µM has been biochemically determined by Hu et al. (2012), we hypothesised that perhaps sufficient YMU1 did not transit the cell membrane. The next set of experiments aimed to delineate the rate-limiting factor of YMU1’s potency in cells in an attempt to obtain conditions for more selective dTYMK inhibition.

Control (0.33 % DMSO) 18F-FLTMP

A 500

400

18F-FLTTP 300

 200 18 18F-FLT F-FLTDP 100

00:00 05:00 10:00 15:00 20:00 25:00 30:00 Time (min)

B C

1 µM YMU1 10 µM YMU1 18 18F-FLTMP F-FLTMP

350 350

300 300 18 250 250 F-FLTTP 18F-FLTTP 200 200   18 150 18F-FLT 150 F-FLT 18F-FLTDP 18F-FLDTP 100 100

50 50

0 0

-50 -50 00:00 05:00 10:00 15:00 20:00 25:00 30:00 00:00 05:00 10:00 15:00 20:00 25:00 30:00 Time (min) Time (min)

Figure 21. Metabolite assessment of 18F-FLT uptake in YMU1 treated HCT-116 cells.

Representative chromatograms of HCT-116 cells treated with A. control DMSO (0.33 %), B. 1 µM YMU1 or C. 10 µM YMU1 for 24 h prior to 18F-FLT uptake. Cells were scraped into # 4:1 Acetonitrile-water and processed for phosphate species determination via an ion

exchange column SAX column.

3.4.1. Investigating the effect of efflux transporters on YMU1

YMU1 presents properties that are likely to be a substrate for the ABC transporter, presenting 9 heteroatoms - N + O = 8; + 1 S – (N + O > 8), polar surface area (PSA) of

82 A2 (PSA >85A2) and a calculated lipophilicity (LogP) value of 1.45 (between 1 and

5)111. In lieu of this, we investigated the possible effects of ABC transporters and whether an inhibition of this efflux can lead to an increase in cellular sensitivity to

YMU1. HCT-116 cells were incubated with either 0.1 µM, 1 µM, µM YMU1 and/or 50

µM Verapamil for 24, 48 and 72 h. Verapamil hydrochloride was so chosen as a highly reviewed non-selective ABC-transporter inhibitor at concentrations of 50 µM – 100

µM112. As it can be appreciated in figure 22, a statistically significant decrease in cell viability was observed at 24 h for 0.1 µM + 50 µM (54 %: p < 0.001), 1 µM + 50 µM

(54 %: p < 0.001) and 10 µM + 50 µM (46 %: p < 0.001) when compared to their respective YMU1 controls. However, cells treated with 50 µM verapamil alone also showed reduced cell viability to 52 % (p < 0.001). From these results, it can be concluded that verapamil was the predominant (if not the sole) source of inhibiting cell proliferation. It is known that verapamil exerts its pharmacological stress by hindering the L-type Ca2+ channel. At concentrations of 0.02 – 1 mM, verapamil has been shown to alter Na+ /Ca2+-exchanger transport activity which drastically affects several processes that are essential for cell survival. This offers a plausible explanation as to why cellular viability was dramatically affected since the threshold between toxic and non-toxic concentrations may have been breached in HCT-116 cells. Furthermore, verapamil has previously been shown by Jensen et al. 1995113, amongst others114,115, to perturb cell proliferation but, the mechanism of action appear to be cell-specific. As the combination treatment of verapamil with the

highest concentration of YMU1 exerted minimal cytotoxicity (p>0.05), the

optimisation of verapamil for this assay was circumvented. In the meantime, we

sought to investigate the potency of YMU1 by extracting crude lysates from HCT-116

and co-incubating them with 18F-FLT under fold concentrations of YMU1. After a 30

min incubation, samples were quenched and processed for rHPLC analysis.

Conversions 18F-FLT to 18F-FLTMP were evident in rHPLC chromatograms (Appendix,

figure 3) however, 18F-FLTDP or 18F-FLTDP could not be detected. It can be concluded

that the conditions for sustaining dTYMK activity were not met for analysis (Appendix

figure 3). Thus, the direct effect of YMU1 on dTYMP activity could not be verified.

H C T -1 1 6

1 1 0 1 0 0 9 0 2 4 h

) 8 0

y o

t 4 8 h i

r 7 0

i n

b 6 0 7 2 h

f 5 0

e 4 0

C % ( 3 0 2 0 1 0 0

l l l l ) 1 1 1 i i i i % U U U m m m m 5 a a a a 2 M M M p p p p 6 Y Y Y a a a a . r r r r (0 M M M e e e e    V V V V O 1 1 0 S . 1 M M M M M O     0 D 0 0 0 5 5 5 5 + + + 1 1 1 U U U M M M Y Y Y M M M    1 1 0 . 1 O

Figure 22. Cell viability assay in HCT-116 cells after treatment with verapamil hydrochloride in a range of concentrations. YMU1’s potency was measured by SRB assay in the presence or absence of verapamil for a period of 24-72 h. The data represented are mean ± SEM of three individual experiments, each done in quintuples.

3.4.2. Liposomal YMU1 drug delivery

Owing to the physiochemical properties of YMU1, we also queried in vitro permeability as a plausible rate-limiting step for the bioavailability of the drug. To our knowledge, a permeability assay is yet to be conducted on YMU1, and its assumed cell penetration was based on Hu et al. ’s (2012) combination study. We postulate the enhanced cytotoxicity observed during YMU1-Doxorubicin combination therapy to be due to the membrane e damaging effect of doxorubicin which, permits YMU1’s entry into cells116. To validate this speculation, YMU1 was encapsulated into a minute spherical sac of phospholipid molecules and incubated with HCT-116 cells for 24 h. Because of the nature of drug encapsulation, the maximal concentration of YMU1 achievable was 1.9 µM. An additional limitation associated with the sigma liposome mix (section 2.6.) was the variability in the number of liposomes delivered at increasing concentrations of YMU1. A fold change of YMU1 meant a parallel fold increase in the number of liposomes administered. The liposomes intensity particle size distribution was calculated as size 392.1 nm ± 14.14

(Appendix, figure 9). As shown in figure 23A, an incubation of 190 nM liposome encapsulated YMU1, known herein as LP-YMU1, resulted in a 2- fold reduction of 18F-

FLT uptake when compared to its liposome DMSO control (LP-DMSO-1). A significant difference (p<0.001) between LP-DMSO-1 and 190 nM LP-YMU1 was observed indicating a reduced nucleoside flux (with respect 18F-FLT) in the presence of YMU1.

Unexpectedly, 18F-FLT uptake remained unchanged between cells treated with 1.9

µM LP-YMU1 and its respective control (LP-DMSO-2) suggesting that, at higher concentrations, liposomes may have confounding effects on the TK1 activity which appeared to influence the level of radiotracer uptake observed. This theory was

supported by Komizu et al. (2014) who created hybrid liposomes, similar to the sigma concoction, and reported its growth inhibitory effect on HCT-116 and A549s through the induction of cell cycle arrest at G0/G1 phase. This was also apparent in the subsequent western blot study of the liposomal conditions (appendix, figure 4) We have shown, in section 3.3.2, that a G0/G1 cell cycle arrest influences TK1 activity and

18F-FLT uptake. In lieu of this, we sought to understand whether YMU1 had further shifted the 18F-FLTMP: 18F-FLTDP ratio in favour of monophosphate accumulation.

Since an incorporation assay only gives a quantitative measure of the overall 18F-FLT trapping and does not differentiate between the distributions of FLT nucleotides in extracts, rHPLC was employed to assess the intracellular conversions of 18F-FLT using the same conditions previously described. All three 18F-FLT-nucleotides were detected in cell extracts (as described in methods and materials) using SAX ion exchange column (figure 23). For simplicity, the distribution of 18F-FLT nucleotides has been summarised in Table 4. The 18F- FLTDP/18F- FLTTP fraction in 190 nM LP-

YMU1 (23, 18%) treated cells were comparable to LP-DMSO-1 (26, 27 %). A modest increase of 18F- FLT/18F- FLTMP fraction in 190nM LP-YMU1 (77, 82 %) treated cells was evident in the chromatogram when compared to LP-DMSO-1 (73, 74 %). Aside from the confounding effects of liposome incorporation, these results suggest that

YMU1 may have exerted additional stress on the nucleotide synthesis pathway. At

1.9 µM LP-YMU1 the 18F- FLTDP/18F- FLTTP fraction was non-existent whilst, LP-

DMSO-2 accumulated 11 and 10 % (n=2) of the phosphorylated species. TK1 activity appears to be attenuated under 1.9 µM LP-YMU1 which resulted in the retention of

18F- FLT at 98 and 95 % with, minimal conversions to monophosphates (2, 5 %).

A B

2 0

i 1 5

t o

r p

1 0

g m

/ D I 5

% A B

) 1 1 % U 2 0 3 U M .3 M Y Y (0 M M O  n n S i 1 5 0 .9 M 9 1 e

D 1 t o

r p 500 1 0

g 350 m / D

C D 400

I 300 5 % 250 300 0 D 200 18  )  18 1 1 F-FLTMP % 150U F-FLTMP 200 3 U M .3 M Y Y (0 M M 100 O n  S 100 0 .9 M 9 1 A D 1 50 B C 500 0 0 600 350 300 C -50 D 400 300 00:00 05:00 10:00 15:00 20:00 25:00 30:00 00:00 05:00 10:00 15:00 20:00 25:00 30:00 500 Time (min) 1 0 *** Time (min) 250 300 D 200 18 200 18  400  F-FLTTP F-FLTTP 8200 150

18 n F-FLT uptake i 100 e 18 t 100 300  F-FLT o  18 50 r 6 F-FLT 18 TP p TP 100 F-FLTDP 018 0 200 (%ID/mgE of protein)g F-FLTDP -50

m 4 / 00:00 05:00 10:00 15:00 20:00 25:00 30:00 00:00 05:00 10:00 15:00 20:00 25:00 30:00 D Time (min) Time (min) I 1 0 100

% 2 TP 0

8 0

i e

0 t o

r 6 00:00 05:00 10:00 15:00 20:00 25:00 30:00 00:00 05:00 10:00 15:00 20:00 25:00 30:00 1 s p e U Time (min) Time (min) m E M g o Y s l m 4 o a /

p D li m I l o 190 nM LP-YMU1 o s % 2 LP-DMSO-1 tr o n ip o L C M 0 n 0 9 s 1 1 18e U m M 18 o F-FLT Y F-FLT s l G H o a F ip m l F El o 18 D o s r o F-FLTMP t n ip 2 .5 o L C M 120 n 0 120 9 50 1

2 .0 100 50 n i G H18 100 e F 40

t 80 F-FLTTP

o NS

r 1 .5 40 2 .5 30 p 80 60 120   g 50 20

m 1 .0 2 .0 40 30 100TP

/ n

18 i 60 D e 40  I F-FLT uptake t 10 80

o 20 % r 1 .5 0 .5 30 20 p 40

60 0   g 0 20 (%ID/mg of protein)m 1 .0 40

0 .0 / 00:00 05:00 10:00 15:00 20:00 25:00 30:00 00:0010 05:00 10:00 15:00 20:00 25:00 30:00 20 D I Time (min) 10 e 1 Time (min) 20 m U % 0 .5 o M s Y 0 0 0 0 o l p a li l m 0 .0 00:00 05:00 10:00 15:00 20:00 25:00 30:00 00:00 05:00 10:00 15:00 20:00 25:00 30:00 o o 00:00 05:00 Time10:00 (min) 15:00 20:00 25:00 30:00 00:00 05:00 10:00 15:00 20:00 25:00 30:00 tr s e 1 Time (min) n o m U o ip o M Time (min) Time (min) L s Y C o l M ip a l  l m 9 o o . tr s 1.9 µM LP-YMU1 1 n o LP-DMSO-2 o ip C L M  .9 1

Table 4. integration of rHPLC 18F-FLT chromatograms under specified conditions

Conditions FLT area of FLTMP area FLTDP area FLTTP area FLTDP + FLTTP peak (%) of peak (%) of peak (%) of peak (%) area of peak (%)

LP-DMSO-1 06, 03 68, 70 02, 02 24, 25 26, 27

190 nM LP-YMU1 07, 09 70, 73 02, 02 21, 16 23, 18

LP-DMSO-2 65, 69 24, 21 00, 00 11, 10 11, 10

1.9 µM LP-YMU1 98, 95 02, 05 00, 00 00, 00 00, 00

Figure 23. The effect of YMU1 liposomal drug delivery on nucleoside flux in HCT-116

cells. A Ƴ counted cell uptake of 3.7 MBq 18F-FLT in LP-DMSO-1 and 190 nM. B and C. Corresponding rHPLC chromatograms for LP-DMSO-1 and 190 nM LP-YMU1. D Ƴ counted cell uptake of 3.7 MBq 18F-FLT in LP-DMSO-2 and 1.9 µM LP-YMU1. E, F Representative 3.5.rHPLC Validation chromatogra of HCTms- 116for LP model-DMSO by-2 siRNAand 1.9 µM LP-YMU1. Chromatogram retention times are as follows; 18F-FLT = 5 min, 18F-FLTMP= 15 min, 18F-FLDP= 17 min and 18F-FLTTP

= 22 min. Table 4. Corresponding integration table of chromatograms.

3.5. siRNA knockdown

Taken together the possible interference of liposomes with the cell cycle, this model was no longer considered for future work in vitro. Nonetheless, the generated data echoes Hu et al. ’s (2012) findings for YMU1’s lack of potency when administered singularly. While the rate-limiting factor of YMU1’s bioavailability remains undetermined, we proceeded to evaluate (by Parallel artificial membrane permeability assay [PAMPA] and Caco-2) the permeability coefficient of YMU1 to elucidate its ability to permeate the cell membrane.

In collaboration with the Cancer Therapeutics section of the Institute of Cancer

Research (ICR), a combination study with PAMPA117 and Caco-2117 was conducted to determine the absorption potential of YMU1 in cells. The PAMPA model, though highly efficient and a high-throughput model, only captures the transcellular passive permeability across the lipoidal membrane barrier without considering the relative contribution of pores or drug transporters in cells. The Caco-2 model, however, incorporates both the transcellular passive and transporter mediated (efflux and influx) permeability along with the paracellular components of transport. The PAMPA permeability assay indicated a low classification score of 0.2 x 106cm/s at physiological pH 7.4 (low = < 5 x 106cm/s) whilst, Caco-2 presented a “moderate” classification score of 2.7 x 10-6 cm/s for apical > basolateral (A>B) under physiological conditions (pH 7.4) (figure 24). Considering that a moderate classification is defined as falling between 2 -20 x 10-6 cm/s, YMU1 appears to have borderline low transcellular permeability. Support for a role for PGP transporter is demonstrated by the effect of verapamil on YMU1 transport. It is also possible that

the accumulation results of Caco-2 may be due to the drug sticking on the cellular membrane. YMU1 can be concluded to have partial/low cellular permeability as per

Caco-2 and PAMPA assays.

Caco-2 Permeability Classifications A low 2> x 10-6 cm/s medium 2-20 x 10-6 cm/s high >20 x 10-6 cm/s <80% <50% Assay A>B B>A Recovery % Efflux Ratio Compound Concentration Average Average AB BA SD SD (B>A/A>B) (uM) (*10-6cm/s) (*10-6cm/s) AVE SD AVE SD CCT352141 10 1.0 0.0 41.2 2.1 42.62 80.8 2.6 97.3 2.1 CCT352141 + inhibitor 10 7.0 0.2 13.4 1.6 1.91 78.0 1.2 96.1 2.8 CCT251981 10 0.4 0.1 1.0 0.0 2.28 82.8 6.7 96.7 0.6 CCT138379 10 31.8 1.8 40.9 12.8 1.29 89.8 5.2 95.5 1.9 YMU1 10 2.7 0.8 20.4 3.2 7.5 19.71 3.48 95.16 3.09 YMU1 + inhibitor 10 4.8 (n=2) 6.9 0.7 1.4 30.70 6.62 90.13 0.43

PAMPA Classifications B low <5 x 10-6 cm/s medium 5-20 x 10-6 cm/s high >20 x 10-6 cm/s

Recovery pH5 pH5 pH6.5 pH6.5 pH7.4 pH7.4 Compound ID Papp Permeability Papp Permeability Papp Permeability pH 5 pH 6.5 pH 7.4 x10-6 cm/s Class x10-6 cm/s Class x10-6 cm/s Class Mean StDev Mean StDev Mean StDev YMU1 <0.3 low <0.3 low 0.2 low 107% 2% 115% 5% 95% 2% Antipyrine 7.8 medium 7.1 medium 7.6 medium 108% - 111% - 121% - Nadolol <0.2 low <0.2 low <0.2 low 103% - 102% - 111% - Verapamil <0.3 low 16.9 medium 90.1 high 69% - 44% - 40% - Coumarin 79.6 high 137.2 high 125.1 high 90% - 94% - 100% -

Figure 24. Caco-2 and PAMPA permeability assay of YMU1. A Results of for YMU1 in Caco -2 assay along with relevant positive controls/ efflux mediators. B Results of for YMU1 in PAMPA assay along with relevant positive controls/ efflux mediators

3.5. Validation of the HCT-116 model by RNA interference

Hypothesising that a knockdown of dTYMK would inhibit dTTP biosynthesis and consequently DNA synthesis, we investigated the effect of transient siRNA-mediated dTYMK knockdown in HCT-116 cells. Of the four independent siRNAs transfected

(varying in dTYMK sequence target site) , three (siRNA 5,6 and 15) resulted in a 53,

80 and 59 % reduction of dTYMK protein when compared to negative control siRNA

Luciferase (figure 25A). A Knockdown of dTYMK was also observed following transfection with siRNA 14; however, GAPDH was similarly reduced and thus implicates siRNA off-target silencing as a plausible cause. Since siRNA specificity is not absolute, off-target gene silencing usually occurs through multiple mechanisms including; the induction of an interferon response and/or mRNA degradation because of partial sequence complementation118. Furthermore, similarities between off-target DNA sequences and siRNA 14 were examined using the basic local alignment search tool (BLAST). Excluding dTYMK, no complimentary matches to

GAPDH or other biological sequences were found. We questioned the possibility of siRNA 14 binding upstream of GAPDH open reading frame (ORF) which causes transcription inhibition and subsequently, mRNA for protein translation. Using the

DNA Strider tool for analysing molecular sequences, siRNA 14 did not align upstream

(N-terminus) nor downstream (C-terminus) of GAPDH. The unanticipated off-target activity of siRNA 14 to GAPDH complicates the interpretation of dTYMK expression and renders the results inconclusive.

A cell survival assay was performed in parallel with the previously described western blot analysis. To our surprise, siRNA 5, 6, and 15 appeared to have little or no effect

on cell survival while, the complicated siRNA 14 sequence revealed a 30 % (p<0.001) reduction in cell growth (figure 25B). This contests the expected decrease in cell proliferation under siRNA 5, 6 and 15 since a reduced supply of dTTP is hypothesised to result in an increase of dUTP misincorporation during nucleotide excision repair32,110.

A B

1 .5

n s n s

l n s

a n s v

i 1 .0

v r

***

l e

C 0 .5

0 .0

l ) 5 6 4 5 l o A 1 1 o tr A r n N N A A t R N N n o i iR o C s s iR iR c s s ( c u L A N iR s

Figure 25. siRNA mediated dTYMK knockdown in HCT-116 cells. A. Representative western blot showing expression of stated proteins in HCT-116 cells. B. Cell survival assay by Sulforhodamine B for cytotoxicity screening. Error bars represent SEM (n=6). Stars represent significant differences compared to control.

Because post-transcriptional regulation of gene expression includes mRNA stability, we speculated the lack of correlation between expression and the observed cell survival phenotype to be as a result of high dTYMK stability and low mRNA decay rate. Changes in the phenotype may be masked if the corresponding transcript is stable. Therefore, it is possible that cells only require a small percentage of the functional transcript to sustain dTTP supply without disrupting cell growth. Sharova

et al. (2015) published a recent report on the mRNA stability of 19 977 genes. In turn, we examined the groups’ database of the mRNA half-life of dTYMK which revealed a moderate stability of 6.27 h and a decay rate (calculated by average decay rate of

200 genes with most stable mRNA) of d= 0.125 h-1. From this, we queried the relationship between transcription translation and dTYMK mRNA degradation in

HCT-116 cells. The intercalating ability of Actinomycin D to DNA at the transcription initiation complex was employed to inhibit the elongation of RNA chain by RNA polymerase119–121. Cycloheximide was also selected for mRNA decay to indirectly inhibit mRNA decapping with its primary effect being a translation inhibitor119,120.

Since AD gives us a measurement of transcription inhibition similar to that elicited by dTYMK siRNA transfected cells, the expression profile of the translated protein, before the addition of actinomycin D, should give an indication of dTYMK stability.

Western blot analysis showed a decrease in dTYMK expression after 2h CHX and AD treatment while, the 12 h time point presented a highly significant reduction of 62 %

(p <0.001) when compared to control (figure 26A). The observed reduction of dTYMK at 6h AD correlated with the half-life reported by Sharova et al. (2015) and, dTYMK expression reduced linearly over time (figure 26C). Using the result of dTYMK expression (38 %) over the 12 h time course, it is reasonable to assume that after 48 h, the translated protein would have turned over 4 x to give a result of ~ 9% expression. Given its presumed essentiality for DNA synthesis, it is possible that this expression profile is enough to sustain the overall dTTP pool for DNA flux.

Actinomycin D inhibition results in p53 accumulation and promotes p53 translocation into mitochondria for apoptosis. An elevation of p53 expression is evident in figure

26A and confirms the successful inhibition of transcription.

B C

120 R2 = 0.9919

)

2 100

- actin (OD/mm actin -



to 20

% dTYMK expression normalised dTYMK expression % 0 2 4 6 8 10 12 Time (h)

Figure 26. Effect of transcription and translation inhibition on dTYMK expression. A. Representative western blot showing expression of stated proteins in HCT-116 cells following a time course treatment of CHX and AD. B. Densitometry of dTYMK expression normalised to control. C. Graph of dTYMK expression over the time course treatment of HCT-116 cells with AD. Error bars represent SEM (n=2 independent repeats). Stars represent significant differences compared to control.

To gain a better understanding of the extrinsic stress exerted on DNA synthesis by siRNA mediated dTYMK knockdown and to assess the relationship between dTYMK knockdown and the changes in the overall nucleoside salvage flux, an incorporation assay with 18F-FLT (a thymidine derivative) was performed. In accordance with figure

25, siRNA 6 and 14 were selected for radiotracer uptake in HCT-116 cells. 18F-FLT uptake decreased by 65 % (p<0.0001) and 11 % (p<0.01) under the transient effect of siRNA 14 and siRNA 6 with respect to control (figure 27).

Taken together, a knockdown of dTYMK alters the salvage nucleotide synthesis pathway; however, to our surprise, cell proliferation was sustained without detrimental effects in vitro (figure 25B). Furthermore, Hu et al. 2012 could not see an increase in gamma H2AX following siRNA knockdown of dTYMK when compared to control. This suggests that a transient knockdown of dTYMK does not necessarily induce a DNA damage response.

10 * 18F-FLT uptake (%ID/mg protein) 5

protein ID/mg ***

0 Control siRNA siRNA 6 siRNA 14

Figure 27. Effect of dTYMK knockdown in HCT-116 cells. Cells were exposed to nontargeting siRNA luciferase or dTYMK 6 and 14 for 48 h. A. Lysates were assessed by

western blot for the expression of indicated proteins. Control indicates untreated. Cell

proliferation was measured by SRB assay. Mean of n=6 in triplicates ± SEM; * P< 0.5, ** P< 0.01, *** p<0.001 as per Bonferroni’s post hoc analysis

100

3.6. Generation of a HCT-116 cell line encoding mutant dTYMK

A knockdown of dTYMK impaired salvage pathway nucleotide synthesis as shown by

18F-FLT uptake (figure 27). To cross validate these findings and to develop stable clones for further analysis, cellular models of dTYMK knockdowns were created by introducing mutations into wild-type (WT) HCT-116 cells using CRISPR/Cas9- mediated gene deletion These experiments were conducted with the kind help of

Professor Laki Buluwela of Imperial College London. HCT-116 cells were selected as an adequate model system since it has been shown, by literature, to have good efficacy in RNA-guided endonuclease gene editing122. CRISPR/Cas9 systems are RNA- programmable adaptive immune mechanisms utilised by bacteria and archaea to degrade foreign nucleic acid82. For our application, type II was employed, as it is the most studied and universally accepted method of editing mammalian genes83. The first experiment sets out to identify, by bioinformatics, an appropriate target site for editing dTYMK. Two 23-bp (20 target + 3 Protospacer Adjacent Motif [PAM]) exon encoded sequences were identified from crispr.mit.edu and filtered to minimise off- target cross-reactivity. According to ENSEMBL, this dTYMK gene has 7 transcripts

(splice variants), but only 3 of these; DTYMK 201, 203, 205 present an open ORF reading frame that may code for an active dTYMK protein (Table 5). Upon sequence assessment, DTYMK 203 and 205 present a 3’ and/or 5’ truncation preventing the annotation of the start and the end of protein-coding region inferring that DTYMK

203 and 205 can be expressed but not necessarily active (Table 5).

101

Table 5. Transcript (Splice variants) of DTYMK as reported by ENSEMBL (protein codable variants have been highlighted in orange)

Generally, DTYMK 201 is considered the full length and active version of dTYMK in humans; it also corresponds to isoform 1 with regards to associated Uniport P23919.

The splice variants DTYMK 202 and 204 have nonsense mediated decay (NMD) biotype. The presence of nonsense mutations is detected by the NMD pathway which selectively degrades mRNAs harbouring premature termination codons (PTCs), and the production of truncated or erroneous proteins are prevented123. Therefore, these splice variants code for non-functional dTYMK mRNA. Taken together, constructs against exon 1 were created not only due to its presence in all the splice variants but, to encourage non-homologous end joining mediated frameshifts which can potentially cause a premature termination in mRNA translation.

102

WT CRISPR

WT CRISP R

Figure 28. DTYMK gene organisation and Location of coding exon1 CRISPR guide RNAs. A. Diagram (not to scale) indicates overall gene structure and positions of the CRISPRS and PCR primers. DNA sequence heterogeneity resulting from Cas9 targeting by CRISPR 25926363 – 25926373. B. Alignment between WT and exon 1 deleted DTYMK Open Reading Frame (ORF).

HCT-116 cells were transfected by nucleofection with dTYMK CRISPR gRNAs

(25926363, 25926373) (figure 28) and Streptococcus pyogenes Engen Cas9NLS

(SpCas9) programmed with tracrRNA to produce a Cas9 Ribonucleoprotein mix.

These were harvested 72 h post-transfection and PCR primers SA4760DTYMKex1f and SA4761DTYMKex1r amplified around the editing site to generate WT and edited

PCR products. Gene editing efficiency was estimated by running the PCR products on

1% agarose gel to resolve full-length DNA and products encoding deletions mediated by the CRISPR gRNAs 25926363- 25926373 combination. Figure 29A shows evidence of CRISPR mediated deletion owing to the appearance of a 247 nt cDNA product.

However, WT (336 nt) is still prominent, reflecting protein transfection efficiencies.

Nonetheless, 24 single cell clones were picked by accutase soaked cloning discs for clonal expansion. They were pooled into four groups, each containing 6 different clones, before being subjected again to a DNA PCR deletion detection assay as

103

described above. Group B (Appendix, figure 5) showed the desired shift in PCR product migration and encompassed the clone with the predicted indel deletion. Of the 6 clones isolated from group B, two exhibited the predicted 89 bp deletion (from

336 nt to 247 nt) as per CRISPR target (figure 29B). However, WT was still present in both clones which suggests an incomplete knockout. For precision genotyping,

Sanger sequencing was performed on PCR products, amplified from gDNA spanning the target site. Clone one (figure 29D) indeed revealed the presence of dTYMK deletion and endogenous WT as per DNA mismatch detection assay. This line hereafter referred to as “B1” cells, is therefore likely to be mutant/ WT. Sanger sequencing of clone two (figure 29E) inferred deletion due to the presence of undetermined base sequences and no clear WT. This clone hereafter referred to as

“B5” cells, is likely to be mutant/mutant, as the deletion appears to have taken out most of Exon 1.

104

A B

Figure 29. DTYMK sequencing in wild type and dTYMK knockdown HCT116 (B1 and B5). A, B. Genomic DNA was prepared from HCT116 cells transfected with expression constructs for the Engen Cas9NLS (SpCas9) preloaded with tRNA and CRISPR 25926363 –

25926373 at a plasmid ratio of 1:1, or from control cells transfected with Cas9-GFP. PCR product encompassing dTYMK exon 1 was derived from these DNAs and used in DNA sequence analysis. Single clone isolation and expansion permitted the identification of mutant dTYMK. C, D, E. The upper panel shows the equivalent region from control and

Cas9-GFP transfected cells whilst the lower panel shows the equivalent region recognised by CRISPR 25926363 – 25926373 (highlighted) in B1 and B5 cells. The position of the predicted CRISPR 25926363 – 25926373 directed cleavage sites is arrowed for B1. Sequence analysis for B5 showed a big spread of undetermined nucleotide base sequences.

105

To confirm the allele sequences, the PCR products of HCT -116 WT, B1 and B5 cells were subjected to plasmid cloning to produce templates representative of individual alleles. For B1, two plasmid preps produced mixed sequences. Of the remaining ten, eight clones showed the sequence for the intended CRISPR engineered deletion and two remained WT. The ratio of mutant to wild-type clones is in the ratio of 5:1, suggesting that the HCT116 line possibly has as many as six copies of the gene (figure

30A). The sequencing results for clone B5 appears to produce three alleles, and the frequency of the clones support the notion of six copies (1:1:4; WT: Allele 2: Allele 3).

One allele is wild-type, the second is mutated through a single base insertion (indel mutation) at the cutting site for CRISPR 25926373 (figure 30B). The third allele

(present as four copies) is a 76 bp deletion that appears to be a consequence of damage from DNA repair at the cutting site by CRISPR 25926373 sequence. While the sequencing result (detailing the site of deletion) appears straight-forward, it still brings to question as to why four copies of the six genes encode the same unexpected deletion. We envisioned the cause to be down to the complexity of DNA repair following CRISPR cutting; however, this effect was beyond the scope of our study and needs further attention in future work.

106

Figure 30. Allele sequencing results for CRISPR/CAS9 generated mutant dTYMK cells. Following PCR amplification of DMTYK exon 1 sequences from genomic DNA using primers

SA4760DTYMKex1f and SA4761DTYMKex1r, PCR products were cloned into cloning vector

(pJet 1.2). Plasmid clones were selected, and DNA extracted by mini-prep. These were then sent for Sanger sequence to compile allele sequences. A. Genotyping results of B1 displaying an allele frequency of 1:5 (WT:Mutant) ratio. B. Allele frequency distribution of WT:Mutant (1:1:4) ratio in B5 cells.

107

3.7. CRISPR-Cas9 cell line characterisation

As a more definitive measure of gene knockdown, the differential protein expression of B1 and B5 cells were assessed by western blot which avoids signal from in-frame deletions or mutations shown by sequencing. WT HCT-116 cells expressed 2.3 to 2.9- fold higher levels of dTYMK when compared to B5 (42 ± 6%) or B1 (33 ± 2 %)

Surprisingly, similar levels of TK1 and TS expression were evident in WT and mutant cells (figure 31).

A B HCT B1 B5

DTYMK 24 kDa

d 1 5 0

e s

β-actin i ***

42 kDa l

) a

2 m

m ***

o m

TK1 26 kDa /

n 1 0 0

(

s n

β-actin i s

42 kDa t

- 5 0

 e

TS 30 kDa

42 kDa Y

β-actin T 0 d H C T -1 1 6 B 1 B 5

Figure 31. Expression profile of the essential enzymes in the thymidine nucleoside

pathway. A. Representative western blot showing expression of stated proteins in HCT116, B1 an. B5 cells. B. Quantification of dTYMK expression normalized to β actin

(n=3). Stars represent significant differences compared to control.

Following the confirmation of gene editing, cells were grown for three population doublings to determine whether a partial knockout of dTYMK will reduce cell growth as a consequence of nucleotide imbalance and in turn, futile cycles of DNA synthesis.

As shown in figure 32, the growth rate of parental HCT-116 to mutant cells were surprisingly comparable. Our result echoed Hu et al. ’s (2012) findings but, contradicts that of Lui et al. (2013) who observed a pronounced reduction in

108

proliferation rate following dTYMK shRNA knockdown. As the growth rate of the mutant cells remained relatively unchanged, it became of interest to assess whether the cell cycle distribution differed between the three groups.

Using flow cytometry, a cell cycle analysis was employed to determine the G1, S and

G2/M profiles of B1, B5 and parental. In figure 33, the proportion of cells at the G1 phase increased from 39.73 % (HCT-116) to 44.30 % and 50.67 % in B1 and B5 cells respectively. According to the two-way ANOVA, there was a statistical difference between the three groups (p<0.001). A Bonferroni post-hoc analysis revealed p values of p=0.003 for HCT-116 vs B1 and p<0.001 for HCT-116 vs B5. It was also observed that the proportion of S-phase in the mutant B5 cells (45.06 %) were significantly increased (p <0.001) when compared with that of wild-type HCT-116

(37.33 %). However, the proportion of cells in S-phase in B1 cells (38.30 %) did not significantly change (p > 0.05) when compared to parental (37.33 %).

6 H C T - 1 1 6 ( 2 1 h D .T .)

) B 1 ( 2 2 h D .T .) 1

y B 5 ( 2 1 h D .T .)

n a

o 4

D t

f e

o t

r 2

a l

r (

0 1 2 3

D a y

Figure 32. Comparison of doubling times in parental HCT-116 and mutant cells. Impact on proliferation as determined by SRB assay. Mean of n =3 in triplicate ± SEM; for statistical analysis all 9 data points were pooled. Doubling time is indicated in brackets next to the cell line name, as calculated on http://www.doubling- time.com/compute.php?lang=en.

109

*** NS NS HCT HC HC - T- T- 116, 11 11 B1 6, 6, and B1 B1 B5 an an Figure 33. Cell cycle analysis of mutant vs wild type cells. A. Representative graph of cell d d cycle distribution after analysisB5 of DNAB5 content with propidium iodide. B. Graph representing the percentages of cells in G1, S and G2/M phase. Bars are mean ± SEM from 3 experiments.

The cell cycle distribution (figure 33) data suggests the presence of genomic stress in theFigure CRISPR/Cas9 20. Cell cycle generated analysis mutantsof mutant al vsthough, wild type the cells.rate ofA. DNAThe distribution synthesis and of cells in turn in the different phases of the cell cycle was analysed in HCT-116, B1 and B5 cells. B. cellSummary proliferation of percentage appears of to cells be inrelatively each phase comparable of the cell incycle figure Significant 32. This differences disparity begsare theindicated question as follows:of what * hpappens ≤ 0.05, **when p ≤ 0.01 the andsalvage *** ppathway ≤ 0.001. Barsis inhibited are mean at ± the SEM level from of 3 experiments. dTYMK. Using a multi-pronged approach involving 18F-FLT metabolite profiling, in vivo growth determination, xenograft immunoblotting and localisation assay, we will attempt to investigate the full spectrum dTYMK regulation for DNA synthesis.

110

3.8. Profiling of 18F-FLT phosphate species in HCT-116 dTYMK mutant cell lines

To assess the early stage of the DNA synthesis pathway, 18F-FLT was incubated with

HCT-116, B1 and B5 cells that had been pre-treated with either 100 µg/ml 5FU or

0.1% DMSO for 1.5 h. 5-fluorouracil was introduced to initiate the well-documented flare effect and investigate the interplay between the de novo and salvage pathway.

Under control conditions, 18F-FLT accumulation is significantly reduced in both B1

(11.0 ± 0.1 %, p<0.001) and B5 cells (12.6 ± 0.2 %, p<0.001) when compared to HCT-

116 (18.1 ± 0.2 %) (figure 34). With 5FU treatment, 18F-FLT flare is evident across all cell lines as the salvage pathway attempts to circumvent inhibition of the de novo pathway. Increased fluorothymidine is available through the flare-related redistribution/ translocation of ENT1 transporter from the intracellular compartment of cells to the plasma membrane. The same phenotypic (WT>B1>B5) can be seen in

5FU treated B1 (41.57 ± 2.8 %, p<0.001) and B5 (51.01 ± 1.1 %, p<0.001) cells indicating a correlative decrease in the intracellular trapping of 18F-FLT with respect to 5FU treated HCT-116 (64.43 ± 2.4 %). The difference, while statistically significant within both conditions, does not give a greater insight into what happens when dTYMK has been knocked down dTYMK.

Lui et al. (2013) reported that a knockdown of dTYMK reduces dTTP while increasing dUTP levels and subsequently, dUTP misincorporation into DNA when the ribonucleotide reductase R2 subunit activates nucleotide excision repair. We hypothesised an increase in the incorporation of chain terminating 18F-FLTTP into

DNA, further potentiated by 5FU, into the DNA of B1 and B5 cells as a consequence of nucleotide imbalance. Keeping with the same conditions, cells were prepared and

111

then subjected to perchloric acid washes to remove negatively charged compounds, i.e. phosphorylated molecules, which are not successfully bound to macromolecules

(i.e. DNA, RNA and protein) 124. Under 0.1% DMSO treatment, the acid insoluble fractions (macromolecule fraction) of B1 and HCT-116 cells show a near complete washout of unbound 18F-FLT. With 5FU, an increase of 18F-FLTTP accumulation into macromolecules is evident in B1 cells. An assessment of B5 cells revealed a statistical increase in the incorporation of FLTTP into the macromolecules under control conditions (1.36 ± 0.1 %, p<0.001). However, 5FU treatment exerted very little difference when compared to parental HCT-116 (figure 34B). Overall, levels of FLT incorporation into macromolecules was low.

112

A 1 0 0 C o n tr o l

1 .5 h 5 F U

) n

i 8 0

p p

6 0

T g

L 4 0

1 2 0

% (

6 1 5 1 B B 1 T C H

B T o ta l A c id in s o lu b le 8 0

6 0 C o n tro l

)

n i

e 1 .5 h 5 F U e

t 4 0

f u 2 0

L F

m 2 .0

1 1 .5 % ( 1 .0

0 .5

0 .0

6 1 5 6 1 5 1 B B 1 B B 1 1 T T C C H H

Figure 34. Assessment of macromolecule incorporation of 18F-FLT with or without 100

μg/ml 5FU treatment in HCT-116, B1 and B5 cells. A. Cells were seeded into 6-well plates at 3.0x105 cells/well. After overnight culture fresh media containing either control (0.33

18 % DMSO) or 100 μg/ml 5FU for 1.5 h prior to the addition of ~0. 74 MBq F-FLT. The graphs represent the average values (n=6) and the standard error of the mean. B.

Representative graphs of 2 experiments with n=3. Following 18F-FLT incubation, control

(0.33 % DMSO) or 100 μg/ml 5FU treated cells were washed three times in PBS and then in 10 % PCA fractions were gamma counted and normalised to protein. Control conditions were lysed in RIPA buffer before gamma counting. Results were normalised to the protein content of parallel plates.

113

The 5FU 18F-FLT flare has been well documented; however, the actual distribution of

18F-FLT phosphate species is yet to be characterised. It became of interest to not only understand the profiling of these metabolites but, to assess the degree of change within the dTYMK mutant cell lines during the flare response. Under control conditions, the 18F-FLTDP+FLTTP fraction decreased from 26.5 ± 3.5 % (HCT-116) to

13 ± 3% and 15 ± 3 % in B1 and B5 cells respectively (Table 6). Unsurprisingly, 18F-

FLTMP accumulated in mutant clones probably due to the reduced expression/activity of dTYMK. An accumulation of 18F-FLTMP was also observed in

5FU treated HCT 116 (80 %) and B5 (83 %) cells and their 18F-FLTDP+FLTTP fraction were also comparable. In contrast, 5FU treated B1 cells displayed a better conversion of 18F-FLTMP to into its di and triphosphate species when compared to its vehicle (0.1

% DMSO). However, the rHPLC 5FU metabolite experiments require further validations in vitro before making conclusive remarks.

114

18F-FLTMP A (15:00)

18F-FLTDP 15:00 (17:06)

18F-FLT 18 F-FLTTP (04:39) (19:09) 17:06

04:39 19:09 3B1 – 0.1 % DMSO 3B1 – 100 µg/ml 5FU HCT 116 - 0.1 % DMSO HCT 116 – 100 µg/ml 5FU 3B5 – 0.1 % DMSO 3B5 – 100 µg/ml 5FU

00:00 05:00 10:00 15:00 20:00 25:00 30:00 Time (mm:ss)

TABLE 6. Integration of 18F-FLT following 5FU treatment

CELL LINE 18F-FLT 18F-FLTMP 18F-FLTDP 18F-FLTTP 18F-FLTDP + 18F-FLTTP HCT-116 1 % 69 %, 76 %, 4%, 4%,6% 26%, 19 %, 30 % 23%,28% (VEHICLE) 72% 22% HCT -116 + 1.5 1 % 80% 4% 15% 19 % H 5FU B1 (VEHICLE) 1 % 84 %, 84 4 %, 3 %, 5% 12 %, 12 %, 16 %, 15 %, %,78% 10% B1 + 1.5 H 5FU 4% 61 % 10 % 25 % 35 % B5 (VEHICLE) 1 % 81 %, 87 %, 4 % 2 %, 2% 14 %, 10 % 18 %, 12 %, 84 % 16% B5 + 1.5 H 5FU 1 % 8 3 % 3 % 14 % 17 %

Figure 35. Metabolite assessment of 18F-FLT uptake of HCT-116, B1 and B5 cells treated with either 100 µg/ml or 0.33 % DMSO. A. Representative chromatograms of specified treated cells for 1.5 h prior to 18F-FLT uptake. Cells were scraped into 4:1 Acetonitrile- water and processed for phosphate species determination via an ion exchange column SAX column. Table 6. Respective integration table of chromatograms. For vehicle

samples, 3 independent repeats where conducted.

To conclude, an altered salvage nucleotide metabolism (reduced conversions to

diphosphates) can be seen in both clones in control and 1.5 h 5FU treatment

conditions. Surprisingly, a dTYMK knockdown resulted in a decrease of 18F-FLTTP

115

incorporation into the macromolecules of B1 cells while, an increase was observed in B5 cells.

From literature, we know that dTYMK is the only known enzyme for the synthesis of dTDP required for DNA synthesis. A knockdown of this convergent enzyme should result in a reduction of cell growth due to the reduced supply of DNA precursors.

However, the in vitro survival assay portrayed a knockdown of dTYMK as dispensable at the time points investigated; whether longer time points will allow the effect of

DNA repair to more effectively affect survival remains to be seen. The discrepancy between growth rate and dTYMK expression may be due to insufficient knockdown and/or, the presence of compensatory mechanisms that mediate nucleotide synthesis to overcome the dTDP shortage. We suspected the components of media, particularly L-glutamine, as a supplier of the building blocks required for pyrimidine synthesis, either via ribonucleotide reductase (RNR) de novo or an unknown pathway, to compensate for the reduced dTTP pool. This hypothesis was derived following Lui et al. (2013) findings on growth retardation rescue upon the addition of exogenous dTTP into the culture medium of shdTYMK knockdown cells. Since dTTP is highly charged, we speculated the possibility of an unknown putative dTTP transporter or regulatory mechanisms that break down exogenous dTTP into thymidine derivatives and reform intracellularly through an undetermined pathway capable of bypassing dTYMK loss of function. Taken together, HCT-116 and B1 cells were removed from their L-glutamine and supplement-rich in vitro setting to a more representative in vivo environment. B1 cells were so chosen for the apparent 1:5 wild-type to dTYMK mutant allele frequency result previously shown in figure 30.

116

3.9. In vivo assessment

Tumour xenograft of HCT-116 (WT) and B1 cells were produced in female BALB/c nude mice aged 6–8 weeks89. Tumour growth was sequentially monitored (every 1-2 days) through calliper measurements for 30 days or until tumour volumes reached approximately 120-140 mm3. In contrast to the in vitro proliferation assay, the

CRISPR/Cas9 knockdown of dTYMK resulted in a marked impairment of growth in B1

(figure 36). Western blot analysis of xenografts revealed a consistent 2.7-fold reduction of dTYMK expression in B1 (0.63 ± 0.22, p<0.001) when compared to parental HCT-116 (1.72 ± 0.39). Furthermore, there were no significant differences between B1 and HCT-116 tumours for TK1 and TS expression (figure 37).

Our result illustrates the contrast between the in vitro and in vivo growth of B1 cells suggesting that media supplementation possibly glutamine release125 and/or other factors including glucose and oxygen126, may confer an advantage in the proliferation of cells in vitro. In an in vitro setting, a steady flux of respiratory substrates is readily available for maintaining the redox status through the glutathione system during periods of oxidative stress125,126. Glutamine metabolism has been well documented as a carbon and bioenergetic source for cancer cells125. In a tumour microenvironment, the availability of these substrates is dramatically reduced. We suspect that a deregulation in the mitochondrial functionality contributes, at least in part, to the tumour growth retardation of dTYMK knockdown.

117

1 6 0

1 4 0

) 3 1 2 0

m H C T -1 1 6 m

( 3 B 1

1 0 0

m u l 8 0

u 6 0

o m

u 4 0 T

2 0

0 0 5 1 0 1 5 2 0 2 5 3 0 3 5 4 0 D a y Figure 36. Tumour growth curves of HCT-116 and B1. BalbC/athymic nude mice were injected subcutaneously with 5x106 of either HCT-116 or B1 cells. Tumour dimensions were determined by caliper measurements and volumes calculated by the ellipsoid formula for estimating tumour mass. Points indicate mean tumour size (n=6) on specified days; bars, S.E.M.

1.5 2.0 *** 1.5 NS NS

1.5 1.0 1.0

1.0

0.5 0.5

- actin (OD/ mm - actin

- actin (OD/ mm - actin 0.5 (OD/ mm - actin



TS expression normalised TS expression 0.0

0.0 normalised expression TK1 0.0 dTYMK expression normalised expression dTYMK HCT-116 B1 HCT-116 B1 HCT-116 B1

Figure 37. Expression profile of the essential enzymes of the thymidine nucleoside pathway in HCT-116 and B1 xenografts. Representative western blot showing expression of stated proteins in xenografts. B, C, D. Quantification of dTYMK, TK1 and TS expression normalized to β actin (n=6). Stars represent significant differences.

118

We proceeded to evaluate the effect of reduced mitochondrial functionality on dTYMK knockdown cells. A ratio between proliferation under hypoxic and normoxic conditions were calculated to assess the effect of altered mitochondrial functionality on dTYMK knockdowns (figure 38). The hypoxia assay was performed over a period of 24-72 h in an attempt to mimic the tumour microenvironment where oxygen diffusion and nutriment supply into tissues is reduced. The lowest ratio at 72 h the lowest ratio of cell number under normoxia relative to hypoxia occurred with HCT-

116 cells (5.1) while B1, and B5 cells presented higher a ratio of 5.6 and 7.0 respectively. In parallel plates, lysates were collected and subjected to western blot analysis for the expression profiles of dTYMK, NDPK, TK1, TS and HIF-1α (a universal marker of hypoxia). All cell lines cultured for 24-72 h in low oxygen, compared to normoxia, showed a marked increase of HIF-1α under hypoxia when compared to normoxia and thus, confirming the successful induction of physiological hypoxia

(<0.1% O2) (Appendix, figure 6). Both dTYMK and NDPK protein expression were modestly reduced in B1 and B5 cells following 72 h hypoxia (Appendix, figure 6). TK1 and TS were also modulated suggesting an overall impact on the nucleotide synthesis pathway. However, further analysis with 18F-FLT radiotracer uptake could not be performed since hypoxic conditions would be lost at the point of tracer injection.

119

A 1.0

0.8 24 h Hypoxia 24 h Normoxia 0.6 72 h Hypoxia 0.4 72 h Normoxia

(492 nm) (492

Cell density Cell 0.2

0.0 HCT-116 B1 B5

B ** 8 ns 6

h 72 change Fold

normoxia vs hypoxia vs normoxia 0 HCT-116 B1 B5

Figure 38. Effect of hypoxia on proliferation following CRISPR/CAS9 mediated dTYMK genomic knockdown. A. Cell density as determined by SRB to investigate the effect of hypoxia on HCT-116 and mutant B1/B5 cells. B. Cell density fold-change in parental HCT-

116 vs mutant cells under hypoxic and normoxic conditions for 24 and 72 h. Data

represents mean of n=2 in quintuplicates ± SEM.

Taken together, the results indicate that overall hypoxia impairs cell proliferation independent of dTYMK status. A subtle differential sensitivity in B1 knockdown cells, when compared to parental at 72h, was observed under hypoxic conditions (figure

38B).

120

3.10. Localisation of dTYMK by fluorescent microscopy.

According to uniport (https://www.uniprot.org/), dTYMK is distributed between the nucleus, cytosol and the mitochondria. It became of interest to understand whether the subcellular localisation of dTYMK is altered following CRISPR/Cas9 genomic editing in the two mutant HCT-116 cell lines. To attain the spatial information of the protein's distribution in situ, HCT-1 16, B1 and B5 cells were subjected to immunofluorescent imaging. Upon antibody titration for optimisation, it became apparent that the Abcam antibody (dTYMKabcam), used in all experiments thus far, showed a cytoplasmic and perinuclear dTYMK distribution (figure 39). This prompted an investigation into the target sequence of the dTYMKATLAS, an antibody previously abandoned in section 3.3.2 (figure 16), to understand whether its predicted dTYMK subcellular localisation, particularly within the mitochondria, differed from that of dTYMKabcam. Since the expression profile of the critical enzymes in the mitochondrial nucleoside synthesis pathway are known to be unchanged during cell cycle modulation, it may also offer some insight as to why an even expression of dTYMK was observed in figure 16, following drug-induced cell cycle arrest. According to the review of dTYMKATLAS antibody (HPA042719), the enzyme appears to be localised in the cytoplasm and the mitochondria. DTYMK is reported to exist as two isoforms

(Isoform 1 = Ascension P23919 and Isoform 2 = Ascension P23919b) with the former being cytoplasmic and the latter speculated to be mitochondrial; however, the existence of a functional isoform of TMPK localised in mitochondria remains undefined. Next, we queried the target specificity of the dTYMKATLAS epitope sequence to potentially two of dTYMK. Indeed, a BLAST of the target sequence revealed a 100% match to both isoforms of dTYMK. In theory, dTYMKATLAS should bind

121

to all splice variants of dTYMK and reflect the total protein expression. In either cases, mutant cells (B1 and B5) localised to the puncta /mitochondrial (as predicted by

ATLAS) or cytoplasm/nuclear/perinuclear (Abcam) with the same distribution as wild-type HCT-116 (figure 39). This was not in accordance with the reduced expression of dTYMK expected in mutant cell lines; however, the IF distribution is qualitative and may not necessarily permit quantitative differences. As these images were obtained from an electron light microscopy, we suspect triplet-singlet transition (that results in photobleaching) to be the limiting factor in adequately distinguishing between signal and background.

122

ATLAS ABCAM

A B DTYMK DTYMK

HCT-116

Nucleus Actin Merge Nucleus Actin Merge

DTYMK DTYMK B1

Nucleus Actin Merge Nucleus Actin Merge

DTYMK DTYMK

Nucleus Actin Merge Nucleus Actin Merge

Figure 39. dTYMK mutants B1 and B5 show unaltered subcellular localization. HCT-116

wild-type cells transfected with empty vector, B1, or B5 were allowed to adhere to microscope slides, fixed with formaldehyde, stained with antibody against dTYMK from

A. Atlas (left panel) or B. Abcam (right panel) and imaged by fluorescent microscopy. Images are representative of three independent experiments.

123

3.11. Discussion

The CERES and project Achilles scoring models revealed a genetic dependency of tumour cells to dTYMK while the Cancer Cell Line Encyclopaedia presented a lack of correlation between dTYMK DNA copy number and mRNA expression suggesting other processes to be the driving factor of dTYMK expression. It was further observed that a knockdown of dTYMK in specific cancer cell lineages (figure 12B), i.e. acute myeloid lymphoma, did not result in reduced cell survival as per the CCLE gene essentiality score. This inferred that genetic alterations/copy number loss of dTYMK may not give rise to the loss of DNA synthesis. Our speculations were supported by

Hu and Chang 2008127 and Hu et al. 2012110 who reported that a dTYMK knockdown does not induce genotoxicity nor inhibit cell proliferation. Gaggriven the crucial role of dTYMK for dTDP formation, it is unclear why cells can endure such insults without decreased cell growth. The authors proposed the production of variant dTYMK isoforms that may account for residual dTDP production following dTYMK inhibition; however further genetic analysis was not conducted by the group.

An extensive characterisation of dTYMK expression in a panel of cell lines in control and treatment conditions were conducted before utilising the CRISPR/Cas9 system to stress the DNA synthesis pathway extrinsically. Two stable clones, B1 and B5, were created by introducing mutations into wild-type HCT-116 cells using CRISPR/Cas9- mediated homology-directed repair. CRISPR constructs 25926363 and 25926373 were raised against exon 1 to target all splice variants and isoforms of dTYMK to create non-homologous end joining mediated frameshifts. Following nucleofection, allele sequencing revealed a mutant to wild-type ratio of 5:1 in B1 and 4:1:1 (Allele

124

2: Allele 3: WT) for B5 suggesting, an incomplete CRISPR knockout (figure 28). As a more direct readout of protein expression, western blot analysis was employed and confirmed notion of an incomplete knockout as B1, and B5 cells still expressed 33 ±

2 % and 42 ± 6% levels of dTYMK, respectively. It is known that a shortage of dTDP initially propagates dTMP accumulation which causes nucleotide imbalance. To circumvent genetic instability that arises from this imbalance, it has been suggested by Munch-Petersen et al. 199578, amongst others128,129, that TK1 expression/activity is regulated by dTTP in a negative feedback mechanism that prevents thymidine induced-cytotoxicity. Surprisingly, in vitro expression levels of TK1 and TS remained relatively even between wild type and mutant cells further echoing the possibility that a dTYMK knockdown does not cause genotoxicity or activate DNA damage responses within the time points examined (figure 31). Whether longer time points would have resulted in genotoxicity and alteration of cell growth remains unknown.

We hypothesised that an inhibition of dTYMK activity should confer a growth disadvantage due to the reduced TDP pool for DNA incorporation. However, this was not evident in figure 32 which showed comparable proliferation rates in wild-type and mutant cells. As indicated in figure 33, mutant cells have smaller 4N peak in DNA content and an accumulation of cells at G1/S-phase that perhaps the ability of this cancer cell line to undergo polyploidy is curtailed. Further work to benchmark dTYMK expression (and localisation) in normal cells will clarify this hypothesis. Owing to this, an 18F-FLT incorporation assay was used to assess the relationship between dTYMK knockdown and the changes in the nucleoside salvage flux. We noted a significant reduction of overall 18F-FLT uptake (figure 35) and FLTMP metabolite accumulation

(Table 6) in mutant cells thus, confirming dTYMK as the rate-limiting step in the

125

salvage nucleotide pathway. This again opens the question as to why cells can tolerate dTYMK knockdown and sustain cell proliferation when dTYMK functionality and in turn, salvage nucleotide synthesis has been deregulated. Indeed, it can be argued that the discrepancy between growth rate and dTYMK expression is solely because of an incomplete knockout but, given its essentiality, we also consider the likelihood of compensatory mechanisms that mediate nucleotide synthesis in the case of an unexpected dTYMK functional loss. Our observations were extended into an in vivo setting which revealed a marked impairment of growth in B1 tumours when compared to parental HCT-116. Western blot analysis confirmed an unchanged dTYMK expression phenotype in B1 xenografts. Considering that the mitochondria has its own pathway for nucleotide synthesis we suspected the contrast between the growth of B1 cells in vitro and in vivo to be due to mitochondrial compensation. In an in vitro setting, a steady flux of substrates are readily available in media which maintain the redox status through the glutathione system during periods of oxidative stress130. In the tumour microenvironment, nutrients including glucose, glutamine and oxygen are limited, and efficiency of the mitochondria and the cell in general. A putative mitochondrial dTYMK, known as dTYMK2, was reported by Chen et al.

200824 as a supplier of dTDP in the mitochondria when cytosolic dTYMK expression is reduced during mitosis. In proliferative cells, dTDP is recruited from the cytoplasm into mitochondrial matrix via dNDP transporter and converted to dTTP by mitochondrial NDP kinase28. Since the mitochondria DNA synthesis is not cell cycle regulated24, a dTTP supply is still needed for mtDNA post-mitotic division. We initially considered this as plausible compensatory route following dTYMK knockdown however, the CRISPR constructs 25926363 and 25926373 were targeted to exon 1

126

which is present in all variants dTYMK whether cytosolic or the putative mitochondrial protein. This prompted an investigation (by Basic Local Alignment

Search Tool) into the protein sequence,

MAFARRLLRGPLSGPLLGRRGVCAGAMAPPRRFVLELPDCTLAHFALGADAPGDADAPDP, proposed by Chen et al. 2008 for mtdTYMK. A BLAST revealed a match with cytidine/ uridine monophosphate kinase 2 (CMPK2) and not dTYMK. Furthermore, the sequence has been referenced and attributed to CMPK2 on the NCBI database

(Reference Sequence: NM_207315.3). On this basis, we hypothesise the duality of

CMPK2, localised in the mitochondria, as the compensatory enzyme that supplies the dTTP pool for both mitochondrial and cytosolic DNA synthesis at points of genomic stress. Although Chen et al. 2008 mistook CMPK2 for dTYMK2, their research showed a great reciprocal relationship in the protein expression pattern of dTYMK1 and

CMPK2 during monocytic/macrophage differentiation. They also noted that an overexpression of CMPK2 increased the steady-state level of cellular dTTP and promoted the conversion of radioactive labelled-thymidine and -dTMP to dTDP and dTTP in mitochondria. This suggests that CMPK2 may substitute for cytosolic dTYMK during dTTP synthesis. Moreover, Floyd et al. 2007131 published the discovery of the first mammalian mitochondrial carrier PNC1, as a transporter for UTP, TTP and CTP further supporting the speculation of mitochondrial and cytosolic dTTP exchange.

In an in vitro setting, a knockdown of dTYMK was portrayed as dispensable (figure

32). We suspected the components of media and in particular, L-glutamine and folic acid as contributors at least in part, to producing precursors that sustained proliferation rate in vitro. In the de novo pathway, deoxynucleoside triphosphates

(dNTPs) are generated as a multistep system involving the use of 5-phosphoribosyl-

127

1-pyrophosphate, glutamine, glycine, aspartate, and folic acid. Since animal cells cannot synthesise folic acid, as reported by Murray et al. 1999, folic supplementation into culture media is essential for activating de novo DNA synthesis. In an in vivo setting, the availability of folic acid132 is reduced and in turn, impedes the compensational ability of the de novo pathway. Glutamine amide nitrogen, readily available in tissue culture media, is known to be utilised by cells for the synthesis of purine nucleotides, CTP from UTP and for the production of carbamoyl phosphate in the de novo synthesis of pyrimidines133. Therefore, it is possible that mutant cells utilise carbamoyl phosphate synthase and its downstream ribonucleotide reductase to produce carbamoyl phosphates and eventually deoxynucleotide diphosphates such as dUDP; following a cascade of events. The dUDP formed in cells are targeted for conversion into thymidylate monophosphate by TMP synthetase. This, in combination with the TMP pool derived via the TK1 cytosolic conversion of thymidine, could supply the mitochondria CMPK2 with TMP via the TOM20 transporter 134. It offers an explanation as to why TK1 and TS expression remained unchanged since an equilibrium of nucleotides may have been established via CMPK2 maintenance of dTDP, thereby averting nucleotide imbalance which can lead to cell death over time. It has also been reported by Hu et al. 2011135, that the level of CMPK can be increased by 500% or decreased by 95–98%. In cells with overexpressed

CMPK, they reported a feedback regulation of different enzymes and their products that balance out the impact of CMPK activity. The group concluded that the CMPK expressed in RKO cells is not critical for the phosphorylation of dCMP and the maintenance of natural nucleotide pools135. We query whether the lack of essentiality noted by the group was due to the kinase remaining dormant and only

128

activated during periods of genomic stress/ nucleotide imbalance. While this study refers to nuclear CMPK, it is possible that CMPK2 exhibits the same phenotype which has been well documented in other cytoplasmic/mitochondrial duo enzymes such as

TK1-TK2. If this is the case, the extreme increase in kinase expression and subsequent decrease to near zero may offer support to the notion of activation during genotoxic insults pertaining to dTYMK loss of function. We proceeded to evaluate the effect of reduced mitochondrial functionality on dTYMK knockdown cells by subjecting the mutant cells to hypoxic and normoxic conditions in vitro. The hypoxia to normoxia ratio of B1 at 72 h showed a modest increase with respect to HCT-116 but, were not statistically significant. It is possible that the 24-72 h hypoxia window was not sufficient in mimicking the tumour microenvironment where tumour vascularisation is usually inadequate causing a limitation of oxygen diffusion and nutrient supply into tissues. However, we also consider the likely activation of mitochondrial genes that overcome the effect of hypoxia and to give the phenotype observed in dTYMK knockdown cells (figure 38). The use of mitochondrial inhibitor inhibitors like metformin may be a better alternative to hypoxia for inhibiting mitochondrial function. If indeed the mitochondria plays a role in the compensatory pathway, a pronounced decrease (following therapeutic intervention) in B1 and B5 cell survival should be observed when compared to WT.

A study by Lui et al. 201332 reported the growth retardation of LKB1-WT cells following dTYMK knockdown. This is contrary to both Hu et al. (2012) and the findings of my study which displayed relative ease of growth under dTYMK knockdown. The group proceeded to evaluate methodologies for rescuing the small hairpin dTYMK

(shdTYMK) growth phenotype by adding dTTP into the culture medium. Indeed,

129

growth retardation reversed, and it was concluded that targeting dTYMK is synthetically lethal in LKB1-WT cells. While the authors did not elucidate how a highly charged molecule like dTTP gains entry into cells, we speculated the possibility dTTP breakdown into its thymidine derivative and its subsequent active transport into cells by ENT1/2 transporters. Here we believe the readily available thymidine is converted into TMP and transported into mitochondria (via TOM20) for the supply of dTTP.

Indeed, if CMPK2 can substitute for dTYMK at points of dTYMK functional stress, it can be argued that the 18F-FLT incorporation and rHPLC metabolite assays (figure 35) should display the same profile as parental HCT-116 since the dTTP pool is maintained by CMPK2. However, the prevalence of 18F-FLTMP over 18F-FLTDP and

18F-FLTTP has been suggested by Grierson et al. (2004) to be a consequence of reduced substrate avidity of FLTMP towards dTYMK due to, the 18F- positioning at the 3’ of ribose sugar16. Therefore, a knockdown should potentiate this phenomenon due to the reduced bioavailability of dTYMK. Since the mitochondrial and cytosolic nucleotide synthesis machinery show the same dynamic behaviour, it is possible that

CMPK2 (with its similarities to dYTMK) may not be avid to FLTMP leading to the radiotracer accumulation in its monophosphate form. Finally, a phylogenetic analysis and multiple sequence alignment of CMPK2 with other thymidylate kinase was conducted by Xu et al. 2008 which indicated that the CMPK2 had a thymidylate kinase domain136.

While many of these conclusions remain speculative one thing is clear, we report the first observation of a growth retardation in vivo following a CRISPR/Cas9 mediated dTYMK knockdown and, shed light on the possible compensatory mechanisms that

130

become more apparent in vitro than in vivo. Given the presumed essentiality of dTYMK, it is not unreasonable to suggest these mechanisms (figure 40) as plausible compensatory pathways which circumvent the genotoxic impact of dTYMK’s reduced functionality. We also put forward the possibility of TS duality as more than just a dUMP methylator or, the presence of an unknown methylator of dUDP or dUTP into dTDP and dTTP. It would be of worth docking other uridine phosphate species into the active site of TS in silico. From Hanson et al. (2009) report, it is acknowledged that around 50 % proteins coded in the human genome remains unknown 137. Finally, appreciating the actinomycin D and cycloheximide in vitro results from section 3.5; figure 26, we do not disregard the possibility that the remaining wild type dTYMK in mutant cells may be the progenitive factor for the phenotype observed both in vitro and in vivo. To validate our hypothesis, it would be of worth re-editing the mutants with the CRISPR/Cas9 system to create a complete knockout. Survival of these cells, following knockout, will significantly support the notion of compensatory pathways and fully disclose dTYMK essentiality since a complete knockout, in theory, should result in complete cell kill. While the discrepancy between in vitro and in vivo proliferation exists, tumour xenografts are better representatives of human tumorigenesis making the models a worthy platform for assessing thymidine monophosphate bioisosteres for the future evaluation of imaging tools that report on dTYMK activity

131

Figure 40. Schema of alternative contribution to dTTP biosynthesis involving the possible import of dUMP into the mitochondria.

132

Chapter 4

Evaluation squaramide thymidine monophosphate derivative for PET imaging of cell proliferation

133

4.1. Impetus for the use of TDR monophosphate

Although 18F-FLT showed great promise and correlation with dTYMK expression under cell cycle modulation (with the exception of nocodazole treatment), it is still dependent upon TK1 activity and disregards the relative contributions of the de novo pathway. Furthermore, tumour specific characteristics such as; thymidine phosphorylase and serum/tissue thymidine levels, the consideration of patient- specific and cancer cell lineage, species dependent glucorination in humans, possible efflux of 18F-FLT or FLTMP from cells, the role of a putative deoxynucleotidase and differences in expression or modulation of ENT-1 are all known to impact the level of tracer accumulation (section 1.2.4). 18F-FLT is, therefore, context-dependent and produces variable, and at times, unreliable proliferative indices that are disproportionate to the rate of DNA synthesis (with respect to Ki67) 77. Therefore, the development of a tracer that demonstrates an independence to TK1 and TS activity, by reporting on dTYMK activity, would overcome these biological limitations would be indispensable for imaging cell proliferation. Such a radiotracer would ideally mimic 18F-FLTMP by presenting its fluorine at the nucleobase as opposed to the pentose sugar (figure 41). However, an incorporation of a phosphate group into

18F-FLT along with fluorination on nucleobase would prove to be challenging during synthesis. More importantly, a tracer presenting a highly charged phosphate group would be the rate-limiting step in the passage of the molecule through the cell membrane. To circumvent these challenges, we investigated bioisosteres of phosphates in literature. Amongst existing bioisosteres, squaric acid diamides have been modelled as adequate phosphate isosteres in DNA. Sato et al. (2002)81 and Seio et al. (2005)80 reported the synthesis of squaryl-containing thymidine and cytosine

134

deoxynucleoside 5’-phosphate analogues and their successful incorporation into an oligonucleotide. It became of interest to explore these findings for the purpose of

PET imaging. The squaramide moiety (derived from squaric acid) present electrostatic potential maps that are indicative of reasonable phosphate mimicry through the charges emulated by the squarate’s key oxygen atoms80. The unique structure of a squaric acid, two acidic hydroxyl groups with calculated pKa values of

0.54 and 3.48, as well as two highly polarised carbonyl groups, provides proton acceptor sites at the carbonyl function for hydrogen bonding with other molecules81.

Although these moieties are useful candidates for generating biologically active anticancer agents, to our surprise no in vitro or in vivo biological evaluation has been reported. Owing to this, we aimed to evaluate the use of a 18F-FLTMP analogue as a potential proliferation radiotracer that reports on thymidylate kinase activity in vitro and in vivo.

5’-squarylamide-3’-[18F]fluoro-3’-deoxythymidine (18F-SqFLT), synthesised by Dr

Diana Brickute of Imperial College London, mimics FLTMP through the addition of a squaramide coupled to the 5’ of the fluorothymidine pentose sugar (figure 41). The squaryl-containing nucleotide is hypothesised to passively diffuse across the membrane through the resonance structure of squaramide derivatives that modulate the charge distribution on the squaramide moiety81. The squaryl group, within milliseconds, can be reversibly stabilised by the formation of a typical 2π aromatic system fulfilling the Hückel's rule of aromaticity between the carbonyl and endocyclic double bonds138. In a biological system, we envision the entry of stabilised species into cells and its subsequent recognition by dTYMK during periods of charge redistribution (figure 41). Since end dTYMK is a convergent enzyme of both the

135

salvage and de novo pathway, it is believed that its rate of activity should be directly proportional to the rate of DNA synthesis. Thus, successful entry and subsequent phosphorylation of 18F-SqFLT by dTYMK, should permit non-invasive imaging and quantification of the tumour proliferative fraction.

A 18F-FLTMP

18F-SqFLT

B C

Figure 41. Structure of 18F-FLTMP and species of 18F-SqFLT during charge redistribution. Representative schema of novel phosphate mimic and resonance structure. A. 18F-FLTMP B. 18F-SqFLT form anticipated to permeate cell membrane. C. 18F-SqFLT form anticipated to be recognised by dTYMK.

4.2. In silico modelling of 18F-SqFLT for target validation

For our application, it was essential to validate the squaric moiety as a substrate to dTYMK in silico before the development of 18F-SqFLT. To further our understanding of the catalytic mechanism of the enzyme, we employed the use of Gaussian 09 and

UCSF Chimera to model high-resolution structures of human dTYMK in complex with

TMP and adenosine diphosphate (ADP) along with 18F-SqFLT and ADP. Molecular docking is a useful technique to study the interactions between a small molecule and

136

a protein at an atomic level139. In silico modelling was carried out in collaboration with Dr Karl Thorley of Imperial College London. An oniom method (using micro- iterations and quadratic coupled algorithm) was employed to optimise the system geometrically. The enzyme was treated through molecular mechanics universal force field (UFF), while the small molecule was modelled with more accurate density functional theory (B3LYP/6-31G*). The binding affinity for each guest molecule into the enzyme and an exposition of the fundamental biochemical processes can be predicted through mathematical simulations139. Binding energies were calculated by the difference in energies between the whole system minus the energies of the enzyme and guest molecule alone, taking into account the differing levels of theory used for each component. Our data indicated the interaction energy (ΔG; delta G) of

18F-SqFLT with the catalytic site of dTYMK (figure 42) to be comparable or slightly increased when likened to TMP. Moreover, the OH group of the 18F-SqFLT appears to be near the phosphate group of ADP (figure 42).

137

A B

ADP ADP

Table 7. B3LYP/6-31G(d): UFF

dTYMK code Compound (ΔG) kJ/mol

1nmy TMP -221.7 1nmy 18F- SqFLT -223.1

Figure 42. Stick representation of human ADP-bound dTYMK residues of the P-loop (13−17) coloured in white and enclosing either TMP or 18F-SqFLT. A Space-filling model of dTMP with dTYMK active site (P-loop) in the presence of ADP and Mg2+. B Space-filling binding interaction model of [18F-SqFLT with dTYMK active site (P-loop) in the presence of ADP and Mg2+. UCSF Chimera used to model high-resolution structures of dTYMK in complex with specified compounds. Table 7. Representative interaction energies (ΔG) kJ/mol were predicted through mathematical simulations using Gaussian 09. Colour code: Carbon, C: black Hydrogen, H: white Oxygen, O: red Nitrogen, N: blue Phosphorus, P: orange. The big green space filler represents 18F whilst smaller filler represents Mg2+.

138

Encouragingly, the structural overlay of 18F-SqFLT showed great flexibility of the squaramide in the catalytic site of dTYMK (figure 43).

We can conclude, at least in silico, the potential recognition of 18F-SqFLT by the target enzyme and the capability of phosphotransfer in the presence of ADP + Mg2+.

Figure 43. Structural overlay of active site residues for dTYMK and 18F-SqFLT. Structures showing rigidity of 18F-SqFLT into the dTYMK pocket and in particular, residues of the P-Loop. Thin sticks = dTYMK, Thick sticks = 5’squarylamide-FLT

139

4.3. In vitro evaluation

4.3.1. Cell uptake

Initial studies of 18F-SqFLT were performed in HCT-116, B1, B5, OST TK1- cell and Kelly cell lines. OST TK1- (TK1-deficient human osteosarcoma cell line) were included as a differential cell model for evaluating the contributions of de novo vs salvage pathway in DNA synthesis while, Kelly was added for its low copy number expression of dTYMK (Kelly = 0.514984173,

HCT=1.985772018) as per Achilles database. All experiments were compared to 18F-FLT. As expected, the uptake of 18F-SqFLT was significantly lower in B1 (0.14 ± 0.04 % ID/mg) cells than HCT-116 (0.24 ± 0.04 %ID/mg). Surprisingly, B5 cells (0.22 ± 0.02 % ID/mg) had comparable uptake to that of HCT-116 with no significant difference (figure 44A). This is in contrast to the highly significant reduction noted with 18F-FLT uptake (figure 44C). Although

B1 cells showed great promise and correlation with 18F-FLT uptake, the absolute uptake of the radiotracer was minimal across all the cell lines (<0.3 % ID/mg). Uptake in differential models OST TK1- and Kelly were also comparable to HCT-116 again with no significant difference between the three groups (Figure 44B). The next step was to verify the uptake characteristics 18F-SqFLT with increasing time and extract the most useful pharmacokinetic (or time-dependent) information about radiotracer retention in cells.

An 18F-SqFLT time-activity curve was generated in HCT-116 cells at 5, 15, 30, 60, 90 and

120 minutes. Figure 44D reveals a peak uptake of 18F-SqFLT at 15 minutes and a subsequent drop to a steady rate between 30 – 120 min (Figure 42D). Paradoxically, 18F-

FLT accumulation increased over time due to the continuous phosphorylation of radiotracer into its mono-di and triphosphate form (Figure 42E)16.

140

A ns B ns 0.3 ** 0.3 ns

0.2 0.2

0.1 0.1

F-SqFLT uptake

F-SqFLT uptake F-SqFLT

(% ID/mg of protein)

(%ID/mg of protein) of (%ID/mg

0.0 0.0 HCT-116 3B1 3B5 HCT-116 OST TK1- KELLY

C *** *** 20

F-FLT uptake F-FLT

18 5

(%ID/mg of protein) of (%ID/mg

0 HCT-116 3B1 3B5

D 0.4 E

0.3

0.2

F-SqFLT uptake

18 0.1

(% ID/mg(% of protein)

0.0 15 30 60 90 120 Time (mins)

Figure 44. The effect of 18F-5’Squaryl-FLT and 18F-FLT cell uptake in specified cell lines. Cells were seeded into 6-well plates at 3.0x105 cells/well. After overnight culture fresh media containing 18F-5’Squary-TdR was added to each well yielding a final concentration of 0.74 MBq/well in a volume of 1ml. A. Uptake in HCT-116, B1 and B5 cells following 1 h incubation of ~0. 74 MBq 18F-SqFLT in cells. B. Graph of 18F-SqFLT in HCT-116, OST TK1- and Kelly cells following 1 h radiotracer incubation. The graphs represent the average values and the standard error of the mean. Statistics were performed using One-way Anova and Bonferroni's multiple comparison test as a post hoc analysis. C. Representative graph of 18F-SqFLT time course in HCT-116. D. Representative time course assay of 18F- FLT uptake in HCT-116. The graphs represent the average values (n=6) and the standard error of the mean.

141

4.3.2. Evaluation of competition and transporter modulation

Appreciating that there may be interactions with nucleotide transporters, we proceeded to investigate the effect of pharmacological doses (1 ng/ml - 10 µg/ml) of cold 19F-SqFLT and thymidine (positive control) on 18F-FLT uptake. As 19F-SqFLT is hypothesised to diffuse into cells passively, our initial experiment aimed to investigate the possible saturation of ENT1/2 transporters by 19F-SqFLT during 18F-

FLT uptake. The effect of dose escalation of thymidine was also assessed as a positive control. As expected, the plot of activity versus log thymidine concentration (Figure

45A) revealed a monophasic competition curve with a 50 % inhibitory concentration of ~ 1 µg/mL in HCT-116 cells while, log 19F-SqFLT concentration (Figure 45B) showed little changes in 18F-FLT uptake. It can be concluded that 19F-SqFLT is not a substrate for TK1 or ENT-1/2 transporters since saturation was not observed. A metabolite rHPLC assay was also conducted to validate 19F-SqFLT as a non-substrate for TK1 and, investigate 19F-SqFLT competition with FLTMP for dTYMK active site. As a more sensitive assay, we utilised rHPLC to analyse 18F-FLT conversions to mono, di, and triphosphate under 100 ng/ml and 1 µg/ml of either 19F-SqFLT or thymidine in situ.

In keeping with literature observations of thymidine-FLT competition for ENT-1/2 transporters and TK1 active site, an increase in the proportion of 18F-FLT from 2.44

% under control conditions to 16.75 % under 1 µg/ml of TdR can be observed (table

8). However, the 19F-SqFLT concentration had little impact on 18F-FLT conversions.

Since 19F-SqFLT is a substrate to dTYMK, it was envisioned to have no impact on ENT-

1-thymidine modulation nor TK1 activity. However, the 18F-FLTMP-18F-FLTTP presented subtle differences between 100 ng/ml, 1 µg/ml and control conditions.

142

In conclusion, our data suggests that 19F-SqFLT had no inhibitory or saturating effects on TK1 and ENT1/2 transporters. The modest change in FLTMP-FLTTP ratio, noted in table 8, suggests a slight competition of 19F-SqFLT with 18F-FLT for dTYMK active site.

Indeed, a drastic shift in FLTMP-FLTTP ratio was anticipated; however, this was not the case and may be due to slow passive diffusion. Nonetheless, 18F-SqFLT (figure

44A) appeared to show some specificity as per effect on B1 cells which correlated well with 18F-FLT uptake. Although 18F-SqFLT passive diffusion is anticipated to be slow, it is specific to dTYMK and, therefore, put forward the possibility of 18F-SqFLT providing a good target – to -background ratio; a prerequisite for in vivo studies.

A B

30 30

25 25

20 20

15 15

F-FLT uptake 10 F-FLT uptake 10

(%ID/mg protein)

(%ID/mg protein) 5 5 0 10 -3 10 -2 10 -1 10 0 10 1 10 2 0 10 -3 10 -2 10 -1 10 0 10 1 10 2 Thymidine (µg/ml) 19F-SqFLT (µg/ml)

Figure 45. Assessment of thymidine or 19F- 5’squaryl-FLT competition with 18F-FLT.

HCT-116 cells (at 60 – 70 % confluence) were incubated with either thymidine or 19F-

18 5’squaryl-FLT at 1ng/ml – 10 µg/ml for 20 min prior to co-incubation with 0.74 MBq F- FLT. A and B. The plot of activity versus log thymidine/19F- 5’squaryl-FLT concentrations with accumulation values representing tracer activity normalised to total protein. Error bars represent SEM for n=5 samples.

143

A B 18F-FLTMP 18F-FLTMP

800 250

200 600 18F-FLTTP 18F-FLTTP 150 18  400 F-FLT  18F-FLTDP 100 18F-FLTDP 200 50

0 0

00:00 05:00 10:00 15:00 20:00 25:00 30:00 00:00 05:00 10:00 15:00 20:00 25:00 30:00 Control Time (min)

C 18F-FLTMP D 18F-FLTMP

1000 1000

800 800

18 18 600 F-FLTTP 600 F-FLTTP   400 400 18 18F-FLTDP F-FLTDP 200 200

0 0

00:00 05:00 10:00 15:00 20:00 25:00 30:00 00:00 05:00 10:00 15:00 20:00 25:00 30:00 Time (min) Time (min)

Table 8. Integration of rHPLC 18F-FLT chromatograms under specified conditions

Conditions FLT area of FLTMP area FLTDP area FLTTP area of peak (%) of peak (%) of peak (%) peak (%)

Control lysates 2.4, 1.3 62.4, 64.8 3.7, 2.2 31.5, 31.7

1 µg/ml Thymidine 16.8, 18.2 52.6, 56.4 3.1, 2.7 27.5, 22.7

100 ng/ml 0.0, 0.0 67.8, 64.1 3.4, 3.1 28.8, 32.8

19F-Sq-FLT 1 µg/ml 0.0, 0.0 68.3, 64.8 4.1, 3.6 27.6, 31.6 19F-SqFLT

Figure 46. Metabolite assessment of thymidine or 19F- SqFLT competition with 18F-FLT by rHPLC. A, B, C and D. Representative chromatograms of control DMSO (0.33 %), 1 µg/ml thymidine, 100 ng/ml 19F-SqFLT and 1 µg/ml 19F- SqFLT following 20 min pre-

18 incubation and subsequent co-incubation with 3.7 MBq F-FLT. Table 8. Integration table of 18F-FLT metabolites under specified conditions.

144

4.3. Thymidine salvage and global metabolism

It is known that cancer cells reprogram their cellular energetics during periods of bioenergetics and redox stress. As a result, we sought to investigate the effect of

CRISPR/Cas9 mediated dTYMK knockdown and determine the correlation between uptake of 18F-FLT, 18F-SqFLT and other markers of metabolism. We focused on the assessment of glucose and choline uptake as measures of initial metabolism of glucose consumption along with changes in the demand for phospholipid building blocks during oncogenic development.

Since the proliferation rate of mutants remained relatively unchanged when compared to parental HCT-116 cells (section 3.7; figure 32), we anticipated an increase in the pentose phosphate shunt of mutant cells to produce ribose 5- phosphate for sustaining cell proliferation in vitro. Interestingly, 18F-FDG significantly decreased in both B1 (2.92 ± 0.21 % ID/mg; p <0.001) and B5 (2.82 ± 0.28 % ID/mg; p<0.001) indicating a decrease in the bioenergetic and anabolic demands or a switch to lactic acid fermentation in mutant cells when compared to parental HCT-116 (4.20

± 0.25 % ID/mg) (figure 47A). Nevertheless, 18F-FDG correlated with 18F-FLT (figure

44C) within the same cell lines. In contrast, 18F-D4-FCH, increased from 10.46 % in

HCT-116 to 12.44 %; p<0.001 and 13.42 %; p<0.001 in B1 and B5 cells respectively

(figure 47B).

145

A B *** ***

5 *** 15 *** 4

10 3

2 5

F-FDG uptake F-FDG

F-D4-choline uptake

1 (% ID/mg of protein)

(%ID/mg of protein) of (%ID/mg

0 0 HCT-116 3B1 3B5 HCT 116 B1 B5

Figure 47. Uptake of 18F-FDG and 18F-D4-Choline in HCT-116 and mutant cells. Cells were seeded at 3 x105 cells/well and incubated overnight. 0.74 MBq radiotracer was subsequently incubated in cells for 1h before being washed thrice and PBS and lysed for gamma counting. A. 18F-FDG uptake. B. 18F-D4-Choline uptake. Data represents mean of n=6 ± SEM. Statistical significance was determined by Bonferonni’s post hoc analysis = p<0.5, ** p<0.01, *** = p<0.001

As a more sensitive assay, we utilised the unprecedented coverage capacity of hydrophilic interaction liquid chromatography high-resolution mass spectrometry

(HRMS) to a measure a broad diversity of polar metabolites including TMP and TTP.

This analysis was performed in collaboration with Dr Alexandros Siskos and Dr Eirini

Kouloura of Imperial College London. We aimed to measure and interpret a range of central carbon metabolites (implicated in nucleotide metabolism, glycolysis, TCA cycle and fatty acid metabolism) however, due to instrumental limitations and an incomplete database, characterisation of matches to the available spectral library were limiting which, in turn, hampered the identification of most metabolites.

Nevertheless, a positive mode analysis under acidic conditions could be conducted to generate metabolites relating to nucleotide metabolism. A runtime 30 min per

146

analysis was applied to the system to preserve the high-throughput and increase chromatographic performance. Figure 48 represents the % change of metabolites in

B1 and B5 compared to HCT-116. For most metabolites, B5 appears to oppose the results of B1. This was an unexpected phenotype since the 2 mutants were anticipated to behave similarly.

There was a modest rise in TMP in B1 (+ 4 ± 1 %) and a moderate decrease of TTP (-

15 ± 10 %) when compared to HCT-116. B5 cells displayed high TMP (+37 ± 14 %) and an unexpected increase in TTP (+20 ± 10 %) (figure 48). It is well understood that an accumulation of the substrate on which the enzyme normally acts, may reflect the loss of function of the enzyme in question. Therefore, an accumulation of TMP suggests an inhibition of its conversion into TDP due to dTYMK knockdown.

Interestingly, the level of dUMP decreased by 21 ± 6 % in B1 cells, possibly due to its consumption. A 24 ± 3 % drop in ATP was also noted in figure 48. It is likely that the reduced uptake of 18F-FLT previously observed in figure 44C, is due to a reduce TK1 enzymatic activity as a consequence of reduced ATP substrate availability.

Correlation between the levels of ATP as a cofactor for TK1 activity has been shown by Barthel et al. 2005 and Schiepers et al. 2007140 during the analysis of 18F-FLT kinetics in glioma patients. The 33% increase of UDP-N glycosylation in B1 cells suggests an upregulation of protein glycosylation as a protective response for cell tolerance to cellular stress141. The remaining metabolites show a high degree of variability and appear to play essential roles in nucleotide metabolism. It would be of worth exploring the modulations of these metabolites, in dTYMK knockdowns, for future validations in vitro.

147

+37 % TMP disodium salt + 5 % NADP NAD glucuronic acid L-glutathione reduced glycerol 3 phosphate lithium salt Dihydroxyacetone phosphate dilithium salt Phosphoenoylpyruvate monosodium hydrate 4-DeoxyL-erythronic acid sodium salt hydrate B1 GTP sodium salt hydrate B5 CTP disodium salt ITP trisodium salt UTP tris salt dCTP disodium salt ATP disodium salt hydrate GDP sodium salt dATP disodium salt +20 % TTP sodium salt -15 % L-glutathione oxidised IDP sodium salt UDP sodium salt hydrate citric acid DL-isocitric acid trisodium hydrate dADP sodium salt IMP disodium salt hydrate UMP UDP-N-acetyl glucosamine sodium salt Aspartic acid +25 % dUMP -21 % dAMP Glutamate -100 -75 -50 -25 0 25 50 75 100 125 150 Metabolite intensity (% of WT HCT-116)

Figure 48. Variation patterns of metabolites involved in purine and pyrimidine metabolism, glycolysis and TCA cycle in B1 and B5 cells. Data was generated by normalising the area under the curve (AUC) of each metabolite to the AUC of internal standard. These were further normalised to cell count before taking a % of intensity with respect to parental HCT-116. Data represents mean of n=3 ± SD.

148

4.4. In vivo PET imaging of 18F-SqFLT and 18F-FLT

In vitro uptake results encouraged a dynamic PET imaging of 18F-SqFLT in BALB/c athymic nudes inoculated with HCT-116 and B1 cells as a pilot study. All images were decay corrected and reconstructed with a 3-dimensional maximum likelihood estimation method (3D ML-EM). From PET measurements, a time-activity analysis was performed by defining regions of interest (ROIs) on selected organ scans. For comparison, dynamic imaging with 18F-FLT was also performed in the same mice 48 h post 18F-SqFLT scan. Similar to in vitro cell uptake (figure 44), both HCT-116 and B1 tumour-associated radioactivity were low for 18F-SqFLT over 60 min scan; at 13 min

(HCT-116: 1.6 ± 0.35, B1: 0.53 ± 3.54 MBq/ml). Although a 3-fold reduction was observed, the absolute uptake values in tumours were similar to that of background and brings to light the inferiority of 18F-SqFLT in depicting the tumour proliferative fraction. In contrast, tumour-associated 18F-FLT radioactivity linearly increased in both HCT-116 and B1 tumours. As expected, and suggested by previous in vitro growth data, 18F-FLT significantly decreased (p<0.1) in B1 tumours when compared to HCT-116. Images from both cohorts (HCT-116 and B1) attributed the highest accumulation of 18F-SqFLT in kidneys, liver and bladder (figure 49). A high uptake of

18F-SqFLT was also observed in the gall bladder which indicates a partial excretion via the hepatobiliary route. Taken together, 18F-SqFLT failed to successfully trace the tumour proliferative fraction with respect to dTYMK activity and, produced a high background signal (figure 49). Figure 50A-C represents the time-activity curves

(TACs), normalised uptake value at 40-60 min (NUV40-60) and area under the curve

149

(AUC) values for 18F-SqFLT uptake in both HCT-116 and B1 xenografts. The TAC showed initial uptake into the tumour which was followed by a timely washout of the radiotracer (figure 50A). Given the rapid washout, we wondered if efflux transporters were involved in the uptake mechanism. However, pre-exposure of HCT-116 cells to

3 µM verapamil resulted in little change in 18F-SqFLT uptake when relative to control

18 (appendix, figure 8). The tumour NUV40-60 of F-SqFLT were comparable between

HCT-116 and B1 xenografts. For most radiotracers, the NUV value is normally analysed between 40-60 minutes since the uptake value, within the ROI at this timepoint, is usually the highest value due to radiotracer accumulation. Given the unique kinetics of 18F-SqFLT uptake in tumours (wash-in and wash-out), the AUC was analysed to give a better indication of the level of radioactivity exposure, over the 60 min period, between the two groups. As shown in figure 50C, the AUC of B1 xenografts were 2-fold less than HCT-116.

As per in vitro results, 18F-FLT uptake was expected to decrease in B1 xenografts

(shown in figure 44C) when compared to HCT-116. Correspondingly, in vivo 18F-FLT-

PET imaging of B1 xenografts resulted in a highly significant decrease (- 47 % NUV40-

60; p<0.001) of radiotracer tumour accumulation when compared to HCT-116 (figure

50E).

150

A HCT-116 Tumour

18F-SqFLT 18F-FLT

A1 Tumour Tumour A3 A4 A2 Tumour Tumour

70.0 mm 100 % 70.0 mm 100 % 70.0 mm 100 % 70.0 mm 100 %

0 mm 0 % 0 mm 0 % 0 mm 0 % 0 mm 0 %

Coronal Sagittal Coronal Sagittal

B B1 Tumour

18F-SqFLT 18F-FLT

B1 TumourTumour B2 Tumour B3 TumourTumou r B4 TumourTumour

70.0 mm 100 % 70.0 mm 100 % 70.0 mm 100 % 70.0 mm 100 %

0 mm 0 % 0 mm 0 % 0 mm 0 % 0 mm 0 %

Coronal Sagittal Coronal Sagittal

Figure 49. Representative PET images selected from a 60-min dynamic PET scan of viable tumour-bearing HCT-116 or B1 mice imaged with 18F-SqFLT and then re-imaged with 18F-FLT 48 h later. A. Panel represents coronal and sagittal images of HCT-116 xenograft imaged first with 18F-FLT and re-imaged with 18F-SqFLT. B. Panel represents coronal and sagittal images of a B1 tumour bearing mouse imaged first with 18F-FLT and re-imaged with 18F-SqFLT. White arrows indicate the loci of tumours.

151

A B 4.0 3.5 HCT-116 (Tumour) 3 3.0 B1 (Tumour) 2.5 HCT-116 (Whole Body) 18 F-SqFLT uptake B1 (Whole body) 2 2.0 NUV40-60 2.0 (MBq/ml of tissue) 40-60 1.5 (MBq/ml of tissue) ns

F-SqFLT uptake F-SqFLT

NUV

(%ID/ml tissue) of 1 1.0

0.5 (%ID/ml tissue) of

0.0 0 0 20 40 60 80 HCT-116 B1 Time (min)

AUC0-60

(MBq/ml of tissue)

D E 30

25 HCT-116 (Tumour) 20 25 B1 (Tumour) 15 *** HCT-116 (Whole body) 20 18 10 F-FLT uptake B1 (Whole body) NUV 5 40-60 15 (MBq/ml of tissue)2.0 40-60

F-FLT uptake F-FLT (MBq/ml of tissue) 10

1.5 NUV

(%ID/ml of tissue) (%ID/ml 1.0

(%ID/ml (%ID/ml tissue) of 5 0.5 0.0 0 0 20 40 60 80 HCT-116 B1 Time (min)

Figure 50. Time activity curves of HCT116 and B1 tumour bearing mice scanned with either 18F-FLT or 18F-SqFLT over a 60 min period. TACs comparison between HCT-116 and B1 tumours following 0.74 MBq injection of either A. 18F-SqFLT or D. 18F-FLT. Average wholebody uptake was included to show unvaried radioactivity concentration. B, E.

Normalized Uptake Value at 40-60 minutes (NUV40-60) expressed as percentage of injected dose per ml for either 18F-SqFLT or 18F-FLT. C. Area under the curve (AUC) calculated as an

integral NUV0-60. Values are expressed as percentage of injected dose per ml per minute. Analysis of variance (ANOVA) was performed to compare the datasets.

152

4.5. Biodistribution of 18F-SqFLT in BALB/c tumour bearing mice

As proof of concept and to predict the clinical utility of nucleoside analogues presenting squaramide phosphate mimics, following 18F-SqFLT imaging, the radiotracer biodistribution in BALB/c nude mice inoculated with HCT-116 and B1 cells was assessed by gamma counting. As shown in figure 51, biodistribution revealed rapid clearance of 18F-SqFLT from most tissues with the urinary tract being the primary route of excretion in both models. Unsurprisingly, the tumour-associated- radioactive did not differ between the two models at the timepoints examined (HCT-

116 = 0.28 ± 0.21 %, B1 = 0.21 ± 0.10 %) since the biodistribution study was conducted at 60 min following PET imaging.

Taken together, the biodistribution study echoes PET imaging and confirms the low tumour associated radioactivity of 18F-SqFLT in both HCT-116 and B1 xenografts at 60 min.

153

A Biodistribution of HCT-116 tumour bearing mice 1200 1000 %ID/g values: 800 600 Tumour = 0.28 ± 0.21 % 400 Tumour to blood ratio = 5.55 ± 5.21 % 200 Tumour to muscle = 1.15 ± 1.03 % 5 4 3 uptake F-SqFLT

18 2

(% ID/g of protein) 1 0

UrineHeart Liver Lung BoneBrain Blood Spleen Kidney Muscle Tumour Large intestine Biodistribution of 3B1 tumour bearing mice B 1000

800 %ID/g values: 600 Tumour = 0.21 ± 0.10 % 400 Tumour to blood ratio = 6.21 ± 0.59 % 200 Tumour to muscle = 1.3 ± 1.1 % 5 4 3

F-SqFLT uptake F-SqFLT

(% ID/g of tissue) 2

18 1 0

BloodUrineHeart Liver Lung BoneBrain Spleen Kidney Muscle Tumour

Large intestine

Figure 51. Biodistribution study of 18F-SqFLT in HCT-116 and 3B1 tumour-bearing BALB/c nude mice following PET imaging. A.B. Tissues were excised 1h after tracer injection and associated radioactivity measured on a gamma counter. Uptake is expressed as %ID/g tissue ±SEM (n =3).

154

18 4.6. Investigation of F-SqFLT LogD7.4

To provide greater insight into the use squaramide as adequate phosphate mimics, an area of research driven by phosphatase inhibitor therapy, and to gain an understanding poor cellular uptake and the rate limiting factor of the rapid kinetics

18 of F-SqFLT in vivo. The LogD7.4 value was assessed using the shake flask method

(previously described in section 2.20). The lipophilicity of 18F-SqFLT was investigated by deriving a partition coefficient of the tracer between water and octanol. The

18 LogD7.4 value was determined to be -2.90 ± 0.24 (Table 9) suggesting that F-SqFLT is unlikely to permeate biological membranes readily without the use of putative transporters. Based on the results obtained thus far, 18F-SqFLT is not substrate for nucleoside transporters.

Table 9. Calculated LogP (cLogP) values and measured LogD7.4 for 18F-FLT, 18F-FLTMP and 18F-SqFLT.

Partition coefficient

Method 18F-FLT 18F-FLTMP 18F-SqFLT

b LogD7.4 ND ND -2.90 ± 0.24

CLogP -0.74 -1.15 -1.20

ND = not determined.

a Calculated using Chemdraw 16.0 (Cambridgesoft, USA) b Performed n = 3 with triplicate measurements, represented as Mean ± SD.

155

Given the LogD7.4 values, it is unlikely that the chosen phosphate mimic can be used for future analysis as a permeable bioisostere of thymidine monophosphate for PET imaging.

4.7. Discussion

Over the past few decades, several tracers have been developed for use with positron emission tomography (PET) to assess cellular proliferation. Because crystalline nucleoside thymidine is incorporated into DNA but not RNA, it has been widely targeted to determine tumour proliferative ability, staging and patient therapeutic response142. More commonly, the exploitation of thymidine kinase-1

(TK1) substrate 18F-FLT uptake for imaging of proliferation has been broadly accepted but, its limitation in determining S-phase fraction has become more apparent over the past few years13,77. One such limitation is the inability to discriminate differential contributions of de novo pathway to versus the salvage-driven thymidine- monophosphate pool. Overall, the magnitude of 18F-FLT uptake is context dependent and does not accurately reflect the rate of cell proliferation. Thus, the development of a tracer that demonstrates an independence to TK1 and TS activity by reporting on dTYMK activity would be indispensable. We present the first use of a novel nucleotide analogue, 18F-SqFLT, combined with the sensitivity of PET imaging to aid the evaluation of dTYMK activity during DNA synthesis. Preliminary in silico modelling of 18F-SqFLT to dTYMK nucleotide binding site revealed a calculated free-energy similar to that of TMP-dTYMK. Moreover, the OH group of the squaramide appeared

156

to be in close proximity to the phosphate group of ADP suggesting, the likelihood of a phosphate donation.

Following mathematical modelling, a detailed investigation of the tracer both in vitro and in vivo was conducted to demonstrate; tracer specificity, its independence of

TK1, its incorporation into cells in an S-phase specific manner along with, PET studies of contrast/sensitivity. Tracer specificity was evaluated in CRISPR/Cas9 generated dTYMK knockdowns, B1 and B5 alongside parental HCT-116 and Kelly as differential cell models of dTYMK expression. All experiments compared uptake with 18F-FLT. OST

TK1- cells were also introduced to assess the contributions of the de novo vs salvage pathway on radiotracer uptake. It was hypothesized that cells lacking TK1 will not incorporate 18F-FLT but 18F-SqFLT due to its independence of TK1 activity.

Unexpectedly, 18F-SqFLT uptake was extremely low and comparable to that of 18F-

FLT. The remaining cell lines, HCT-116, KELLY and dTYMK mutants, also presented a significantly low 18F-SqFLT uptake when compared to 18F-FLT (p<0.001; ~ 100 - fold lower). The contrast of results confirms that 18F-SqFLT is not a substrate to TK1 and incriminates radiotracer diffusion rate to be the limiting factor.

A knockdown of dTYMK was anticipated to have a broader effect on the global metabolism (i.e. glucose and choline consumption) of B1 and B5 cells when compared to HCT-116 since cancer cells are known to reprogram their cellular energetics/metabolism during periods stress (i.e. bioenergetics and redox stress)126.

Given the relatively unchanged doubling times of B1 (22h), B5 (21 h) and HCT-116

(21h) cells, we anticipated an increase in the glucose consumption of B1 cells as a source of energy for supporting cell proliferation. On the contrary, glucose

157

metabolism (as reported by 18F-FDG) significantly decreased in both B1 and B5 cells.

Had a dTYMK knockdown resulted in a drastic increase of cellular doubling time, it could have been argued that the reduction observed was as a consequence of a reduced need for glucose consumption due to delayed cell growth143. As this was not the case, the result suggests that glucose metabolism may not be a major player in the compensatory pathway that appears to sustain the proliferation rate of dTYMK knockdowns in vitro. In all, the uptake level of 18F-FLT, 18F-FDG and 18F-SqFLT (albeit to a lesser extent) in HCT-116, B1 and B5 cells were correlated and showed reduced tracer accumulation in B1 and B5 cells indicating, a change in the anabolic demands of mutant cells. Choline is an essential component of the phospholipid bilayer for generating membranes for new cells created during mitosis, mitochondrial fission and fusion144. To an extent, an increase of choline metabolism is widely associated with an increase in cell proliferation141. Given the slightly lower doubling time of B1, we hypothesised a modest decrease in the retention of 18F-D4-FCH and comparable uptake values in B5 and HCT-116 cells. In contrast, figure 47B revealed an increase of

18F-D4-FCH retention in both B1 and B5 cells. A study by Trousil et al. 2016 noted that an inhibition of choline kinase and therefore choline metabolism, with ICL-CCIC-0019, affected mitochondrial function and activated glycolysis (possibly through AMPK)144.

Therefore, the upregulation of 18F-D4-FCH in dTYMK knockdown cells may indicate an increase in the mitochondrial functionality and demand of choline since choline can be oxidised in the mitochondria to form betaine. Interestingly, the solute carrier family 25 member 46 (SLC25A46) gene (suggested by Abrams et al. (2015)145 to play a role in mitochondrial dynamics by controlling mitochondrial membrane fission) was found, in figure 13C, to be negatively correlated to dTYMK. This offers support for

158

the hypothesis of an increase in the mitochondrial functional capacity (to aid nucleotide synthesis following dTYMK knockdown) and, offers an explanation as to why an increase in choline uptake was observed. Figure 52 highlights the interaction of betaine and methionine with the folate cycle that provides 5,10-

Methylenetetrahydrofolate as methyl donors for the de novo pathway of DNA synthesis.

Figure 52. Schematic diagram of the interactions of the methionine and folate cycle for dTMP synthesis. Image adapted from Epgui - CC BY-SA 3.0, https://commons.wikimedia.org/w/index.php?curid=31804170

The absence of death, due to thymidine starvation, in dTYMK knockdowns suggests a compensatory pathway with the capability of sustaining the dTTP pool required for

DNA synthesis. In general, an impairment of dTYMK should result in accumulation of

TMP32 owing to dTYMK inhibition and the relative contributions of ribonucleotide reductase (RNR), a heteromultimer that comprises a large (R1) and small (R2) subunits, for reducing NDPs to supply dNDPs during DNA repair146. High-resolution mass spectrometry-based (HRMS) permitted a more sensitive investigation of the

159

central carbon metabolites implicated in nucleotide metabolism. Compared to HCT-

116, B5 cells presented a moderate increase of TMP accumulation indicating an inhibition of dTYMK. This is in line with literature reports of TMP retention followed by a subsequent decrease of dTDP in dTYMK knockdowns32. B1 cells, however, produced a less convincing increase of TMP suggesting either TMP is being effectively utilised through an alternative route or, the level dTYMK knockdown (- 73 %) was insufficient in generating a pronounced response. Interestingly, a decrease in dUMP was observed in B1 cells perhaps suggesting an upregulation in the consumption the metabolite. Here we speculate a dual function of TS in methylating not only dUMP but, dUDP (converted from dUMP or supplied via RNR110) to produce dTDP for sustaining the dTTP flux into DNA.

It has been reported by Blount et al. 1997147 that dTTP synthesis occurs in an “on-site and on-demand” manor. A loss of the capacity to synthesize dTTP results in an increase of dUTP mis-incorporation into DNA, causing genomic instability during DNA damage repair31,148. Interestingly, UTP levels in B1 cells were comparable to HCT-116 whilst, an increase of 10 % was observed in B5. Given unchanged growth rate dTYMK knockdowns, it can be concluded that UTP mis-incorporation into DNA which results in futile cycles of DNA synthesis and repair, is not a predominant factor. A drop in the accessibility of ATP in B1 cells was also noted in figure 48. It is likely that the reduced substrate availability of ATP exerted cofounding effects on TK1 activity69 thereby, reducing salvage pathway TMP contributions and preventing nucleotide imbalance69. The observed decrease of 18F-FLT uptake in dTYMK mutants (figure 44C) despite TK1 expression remaining unchanged (section 3.9; figure 31) offers support to this theory. Taken together, our results provide an insight into the possible duality

160

of TS in methylating dUDP for dTDP production through a bypass mechanism both in vitro and to a lesser extent, in vivo. In a recent report by Siddiqui et al. 2019, TS was strongly associated with the de-differentiation phenotype of triple negative breast cancers, reducing CD44+CD24− cells, and suppressing migratory and sphere- forming ability149. These findings together with the results obtained thus far, suggests that TS activity could be implicated in other cellular functions and not just dUMP methylation. As a more definitive assay, it would be of worth introducing a chronic (24+ h) treatment of dTYMK knockdowns with 5FU to evaluate its effect on cell proliferation and in turn, shed light on TS duality for compensation.

Although the in vitro uptake results of 18F-SqFLT is significantly lower than traditional radiotracers like 18F-FLT, we hypothesise the radiotracer to be dTYMK specific and should, therefore, give a high tumour to noise ratio resulting in high image contrast, due to the low proliferation rate (and therefore lower uptake) of normal tissue. Our observations were extended into an in vivo setting which revealed a substantial difference in the tumour retention profiles of 18F-SqFLT and 18F-FLT (as shown by the

TACs; figure 50). The rapid washout of 18F-SqFLT tumour-associated radioactivity after 5 min contrasted that of 18F-FLT uptake which displayed an increase in the tumour-associated radioactivity over time. The difference in the kinetics of 18F-SqFLT from 18F-FLT in spite of their similarly low lipophilicity suggested the involvement of transporters in the case of 18F-FLT but not 18F-SqFLT; and/or existence of efflux mechanisms for the latter. Initially, efflux transporters were implicated as the cause of low radiotracer retention in tissue since 18F-SqFLT presented properties that are likely to be a substrate for ABC transporters111, presenting 9 heteroatoms - N + O =

10 (N + O > 8), polar surface area (PSA) of 125.04 A2 (PSA > 85A2). However,

161

verapamil exposure only resulted in minute and non-significant changes of radiotracer uptake (Appendix, figure 8). At this point, it is important to note that our study with 18F-FLT or 18F-SqFLT (both biodistribution and imaging) were done under anaesthesia which should be kept in mind during the interpretation results. Taken together, it was concluded that, 18F-SqFLT is not a substrate for ABC transporter and the low accumulation radiotracer may be charge specific.

18 Nonetheless, F-SqFLT was still able to differentiate (as per AUC0 -60 min) between the two tumour models albeit to a lesser extent when compared to 18F-FLT. It can be concluded that 18F-SqFLT is likely to be dTYMK specific but not PET sensitive due to the low overall uptake.

A pronounced difference of 18F-FLT tumour uptake can be observed in the PET images and TACs of B1 xenografts; further echoing in vitro results. The reduced uptake of

18F-FLT in B1 cells suggests a reduction in the tumour proliferative fraction possibly owing to, reduced TDP availability due to dTYMK knockdown and the reduced functional capacity of the mitochondria to aid DNA synthesis in vivo. Although the primary cellular determinant of 18F-FLT accumulation is thymidine kinase 1 (TK1), other factors like thymidine or ATP have been reported as major modulators of the radiotracer uptake. We have shown, in vitro, a decrease in the level of ATP and a concomitant decrease of 18F-FLT in B1 cells. In an in vivo setting, the difference of 18F-

FLT uptake between the two models is more pronounced possibly due to the absence of media supplementation which confer an advantage in cell proliferation in vitro.

The biodistribution study confirmed organ-specific variations in 18F-SqFLT retention and pharmacokinetics with clearance primarily through the urinary tract. In our chosen tumour models HCT-116 and B1, 18F-SqFLT tumour signal-to-background

162

contrast was qualitatively inferior to that of 18F-FLT. Although the distribution pattern

(tumour/muscle ratio) suggested that 18F-SqFLT is unsuitable in assessing dTYMK expression-activity, it is important to note that these values were generated at time

= 60 min (post PET imaging). Given the kinetics of 18F-SqFLT (TAC graph, figure 50A) emphasis should, therefore, be on the AUC results as a more relevant measure of the tumour radiotracer (parental and metabolites) exposure (over 60 min) and the contrast between the two tumour models.

Taken together, our data suggests the rate limiting factor in the sensitivity of 18F-

SqFLT with PET imaging to be its inability to adequately permeate biological membranes due to its overall charge. This was supported by the LogD7.4 value of -

2.90  0.24 and aided the conclusion that 18F-SqFLT would exhibit high aqueous solubility with low lipophilicity.

Although the use of phosphate bioisosteres present the capacity to be revolutionary in both PET imaging and development of antiviral/anticancer agents, it still remains a challenge to balance bioisostere pKa and polarity against membrane permeability and metabolic stability150. Nonetheless, our research brings biological context for the use of squaramide in the synthesis of sugar-nucleotide mimics observing that, a high radiotracer background specifically in the liver, gall bladder and kidneys, could limit the utility of squaramide bearing nucleotide for imaging primary or metastasised tumours in these organs.

163

5.0. Summary and concluding remarks

18F-FLT is the universally accepted marker of cell proliferation which, except in a few instances strongly correlates with immunohistochemical markers of proliferation including PCNA and Ki67. However, the rate of tracer uptake is context dependent and can produce false negatives or positives owing to factors such as tissue thymidine phosphorylase and serum/tissue thymidine levels in specific cancer lineages. These limitations led to further interest in developing derivatives of 18F-FLT, on the basis of the pharmacophore of thymidine, as radiotracers to image different segments of the

DNA synthesis pathway. This project aimed to develop and validate a new approach of utilising a first generation bioisostere of thymidine phosphate for imaging purposes and improve on the knowledge of dTYMK gene essentiality. Guided by the in silico modelling studies of structure-activity relationships between TMP and the homodimeric enzyme dTYMK, we developed a novel squaramide- thymidine radiotracer, 18F-SqFLT, as a potential proliferation marker hypothesised to derive better quantitative estimates of tumour proliferation than 18F-FLT. Since the success of 18F-SqFLT is dependent on its avidity to dTYMK, the relationship between dTYMK gene expression, its enzymatic activities and the growth rate of proliferating cells was first explicated before radiotracer assays could be performed. The first part of the thesis focused on evaluating the impact of dTYMK on cell proliferation following therapeutic and genotoxic interrogation. Under the influence of cell cycle arresting agents such as 5FU and APH, we noted a fluctuation in dTYMK expression throughout the phases of the cell cycle which echoed the results of Huang et al. (1994). This evidence of cell cycle regulation further supports the notion of dTYMK as an attractive biomarker of cell proliferation. The use of the only commercially available

164

inhibitor of dTYMK (YMU1) was found to have little or no impact on cell viability when administered singularly. On the basis of our PAMPA and Caco-2 assays, it was concluded that the low permeability characteristic of the drug may be the rate limiting step. An optimisation of YMU1 structure, to increase permeability, would therefore heighten the sensitivity of cancer cells to even lower doses of chemotherapy than those previously reported9. As an essential enzyme in the dTTP biosynthesis pathway, an inhibition of dTYMK activity should confer a growth disadvantage due to the reduced TDP pool for the DNA flux. However, the two stable clones, B1 and B5, created by CRISPR/Cas9-mediated gene editing portrayed a knockdown of dTYMK as dispensable in vitro. Extending the observations into an in vivo setting revealed a marked impairment of growth in B1 tumours when compared to parental xenografts. It was then speculated that a bypass mechanism may exist that becomes more apparent in vitro than in vivo. Components of media i.e. L- glutamine and folic acid were also suspected as contributors, at least in part, to producing precursors that sustained cell proliferation in vitro. Our results also inferred the possible duality of TS in methylating dUDP for dTDP production through a bypass mechanism both in vitro and to a lesser extent, in vivo. As the CRISPR/Cas9 models were incomplete knockouts of dTYMK, future work will therefore include re- editing to see if the remaining WT allele can be removed and offer an insight into cellular gene dependency on dTYMK. Figure 53 represents the proposed bypass mechanisms that may account for sustained cell proliferation of dTYMK mutants in vitro.

165

Figure 53. Suspected bypass mechanism for dTDP synthesis following dTYMK inhibition. During genomic stress, the de novo and RNR pathway may become upregulated to supply TS with sufficient dUDP for the production of dTDP. Since TS is a methylating enzyme, we speculate its duality in methylating other deoxyuridine nucleotides i.e. dUDP as a failsafe mechanism for dTYMK during dTTP biosynthesis. Image was modified from Strum (2014)

The second part of the thesis explores the potential of 18F-SqFLT in reporting dTYMK activity as a better method of monitoring the tumour proliferative fraction and, provide an insight into the biodistribution of squaramide deoxynucleotides that are of current interest in medicinal chemistry. Although there was a discrepancy between in vitro and in vivo proliferation dTYMK mutants, the tumour xenografts generated were better representatives of human tumorigenesis making the models a worthy platform for assessing thymidine monophosphate bioisosteres for the

166

evaluation of imaging tools that report on dTYMK activity. Initial experiments with

18F-FLT under pharmacological doses of cold 19F-SqFLT resulted in an unhindered uptake of 18F-FLT suggesting that 19F-SqFLT is not a substrate for TK1 or ENT-1/2 transporters. As a proof of concept, molecular imaging with PET revealed a clearance of 18F-SqFLT predominantly through the renal system, resulting in a high activity in the urine. The appreciable difference between the AUC0-60min in WT and B1 knockdown cells suggested that 18F-SqFLT is specific to dTYMK but not sensitive since the overall tumour was significantly low (figure 50C). A high accumulation of radioactivity was also observed in the main excretory organs such as; liver, kidneys and large intestine which provided valuable pharmacodynamic information on squaramide deoxynucleosides that have become of interest in the development of novel anticancer agents. The successful differentiation of dTYMK knockdown from

WT tumours indicated tracer specificity; however, the low overall uptake led to the

18 conclusion that F-SqFLT permeability was rate limiting step. LogD7.4 measurements confirmed this notion suggesting that the radiotracer emulated the charge of its

18 analogue, TMP. Improving the LogD7.4 value of F-SqFLT will significantly enhance the sensitivity of the radiotracer and increase its efficiency in accurately depicting the tumour proliferative fraction with respect to dTYMK. Imaging with 18F-FLT revealed a stark difference between the rate of DNA nucleotide flux of WT tumour and dTYMK knockdowns (figure 49) thereby, confirming a change in the dTTP biosynthesis pathway with respect to TK1 activity.

167

5.1. Future directions

Preliminary work described in this thesis introduced the use of squaramide moieties in PET imaging and, shed light on a possible bypass mechanism for DNA synthesis following dTYMK knockdown. Previous investigation into the genetic dependency of tumour cells to dTYMK, following copy number loss, revealed 23 candidate genes that were negatively correlated to dTYMK. Although a screen of these candidate genes showed know apparent correlation with thymidine DNA synthesis, they are still biologically meaningful and are worth future validations in vitro. Introducing shorthair pin RNA mutations of these genes into dTYMK mutant clones may be worth investigating to see if, these alterations confer a growth disadvantage and expand on regulatory mechanisms that become heightened during dTYMK genomic stress. Of particular interest is the SLC25A17 gene (a peroxisomal transporter for multiple cofactors like coenzyme A (CoA) and nucleotide adenosine monophosphate (AMP)) along with SLC25A46 (. Monitoring, i.e. through metabolomics, the metabolism of their encoded protein along with their respective downstream effectors, may give a greater insight into the interplay of the mitochondria with other activated pathways that may help to sustain cell proliferation following dTYMK knockdown. An upregulation of SLC25A17 gene may indicate an increase in the demand of adenosine monophosphate kinase (AMPK) for adenosine monophosphate. AMPK, a target of

LKB1/STK11, has been implicated by Lui et al.. (2003) who reported a deficit in the nucleotide metabolism of LKB1-mutant lung cancer cells that increased the sensitivity of cancer cells to dTYMK inhibition. It was suggested that the synthetic lethality of LKB1 loss and dTYMK knockdown was partly due to the lower expression of dTYMK in LKB1 -null cell lines that resulted in a greater dependency of cells to the

168

dTTP synthesis pathway. Introducing LKB1 -null and proficient cell lines to our generated B1 and B5 panel may provide a good platform for understanding the role of SLC25A17 and, AMPK (in the context of mitochondria biogenesis) for the dTTP synthesis pathway and their modulation following dTYMK knockdown. Initial characterisation of LKB1 -null A549 and H23 along with LKB1 -proficient PC9, PC9ER and H1975 cells have been presented in Appendix figure 1 and 2. Noteworthy is the decrease of ATP in B1 cells which increases AMPK activation since the enzyme is known to be directly phosphorylated and, in turn, activated by LKB1 in the context of low cellular ATP levels32.

Success, in identifying the critical pathways compensating for dTYMK knockdown, will permit the future development of anti-cancer agents targeted to the most relevant candidate genes as part of an adjuvant therapy with inhibitors of dTYMK. To our knowledge, we presented the first metabolomic analysis on the effects of dTYMK knockdown on global metabolism. While our preliminary work was centred around changes in pyrimidine and purine metabolism, expanding on other areas such as changes in the glycolysis, citric acid cycle and Pentose phosphate pathway, may add to the literature knowledge of the intricate network system between nucleotide and global metabolism. Our observations implicated TS as a plausible dTYMK bypass route by methylated dUDP or dUTP into dTDP and dTTP respectively. To validate this theory, a study must be conducted to: (1) distinguish deoxyuridine from deoxyruridine diphosphate and deoxythymidine diphosphate on HPLC under condition when ATP is present to catalyse the reaction and if (1) is feasible then, (2) incubate deoxyruridine, and deoxyuridine diphosphate with TS with required co-

169

factors to observe the possible conversation of deoxyuridine diphosphate to deoxythymidine diphosphate by TS and its rate.

170

Appendix

PC9 ER A549H23 PC9 H1975

dTYMK Atlas 24 kDa

β-actin 42 kDa

dTYMK Abcam 24 kDa

β-actin 42 kDa

TK1 26 kDa

β-actin 42 kDa

TS 30 kDa

β-actin 42 kDa

TP 50 kDa

β-actin 42 kDa

Figure 1. Expression profile the key enzymes involved in the dTTP biosynthesis of LKB1 null and proficient cells

171

Figure 2. Metabolite assessment of 18F-FLT uptake in LKB1 null and proficient cells.

7000 6000 2000

FLT - standard Control 5000 1500  4000  1000 3000

2000 500 1000

0 0 00:00 05:00 10:00 15:00 20:00 25:00 30:00 00:00 05:00 10:00 15:00 20:00 25:00 30:00 Time (min) Time (min)

1400 1000 1200 1.9 uM YMU1 190nM YMU1 1000 800

800   600 600 400 400

200 200

0 0

00:00 05:00 10:00 15:00 20:00 25:00 30:00 00:00 05:00 10:00 15:00 20:00 25:00 30:00 Time (min) Time (min) Figure 3. Enzymatic assay of extracted lysates incubated with 18F-FLT and gamma counted by rHPLC

172

Figure 4. Representative western blot analysis of cells with control liposomes and liposomal YMU1. Cells were treated for 24 h with specified conditions before being washed 3x in PBS and then lysed for protein quantification.

173

Figure 5. Clone selection process of mutant dTYMK HCT-116 following nucleofection with gRNAs

174

24 h 48 h

HCT-116 B1 B5 HCT-116 B1 B5 N H N H N H N H N H N H dTYMK 24 kDa

NDPK 18 kDa β-actin 42 kDa

TK1 26 kDa β-actin 42 kDa

TS 30 KDa

β-actin 42 kDa

COX IV 17 kDa

β-actin 42 kDa

HIF 1 α 17 kDa

β-actin 42 kDa

Figure 6. Expression profile of the key enzymes involved in the dTTP biosynthesis. HCT-116 and dTYMK mutants were subjected to 24 and 48 h hypoxic and normoxic conditions prior to protein evaluation with western blot.

Figure 7. Radiotracer uptake of [18F-FLT] in HCT-116 and BT474 following 5FU treatment for acute (1.5 h) and chronic (24 h) treatment with 5FU.

175

e 0 .1 0 n s 1 5 *

)

n a

0 .0 8 i

T e

o F

o 1 0

e p

p 0 .0 6

g a

h l

0 .0 4 c

/ r

4 5

% %

0 .0 2 F

(

8 '

F 8

1 0 .0 0 0 C o n tro l 1 h V e ra p a m il H C T 1 1 6 1 h V e r a p a m il

Figure 8. Uptake analysis of HCT-116 following verapamil interrogation. HCT 116 cells were pre- incubated with 3 µM Verapamil for 1 and then co-incubated with 0.74 MBq of either [18F-]-SqFLT or 18F-D4-Choline. One-way Anova statistical analysis and Dunnets post hoc was performed for n=5 samples.

Liposomal size (d. nm) = 392.1 SD (d. nm) = 14.14

Figure 9. Intensity particle size distribution YMU1 infused liposomes obtained from 3 repeat measurements. Measurements were performed at 25° C in phosphate buffer (50mM Na- phosphate, 150 Mm NaCl at pH 7.4.

176

Table 10. A list (1-13) of the 23 negatively correlated candidate genes to dTYMK and, their function as defined by Uniprot (https://www.uniprot.org) Gene Function 1. HLA-DQA1 A member of a family of human leukocyte antigen (HLA) complex that plays a central role in the immune system by presenting peptides derived from extracellular proteins. 2. MET Regulates many physiological processes including proliferation (by interacting PI3-kinase subunit PIK3R1, PLCG1, SRC, GRB2, STAT3 or the adapter GAB1), scattering, morphogenesis and survival. MET is essential for invasive growth and helps to mediate epithelial- mesenchymal transition (EMT). 3. TNFRSF14 Important for lymphocyte activation and plays a vital role in the pathogenesis of herpes simplex virus (HSV) 4. EFNA5 A cell surface GPI-bound ligand for Eph receptors essential for migration, repulsion and adhesion during neuronal, vascular and epithelial development 5. FAM50B Contains an intronless open reading frame that ascended from ancestral retroposition and plays a role in the circadian clock; an internal time-keeping system, regulates various physiological processes through the generation of approximately 24 hour circadian rhythms in gene expression, which are translated into rhythms in metabolism and behaviour). This gene is imprinted and paternally expressed in many tissues. 6. CALM2 Helps to mediate the control of a large number of enzymes, ion channels, aquaporins and other proteins through calcium-binding. 7. FAM43A Function is currently unknown but has been identified by Wan et al. (2004) as a possible cancer development and progression candidate gene. 8. LIFR Signal-transducing polyfunctional cytokine that plays a role in differentiation, survival, and proliferation of a wide variety of cells. It mediates the biological activity of LIF by blocking its binding to receptors on target cells 9. IFNAR1 - Activation of various innate immune signalling pathways such as Interferon- Toll-like receptor 3, Toll-like receptor 4, Toll-like receptor 7 and α/β receptor 1 Melanoma Differentiation-Associated protein 5 (MDA-5). 10. JUN A proto-oncogene that encodes a protein which interacts with the enhancer heptamer motif 5'-TGA[CG]TCA-3'. Promotes activity of NR5A1 when phosphorylated by HIPK3 leading to increased steroidogenic gene expression upon cAMP signalling pathway stimulation 11. ARFGEF1 Promotes guanine-nucleotide exchange on ARF1 and ARF3. Promotes the activation of ARF1/ARF3 through replacement of GDP with GTP. 12. SLC25A17 Peroxisomal transporter for multiple cofactors like coenzyme A (CoA), flavin adenine dinucleotide (FAD), flavin mononucleotide (FMN) and nucleotide adenosine monophosphate (AMP), and to a lesser extent for nicotinamide adenine dinucleotide (NAD+), adenosine diphosphate (ADP) and adenosine 3',5'-diphosphate (PAP) 13. PDZRN3 Plays an important role in regulating the surface level of MUSK on myotubes.

177

Table 11. The remaining 14-23 negatively correlated candidate genes to dTYMK and, their function as defined by Uniprot (https://www.uniprot.org) Gene Function 14. DOCK10 Dedicator of cytokinesis protein 10 encodes a guanine nucleotide- exchange factor (GEF) that activates CDC42 and RAC1 by exchanging bound GDP for free GTP. Essential for dendritic spine morphogenesis in Purkinje cells and in hippocampal neurons, via a CDC42-mediated pathway 15. HOXC9 Homeobox protein Hox-C9 encodes a sequence-specific transcription factor which is part of a developmental regulatory system that provides cells with specific positional identities on the anterior-posterior axis. 16. PIGX Phosphatidylinositol-glycan biosynthesis class X encodes protein that is an essential component of glycosylphosphatidylinositol- mannosyltransferase 1 which transfers the first of the 4 mannoses in the GPI-anchor precursors during GPI-anchor biosynthesis. 17. FMO4 Dimethylaniline monooxygenase [N-oxide-forming] 4 encodes a protein involved in the oxidative metabolism of a variety of xenobiotics such as drugs and pesticides. 18. ARNTL2 Aryl hydrocarbon receptor nuclear translocator-like protein 2 encodes a protein transcriptional activator that forms a core component of the circadian clock. 19. MPPE1 Metallophosphoesterase 1 is required for transport of GPI-anchor proteins from the endoplasmic reticulum to the Golgi. 20. SLC25A46 Solute carrier family 25 member 46 has been suggested by Abrams et al. (2015)2 to play a role in mitochondrial dynamics by controlling mitochondrial membrane fission. 21. SRCIN1 SRC kinase signalling inhibitor 1 acts as a negative regulator of SRC by activating CSK which inhibits SRC activity and downstream signalling, leading to impaired cell spreading and migration 22. ASAP1 Arf-GAP with SH3 domain, ANK repeat and PH domain-containing protein 1 encodes phosphatidylinositol 4,5-bisphosphate- dependent GTPase-activating protein activity for ARF1 (ADP ribosylation factor 1) and ARF5 and a lesser activity towards ARF6 23. LMNB1 Lamin-B1 a component of the nuclear lamina, a fibrous layer on the nucleoplasmic side of the inner nuclear membrane, thought to provide a framework for the nuclear envelope and may also interact with chromatin.

178

Table 11. Evaluation of different isosteres of phosphoric acid

Phosphate Structure Positive Negative Bioisosteres Phosphonates 1. They are approximately Overall molecule isosteric with phosphates is charged 2. Phosphorus–carbon bond therefore cell found in SAT is more stable permeability is to hydrolysis than the inhibited. phosphorus–oxygen bond found in phosphates82. Heterocyclic- 1. They are approximately Unlikely to be based isosteric with phosphates. phosphorylated. 2. Overall charge is dispersed Formulated to over the whole molecule mimic only a therefore enabling improved phosphate but cell permeability. unable to for phosphorus- oxygen bonds Boranophosphate 1. Isosteric to neutral methyl 1. Permanent phosphonates but, like the negative charge phosphate group, possesses on the borate a negative charge. preludes poor 2. Boranophosphates are permeability. more lipophilic than 2. Increasing phosphate groups but remain concerns with water soluble82. pharmacology and toxicology of boron Boronic acid- 1. Potentially useful as Increasing based phosphate bioisosteres given concerns with that the trigonal-planar pharmacology boronic acid forms dative and toxicology of bonds with nucleophiles to boron form tetrahedral structures with a formal negative charge on the boron atom Squaric acid and 1. They are approximately 1. Anticipated to squaramide isosteric with phosphates be cell permeable 2. Overall charge is dispersed as per Hückel's over the whole molecule rule of therefore enabling improved aromaticity but cell permeability. limited literature 3. Evidence from Sato et al. on how (2002) of successful mechanics. incorporation into oligonucleotide.

179

Appendix B – Spectra analysis

ATP

180

dUMP

181

TMP

182

TTP

183

References

1. Hanahan, D. & Weinberg, R. A. Hallmarks of cancer: the next generation. Cell 144, 646–74 (2011). 2. Evan, G. I., Evan, G. I., Vousden, K. H. & Vousden, K. H. Proliferation, cell cycle and apoptosis in cancer. Nature 411, 342–348 (2001). 3. Sadikovic, B., Al-Romaih, K., Squire, J. & Zielenska, M. Cause and consequences of genetic and epigenetic alterations in human cancer. Curr. Genomics 9, 394–408 (2008). 4. Soloviev, D., Lewis, D., Honess, D. & Aboagye, E. [18F]FLT: An imaging biomarker of tumour proliferation for assessment of tumour response to treatment. Eur. J. Cancer 48, 416–424 (2012). 5. Takeuchi, K. & Ito, F. Receptor Tyrosine Kinases and Targeted Cancer Therapeutics. Biol. Pharm. Bull. 34, 1774–1780 (2011). 6. Longley, D. B., Harkin, D. P. & Johnston, P. G. 5-Fluorouracil: Mechanisms of Action and Clinical Strategies. Nat. Rev. Cancer 3, 330–338 (2003). 7. Stathopoulos, G. P., Clinic, F. O. & Hospital, E. D. Survival of untreated advanced colorectal cancer patients. 731–733 (2011). doi:10.3892/ol.2011.310 8. Weinerman, B., Maroun, J., Stewart, D. & Cripps, C. J. J. Sequential methotrexate, 5- fluorouracil, and leucovorin in metastatic measurable colorectal cancer. Does it work? Cancer Invest. 8, 399–43 (1990). 9. Chen, Y. H. et al. Chemical Inhibition of Human Thymidylate Kinase and Structural Insights into the Phosphate Binding Loop and Ligand-Induced Degradation. J. Med. Chem. 59, 9906–9918 (2016). 10. Kenny, L. M., Aboagye, E. O. & Price, P. M. Positron Emission Tomography Imaging of Cell Proliferation in Oncology. Clin. Oncol. 16, 176–185 (2004). 11. Heinzmann, K. et al. The relationship between endogenous thymidine concentrations and [18F]FLT uptake in a range of preclinical tumour models. EJNMMI Res. 6, (2016). 12. Tehrani, O. S. & Shields, A. F. PET imaging of proliferation with pyrimidines. J. Nucl. Med. 54, 903–12 (2013). 13. Bading, J. R. & Shields, A. F. Imaging of cell proliferation: status and prospects. J. Nucl. Med. 49 Suppl 2, 64S–80S (2008). 14. Whisenant, J. G., McIntyre, J. O., Peterson, T. E., Kang, K., Sánchez, V., Manning, H. C., Arteaga, C. L., & Yankeelov, T.E. Utility of 18 FLT-PET to Assess Treatment Response in Trastuzumab-resistant and Sensitive HER2- overexpressingHuman Breast Cancer Xenografts. Mol imaging Biol. 71, 3831–3840 (2014). 15. Jackman, A. L. Thymineless Death. in Antifolate Drugs in Cancer Therapy 423–435 (1999).

184

16. Grierson, J. R., Schwartz, J. L., Muzi, M., Jordan, R. & Krohn, K. A. Metabolism of 3’- deoxy-3’-[F-18]fluorothymidine in proliferating A549 cells: validations for positron emission tomography. Nucl. Med. Biol. 31, 829–837 (2004). 17. Krohn, K. A., Mankoff, D. A. & Eary, J. F. Imaging cellular proliferation as a measure of response to therapy. J. Clin. Pharmacol. 41, (2001). 18. Cui, Q. et al. Thymidylate Kinase: An Old Topic Brings New Perspectives. Curr. Med. Chem. 20, 1286–1305 (2013). 19. Koseki, Y., Kinjo, T., Kobayashi, M. & Aoki, S. Identification of novel antimycobacterial chemical agents through the in silico multi-conformational structure-based drug screening of a large-scale chemical library. Eur. J. Med. Chem. 60, 333–339 (2013). 20. Brundiers, R. et al. Modifying human thymidylate kinase to potentiate azidothymidine activation. J. Biol. Chem. 274, 35289–35292 (1999). 21. Ostermann, N. et al. Potentiating AZT activation: Structures of wild-type and mutant human thymidylate kinase suggest reasons for the mutants’ improved kinetics with the HIV prodrug metabolite AZTMP. J. Mol. Biol. 304, 43–53 (2000). 22. Huang, S. H. et al. Human dTMP kinase: gene expression and enzymatic activity coinciding with cell cycle progression and cell growth. DNA Cell Biol. 13, 461–471 (1994). 23. Zimmermann, W., Chen, S. M., Bolden, A. & Weissbach, A. Mitochondrial DNA replication does not involve DNA polymerase alpha. J. Biol. Chem. 255, 11847– 11852 (1980). 24. Chen, Y. L., Lin, D. W. & Chang, Z. F. Identification of a putative human mitochondrial thymidine monophosphate kinase associated with monocytic/macrophage terminal differentiation. Genes to Cells 13, 679–689 (2008). 25. Rampazzo, C. et al. A deoxyribonucleotidase in mitochondria : Involvement in regulation of dNTP pools and possible link to genetic disease. 97, 8239–8244 (2000). 26. Rampazzo, C. et al. Mitochondrial Deoxyribonucleotides, Pool Sizes, Synthesis, and Regulation. J. Biol. Chem. 279, 17019–17026 (2004). 27. Ferraro, P., Nicolosi, L., Bernardi, P., Reichard, P. & Bianchi, V. Mitochondrial deoxynucleotide pool sizes in mouse liver and evidence for a transport mechanism for thymidine monophosphate. Proc. Natl. Acad. Sci. 103, 18586–18591 (2006). 28. Pontarin, G., Gallinaro, L., Ferraro, P., Reichard, P. & Bianchi, V. Origins of mitochondrial thymidine triphosphate: Dynamic relations to cytosolic pools. Proc. Natl. Acad. Sci. 100, 12159–12164 (2003). 29. Bridges, E. G., Jiang, Z. & Cheng, Y. C. Characterization of a dCTP transport activity reconstituted from human mitochondria. J. Biol. Chem. 274, 4620–4625 (1999). 30. Traut, T. W. Physiological concentrations of purines and pyrimidines. Mol. Cell. Biochem. 140, 1–22 (1994). 31. Hu, C. et al. Article Tumor Cells Require Thymidylate Kinase to Prevent dUTP Incorporation during DNA Repair. Cancer Cell 22, 36–50 (2012). 32. Liu, Y. et al. Metabolic and functional genomic studies identify deoxythymidylate

185

kinase as a target in LKB1 -mutant lung cancer. Cancer Discov. 3, 870–879 (2013). 33. Sharma, R. & Aboagye, E. Development of radiotracers for oncology - The interface with pharmacology. Br. J. Pharmacol. 163, 1565–1585 (2011). 34. Kapoor, V., McCook, B. M. & Torok, F. S. An introduction to PET-CT imaging. Radiographics 24, 523–543 (2004). 35. Farwell, M. D., Pryma, D. A. & Mankoff, D. A. PET / CT Imaging in Cancer : Current Applications and Future Directions. 3433–3445 (2014). doi:10.1002/cncr.28860 36. Basu, S. et al. Fundamentals of PET and PET/CT imaging. Ann. N. Y. Acad. Sci. 1228, 1–18 (2011). 37. England, C. G., Hernandez, R., Bou, S., Eddine, Z. & Cai, W. Molecular Imaging of Pancreatic Cancer with Antibodies. (2016). doi:10.1021/acs.molpharmaceut.5b00626 38. Langley, G., International, H. S. & Farnaud, S. J. Microdosing : safer clinical trials and fewer animal tests. (2014). doi:10.4155/bio.09.168 39. Hatt, M. et al. Reproducibility of 18F-FDG and 3’-deoxy-3’-18F-fluorothymidine PET tumor volume measurements. J. Nucl. Med. 51, 1368–1376 (2010). 40. Nelson, C. A., Wang, J. Q., Leav, I. & Crane, P. D. The interaction among glucose transport, hexokinase, and glucose-6-phosphatase with respect to 3H-2- deoxyglucose retention in murine tumor models. Nucl. Med. Biol. 23, 533–41 (1996). 41. Verma, N., Cowperthwaite, M. C., Burnett, M. G. & Markey, M. K. Differentiating tumor recurrence from treatment necrosis: A review of neuro-oncologic imaging strategies. Neuro. Oncol. 15, 515–534 (2013). 42. Suzuki, H. et al. Limitations of FDG-PET and FDG-PET with computed tomography for detecting synchronous cancer in pharyngeal cancer. Arch. Otolaryngol. Head. Neck Surg. 134, 1191–1195 (2008). 43. Zhang, W. et al. A highly selective and specific PET tracer for imaging of tau pathologies. J. Alzheimer’s Dis. 31, 601–612 (2012). 44. Witney, T. H., Fortt, R. R. & Aboagye, E. O. Preclinical Assessment of Carboplatin Treatment Efficacy in Lung Cancer by 18F-ICMT-11-Positron Emission Tomography. PLoS One 9, e91694 (2014). 45. Witney, T. H. et al. Preclinical evaluation of [18F]fluoro-pivalic acid: A novel metabolic radiotracer for tumor detection. J. Nucl. Med. 55, 166–166 (2014). 46. Challapalli, A. et al. 18F-ICMT-11, a caspase-3-specific PET tracer for apoptosis: biodistribution and radiation dosimetry. J. Nucl. Med. 54, 1551–6 (2013). 47. Zhang, C. C. et al. [18F]FLT-PET Imaging Does Not Always ‘Light Up’ Proliferating Tumor Cells. Clin. Cancer Res. 18, 1303–1312 (2012). 48. Kenny, L. The Use of Novel PET Tracers to Image Breast Cancer Biologic Processes Such as Proliferation, DNA Damage and Repair, and Angiogenesis. J. Nucl. Med. 57, 89S–95S (2016). 49. Gardelle, O. et al. [76Br]Bromodeoxyuridine PET in tumor-bearing animals. Nucl. Med. Biol. 28, 51–57 (2001).

186

50. Lee, E. J. & Lee, S. J. Etoposide increases equilibrative nucleoside transporter 1 activity and fluorothymidine uptake: screening of 60 cytotoxic agents. Oncol. Rep. 29, 763–770 (2013). 51. Wantzin, G. L. & Killmann, S.-A. Nuclear labelling of leukaemic blast cells with tritiated thymidine triphosphate after daunomycin. Eur. J. Cancer 13, 647–655 (1977). 52. Lonneux, M., Labar, D., Bol, A., Jamar, F., and Pauwels, S. Uptake of 2–11-C thymidine in colonic and bronchial tumors. In: Metabolic Imaging of Cancer: Proceedings of the European Conference on Research and Application of Positron Emission Tomography in Oncology. PET-Centrum AZG (1996). 53. Christman, D., Crawford, E. J., Friedkin, M. & Wolf, A. P. Detection of DNA synthesis in intact organisms with positron-emitting (methyl- 11 C)thymidine. Proc. Natl. Acad. Sci. U. S. A. 69, 988–92 (1972). 54. Mankoff, D. A. & Eary, J. F. Proliferation imaging to measure early cancer response to targeted therapy. Clin. Cancer Res. 14, 7159–7160 (2008). 55. Eary, J. F. et al. 2-[C-11]thymidine imaging of malignant brain tumors. Cancer Res. 59, 615–621 (1999). 56. Waarde, A. Van et al. Selectivity of 18F-FLT and 18F-FDG for differentiating tumor from inflammation in a rodent model. J. Nucl. Med. 45, 695–700 (2004). 57. Pinilla, I., Rodríguez-Vigil, B. & Gómez-León, N. Integrated FDG PET/CT: utility and applications in clinical oncology. Clin. Med. Oncol. 2, 181–198 (2008). 58. Wells, P. et al. Assessment of proliferation in vivo using 2-[11C]thymidine positron emission tomography in advanced intra-abdominal malignancies. Cancer Res. 62, 5698–5702 (2002). 59. Saito, Y. et al. Quantitative autoradiographic mapping of herpes simplex virus encephalitis with a radiolabeled antiviral drug. Science (80-. ). 217, 1151 LP-1153 (1982). 60. Sun, H. et al. Imaging DNA Synthesis In Vivo with. J. Nucl. Med. 46, 292–296 (2005). 61. Conti, P. S. & Watanabe, A. Synthesis of ([ I ’ Cl-FMAU ): A Potential Nucleoside Analog for In Viva Study of Cellular Proliferation with PET. Nucl. Med. Biol. 22, 783– 789 (1995). 62. Tehrani OS, Douglas KA, Lawhorn-Crews JM, S. A. Tracking cellular stress with labeled FMAU reﬂects changes in mitochondrial TK2. Eur J Nucl Med Mol Imaging 35, 1480–1488 (2008). 63. Arnér, E. S. J., Spasokoukotskaja, T. & Eriksson, S. Selective assays for thymidine kinase 1 and 2 and deoxycytidine kinase and their activities in extracts from human cells and tissues. Biochem. Biophys. Res. Commun. 188, 712–718 (1992). 64. Berk, A. J., Meyer, B. J. & Clayton, D. A. Mitochondrial-specific thymidine kinase. Arch. Biochem. Biophys. 154, 563–565 (1973). 65. Bergstrom, M. et al. In Vitro and Animal Validation of Bromine-76 . - Bromodeoxyuridine as a Proliferation Marker. (1998). 66. Lu, L. et al. 76 Br-BFU as Proliferation Markers. 43, 1688–1698 (2002).

187

67. Blau, I.W., Elstner, E., Waechter, M., Ihle, R., von Janta-Lipinski M, L. P. Sensitivity of CFU-GM from normal human bone marrow and leukaemic clonogenic cells (CFU-L) from blood of patients with myelogenous leukaemia to 3’-deoxy-3’-fluorothymidine in comparison to 3’-azido-3’-deoxythymidine. Blut. 59, 455–457. (1989). 68. Kong, X. B. et al. Comparisons of anti-human immunodeficiency virus activities, cellular transport, and plasma and intracellular pharmacokinetics of 3’-fluoro-3’- deoxythymidine and 3’-azido-3’-deoxythymidine. Antimicrob. Agents Chemother. 36, 808–818 (1992). 69. Barthel, H. et al. The uptake of 3′-deoxy-3′-[18F]fluorothymidine into L5178Y tumours in vivo is dependent on thymidine kinase 1 protein levels. Eur. J. Nucl. Med. Mol. Imaging 32, 257–263 (2005). 70. Grierson, J. R., Schwartz, J. L., Muzi, M., Jordan, R. & Krohn, K. A. Metabolism of 3′- deoxy-3′-[F-18]fluorothymidine in proliferating A549 cells: Validations for positron emission tomography. Nucl. Med. Biol. 31, 829–837 (2004). 71. Muzi, M; Mankoff, D. A; Grierson, J. R; Wells, J. M; Vesselle, H and Krohn, K. A. Kinetic Modeling of 3 Ј -Deoxy-3 Ј - Mathematical Studies. J. Nucl. Med. 371–381 (2015). 72. Muzi, M. et al. Kinetic analysis of 3’-deoxy-3’-fluorothymidine PET studies: validation studies in patients with lung cancer. J. Nucl. Med. 46, 274–282 (2005). 73. Lu, L., Jiang, L., Guan, H., Gao, Y. and Lu, H. Imaging proliferation in human leukemia-tumor bearing mice with 18 F-FLT: Comparison with 18 F-FDG PET. Hell. J. Nucl. Med. 6–9 (2012). 74. Ekshyyan, O., Sibley, D., Caldito, G.C., Sunderland, J., Vascoe, C. & Nathan, C. A. . 18F-fluorodeoxythymidine micro–positron-emission tomography versus 18F- fluorodeoxyglucose micro–positron-emission tomography for in vivo minimal residual disease imaging. 123, 107–111 (2013). 75. Valentiner, U. et al. [18F]FDG and [18F]FLT PET-CT and MR imaging of human neuroblastomas in a SCID mouse xenograft model. Anticancer Res. 28, 2561–2568 (2008). 76. Pantaleo, M. A. et al. Advances in preclinical therapeutics development using small animal imaging and molecular analyses: The gastrointestinal stromal tumors model. Clin. Exp. Med. 9, 199–205 (2009). 77. McKinley, E. T. et al. Limits of [18F]-FLT PET as a Biomarker of Proliferation in Oncology. PLoS One 8, 1–9 (2013). 78. Munch-Perterson, B; Cloos, L; Jenson, H. Human Thymidine Kinase 1 Regulation in Malignant Normal and. Adv. enyme Regul. 35, 69–89 (1995). 79. Schelhaas, S. et al. Preclinical applications of 3’-deoxy-3’-[18F]fluorothymidine in oncology - A systematic review. Theranostics 7, 40–50 (2017). 80. Seio, K., Miyashita, T., Sato, K. & Sekine, M. Synthesis and properties of new nucleotide analogues possessing squaramide moieties as new phosphate isosters. European J. Org. Chem. 5163–5170 (2005). doi:10.1002/ejoc.200500520 81. Sato, K., Seio, K. & Sekine, M. Squaryl group as a new mimic of phosphate group in modified oligodeoxynucleotides: Synthesis and properties of new

188

oligodeoxynucleotide analogues containing an internucleotidic squaryldiamide linkage. J. Am. Chem. Soc. 124, 12715–12724 (2002). 82. Hsu, P. D., Lander, E. S. & Zhang, F. Development and Applications of CRISPR-Cas9 for Genome Engineering. Cell 157, 1262–1278 83. Stella, S., Alcón, P. & Montoya, G. Class 2 CRISPR–Cas RNA-guided endonucleases: Swiss Army knives of genome editing. Nat. Struct. &Amp; Mol. Biol. 24, 882 (2017). 84. Workman, P. et al. Guidelines for the welfare and use of animals in cancer research. Br. J. Cancer 102, 1555–1577 (2010). 85. Frisch, M.J.E.A., Trucks, G.W., Schlegel, H.B., Scuseria, G.E., Robb, M.A., Cheeseman, J.R., Scalmani, G., Barone, V., Mennucci, B., Petersson, G. and Nakatsuji, H. Gaussian 09, revision a. 02, gaussian. Wallingford, CT (2009). 86. Becke, A. D. Density-functional thermochemistry. III. The role of exact exchange. J. Chem. Phys. 98, 5648–5652 (1993). 87. Chengteh Lee, Weitao Yang, and R. G. P. Development of the Colle-Salvetti correlation-energy formula into a functional of the electron density. APS 37, (1988). 88. Ferrari, S. et al. Human Thymidylate Synthase Inhibitors Halting Ovarian Cancer Growth. Vitam. Horm. 107, 473–513 (2018). 89. Leyton, J. et al. Noninvasive imaging of cell proliferation following mitogenic extracellular kinase inhibition by PD0325901. Mol. Cancer Ther. 7, 3112–3121 (2008). 90. Liu, X., Abdelrahim, M., Lei, P., Abudayyeh, A. & Safe, S. The nonsteroidal anti- inflammatory drug tolfenamic acid inhibits BT474 and SKBR3 breast cancer cell and tumor growth by repressing erbB2 expression. Mol. Cancer Ther. 8, 1207–1217 (2009). 91. Leinster, D. A. et al. The peritoneal tumour microenvironment of high-grade serous ovarian cancer. J. Pathol. 227, 136–145 (2012). 92. Sala, R. et al. Phosphorylation status of thymidine kinase 1 following antiproliferative drug treatment mediates 3’-deoxy-3’-[18F]-fluorothymidine cellular retention. PLoS One 9, e101366 (2014). 93. Long, B. H. & Fairchild, C. R. Paclitaxel Inhibits Progression of Mitotic Cells to G1Phase by Interference with Spindle Formation without Affecting Other Microtubule Functions during Anaphase and Telephase. Cancer Res. 54, 4355–4361 (1994). 94. Khammanit, R., Chantakru, S., Kitiyanant, Y. & Saikhun, J. Effect of serum starvation and chemical inhibitors on cell cycle synchronization of canine dermal fibroblasts. Theriogenology 70, 27–34 (2008). 95. Wickremasinghe, R. G. & Hoffbrand, A. V. Inhibition by aphidicolin and dideoxythymidine triphosphate of a multienzyme complex of DNA synthesis from human cells. FEBS Lett. 159, 175–179 (1983). 96. Zhang, J. et al. Thymidine kinase 1: a proliferation marker for determining prognosis and monitoring the surgical outcome of primary bladder carcinoma patients. Oncol Rep 15, 455–461 (2006).

189

97. Kyu, Y., Wells, S., Lauo, Y. C. & Lee, A. M. Y. S. Sequences contained within the promoter of the human thymidine kinase gene can direct cell-cycle regulation of heterologous fusion genes. 85, 5894–5898 (1988). 98. White, J. H. M., Green, S. R., Barker, D. G., Dumas, L. B. & Johnston, L. H. The CDC8 transcript is cell cycle regulated in yeast and is expressed coordinately with CDC9 and CDC21 at a point preceding histone transcription. Exp. Cell Res. 171, 223–231 (1987). 99. Alexander, J. L. & Orr-Weaver, T. L. Replication fork instability and the consequences of fork collisions from rereplication. Genes Dev. 30, 2241–2252 (2016). 100. Horii, T. et al. 5-Hydroxymethylcytosine Marks Sites of DNA Damage and Promotes Genome Stability. Cell Rep. 14, 1283–1292 (2016). 101. Coppock, D. L. & Pardee, A. B. Control of thymidine kinase mRNA during the cell cycle. Mol. Cell. Biol. 7, 2925–2932 (1987). 102. Sherley, J. L. & Kelly, T. J. Regulation of human thymidine kinase during the cell cycle. J. Biol. Chem. 263, 8350–8358 (1988). 103. Huang, S. et al. Human dTMP Kinase : Gene Expression and Enzymatic Activity Coinciding with Cell Cycle Progression and Cell Growth. 13, 461–471 (1994). 104. Ke, P., Kuo, Y., Hu, C. & Chang, Z. Control of dTTP pool size by anaphase promoting complex / cyclosome is essential for the maintenance of genetic stability. 1, 1920– 1933 (2005). 105. Marverti, G. et al. Collateral sensitivity to novel thymidylate synthase inhibitors correlates with folate cycle enzymes impairment in cisplatin-resistant human ovarian cancer cells. Eur. J. Pharmacol. 615, 17–26 (2009). 106. Ligabue, A., Marverti, G., Liebl, U. & Myllykallio, H. Transcriptional Activation and Cell Cycle Block Are the Keys for 5-Fluorouracil Induced Up-Regulation of Human Thymidylate Synthase Expression. 7, (2012). 107. Fukuda, K. et al. Antitumor effect of cetuximab in combination with S-1 in EGFR- amplified gastric cancer cells. Int. J. Oncol. 40, 975–982 (2012). 108. PEREA, S. E. et al. Synergistic interactions of the anti-casein kinase 2 CIGB-300 peptide and chemotherapeutic agents in lung and cervical preclinical cancer models. Mol. Clin. Oncol. 2, 935–944 (2014). 109. Chen, Y.-L., Eriksson, S. & Chang, Z.-F. Regulation and functional contribution of thymidine kinase 1 in repair of DNA damage. J. Biol. Chem. 285, 27327–35 (2010). 110. Hu, C. M. et al. Tumor Cells Require Thymidylate Kinase to Prevent dUTP Incorporation during DNA Repair. Cancer Cell 22, 36–50 (2012). 111. Kaliszczak, M. et al. Development of a cyclin-dependent kinase inhibitor devoid of ABC transporter-dependent drug resistance. Br. J. Cancer 109, 2356–2367 (2013). 112. Slade, R. L. et al. Identification of ABC transporter interaction of a novel cyanoquinoline radiotracer and implications for tumour imaging by positron emission tomography. PLoS One 11, 1–21 (2016). 113. Jensen, R. L., Origitano, T. C., Lee, Y., Weber, M. & Wurster, R. D. In Vitro Growth Inhibition of Growth Factor-stimulated Meningioma Cells by Calcium Channel

190

Antagonists. Neurosurgery 36, 365–374 (1995). 114. Lee, Y. S., Sayeed, M. M. & Wurster, R. D. Inhibition of cell growth and intracellular Ca2+ mobilization in human brain tumor cells by Ca2+ channel antagonists. Mol. Chem. Neuropathol. 22, 81–95 (1994). 115. Taylor, J.M. & Simpson, R. U. Inhibition of Cancer Cell Growth by Calcium Channel Antagonists in the Athymic Mouse. Cancer Res. 52, 2413–2418 (1992). 116. Thorn, C.F., Oshiro, C., Marsh, S., Hernandez-Boussard, T., McLeod, H., Klein, T.E. and Altman, R. B. Doxorubicin pathways: pharmacodynamics and adverse effects. Pharmacogenet. Genomics 21, 440–446 (2012). 117. Fortuna, A., Alves, G. & Falcão, A. The importance of permeability screening in drug discovery process: PAMPA, Caco-2 and rat everted gut assays. Curr. Top. Pharmacol. 11, 63–86 (2007). 118. Fedorov, Y. et al. Off-target effects by siRNA can induce toxic phenotype. Rna 12, 1188–1196 (2006). 119. Millar, B. Y. J. A., Cumming, R. L. C., Smith, J. A. & Goldberg, A. Effect of Actinomycin D , Cycloheximide , and Acute Blood Loss on Ferritin Synthesis in Rat Liver. 643–649 (1970). 120. Singh, D., Sharma, A. & Singh, M. Effect of actinomycin D and cycloheximide on ischemic preconditioning- induced delayed cardioprotective effect in rats. 38, 982– 987 (2000). 121. Giannoni, F., Nguyen, V. T. & Bensaude, O. The Transcriptional Inhibitors , Actinomycin D and ␣ -Amanitin , Activate the HIV-1 Promoter and Favor Phosphorylation of the RNA Polymerase II C-terminal Domain *. 274, 16097–16106 (1999). 122. Chang, H. et al. CRISPR/cas9, a novel genomic tool to knock down microRNA in vitro and in vivo. Sci. Rep. 6, 1–12 (2016). 123. Patton, E. E., Anastasaki, C., Cáceres, J. F., Longman, D. & Capper, A. Dhx34 and Nbas function in the NMD pathway and are required for embryonic development in zebrafish. Nucleic Acids Res. 39, 3686–3694 (2011). 124. Hurlbert, R.B. ;Schmitz, H.;Brumm, A. F. Nucleotide metabolism II. Chromatographic Seperation Of Acid-Soluble. J. Biol. Chem. (1953). 125. Deberardinis, R. J. et al. Beyond aerobic glycolysis : Transformed cells can engage in glutamine metabolism that exceeds the requirement for protein and nucleotide synthesis. (2007). 126. Campbell, M. D. & Marcinek, D. J. Biochimica et Biophysica Acta Evaluation of in vivo mitochondrial bioenergetics in skeletal muscle using NMR and optical methods. BBA - Mol. Basis Dis. 1862, 716–724 (2016). 127. Hu, C. M. & Chang, Z. F. Synthetic lethality by lentiviral short hairpin RNA silencing of thymidylate kinase and doxorubicin in colon cancer cells regardless of the p53 status. Cancer Res. 68, 2831–2840 (2008). 128. Carnrot, C., Wehelie, R., Eriksson, S., Bölske, G. & Wang, L. Molecular characterization of thymidine kinase from Ureaplasma urealyticum: Nucleoside analogues as potent inhibitors of mycoplasma growth. Mol. Microbiol. 50, 771–780

191

(2003). 129. Lee, Syng-lih; Cheng, Y.-C. Human Deoxythymidine Kinase. Biol. Chem. 251, 2600– 2604 (1976). 130. Aquilano, K., Baldelli, S. & Ciriolo, M. R. Glutathione: New roles in redox signalling for an old antioxidant. Front. Pharmacol. 5 AUG, 1–12 (2014). 131. Floyd, S., Favre, C., Lasorsa, F.M., Leahy, M., Trugiante, G., Stroebel, P., Marx, A., Loughran, G., O’Callaghan, K., Marobbio, C. M.T., Slotboom, D. J., Kunji, E. R.S., Palmieri, F. & O’Connor, R. The Insulin-like Growth Factor-I–mTOR Signaling Pathway Induces the Mitochondrial Pyrimidine Nucleotide Carrier to Promote Cell Growth. Mol. Biol. Cell 18, 3545–3555 (2007). 132. Sato, K., Kanno, J., Tominaga, T., Matsubara, Y. & Kure, S. De novo and salvage pathways of DNA synthesis in primary cultured neurall stem cells. 71, (2006). 133. Cory, J. G. & Cory, A. N. N. H. (Cory (2006) Role of Glutamine in nucleotide synthesis. 590, 587–589 (2006). 134. Carcamo, W. C., Calise, S. J., von Mühlen, C. A., Satoh, M. & Chan, E. K. L. Molecular Cell Biology and Immunobiology of Mammalian Rod/Ring Structures. Int. Rev. Cell Mol. Biol. 308, 35–74 (2014). 135. Hu, R., Lam, W., Hsu, C. H. & Cheng, Y. C. Ump/cmpk is not the critical enzyme in the metabolism of pyrimidine ribonucleotide and activation of deoxycytidine analogs in human rko cells. PLoS One 6, (2011). 136. Xu, Y., Johansson, M. & Karlsson, A. Human UMP-CMP kinase 2, a novel nucleoside monophosphate kinase localized in mitochondria. J. Biol. Chem. 283, 1563–1571 (2008). 137. Hanson, A.D., Pribat, A., Waller, J., De Crécy-largard, V. 'Unknown’ proteins and ‘orphan’ enzymes: the missing half of the engineering parts list – and how to find it. Biochem J 425, 1–11 (2011). 138. Randic, M. Aromaticity and conjugation. J. Am. Chem. Soc. 99, 444–450 (1977). 139. Genheden, S. & Ryde, U. The MM/PBSA and MM/GBSA methods to estimate ligand- binding affinities. Expert Opin. Drug Discov. 10, 449–461 (2015). 140. Schiepers, C. et al. 18F-fluorothymidine kinetics of malignant brain tumors. Eur. J. Nucl. Med. Mol. Imaging 34, 1003–1011 (2007). 141. McKnight, T. R. et al. Choline metabolism, proliferation, and angiogenesis in nonenhancing grades 2 and 3 astrocytoma. J. Magn. Reson. Imaging 33, 808–816 (2011). 142. Shields, A. F., Grierson, J.R., Dohmen, B.M., Machulla, H.J., Stayanoff, JC., Lawhorn- Crews, J.M. Obradovich, J.E., Muzik, O. & Mangner, T.J. Imaging proliferation in vivo with [F-18]FLT and positron emission tomography. 4, 1334–1336 (1998). 143. Han, J. et al. Glucose promotes cell proliferation, glucose uptake and invasion in endometrial cancer cells via AMPK/mTOR/S6 and MAPK signaling. 138, 668–675 (2016). 144. Trousil, S. et al. The novel choline kinase inhibitor ICL-CCIC-0019 reprograms cellular metabolism and inhibits cancer cell growth. (2016).

192

145. Abrams, A. J. et al. Mutations in SLC25A46, encoding a UGO1-like protein, cause an optic atrophy spectrum disorder. Nat. Genet. 47, 926 (2015). 146. Niida, H. et al. Essential role of Tip60-dependent recruitment of ribonucleotide reductase at DNA damage sites in DNA repair during G1 phase. Genes Dev. 24, 333– 338 (2010). 147. Blount, B. C. et al. Folate deficiency causes uracil misincorporation into human DNA and chromosome breakage: Implications for cancer and neuronal damage. Proc. Natl. Acad. Sci. 94, 3290–3295 (1997). 148. Bauer, S. & Patterson, P. H. The cell cycle-apoptosis connection revisited in the adult brain. J. Cell Biol. 171, 641–650 (2005). 149. Siddiqui, A. et al. Thymidylate synthase maintains the de-differentiated state of triple negative breast cancers. Cell Death Differ. doi:10.1038/s41418-019-0289-6 150. Elliott, T. S., Slowey, A., Ye, Y. & Conway, S. J. The use of phosphate bioisosteres in medicinal chemistry and chemical biology. Medchemcomm 3, 735–751 (2012).

193