Global initiatives applying next generation sequencing to thousands of primary and metastatic tumour samples have detailed the genomic landscape of breast cancer. These unique insights have transformed our understanding of the biological processes underlying these tumours. Despite this, the clinical relevance of these findings remains poorly understood. This is because these datasets are frequently heterogeneous, include patients that may not have received contemporary treatments, often have limited clinical annotation, and commonly lack long-term follow-up for patient survival outcomes. In order to better assess the clinical relevance of genomic driver alterations in breast cancer, this research project performed genomic sequencing using primary tumour samples from patients enrolled in two pivotal phase 3 clinical trials – the Breast International Group 1-98 (BIG 1-98) study and the Suppression of Ovarian Function Trial (SOFT). This provided a well-defined population with hormone receptor-positive, human epidermal growth factor receptor (HER) 2-negative early breast cancer in both postmenopausal and premenopausal populations, as well as long-term survival outcome data collection. Crucially, this allowed for further investigation of late distant recurrence, a unique feature of hormone receptor-positive, HER2-negative breast cancers, as well as the ability to focus on distinct high-risk subpopulations, such as breast cancer arising in young women. In the postmenopausal BIG 1-98 study, comprehensive panel sequencing was performed using 538 tumour samples from patients with hormone receptor-positive, HER2-negative early breast cancer. PIK3CA mutations were the most common driver (49%) and were associated with a reduction in the risk of distant recurrence. Contrastingly, TP53 mutations, amplifications on 11q13 and 8p11, and increasing number of driver alterations were associated with an increase in the risk of distant recurrence. Additionally, PIK3CA mutations were predictive of a greater magnitude of benefit from letrozole compared with tamoxifen. Using the same dataset, we further investigated the association of oncogenic drivers with late distant recurrence (5 years or greater from the time of randomisation). PIK3CA mutations were associated with reduced risk of late distant recurrence, whereas amplifications on 8p11 and BRCA2 mutations were associated with increased risk of late distant recurrence. In the premenopausal SOFT study, comprehensive panel sequencing was performed using 1258 tumour samples from the entire study cohort, and whole exome sequencing was performed using 82 tumour samples from very young patients with high-risk, hormone receptor-positive, HER2-negative early breast cancers. Subgroups of tumours with distinct copy number-amplifications had a higher risk of distant recurrence than tumours that were amplification-devoid, validating previously reported data. Importantly, the study further defined unique molecular characteristics that were enriched in very young women and may explain their poor prognosis. This includes subgroups of tumours with homologous recombination deficiency, a high-risk PIK3CA mutated subgroup, amplification-enriched subsets, and a low estrogen receptor expressing subtype with immune infiltration. Taken together, these results highlight the clinical relevance of genomic sequencing of hormone receptor-positive, HER2-negative early breast cancers. They provide a new prognostic framework using genomic stratification, and highlight molecular targets that should be prioritised for future clinical trial research.

Malignant mesothelioma is an aggressive cancer of the mesothelium for which asbestos-exposure is the predominant etiological factor. The standard treatment regime for mesothelioma is systemic chemotherapy, which exhibits poor efficacy with a median post-diagnosis survival of 12 months. This highlights the need for alternate therapies, such as targeted therapies, to improve patient outcomes. Genomic and cell biological analyses of mesotheliomas indicate that deficiency of key tumour suppressor genes including BAP1 or members of the Hippo tumour suppressor pathway such as NF2, are frequent genetic aberrations in this cancer. Inhibition of cancer-promoting events that occur downstream of these genetic aberrations represents a potential mechanism by which mesothelioma patient outcomes can be improved. Originally identified in Drosophila, the Hippo pathway is highly conserved in mammals and regulates the activity of the oncoproteins and transcriptional coactivators YAP and TAZ and their nuclear interaction partners, the TEAD family of transcription factors. YAP/TAZ-TEAD hyperactivation is an oncogenic driver in Hippo pathway deregulated mesotheliomas, indicating that inhibition of these proteins could represent a new treatment strategy for this disease. An emerging class of small molecule inhibitors of YAP/TAZ-TEAD target the recently identified post-translational modification of TEAD palmitoylation. One such compound, VT107, has shown promising pre-clinical activity in in vitro and in vivo models of Hippo pathway deregulated mesothelioma. Whilst VT107 is a promising therapeutic, the current understanding of its molecular impact is incomplete. Moreover, further identification of therapeutic targets in BAP1 or NF2 deficient mesothelioma cells might enable development of additional targeted therapies to treat candidate tumours. To address these considerations, I aimed to answer three key questions: 1) Which genes are potential therapeutic targets for treatment of NF2 or BAP1 mutant mesothelioma?; 2) How does VT10- 7 modulate the behaviour and transcription regulatory activity of YAP/TAZ-TEAD?; and 3) How does VT107 inhibit the proliferation of Hippo pathway mutant mesothelioma cells? 1) To identify candidate therapeutic targets for treatment of NF2 or BAP1 deficient mesothelioma, I conducted a genome-wide synthetic lethality screen. Due to technical limitations with BAP1 deficient cells, I conducted the screen with only NF2 deficient cells. While I identified a mild synthetic lethal interaction between NF2 and CREB3, I concluded that the screen was unsuccessful in identifying robust synthetic lethal interaction partners of NF2. 2) I assessed the molecular impact of VT107 in a Hippo pathway mutant mesothelioma cell line (NCI-H2052) using single molecule tracking of YAP and TEAD1. With this advanced live microscopy approach, I observed that VT107 increased the nuclear mobility, and reduced the DNA residence-time, of both YAP and TEAD1. Thus, I provided a high resolution molecular understanding of how VT107 influences the ability of YAP and TEAD1 to regulate transcription. 3) I conducted a genome-wide CRISPR screen to investigate the mechanism of action of VT107 in NCI-H2052 cells. I identified that loss of SOCS3, VGLL4, BTAF1 and DR1, among other genes, increased cellular resistance to VT107. I also found that loss of several genes, including TCEB2, could confer VT107 sensitivity. These findings provide unbiased insights into the mechanism by which YAP/TAZ-TEAD inhibition modulates the proliferation of mesothelioma cells. Furthermore, I identified potential drivers of resistance to VT107- information which might guide its optimal application in the clinic for the treatment of Hippo pathway mutant mesothelioma.

Cytotoxic secretory granules (CGs) are unique and specialised lysosomal organelles found specifically in cytotoxic lymphocytes: Natural Killer (NK) cells and CD8+ T cells or Cytotoxic T-Lymphocytes (CTLs). CGs precisely store and deliver perforin and granzymes; molecules that synergistically induce the controlled death of infected and neoplastic cells. The precise molecular mechanisms that govern CG formation remain poorly understood and this work aimed to explore the potential role of the MiT/TFE family of proteins in transcriptionally regulating this process. Three members of the MiT/TFE family (MITF, TFEB and TFE3) are largely recognised as the master regulators of lysosome biogenesis and as the main drivers in the formation of organelles with a lysosomal nature. Under this hypothesis the role of the microphthalmia associated transcription factor (MITF) was studied. CTLs expressing mutant MITF showed a reduced ability to expand and kill target cells in vitro in response to antigen-specific stimulation. However, it was determined that the defect was not intrinsic to the lymphocytes but associated to antigen presenting cells required for CTLs activation. Having ruled out MITF in the formation of CGs, the role of transcription factor EB (TFEB) and transcription factor E3 (TFE3) was explored next. By studying the subcellular trafficking of these two proteins during mouse CTL activation, this work described that TFEB, but not TFE3, translocated from the cytosol into the nucleus. Despite the nuclear TFEB presence, RNA-sequencing of CTLs did not conclusively showed upregulation of canonical TFEB target genes. To gain further understanding on the role of TFEB in CTLs, the lysosomal storage disorder (LSD) Niemann-Pick type C1 (NP-C1) was implemented as a model. This devastating disease, caused by mutations in the lysosomal cholesterol transporter, NPC1 proved to be suitable for this purpose, given that it was shown here that TFEB is retained in the cytosol of CTLs in a mouse model of the disease and this was accompanied by a significant reduction of their cytotoxic function. The implementation of various microscopy techniques led to the discovery that NPC1-defficient CTLs accumulate aberrant autophagic bodies that fuse, or embed, lipid-dense CGs. These observations were corroborated in CTLs derived from a cohort NP-C1 patients. Finally, it was established that reconstituting nuclear TFEB was sufficient to restore autophagic dysregulation, but it was not linked to cytotoxic improvement. Contrastingly, pharmacological depletion of aberrantly accumulated lipids within CGs of NPC1-defficient CTLs, completely restored their cytotoxic function. The work presented here shows that the mechanism behind it is linked to lipid inactivation of the pore forming protein, perforin, and the consequent inability of NPC1-defficient CTLs to kill target cells. Overall, the evidence presented here suggests that TFEB plays a critical role in CTL activation which might be unrelated to its canonical lysosomal biogenesis transcriptional regulation. Moreover, while studying this transcription factor in CTLs in the context of the LSD, NP-C1, a number of phenotypical abnormalities were discovered, including accumulation of autophagic bodies and reduced cytotoxicity. Importantly, the impairment in the autophagic flux in NPC1-defficient CTLs was linked to limited TFEB nuclear translocation while the diminished capacity to kill target cells was associated with lipid accumulation within CGs.

Mutations in the FMS-like tyrosine kinase 3 (FLT3) gene are among the most frequently occurring somatic mutations in acute myeloid leukaemia (AML), a disease that presents with devastating prognosis. FLT3 is primarily expressed on hematopoietic progenitor cells, and during early haematopoiesis coordinates a ligand-dependent signalling cascade that regulates the proliferation and maturation of the progenitor pool. FLT3 internal tandem duplication (FLT3-ITD), the most common type of FLT3 mutation, promotes constitutive FLT3 kinase activity and hyperactivation of downstream signalling pathways including STAT5, PI3K/mTOR and MAPK. Though our understanding of the molecular pathways downstream of mutant FLT3 have greatly improved in the modern genomic sequencing era, the repertoire of molecular signalling events induced by mutant-FLT3 to drive leukaemogenesis has not been fully characterised. In this thesis, a murine model of MLL-rearranged AML harbouring inducible FLT3-ITD expression was developed and used, along with human AML cell lines, to demonstrate that FLT3-ITD promotes serine uptake and serine biosynthesis via transcriptional regulation of neutral amino acid transporters (SLC1A4 and SLC1A5) and genes in the de novo serine biosynthesis pathway (PHGDH and PSAT1). Mechanistically, dysregulation of serine metabolism in FLT3-ITD-driven AML is dependent on the mTORC1-ATF4 axis, which drives RNA-Pol II occupancy at PHGDH, PSAT1, SLC1A4 and SLC1A5. Genetic or pharmacological inhibition of the de novo serine biosynthesis pathway, in vitro and in vivo, selectively inhibited the proliferation of FLT3-ITD-driven AML cells. Pharmacological inhibition of the de novo serine biosynthesis pathway using WQ-2101, an inhibitor of PHGDH, the first rate-limiting enzyme of the de novo serine biosynthesis pathway, sensitises FLT3-ITD-driven AML cells, including primary patient samples, to the standard of care chemotherapy agent cytarabine via exacerbation of DNA damage. Given that transcriptional activation of de novo serine biosynthesis and serine uptake was mediated by mTORC1 (a master regulator of biomass production and cellular metabolism), and that therapeutic responses in vitro and in vivo to mTORC1 inhibitors are poor, a small molecule compound screen utilising 181 epigenetic inhibitory compounds to determine novel synthetic lethal interactions between epigenetic regulators and mTORC1 inhibition was performed. This analysis revealed that inhibition of lysine-specific methyltransferase SETD8, the only known enzyme that catalyses H4K20 monomethylation (H4K20me1), synergised with mTORC1 inhibition. Transcriptional profiling suggested dual SETD8/mTORC1 inhibition preferentially suppressed mTORC1 target genes mediating amino acid biosynthesis and transamination (including de novo serine biosynthesis) to a greater extent than either single agent SETD8 or mTORC1 inhibition alone. Importantly, these observations were independent of global transcriptional repression induced by impaired cell viability or suppression of global transcription. Thus, this preliminary work suggests mTORC1 and/or its target genes and pathways may be dependent on SETD8 and/or H4K20me1 or, alternatively, mTORC1 functionally regulates SETD8. Collectively, the results presented herein provide novel insights into FLT3-ITD-induced metabolic reprogramming events in AML and identify a targetable metabolic dependency in this poor prognosis subtype of disease. In addition, these results provide the preliminary basis of a SETD8/mTORC1 synthetic dependency that can be exploited in FLT3-ITD-driven AML.

Multiple myeloma (MM) is an incurable plasma cell neoplasm. The incorporation of immunomodulatory imide drugs (IMiDs) – thalidomide, lenalidomide and pomalidomide - into clinical practice has improved survival and quality of life of MM patients. However, most patients develop therapeutic resistance to IMiDs, representing a significant clinical problem. IMiDs exert anti-MM effects by engaging Cereblon (CRBN), a ligand of the E3 ligase complex CUL4-DDB1-RBX1. By binding CRBN, IMiDs promote ubiquitination and proteasomal degradation of important MM cell transcription factors, IKZF1 (Ikaros) and IKZF3 (Aiolos). Degradation of such neosubstrates has also been shown to augment an anti-cancer immune response by activation of cytotoxic effector cells against MM plasma cells. Finally, it has been demonstrated that IMiDs modulate events occurring within the bone marrow microenvironment with negative impact on the MM-sustaining niche. Despite these recent advances in the characterisation of IMiD mechanism of action, the precise molecular events underpinning the anti-MM activity of IMiDs remain incompletely understood. Further insight is required to improve clinical responses to IMiDs and overcome therapy resistance in MM. In this work, genome-wide approaches such as RNA-sequencing and CRISPR-based functional genomics screening were employed to dissect the molecular mechanisms of IMiD-sensitivity and -resistance in models of acquired resistance to lenalidomide. RNA-sequencing revealed a strong interferon-like transcriptional response and increased transcription of genes downstream the MAPK pathway, together with downregulation of MYC-responsive genes, following IMiD treatment of IMiD-sensitive MM cells. Conversely, IMiD-resistant cells displayed reduced transcriptional changes following IMiD exposure as a likely consequence of CRBN downregulation and subsequent attenuation of neosubstrate degradation. Genome-wide CRISPR screening demonstrated that loss of CRBN and other genes regulating protein degradation, such as subunits of the COP9 signalosome, but also loss of genes that are not implicated in protein turnover, such as NCOR1 and EDC4, is sufficient to cause resistance to IMiDs in IMiD-sensitive cells. CRIPSR screening also demonstrated that resensitisation of IMiD-resistant cells to IMiDs can occur through loss of ATXN7, TOP2B and other genes implicated in several distinct biological processes, such as glucose metabolism, cell cycle progression and RNA processing. Dexrazoxane, a pan-TOP2 inhibitor and TOP2B-selective degrader, was discovered to possess anti-proliferative properties and combinatorial activity with lenalidomide in certain MM cell lines. To develop a deeper understanding of the biology of CRBN and IMiDs, proximity labelling with BioID2 was employed to characterise the interactome of CRBN in the presence and absence of an IMiD. Co-treatment of BioID2-CRBN-expressing OPM2 cells with IMiDs and/or the proteasome inhibitor bortezomib, followed by liquid chromatography mass spectrometry proteomics, allowed identification of known CRBN interactors (COP9 signalosome subunits) and neosubstrates (IKZF1, IKZF3 and CK1alpha). This approach enabled further characterisation of CRBN-associated proteome and identification of new CRBN and CRBN-IMiD interacting partners, such as MEF2C, DVL1, DVL2, MYH9, PRRC2C and EDC4. This data led to the hypothesis that CRBN may be implicated in several novel biological processes, such as RNA processing and protein translation regulation. Additionally, some of the newly identified interacting partners may modulate the MM niche in the bone marrow and may help understand the occurrence of side effects to IMiD therapy. This work has further defined transcriptional signatures of IMiD treatment and of acquired IMiD-resistance. TOP2B was found to be a potential druggable vulnerability of MM cells with acquired IMiD-resistance. The unbiased study of the CRBN-interacting proteome has helped broaden the knowledge on its biological activities both in presence and absence of an IMiD. Collectively, this thesis provides new and important insight into IMiD mechanism of action and further expands knowledge of IMiD biology.

Intra and inter-tumour heterogeneity poses a challenge for associating molecular and immunohistochemical markers with clinical outcomes. Sequencing technologies has enabled detailed assessment of tumour heterogeneity, facilitating the genomic characterisation of tumours. Whilst such technologies have revealed mutational landscapes and have identified key driver alterations for tumorigenesis, pan-cancer clonal evolution reconstructions are lacking. In order to bridge this gap, I used discrete-time branching models to derive biological insights into tumour progression and reconstructed the clonal evolution in 1,800 patients, successfully linking mutations with growth patterns of disease progression. I first modified a discrete time-branching process to account for individual clonal subpopulations and derived analytical solutions for expectation and variance of both clonal and tumour expansions. Additionally, I derived the expected time for any given clone to successfully expand as \hat{\tau}, and with the use of these analytical solutions, I showed the likely driver and clonal compositions of the tumours and their phylogenies. Secondly, I generated a database of results from four different versions of time-branching process models that covered multiple parameters. Here I identified how an increase in diversity arises by both increased mutation rate and reduced fitness. I further corroborated that total number of drug resistant cells is directly proportional to lineage extinction probability (\delta) and tumour size as shown in previous studies. I also showed that this effect can be extrapolated to other types of functional passenger mutations involved in cancer-specific mortality. Moreover, I showed how commonly used sequencing cut-offs limit the accurate inference of tumour’s average selective advantage and driver mutation rate. Thirdly, I identified that a minimum distance metric can provide accurate fits of simulated cancer cell fractions to real patient tumour data. This metric showed at least 80% accuracy to identify the initial parameters of s and u and at least 40% accuracy to recover the correct evolutionary trajectory. Fourthly, I applied this fitting procedure to reconstruct the evolutionary trajectories of 1,800 tumours from different cancer sequencing studies. The best fits derived suggests that the most likely parameters for the evolution of solid tumours are high driver mutation rates and weak driver effects of fitness. Fifthly, using The Cancer Genome Atlas cohort, I identified an association between predicted degree of clonality and survival, and found branched topologies are common in malignancies with adverse prognosis. In the TRACERx non-small-cell lung cancer cohort, I identified that clonal reconstructions agreed with previously reported phylogenies. Additionally, using data from the Breast International Group 1-98, I identified the role of tumour fitness in determining clinical outcome, and the evolutionary dynamics of TP53 and PIK3CA mutations conducive to distant metastasis. Finally, using data from a metastatic melanoma patient collected through the CASCADE melanoma study, I was able to propose a pattern of dissemination from the primary to metastatic sites in the liver and brain based on the phylogenies recovered from my data-fitting procedure. This study demonstrates the power of the discrete-time branching process in reconstructing tumour evolution, and its potential to uncover insights in the dynamics of tumour growth that are missed by current methods.

Phenotypic diversity of breast cancers poses insurmountable challenges in the treatment of this lethal disease. Recent advances in next generation sequencing have led to unprecedented insight into the genomic landscape underlying breast tumours. This has resulted in burgeoning development of targeted treatments and predictive biomarkers, several of which have recently demonstrated clinical activity. However, key challenges hinder optimal application. On the background of extensive molecular heterogeneity, most biomarkers represent minority patient subpopulations, hampering clinical development. Furthermore, considerable genomic evolution of breast tumours impacts accuracy of genomic characterisation that is thus far heavily reliant on the sequencing of non-contemporaneous and invasive tumour tissue biopsies. Finally, stratification to genomically-matched targeted therapies also fails to fulfil the extent of its promise. In many cases relentless tumour growth remains unperturbed, while in others resistance ultimately develops. Crucially, molecular mechanisms underlying resistance remain poorly understood, while follow-on treatment options are often poorly defined. Central to the promise of personalised medicine is the robust and accurate characterisation of the tumour genome. Minimal invasiveness and convenience of circulating tumour DNA (ctDNA) analysis, with ability to detect tumour genomic aberrations from a blood draw, highly recommends this approach. Recent technological advances have paved the way to a range of clinical applications, with evolving potential for ctDNA analysis to address the continuum of challenges posed to precision medicine throughout patient management. Toward this end, extensive clinical development is required, while prevailing technological hurdles need to be addressed. This thesis explores a multi-faceted and rigorous approach towards the integration of ctDNA analysis in the management of breast cancer patients. Firstly, the development and validation of multiple assays (allele-specific and NGS-based) tailored to breast cancers, enabled comprehensive genomic analysis with in-built flexibility to be readily applicable to a variety of clinical scenarios. Subsequent establishment of a prospective ctDNA-based molecular profiling program across a large cohort of metastatic breast cancer (mBC) patients demonstrated feasibility of real-time analysis in the clinical setting across a range of genomic targets of variable abundance. Importantly, integration of longitudinal testing in this program throughout patient management demonstrated capacity for ctDNA analysis to reflect genomic evolution in real-time to optimise precision-guided patient management. Finally, exploratory longitudinal ctDNA analysis for the study of resistance mechanisms to CDK4/6i, constituting a novel class of targeted compounds for breast cancer, highlights established and novel resistance markers. Indeed, this study also serves to demonstrate a workable framework for ctDNA analysis as a highly effective approach for the de novo elucidation of resistance mechanisms to novel targeted agents that is relevant across cancer types.

Rapid advancement in genomic technologies has driven down the cost of sequencing significantly. This efficiency has enabled large-scale cancer genomic studies to be conducted, generating a vast amount of data across different levels of omics variables. However, the tasks to extract new knowledge and information from this enormous volume of data present unique challenges. These analyses often require the application of specialised techniques for data mining, integration and interpretation to provide valuable insights. With the rise of machine learning adoption in recent decades, many advanced computational algorithms based on artificial intelligence techniques have also been proposed to analyse these genomics data. Although some of these applications have led to clinically relevant conclusions, many others are still relying on incomplete prior knowledge, or limited to only a selected number of features. These limitations raise the general question about the broader applicability of machine learning in the field of cancer genomics. This research addresses this question by assessing the application of machine learning techniques in the context of breast cancer genomics data. This assessment includes a comprehensive evaluation of computational methods for predicting cancer driver genes and the development of a novel deep learning approach for identifying breast cancer subtypes. The evaluation result of driver gene prediction algorithms suggests that the selection of the best method to be applied to a dataset will primarily be driven by the objectives of the study and the characteristics of the dataset. All of the evaluated approaches could identify well- studied genes, but not all of them performed as well on smaller datasets, subtype-specific cohorts, and in discovering novel genes. To examine the benefit of a more complex machine learning model, this thesis also presents a novel deep learning approach that integrates multi-omics data for predicting various breast cancer’ biomarkers and molecular subtypes. This method combines a semi-supervised autoencoder for dimensionality reduction, and a supervised multitask learning setup for the classifications. Taking an input of gene expression, somatic point mutation and copy number data, the algorithm predicts the ER-Status, HER2-Status and molecular subtypes of breast cancer samples. Further survival analysis of the outputs from this deep learning approach indicates that the predicted subtypes show a stronger correlation with patient prognosis compared to the original PAM50 label. While the outputs from machine learning algorithms still require further validation, the adoption of these complex computational methods in cancer genomics will become increasingly common. Collectively, the results from this thesis suggest that the machine learning analysis of ‘omics data hold great potential in automating the discovery of clinically- relevant molecular features.

Prostate cancer (PCa) is one of the most diagnosed tumours in Australia, pertaining to more than 30% of all cancer detection (AIHW 2020). Most of these diagnoses are correlated to men with early, localised stages of the disease that are easily treated with radical proctectomy surgery and/or radiation. One third of these patients will relapse with their disease progressing to the either the recurrent localised stage or the metastatic stage. These advanced cases will have to undergo Androgen Deprivation Therapy (ADT), as hormone dependence is a key driver of prostate tumorigenesis. ADT will help increase patient survival by halting cancer progression, while minimising symptoms. Although this type of therapy can initially alleviate tumour burden, patients will inevitably be subjected to cancer recurrence where their tumours continue to progress despite castrate levels of androgens. An observed second-order effect of ADT is that it creates a therapeutic selection pressure within PCa, leading to the proliferation and establishment of cells that are resistant to androgen ablation. These hormone-refractory cells lead to the development of Castration Resistant Prostate Cancer (CRPC), a stage of the disease that ultimately leads to death. Despite the development of second-generation ADT or chemotherapeutic agents, no effective therapies have been developed yet to combat this lethal stage of the disease. Therefore, there is a need for therapies that can target pathways and mechanisms that are not related to androgen signalling. MYC overexpression is one of the most frequently characterised genomic aberrations within men with CRPC. MYC mRNA overexpression is found in 30% of men with localised disease and in 85% of men with CRPC. Although MYC seems like a prime therapeutic target to combat PCa, previous attempts at developing MYC inhibitors have failed due to major off-target effects making these therapies dangerous. The overexpression of MYC is tightly correlated with increased ribosome biogenesis because MYC-driven tumours have an inherent need for protein synthesis to sustain infinite cell proliferation. The rate limiting step of ribosome biogenesis is the transcription of a ribosome RNA precursor by RNA polymerase I within the nucleolus. MYC orchestrates the over-transcription of this precursor by tightly interacting with Pol I associated components. PIM kinases and the PI3K pathways also tightly interact with MYC in order to increase its stability and prevent its degradation. It can be inferred from this that MYC-driven tumours are highly proliferative, lack any type of cell cycle regulation, and are constitutively producing proteins to sustain their uncontrolled growth. Administration of CX-5461, an RNA Pol I inhibitor, was followed by potent anti-proliferative activity and cell cycle arrest within various in vitro and in vivo models of prostate cancer. Strikingly, the combination of Pol I inhibition along with CX-6258, a PIM kinase inhibitor, was characterised by potent reduction in invasive lesions within two mouse models of prostate cancer. Although both drugs combined seemed to have potential additive effect, little is understood about the mechanism of action of these two inhibitors together. Therefore, one of the main aims of this thesis is to understand the efficacy of this combination. Both drugs will be tested on different subtypes of human prostate cancer in the explant settings in order to correlate patient-drug response with specific genomic or pathological profiles. A further goal for this project will be to utilize mass spectrometry to characterise the main phosphorylation events that occur as a result of the administration of these inhibitors, both alone and in combination. This will further the understanding of the mechanistic action of both drugs combined. Due to the limited efficacy of CX-5461 as a single agent, PMR-116, a novel Pol I inhibitor, will also be tested within different models of prostate cancer to evaluate as a potential anti-cancer drug. The final aim of this thesis will be to combine CX-5461 with targeted 177Lu-PSMA-617 therapy, as the latter has recently been shown to potentiate radiation-induced DNA damage. Through the various aims of this thesis, we hope to find and characterise the efficacy of RNA polymerase I therapies on prostate cancer both as single agents or combined therapies.

The phosphoinositide 3-kinase (PI3K) pathway is one of the most commonly activated pathways in a variety of cancers and it has recently been highlighted as one of the primary modulators of cell metabolism. This has opened promises and challenges for the development of therapeutic strategies to target metabolism in cancer cells harbouring mutations in the PI3K pathway. Through an inducible “exon switch” approach, our laboratory has previously generated mice ubiquitously expressing a mutation in Pik3ca, the gene coding for the subunit p110alpha of PI3K. By using this mouse model (UbCreERPik3caH1047R) our laboratory has shown that mutations in the PI3K pathway lead to dramatic severe defects in glucose homeostasis resulting in hypoglycaemia and hypoinsulinemia. In this thesis the causes responsible for the metabolic dysfunction observed in these mutant mice are investigated. This thesis provides evidence that mutations in the PI3K pathway lead to increased glucose uptake by the tissues, inhibition of hepatic gluconeogenesis and inhibition of insulin release from the pancreas. Previous studies performed on the UbCreERPik3caH1047R mouse model have also shown increased body weight and organomegaly. This thesis demonstrates that the increase in body weight, resulting from the activation of the Pik3caH1047R mutation was not associated with adiposity. On the contrary, mutations in the PI3K pathway determine loss of body fat and increased lipolysis in the adipose tissue of mice, whilst the tissue growth is associated to hypertrophy or hyperplasia. Furthermore, oncogenic activation of the Pik3caH1047R mutation in vivo leads to alteration of the respiratory exchange rate and energy expenditure of the mice and stimulates browning of the adipose tissue. This thesis also shows that activation of the PI3K pathway alters the expression of genes and proteins involved in metabolic pathways, and that these alterations are organ-specific, therefore opening promises for customising treatments to individual patients.

Head and neck cancer is the 6th most common malignancy, accounting for approximately 4% of malignancies, and 1 – 2% of cancer-related deaths. Radiation therapy utilises high energy radiation to kill cancer cells. In head and neck cancer, radiotherapy is one of the main treatment modalities, particularly in curative-intent treatments. Despite advancements in imaging and radiation treatment planning and delivery, the prescribed dose and radiation treatment workflow remained unchanged and is largely ‘one size fits all’. Similarly, the survivorship program for patients with head and neck cancer is ‘one size fits all’, often one standard institutional follow up schedule for all patients treated for head and neck cancer, regardless of expected risk of treatment-related late toxicities, patients’ subsequent risk of recurrence and second malignancy. This thesis focuses on the value and efficacy of imaging and blood biomarkers in improving treatment personalisation in patients with head and neck cancer. In chapters 1 and 2, I explored the use of imaging and blood markers in the pre-, during, and post-radiotherapy settings to further improve risk stratification. In chapter 1, I investigated the potential use of readily available and ‘cheap’ blood biomarkers (neutrophil and lymphocyte counts) as predictors of subsequent outcomes in a large cohort of patients with oropharyngeal cancer. In chapter 2, I designed and conducted a prospective observational study to systematically characterize the kinetics of gross tumour volume and apparent diffusion coefficient (ADC) changes observed in magnetic resonance imaging (MRI) and circulating tumour cells (CTCs) counts during radiotherapy in patients with head and neck squamous cell cancer. In the survivorship period (Chapter 3), I evaluated the effectiveness of current surveillance program and investigated the potential use of PET imaging and alternative imaging frequencies to improve the cost-effectiveness of the survivorship program. In this chapter, I found that 70% of disease recurrence occur within 2 years and the probability of a surveillance imaging detecting a recurrence in an asymptomatic patient with no adverse clinical finding is very low. Furthermore, in patients with human papillomavirus (HPV)-related oropharyngeal cancer, achieving a complete response on post-treatment PET imaging has a negative predictive value of any subsequent recurrence of 92%, so the yield of surveillance imaging is very low in this group. Using a partially observed Markov decision model, a potentially effective surveillance program with less frequent imaging was propositioned in this chapter. Finally, in chapter 4, I assessed the potential use of re-irradiation in the era of modern imaging and new radiation treatment techniques including intensity-modulated radiotherapy (IMRT), proton therapy and stereotactic body radiotherapy (SBRT). I showed the value of different imaging modality (dual energy CT and MRI) in target delineation in patients who had previous radiation. In addition, I demonstrated that the local control rate for each treatment technique is similar. Although wide field radiotherapy (IMRT and proton therapy) had improve disease-specific survival, treatment with these techniques are longer (typically 6 to 7 weeks) and had higher toxicity rates than SBRT (delivered over 5 treatments).

Gastric cancer (GC) patients are mainly asymptomatic and present at an advanced stage with a 5-year survival rate of only 20-30% in most countries. It is crucial to gain greater insight into the process of gastric carcinogenesis in order to develop tools that will allow improved patient stratification and targeted surveillance. The relationship between intestinal metaplasia (IM), a key premalignant lesion in gastric carcinogenesis, and the diffuse type of gastric cancer (DGC) was investigated using data mining of GC patient pathology reports and existing gene expression data. IM and DGC were shown to be associated both histologically and molecularly in a proportion of cases suggesting that IM might be a precursor lesion to a subset of DGC cases. Complete and incomplete IM subtypes, the latter being associated with a greater risk of progression, were characterised at the molecular level using gene expression data acquired from macro-dissected IM tissue. Complete IM was associated with genes highly expressed in the small intestine whereas incomplete IM was associated with genes expressed both in the colon and in GC suggesting it is molecularly closer to a state of malignancy. Using OPAL multiplex immunohistochemistry, the macrophage and T cell landscape in IM was investigated. In addition to traditional “M1-like” (IRF8) and “M2-like” (CD163, CD206) macrophage populations, a novel hybrid subgroup “M1/2” containing macrophages expressing both M1 and M2 markers in different combinations was identified. Overall, complete IM was characterised by M2 macrophages and greater levels of T cell infiltration whereas incomplete IM was characterised by M1/2 macrophages and fewer CD8 and double negative (DN) T cells. Spatial cell analysis showed significantly fewer CD8 and DN T cells in the vicinity of incomplete IM epithelial cells suggesting reduced immune-surveillance may play a key role in progression to dysplasia. IM subtyping is affected by intra-observer and inter-observer variation. To address this, the potential of CD10 and Das1 as biomarkers for subtyping complete and incomplete IM was investigated. Overall CD10 was shown to be an outstanding biomarker for complete IM and Das1 was shown to have potential as an additional risk associated biomarker. GC animal models are associated with long duration and high cost. Optimised protocols for growing and characterising gastrointestinal organoids from gastroscopy biopsies were developed. IM organoid cultures were successfully established and characterised. Additionally a human gastrointestinal organoid biobank was created. In conclusion this thesis offers potential evidence of IM as a precursor lesion to DGC, characterises the molecular differences between complete and incomplete IM and shows key differences in the innate and adaptive immune system between IM subtypes and how these may play a role in progression to GC. It also identifies two biomarkers with potential clinical utility for subtyping IM and describes the methodology for growing IM organoids with future potential as a model for studying gastric carcinogenesis. Finally future studies are required to gain further insight into the cellular and molecular evolution of gastric carcinogenesis which should lead to better management of patients with premalignant lesions and ultimately to the prevention of GC.