The commonest childhood Non-Hodgkin Lymphomas (NHLs) form three major categories: the aggressive mature B-cell NHLs including Burkitt lymphoma (BL), Anaplastic Large Cell Lymphomas (ALCL), and lymphoblastic lymphomas. Most childhood NHL is high-grade and generally responds well to high-intensity chemotherapy. Unfortunately, many patients who relapse have a dismal prognosis, with long-term survival less than 30% for relapsed BL patients in particular. Furthermore, the side effects of chemotherapy on children are often severe, rendering intensive chemotherapy infeasible in less economically-developed regions. Significant progress has been made with the addition of Rituximab to BL therapy further improving overall survival, and the use of ALK inhibitors such as crizotinib have been used to successfully bridge relapsed or refractory ALCL patients to stem cell transplantation---however, resistance to both ALK inhibitors and Rituximab occurs in the clinic. Therefore, there is urgent need for less toxic therapies in primary NHL to improve the quality of life for curable patients, and to reduce the treatment-related mortality for patients without access to high-quality supportive care. Secondly, more effective therapies or strategies to treat relapsed/refractory disease are required to improve survival rates even in well-resourced settings.

It is now recognised that genetic intratumour heterogeneity (ITH) contributes to drug resistance in malignant disease. However, many questions remain unresolved, and improved theoretical and experimental model systems will likely be required to resolve them. To this latter end, a novel relapsed BL patient-derived xenograft (PDX) was established---to our knowledge the first described in the literature---and the maintenance of genetic ITH through engraftment in immune-compromised mice was demonstrated through whole genome sequencing (WES). In addition, pervasive genetic ITH across time and space was detected by multiregion WES of BL cases, meaning that single biopsies cannot rule out the presence of a given variant, regardless of sequencing depth. Furthermore, analysis of allele frequencies in publicly-available data indicated that some BL patient samples harbour genetic evidence of recent subclonal selection, while others appear to have evolved neutrally or near-neutrally. These data highlight the pervasive nature of ITH and thus motivate the use of patient-derived xenograft models which better capture patient tumour heterogeneity than the long-established cell lines often used in preclinical studies.

Similarly, a resource of relapsed ALCL PDX was generated and assessed for drug sensitivity. Two novel ALCL PDXs were established from patients at or before CNS replase, and a third was acquired from collaborators. All three patients had failed chemotherapy and crizotinib treatment, and two patients were resistant to crizotinib. WES of PDX revealed likely mechanisms of crizotinib resistance, with one patient harbouring concomitant ALK mutation and amplification, and the other showing amplification of several oncogenes including IRF4, which has previously been implicated in ALCL pathogenesis. Next, I tested the second generation ALK inhibitor brigatinib, demonstrating the utility of our models as a platform to test targeted therapies. Brigatinib was highly effective in the treatment of all three PDX, and superior to crizotinib in the crizotinib-sensitive PDX. Several brigatinib-treated mice demonstrated complete tumour remission after one to three weeks of treatment, pointing to brigatinib as a therapy option for high-risk patients.

While increasing understanding of the molecular underpinnings of cancer is yielding promising new drug targets, translation of novel therapeutics into clinical practice is costly and time-consuming, with the 'bench to bedside' process rarely being completed within a decade. A major component of this process is testing for safety in animal and human subjects. Drug repurposing refers to the application of previously approved or well-understood drugs to new clinical settings, and has the potential to yield useful treatments in less time by selecting from drugs for which safety profiles have already been determined. Therefore, an /in vitro/ high-throughput screen using approximately 1400 regulatory body-approved drugs applied to BL and ALCL cell lines was undertaken, with several promising hits identified. However, the extent to which established cell lines faithfully represent patient disease is questionable, as /in vitro/ culture conditions can cause rapid divergence from the originating tumour, limiting translational relevance. Therefore, a proof-of-concept drug screen platform for BL PDX cells supported by a feeder cell line was established, offering a relatively low-cost, high-throughput platform to screen drugs for a given BL patient tumour.

In conclusion, this thesis demonstrates extensive genetic ITH in BL, with implications for tumour sampling and resistance to treatment. Novel ALCL PDXs were generated, representing a patient group of unmet clinical need, and providing a resource to test drugs /in vivo/ with tumours of known genetic background. Finally, high-throughput drug screening was undertaken in order to identify known therapeutics with efficacy against NHL cell lines, and a co-culture system was established in order to extend screening to patient-derived cells.