Search

Introduction Medulloblastoma is the most common type of malignant paediatric brain tumour, accounting for 10% of cancer deaths in children. While survival rates have improved in recent years, metastatic spread poses a significant challenge to therapy success. Current therapies also cause irreparable damage to the surrounding brain leading to a decreased quality of life in surviving patients, consequently there is a need for new targeted therapy options. One of the routes for dissemination of metastatic cells is invasion into the leptomeninges, a process during which cells cross basement membrane barriers as they travel from the primary tumour niche into the metastatic niche in the cerebrospinal fluid (CSF). The pathways involved in dissemination and metastasis remain elusive, and a better understanding is crucial to developing targeted therapies. This project aimed to develop an in vitro model system to mimic key characteristics of cell migration from the brain to the CSF, and identify novel candidate genes driving medulloblastoma metastasis. Methods An in vitro culture model representing certain features of the basement membrane interface between the brain parenchyma and the CSF was developed in order to isolate cells for transcriptomic analysis from three compartments: (a) non-metastatic suspension cells, (b) early metastatic adherent cells, and (c) metastatic migrated cells. Differentially expressed genes in these three compartments were identified by Next Generation 3ˡ UPX RNA Seq. The prognostic role of candidates was assessed by analysis of a large medulloblastoma gene expression dataset (Cavalli et al., 2017) on the R2: Genomics Analysis and Visualization Platform (http://r2.amc.nl). SORCS2 expression was validated in medulloblastoma cell lines by qRT-PCR and western blotting, and a possible function assessed in vitro after siRNA knockdown. Results A trans-well migration assay format was modified to represent certain features of the basement membrane, collagen IV and laminin 111 were used as a basement membrane-like coating, and cells were induced to migrate towards artificial CSF supplemented with FBS. The transcripts SORCS2, CORO2B and LMNB1-DT were identified as candidates that may drive the process of migration. The gene SORCS2 demonstrated a significant increase in expression as Group 3 medulloblastoma cells progressed through the stages of migration, from the suspension compartment to the adherent compartment and then to the migration compartment. The transcripts CORO2B and LMNB1-DT were upregulated in cells in the adherent compartment. The expression of SORCS2, CORO2B and LMNB1-DT correlated with poor patient outcome and metastatic recurrence in medulloblastoma patients. SORCS2 was selected for further in vitro investigation as it most significantly correlated with an early poor survival trend in patients with metastatic tumours. SORCS2 gene and protein expression was confirmed in Group 3 and Group 4 cell lines. The results from in vitro experiments did not allow significant conclusions relating to SORCS2 function in migration to be drawn, however they offer preliminary data encouraging further investigation. Conclusions This non-biased, model-led, target identification approach identified genes of interest SORCS2 and CORO2B, and lncRNA LMNB1-DT, in migrating medulloblastoma cells. SORCS2 encodes a brain-specific transmembrane receptor in the VPS10 family of sorting proteins previously linked to actin dynamics in neuronal growth cones, and was implicated in the migratory behaviour of basal-like breast carcinoma. The analyses in this project suggested a link between SORCS2, metastasis, and tumour recurrence in medulloblastoma. The other candidates identified by these analyses warrant further validation and may shed light on the metastatic process in medulloblastoma. A better understanding of the processes and pathways that drive malignancy is a starting point for a push towards improving efficacy of our treatment options while reducing toxicity.

DNA methylation (5-methylcytosine, 5mC) is the major epigenetic modification involved in transcriptional regulation. Ten-eleven translocation (TET) proteins can enzymatically oxidise 5mC producing 5-hydroxymethylcytosine (5hmC), 5-formylcytosine (5fC) and 5-carboxylcytosine (5caC). According to several reports, 5hmC levels are reduced in human tumours; however, the distribution and the exact biological role of TET-dependent oxidation to 5fC and 5caC in cancers is poorly studied. Here, using a range of techniques including, immunocytochemistry, 2 dimensional ultra-performance liquid chromatography coupled with tandem mass-spectrometry (2D-UPLC-MS/MS) and CRISPR/Cas9 gene editing we studied the presence and the distribution of 5caC/5hmC in brain tumours cell lines. Our ultimate goal was to examine the role of TET2 in the oxidation of 5mC to 5hmC/5caC in glioblastoma multiforme (GBM) pathogenesis and the role of 5caC potential readers in the interpretation of active DNA demethylation in medulloblastoma (MB) (paediatric brain tumour) cell lines. We found that, while GBM cell lines exhibit low levels of 5hmC, they are, rather unexpectedly, characterised by detectable 5caC levels. Remarkably, 5caC content in GBM corresponds to elevated levels of TET2 transcript. We next used CRISPR/Cas9 to knockout (KO) TET2 in the LN18 GBM cell line. 5hmC levels were significantly reduced in TET2 KO cell line. Moreover, we showed significantly reduced tumorigenic ability of TET2 KO cells in parallel with the reduction of the transcript levels of key glial cancer stem cell markers. We also showed that SMARCC2 and RCOR2 transcription factors with unique roles in neurogenesis, are highly expressed in paediatric brain tumours, while their transient knockdown resulted at impaired 5caC levels in MB cell lines; indicating their role as 5caC readers. Our compiled data show the unique epigenetic signatures of adult and paediatric brain tumours, the role of TET2 in active DNA demethylation and its potential contribution to tumorigenesis.

5-methylcytosine (5mC) is an epigenetic modification usually associated with transcriptional repression. The Ten–Eleven Translocation proteins (Tet1/2/3) can oxidise 5mC to 5-hydroxymethylcytosine (5hmC), 5-formylcytosine (5fC) and 5-carboxylcytosine (5caC) in vertebrate DNA. These modifications have not been well characterised but have been detected during embryogenesis, in somatic adult healthy and malignant tissues. Transcription factor and tumour suppressor/ oncogene Wilm’s Tumour protein 1 (WT1), though not expressed in healthy brains displays over-expression in high grade neoplasms including Glioblastoma multiforme (GBM) and has been associated with poor patient prognosis. Interestingly, WT1 possesses preferential binding affinity for 5mC and 5caC over unmodified cytosines, 5hmC and 5fC in vitro. Thus we speculate that a WT1-5caC relationship may exist within a DNA demethylation associated and possible a transcriptional regulatory context in brain tumours. Utilising a range of techniques including genome editing and Next Generation Sequencing, we demonstrate that 5caC is enriched in brain tumour cell lines, is depleted upon WT1 mutagenesis and may share genomic occupancy on gene regulatory and intragenic sequences. Additionally, we demonstrate a compromised neurosphere formation phenotype evident in WT1 zinc finger mutant GBM cell lines which correlates with perturbed extracellular matrix and cell adhesion gene expression, and is correlated with reduced tumour formation in vivo.

Introduction: Medulloblastoma, the most common malignant paediatric brain tumour, is a heterogeneous disease with tumours classified based on their histological and molecular profiles. Studying medulloblastoma in vitro is challenging, as traditional two-dimensional (2D) cell culture models fail to recapitulate the multi-dimensional growth, cell-cell interactions, and complex microenvironment that exists in vivo and in patient tumours. This project therefore aimed to respond to the need for a robust, reliable, and optimised three-dimensional (3D) in vitro culture method for medulloblastoma, which could be applied to study drug response and metastatic dissemination. Methods: A panel of medulloblastoma cell lines representing three of the four molecular subgroups (SHH, Group 3, and Group 4) were tested for use in the 3D spheroid model. Standardised growth conditions, including spheroid size and culture medium, were adopted to allow direct comparisons across cell lines and treatments. Spheroids were characterised by imaging analysis, luminescence-based assays, and immunohistochemical staining. Drug response of 3D spheroids was compared to 2D monolayer culture, with additional long-term and dual-inhibitor studies conducted to exemplify the use of the 3D spheroid model in drug screening approaches. SHH medulloblastoma spheroids were applied to migration and invasion models using basement membrane matrices and hyaluronan (HA) hydrogels containing additional extracellular matrix (ECM) components. Next-generation sequencing (NGS) technologies were adopted to gain an insight into the genes and pathways involved in SHH medulloblastoma leptomeningeal metastatic dissemination using the HA hydrogel model. Results: Medulloblastoma cell lines formed spheroids of different morphologies. The SHH subgroup cell lines (DAOY, ONS76, and UW228-3) formed tight, highly reproducible spheroids, and were therefore chosen as suitable models for downstream analysis. SHH medulloblastoma spheroids could be maintained for up to 3 weeks in culture, forming physiological gradients within the 3D structure over time. 3D spheroid culture increased resistance to standard-of-care chemotherapeutic drugs compared to 2D monolayer culture and a subpopulation of viable cells was present after treatment. Dual-inhibitor studies with vardenafil, an inhibitor of the multidrug transporter ABCB1, showed synergistic effects which could facilitate the use of lower doses of chemotherapy. The HA hydrogels, containing ECM components highly expressed in SHH patient tumours, were more suitable, physiologically-relevant models of SHH medulloblastoma metastasis than commercially-available basement membrane matrices. NGS revealed the downregulation of cell cycle genes and upregulation of cell movement genes in DAOY and ONS76 spheroids migrating on the HA hydrogel. Overexpression of the insulin pathway, consistent with in vivo SHH medulloblastoma metastasis models and patient tumours, was also observed. Thus our model allows us to recapitulate known, and identify novel, mechanisms of cell dissemination from the primary tumour. Conclusion: The 3D spheroid model of medulloblastoma described in this study can be used to analyse drug response and model metastatic dissemination. It is a significant improvement over current in vitro techniques and represents an important breakthrough in medulloblastoma research.

Glioblastoma (GBM) is the most common type of adult brain tumour, representing 45.2% of all malignant brain and central nervous system tumours. It is also the most lethal, with a median post-therapy survival of 14 months, due to the high likelihood of tumour recurrence. Intratumour heterogeneity is thought to be a key contributor to GBM tumour recurrence, because treatments such as radiotherapy and chemotherapy typically only successfully target the core of the tumour. Due to its high visibility, this core is excised during surgery, but there is another more subtle region of the tumour, the invasive margin, which generally remains behind. This invasive margin is a mixture between normal brain and highly invasive tumour cells. If left over from surgery, these remaining tumour cells infiltrate the normal brain, and cause recurrence within 2cm of the original site. However, if surgeons are too vigorous in removing the invasive margin, they may cause unnecessary removal of healthy brain tissue, thereby causing other complications. This make developing specific methods to target the cancer cells in the invasive margin a priority area in GBM research. In order to specifically target the invasive margin, we need to have a way of differentiating it reliably from the other regions. In this thesis, I investigated the role that micro RNAs (miRNAs), a type of small non-coding RNA, play in this intratumour heterogeneity. miRNAs act as post-transcriptional regulators of gene expression and can simultaneously suppress multiple genes within the same pathway, which makes them potentially valuable routes for therapy development. I performed a miRNA microarray on tissue samples from three tumour regions (tumour core, rim, and invasive margin) and validated the results with qRT-PCR. I then performed computational predictions of gene targets of the differentially expressed miRNAs and validated them in cell lines and patient samples at both the mRNA (qRT-PCR) and protein (Western blot and immunohistochemistry) levels. I then performed a bi-phasic extraction of polar and non-polar metabolites and employed liquid chromatography-mass spectrometry (LC-MS) and tandem mass spectroscopy techniques to profile the intracellular metabolome and lipidome of U87 cells overexpressing miR-619-5p and miR-4793-3p. Transwell invasion assays were performed to investigate the effect of each miRNA and its target on GBM cell invasion. In this work, I found that two miRNAs, miR-619-5p and miR-4793-3p, are significantly downregulated in the invasive margin. Moreover, these miRNAs control the expression of genes that are involved in regulating lipid metabolism pathways. A second key finding was that the knockdown or pharmacological inhibition of the miR-619-5p target, CPT2, also significantly reduced the invasion of the cells in the invasive margin. Because CPT2 lies in the lipid metabolism pathway, my findings indicate that lipid metabolism may present a possible vulnerability of the GBM invasive cell population, and the understanding the function of miRNAs in regulating lipid homeostasis in GBM may provide novel avenues for GBM therapy.

Introduction: Medulloblastoma is an aggressive malignant neuro-ectodermal tumour of the cerebellum which accounts for 15-20% of childhood central nervous system tumours and frequently disseminates to the leptomeningeal spaces of the brain and spinal cord. Patients with disseminated disease respond poorly to post-surgical multimodal treatment which is thought to be explained by the intrinsic drug-resistant nature of these tumours. Here, we hypothesized that cells gain migratory and invasive capabilities by undergoing an Epithelial-Mesenchymal Transition (EMT)-like process whereby cells alter phenotypically and acquire stem cell-like properties during tumour dissemination. In this study, metastatic genes were identified and tested using 3D in vitro model systems which incorporated important components of the extracellular matrix. Using these markers, evidence for an EMT-like process and the role of the multi-drug transporter, ABCB1, was investigated in metastatic medulloblastomas. Small molecule inhibitors were also used to investigate whether metastatic processes could be targeted and drug resistance mechanisms could be circumvented in both in vitro and/or in vivo settings. Materials and Methods: Growth, morphology and biological processes (e.g. cell migration) were assessed in a panel of non-metastatic and metastatic medulloblastoma cell lines, as well as in non-tumourigenic neural stem cells cultured in a 3D basement membrane extract (BME) using the alamar blue assay (measure of metabolic activity) and time-lapse imaging. Putative metastatic markers were identified through literature review analysis of gene expression datasets or immunohistochemistry of tissue micro-arrays (TMA). These markers were then assessed in samples obtained from medulloblastoma cell lines cultured as 2D monolayers and grown in BME (for 3 and 6 days) by QRT-PCR analysis. Protein expression of selected markers were also assessed in mouse orthotopic xenograft samples by immunohistochemistry or in cell lines using immunofluorescence analysis. Finally, small molecule inhibitors (WIP1 and ABCB1 inhibitors) were used in 3D culture systems (3D spheroid and 3D BME assays) and in an orthotopic metastatic mouse model. Results: Medulloblastoma cell lines demonstrated different growth patterns in 3D. Metastatic cell lines formed metabolically active aggregates which sustained continual cell migration for at least 6 days; whilst non-metastatic and non-tumourigenic cells showed low metabolic activity and rapidly differentiated. Metastatic cell lines which were sustained longest in BME (D283 Med and MED1) demonstrated upregulation of the EMT transcription factor, TWIST1, along with several other EMT and TWIST1-related factors. Further analysis included overexpressing TWIST1 in a non-metastatic cell line (MED6 TWIST1) which induced a dispersed phenotype in 2D and cell aggregation in the 3D BME model which phenotypically resembled metastatic cell lines. TWIST1 and ABCB1 expression correlated with metastasis in patients and was upregulated in the invasive edge of primary tumours and in spinal metastases in an orthotopic metastatic mouse model. Small molecule inhibitors targeting WIP1 (a published metastatic marker) and ABCB1 inhibited cell migration of metastatic cell lines grown in 3D including the MED6 TWIST1 cell line. ABCB1 inhibition also increased sensitivity to etoposide treatment in 3D spheroid models and an orthotopic metastatic in vivo model. Conclusion: The 3D BME model utilised in this study can be used to distinguish metastatic capacity and transcriptional changes of medulloblastoma cell lines. Furthermore, data from this study supports a role for a TWIST1 driven EMT-like process in metastatic medulloblastoma and supports the use of ABCB1 inhibition to overcome chemoresistance.