Search

Our understanding of the molecular pathogenesis of acute and chronic leukaemia has been greatly advanced by high-throughput next generation DNA sequencing. Nowadays, lymphoid, and myeloid malignancies are characterised by many somatically acquired genomic structural rearrangements including chromosomal translocations, inversions, large deletions or insertions, and internal tandem duplications, as well as chromosomal copy number changes. The ultimate effect of these is the deregulation of oncogenes or tumour suppressor genes which promotes transformation to disease. Clinically, the presence of structural abnormalities in a leukaemia patient's tumour genome predict treatment response and patient survival, so it is necessary to conduct investigations at presentation to better understand a patient's tumour genome and inform patient risk stratification and care. The detection of the genomic structural rearrangements vital for clinical management are currently detected using conventional cytogenetic analysis (G-banding), and many structural rearrangements can be detected in this manner, but many others are too subtle and evade detection. Gene-centric techniques such as PCR or FISH can focus detection on specific rearrangements to confirm or refute their presence in a genome, but other less common potentially significant rearrangements will be missed simply because they are not known to be looked for. Overall, conventional cytogenetic techniques are laborious and time consuming, and are of inadequate through-put to determine all genomic rearrangements present in a sample with ease. With the development of 2 x 150 bp paired-end whole genome sequencing however, it is theoretically possible to identify all disease-relevant structural abnormalities with base-pair precision and resolution in a single experiment, reducing the cost and time spent per mutation to generate a complete genomic profile from which clinical decisions can be made. Effective whole genome sequencing analysis is not trivial however, and many approaches outside of clinical practice rely on sequencing both a tumour DNA sample and patient-matched germline DNA sample to discriminate the somatic rearrangements from polymorphic structural variants restricted to the tumour. Similarly, paired tumour and germline DNA sequencing is relied upon to limit the number of technical artefacts that otherwise masquerade as structural rearrangements when analysed by the current publicly available structural variant calling algorithms. Without paired-sample DNA sequencing, current computational approaches detect thousands of structural abnormalities in a single patient by comparison to the reference genome. Most of which are simply not real somatic changes and have the potential to mislead and waste resources in their confirmation. However, the collection of a matched germline DNA sample is simply impractical in haematology-oncology as there is no suitable tissue in which to easily sample, limiting the uptake of whole genome sequencing as a suitable replacement for cytogenetics. The associated costs of sequencing a second DNA sample further limits the use of whole genome sequencing in clinical practice also. Therefore, novel methods and analytical approaches for the detection of structural rearrangements in leukaemia without a germline DNA sample in which to compare to is an unmet clinical need. To that end, we have developed a computational method to detect and identify somatic structural rearrangements of clinical significance in leukaemia, using a tumour-only paired-end whole genome sequencing approach. This approach was developed using a 64-core computer server to enable massively parallel data processing, but the algorithm is scalable and can be implemented on systems with fewer resources. This approach uses discordant and split reads to identify and model structural rearrangements in paired-end whole genome sequencing data but differs from other less specific algorithms because it performs a series of internal validation steps to identify only those discordant split reads that give reliable evidence of structural rearrangement. To identify structural polymorphisms which persist after structural variant detection, each putative breakpoint sequence is compared with publicly available databases of polymorphic structural variation. This approach has been tested on 17 tumour-only leukaemia samples obtained at presentation and demonstrates a higher rate of specificity than the current structural variant algorithms available. It also exceeds the level of sensitivity of cytogenetics. This work demonstrates the superiority of whole genome sequencing in a clinical setting, and the ease of which tumour-only whole genome sequencing analysis can be conducted routinely in clinical haematology-oncology.

I have investigated epigenetic mechanisms of acquired endocrine-resistance in breast cancer using an in vitro model system based on estrogen-dependent MCF7 cells and their derivatives, LCC1 and LCC9. LCC1 cells, derived from MCF7 after passage in ovariectomised mice and routinely cultured in vitro in the absence of estrogen, exhibit estrogen-independent growth. They retain sensitivity to tamoxifen and fulvestrant. LCC9 cells, derived from LCC1 cells by growing them in increasing concentrations of fulvestrant, are completely estrogen-independent and are resistant to fulvestrant and cross-resistant to tamoxifen. When compared to MCF7 cells, LCC1 cells have marked up-regulation of the estrogen receptor α (ERα) protein that is not concomitant with increased estrogen receptor 1 (ESR1) transcription, suggesting a role for estrogen in controlling the proteasomal degradation of ERα. However, despite being grown in the same estrogen-deprived conditions, LCC9 cells do not have up-regulated ERα levels. As LCC1 cells retain sensitivity to tamoxifen and fulvestrant, these data suggest that LCC1 have developed estrogen-independence through ERα uncoupled from its ligand. However, LCC9 cells appear to have developed an alternative mechanism which is not dependent on ERα, presumably explaining their resistance to fulvestrant. I have studied global gene expression changes in the presence and absence of estrogen in these cell lines, using oligonucleotide microarrays, and correlated these data with global DNA methylation data derived from methylation arrays, which interrogate the methylation status of approximately 27,000 CpG dinucleotides in the genome. The analysis led to the discovery of more than 5,000 genes that were potentially either up-regulated or down-regulated by estrogen in MCF7 cells, either directly or indirectly. The transcriptional response to estrogen was generally muted in LCC1 and LCC9 compared with MCF7, but was not completely absent. I used various methods based on differential gene expression to parse the data, including gene ontology analysis, aiming to select genes for further mechanistic study. However, none of these methods led to the conclusive identification of a specific gene (or set of genes) that might have accounted for the physiological differences between the cell lines. In one strategy, I reasoned that, as the endocrine-resistant cells had lost their estrogen-dependence, genes involved might be regulated in an estrogen-dependent manner in MCF7 cells, without exhibiting misregulation in LCC9. This led to the identification of DUSP1 as a candidate gene, which was taken forward for mechanistic study because of its potential role in regulating ERα expression. However, when over-expressing DUSP1 in LCC9 cells, I could not demonstrate any effect on ERα levels. The final approach taken was to identify genes that might have been epigenetically deregulated, being both estrogen-regulated and deregulated in association with aberrant DNA methylation in the estrogen-independent cell lines. Surprisingly, given the phenotypic differences between the cell lines, only a very few genes were significantly methylated between cell lines. Of those that were differentially methylated between MCF7 cells and LCC1/9, only three exhibited the expected inverse correlation between methylation and expression. Of these, the gene CYBA was selected for further investigation. CYBA is a critical component of the NAPDH oxidase complex which is involved in generating oxygen free-radicals. My work suggests CYBA expression is estrogen-dependent, and that chronic estrogen deprivation leads to the epigenetic inactivation of CYBA in breast cancer cells. I speculate that the epigenetic suppression of CYBA may protect cells from the oxidant damage that results from estrogen deprivation and may be part of the mechanism that leads to acquired endocrine-resistance in previously sensitive cells.

Breast cancer is a common disease in women and has major impacts on health and quality of life. About 70% of breast cancers over express ERα, and are classified as ER positive breast cancer. Oestrogen receptor alpha (ERα) belongs to the nuclear receptor superfamily and is responsible for many effects of oestrogen on normal and cancerous breast tissue. Endocrine therapies that block the function of ERα or the synthesis of oestrogen have been a mainstay of ERα positive breast cancer treatment. However, their efficacy is limited by intrinsic and acquired drug resistance overtime, and endocrine resistance remains one of the biggest challenges in breast cancer treatment. In order to investigate the underlying mechanisms of acquired drug resistance, and to develop new strategies for breast cancer therapy, I generated a novel long-term oestrogen deprived cell line (DH) in serum-free condition. As DH cells are cultured in a defined media with known concentrations of growth factors, it provides an ideal system to identify and dissect changes in signalling pathways in response to hormones and inhibitors in vitro. At the same time, DH cells are representative of ER positive breast cancers treated with drugs that reduce the level of oestrogen. It enables the identification of survival pathways that could be activated during oestrogen deprivation. By using this cell model, I find that oestrogen stimulation enables cells to up-regulate the EGFR level and simultaneously reduces ERα expression at both mRNA and protein levels. Once up-regulated, EGFR expression is maintained despite oestrogen withdrawal indicating a stable transcriptional re-programming at the EGFR promoter. By using the whole genome expression microarrays, I identified a list of genes that also show stable changes in gene expression in response to oestrogen, suggesting that the oestrogen promotes transcriptional re-programming at multiple pathways in cells. In terms of signalling pathways, oestrogen activates the growth promoting MAPK pathway in an EGFR dependent manner and a 5-day oestrogen pulse substantially increases the resistance of cells to tamoxifen, while cells remain sensitive to the EGFR inhibitor, demonstrating a functional switch between ERα and EGFR survival pathways. Furthermore, microarray analysis of ERα and EGFR downstream target genes shows that there is a general activation of MAPK gene signature after 5 days of oestrogen stimulation in DH cells. In this thesis, I also investigate the molecular mechanism of oestrogen induced EGFR up-regulation in ER positive breast cancer cells. c-Myb is an oestrogen responsive transcription factor whose expression is regulated by ERα in breast cancer cells. I demonstrate that oestrogen treatment leads to ERα dependent c-Myb up-regulation in DH cells. I also find that c-Myb transiently locates upstream of the EGFR promoter to enhance its expression. As the up-regulation of EGFR in ER positive breast cancer could lead to survival pathway switching and endocrine therapy resistance, c-Myb could be a good drug target to prevent the likelihood these switches and subsequent relapse on endocrine therapies. The expression of EGFR remains high after the removal of oestrogen suggesting there may be epigenetic changes, which maintain the transcriptional re-programming stimulated by c-Myb. Bisulphite sequencing however demonstrates EGFR promoter DNA methylation pattern is not affected by oestrogen. Meanwhile, ChIP microarrays with four different histone modifications show no significant changes around the promoter area of EGFR in response to oestrogen. These observations suggest that alternative epigenetic modifications or epigenetic alternations at other genes may subsequently lead to the stable expression of EGFR in response to oestrogen.

Aberrant methylation of CpG island promoters is a frequent observation in cancer and is known to affect many genes, including tumour suppressor genes. Genes with methylated promoters are usually repressed and inactive, and there is good evidence that most genes that become methylated in cancer are already repressed in the normal tissues from which tumours arise. However, the methylation of some genes appears to arise at previously active loci, suggesting either a stochastic epigenetic event or that these genes are somehow predisposed to becoming methylated. The DNA mismatch repair gene MLH1 is expressed in normal colonic epithelial cells but methylated and down-regulated in some sporadic mismatch repair-deficient colon tumours. These tumours are almost invariably associated with the simultaneous methylation of multiple cancer-specific loci, termed the CpG island methylator phenotype (CIMP) and an activating mutation of BRAF (V600E), raising the possibility that a hypermethylator phenotype may arise in cancer in direct association with a specific genetic alteration. The possibility that MLH1-deficiency caused BRAF mutation was discounted as genetic deficiency of MLH1 is not associated with BRAFV600E. I explored the possibility that BRAFV600E might induce MLH1 methylation but found no evidence in support of this. I then focused on factors that might mediate CIMP gene methylation, of which MLH1 methylation is known to be a part. Bioinformatic analysis of the genes methylated in BRAFV600E colon tumours indicated a significant enrichment in binding sites for the transcription factor MAZ (MYC-associated zinc finger protein). I hypothesised that loss of MAZ might lead to MLH1 down-regulation and its subsequent methylation. In this thesis I provide evidence that both MAZ and MLH1 expression are deregulated during normal colonic epithelial differentiation. The down-regulation of MAZ by RNA interference led to a reduction in MLH1 expression and methylation of its promoter. I speculate that MLH1 methylation may be associated with BRAF mutation because transformation by BRAFV600E allows progenitor cells to undergo a degree of differentiation whilst maintaining their malignant proliferation. I speculate that it is during this process of differentiation that MLH1 becomes susceptible to methylation.