Dissolving heterogeneity in cancer

Cancer - this one word stands for a vast variety of complex diseases. Cancers are named according to the organ they originate from, like breast cancer, colon cancer or brain cancer. However, when analyzing the molecular phenotype of the cells of theses different cancer types one can observe that each type itself consists of many different diseases. The genetic aberrations leading to malignant growth can vary from case to case and so can the molecular mechanisms that a tumour cell uses to facilitate de-differentiation and progression. To the eye of a molecular biologist, cancers are very heterogeneous. Modern treatment strategies target specific oncogenic mechanisms, like the specific inhibition of a signaling pathway. In view of the molecular heterogeneity of cancers, it cannot be expected that all patients will benefit from this treatment. Some cancers depend on the activity of a pathway and will hence respond to its inhibition, while others might not. To use targeted therapies effectively, we must first understand the heterogeneity of tumours. A modern genomics based approach to classify cancers is gene expression profiling using microarrays or high throughput sequencing. Expression profiles monitor the expression of virtually all genes in a tumour simultaneously. They thus characterize its molecular phenotype at a high resolution. Importantly, the oncogenic events, like aberrations in the genome and associated perturbations of signaling pathways, shape these profiles. Hence, classifying tumours by expression profiles means classifying them by the underlying mutational changes that distinguishes them from healthy tissue. The tumour profiling projects, covered in my thesis, are carried out in close collaboration with various experimental groups: Cell of origin expression signatures were developed in collaboration with the group of Christoph Beier from the Department of Neurology, University of Regensburg. The finding that specific genetic perturbations direct the development of different brain tumour types was discovered in collaboration with the group of Ulrike Nuber from the Department of Laboratory Medicine of Lund Strategic Research Center for Stem Cell Biology. The characterization of signaling pathways in human transformed GC B cells and their activity in lymphomas was conducted in collaboration with the group of Dieter Kube from the Department of Haematology and Oncology, University Medical Center Göttingen. My contribution was the statistical analysis of the expression profiles. I designed experiments, analyzed the quality of expression data, identified biomarkers and developed gene expression signatures allowing for clinical distinction of different cancer types. For building gene expression signatures I added a novel statistical method to the pool of existing tools. Gene expression signatures combine the expression levels of multiple genes to allow for highly resolved and robust molecular tumour classification. In this work I defined multiple gene expression signatures: One that allows the clinical distinction of brain tumour types and others that allow the measurement of pathway activities in lymphomas. For building gene expression signatures I developed a novel statistical method, which combines the expression of multiple co-expressed genes to a single real valued index per tumour sample. My thesis addresses the molecular heterogeneity of two tumour types. The first type are brain tumours. I describe how different cellular origins of glioblastomas are reflected in their expression profiles and I develop a cell of origin gene expression signature. Furthermore I describe how the temporal order of these genetic events contributes to the heterogeneity of brain tumours. The second tumour type are lymphomas. Lymphomas represent a heterogeneous group of cancers affecting cells of the immune system. 95% of lymphomas are B cell lymphomas and evolve during B cell differentiation. Differentiation includes crosstalk of different signaling pathways. Most malignant B cell lymphoma use key factors of these pathways to generate survival and proliferation signals. The activity of the dysregulated pathways differs between lymphoma types and thus shapes their expression profiles differently. I characterize different patterns of gene expression changes by B cell specific stimuli and associate these with different pathway activities in DLBCL and BL