Page 22 - Molecular features of low-grade developmental brain tumours

1
20
CHAPTER 1
to complementary probes. For each potential methylation site of interest there are two probes, one that hybridizes with the converted CpG site and one which hybridizes with the original sequence. By calculating the ratio between the intensity of these two probes it is possible to calculate the methylation level of at each site and compare across conditions. The 450K microarray contains 1.5% of the CpGs in the human genome, whereas the 850K microarray contains 90% of the CpGs on the 450K microarray plus an additional 350K CpGs, including CpGs in the promotor regions as well as gene body regions 216. The array output data can be normalized using various packages in R, including minfi and methylumi, which include options for normalizing for probe intensity as well as background corrections using the control probes 217,218. After normalization either the β-values (which are equivalent to the absolute DNA methylation levels) or the M-Value (a Logit transformation of the β-values) can be used for further analysis 219. However, despite all these normalization tools bias can occur due to batch effects, the presence of SNPs and unspecific binding of probes 217,218.
Recently, a machine learning approach for classification of CNS tumours based on the analysis of genome-wide DNA methylation patterns has been developed 220. Using this classifier, SEGAs were classified as low-grade glioma. Furthermore, no differences were found in TSC1 and TSC2 mutated SEGAs and no epigenetic silencing of TSC1 or TSC2 has been seen in TSC related tumours 114-220. Further, for LEATs methylation profiling using the 450K methylation array has been shown to be useful in classifying subtypes 221. The study by Stone et al. showed that BRAF and FGFR1 altered tumours have different DNA methylation profiles. In accordance with this, another study showed that DNT and GG have distinct methylation signatures and that tumours with diffuse growth patterns and immunopositivity for CD34, are more similar to GGs than DNT 222. Furthermore, methylation profiling has also been used to identify two subtypes of DLGNT, with one group appearing to be less aggressive in clinical outcome 223. Taken together, these studies highlight the importance of DNA methylation in integrating molecular diagnostics and classification.
Scope and outline of this thesis
In this thesis, we aimed to investigate the molecular mechanisms involved in the pathology of SEGAs in TSC and GNTs, in order to find potential targets for the development of novel treatments. We therefore examined the molecular pathways and miRNAs in SEGAs and GNTs, by investigating the (epi)genomic, transcriptomic and proteomic profile of SEGAs and the genetic abnormalities in GNTs.
In chapter 2, the prevalence of the BRAFV600E mutation in a large cohort of TSC related SEGAs was investigated. Additionally, massively parallel sequencing of TSC1/TSC2 was performed to confirm that SEGAs fit the classic model of two hit TSC1 or TSC2 inactivation.
In chapter 3, the methylation profile of SEGAs was investigated using the Illumina Infinium HumanMethylation450 BeadChip, where we attempted to identify subgroups and pathways that could play a role in SEGA pathogenesis. Additionally, we tried to link certain methylation changes to the expression of inflammation, mTOR activation, glial and neuronal markers in SEGAs.