Page 73 - Molecular features of low-grade developmental brain tumours

DISTINCT DNA METHYLATION PATTERNS IN SEGA IN TSC
Expression of inflammation, mTOR activation, glial and neuronal markers in SEGAs
SEGAs are considered mixed glio-neuronal tumours, with mTOR activity and presence of inflammation markers. Therefore, we wanted to evaluate the commonalities and differences in the expression of CD3, HLA-DP/DQ/DR, GFAP, MAP2 and pS6 in 42 SEGAs and 8 location- matched controls. In periventricular control tissue CD3, MAP2 and pS6 were not detected, whereas a moderate expression of HLA-DP/DQ/DR and high expression of GFAP was seen (Figure 3a). In SEGA we found several positive CD3 cells and observed an overall increase in positive area for CD3 in SEGA compared to control tissue (Figure 3b; p< 0.0001). HLA-DP/DQ/ DR, GFAP, MAP2 and pS6 were expressed in a heterogeneous manner in SEGAs (Figure 3a). The percentage of positive area of HLA-DP/DQ/DR (p< 0.0001), MAP2 (p= 0.0114) and PS6 (p< 0.0001) were increased in SEGA compared to control, whereas the positive area for GFAP was decreased in SEGA (p= 0.0016). Spearmans rank correlation revealed a weak positive correlation between the expression of CD3 and HLA-DP/DQ/DR (r=0.347; p=0.026), pS6 and HLA-DP/DQ/DR (r=0.368; p=0.016), and GFAP and HLA-DP/DQ/DR (r=0.325; p=0.036) in SEGA. Spearmans rank correlation with clinical data revealed a weak positive correlation between age at surgery and CD3 (r=0.3197; p=0.0416) and a negative correlation between tumour size and CD3 (r=-0,4331; p=0.0118), HLA-DP/DQ/DR (r=-0,4370; p=0.0098), MAP2 (r=-0,4746; p=0.0046) and pS6 (r=-0,4884; p=0.0034).
Two distinct methylation groups in SEGAs
To evaluate potential subgroups within the SEGA samples the top 5% most variable CpGs were analysed with hierarchical clustering, consensus clustering and silhouette. Hierarchical clustering indicated 2 major groups with one group subdividing into two smaller groups, which could not be explained by age at SEGA surgery, gender or TSC mutation (Figure 4a). This was confirmed by both consensus clustering (Figure 4c-d) and silhouette (Figure 4e-g), which indicated that k=3 was most robust.
We first wanted to further investigate the two largest groups identified and performed differential testing between group 1 compared to control (SEGA1-control), group 2 compared to control (SEGA2-control) and group 1 compared to group 2 (SEGA1-SEGA2). We found 4377 hypomethylated and 1411 hypermethylated CpGs in SEGA1-control (Figure 5a), 4883 hypomethylated and 3132 hypermethylated CpGs in SEGA2-control (Figure 5b) and 321 hypomethylated and 70 hypermethylated CpGs in SEGA1-SEGA2 (Figure 5c; adjusted p-value 0.01, β-value difference of >0.2, promoter region). In order to identify differentially methylated genes that were unique for each group, genes corresponding to the differentially methylated CpGs were extracted. Genes that were overlapping between SEGA1-control and SEGA1-SEGA2 and did not overlap with SEGA2-control were considered unique for SEGA1, whereas genes that overlapped between SEGA2-control and SEGA1-SEGA2 but did not overlap with SEGA1-control were considered unique for SEGA2 (70 SEGA1 unique genes and 58 SEGA2 unique genes; Figure 5d). GO analysis revealed 15 GO terms enriched for these 128 unique genes, which were related mainly to the MAPK cascade and adaptive immune response (P-adjusted<0.05; Figure 5e). We further evaluated the RNA expression of these
71
3