DNA methylation analysis of paediatric low-grade astrocytomas identifies a tumour-specific hypomethylation signature in pilocytic astrocytomas

Low-grade gliomas (LGGs) account for about a third of all brain tumours in children. We conducted a detailed study of DNA methylation and gene expression to improve our understanding of the biology of pilocytic and diffuse astrocytomas. Pilocytic astrocytomas were found to have a distinctive signature at 315 CpG sites, of which 312 were hypomethylated and 3 were hypermethylated. Genomic analysis revealed that 182 of these sites are within annotated enhancers. The signature was not present in diffuse astrocytomas, or in published profiles of other brain tumours and normal brain tissue. The AP-1 transcription factor was predicted to bind within 200 bp of a subset of the 315 differentially methylated CpG sites; the AP-1 factors, FOS and FOSL1 were found to be up-regulated in pilocytic astrocytomas. We also analysed splice variants of the AP-1 target gene, CCND1, which encodes cell cycle regulator cyclin D1. CCND1a was found to be highly expressed in both pilocytic and diffuse astrocytomas, but diffuse astrocytomas have far higher expression of the oncogenic variant, CCND1b. These findings highlight novel genetic and epigenetic differences between pilocytic and diffuse astrocytoma, in addition to well-described alterations involving BRAF, MYB and FGFR1. Electronic supplementary material The online version of this article (doi:10.1186/s40478-016-0323-6) contains supplementary material, which is available to authorized users.

that were computed between the tumours and controls, MethLAB analysis was not able to be performed on the full dataset. Therefore differential changes of >0.3 between the averaged sample groups were firstly calculated. The independent variable was sample type (pilocytic or foetal cerebellum or adult cerebellum). Analysis identified 11,671 CpG sites unique to the 3 groups (FDR-corrected p-values <0.05). 4) Supratentorial pilocytic astrocytomas (n=6) v diffuse astrocytomas (n=10) v each normal cerebral cortex group (foetal n=1, adult n=2, glial component n=4, neuronal component n=4) Normal cerebral cortex controls for the adult cortex were taken from GSE41826 [3]. The independent variable was sample type (pilocytic, diffuse, foetal cortex, adult cortex, glial and neuronal components). The co-variables included are bead chip and sex. Analysis identified 63,065 CpG sites that were unique to each of the 3 groups (FDR-corrected p-values <0.05).

Ingenuity Pathway Analysis (IPA) for genes associated with 993 differentially methylated sites identified in the three-way comparison of the test tumour set
IPA analysis shows significant biological and disease categories including functional annotation (p-value <0.05), for 537 genes sorted by p-value.

Additional file 10: Table S9
Comparison of pilocytic astrocytomas with diffuse astrocytomas in the test tumour set 315 CpG sites are shown where δβ ≥ 0.3 (Benjamini-Hochberg corrected p-value <0.05), assessed by MethLAB.
Illumina annotation is included. Rows highlighted in grey (42/315 CpG sites) were absent in the 993 differentially methylated CpG sites between tumours and normal brain control. Ordered by chromosomal location.

Additional file 11: Table S10
Ingenuity Pathway Analysis (IPA) for genes associated with 315 differentially methylated sites in the test pilocytic and diffuse astrocytomas 1_ The 185 genes that were directly associated with 217/315 CpG sites. Enhancers (grey) and CpG probes that are present within more than one gene (#) are highlighted. 2_ IPA analysis shows significant biological and disease categories which includes their functional annotations (p-value <0.05), sorted by p-value.

Additional file 12: Table S11
Normalised log2 expression values for the genes associated with the 315 CpG sites The δβ values between the pilocytic and diffuse astrocytoma groups were obtained from the grouped averaged beta values. Genes with a differential gene expression ≥2 fold are highlighted in grey. Sorted alphabetically by gene symbol.

Comparison of methylation and expression for genes associated with 315 differentially methylated sites in pilocytic and diffuse astrocytomas
Correlation coefficients are shown for differentially methylated CpG sites that show a difference of ≥2-fold in expression of the associated gene. Expression probes were only assessed if they were located within the mRNA transcript. A) CpG sites located within the promoter of the gene. B) CpG sites located within the gene body, 3'UTR and intergenic regions. As the CpG is not within a promoter, correlation was also calculated for a CpG site within the promoter (where possible within the TSS200; brown panel). For each correlation p-values and Benjamini-Hochberg corrected p-values are shown. The table is ordered by FDR-corrected p-values (sites with a p-value <0.05 are shown in blue). Positive correlation is highlighted in yellow. Up-regulated genes in diffuse astrocytomas are highlighted in green. When the expression probe was located in an exon, the probe for the full transcript was used for the correlation (indicated by the asterisk).

Additional file 14: Table S13
Functions of genes which show a significant correlation between DNA methylation and gene expression.
The majority of genes were hypomethylated in pilocytic astrocytomas except the genes highlighted in bold, which were hypermethylated in the pilocytic astrocytomas.

Additional file 15: Table S14
Predicted transcription factor binding sites for the 315 differentially methylated signature in low-grade astrocytomas Table to show TRANSFAC predicted transcription factor binding sites within a 200bp region (100bp either side of the CpG site detected by the Illumina probe). Illumina annotation highlighted in grey and transcription factor binding predictions plus sequence highlighted in blue. List of the top predicted transcription factors and gene expression levels of the top predicted transcription factors for the low-grade astrocytomas are also shown.

Additional file 16: Table S15
Differential expression of AP-1 target genes in pilocytic astrocytomas compared to normal brain controls and diffuse astrocytomas  Table showing AP-1 target genes that are differentially expressed in pilocytic astrocytomas compared to normal brain controls and diffuse astrocytomas. For each gene, methylation and expression status, plus the normalized log2 transformed expression values for each sample are included. The data is ordered by correlation with FOS expression (positive to negative correlation). The colour coordination shows which genes were taken from the 3 lists.

Additional file 17: Table S16
The differentially methylated CpG sites within de-regulated genes from the comparison between infratentorial pilocytic astrocytomas and normal cerebellum controls 1_ Features of 11, 671 probes that are significantly differentially methylated in infratentorial pilocytic astrocytomas compared to published foetal and adult normal cerebellum datasets (averaged beta values with a difference of δβ ≥ 0.3, Benjamini-Hochberg corrected p-value <0.05 [Lambert et al, Acta Neuropathologica 126:291-301, 2013]). 2_ Normalized log2 transformed expression analysis identifies differentially expressed genes between infratentorial pilocytic astrocytomas (n=4) compared to normal cerebellum (n=2, foetal and adult cerebellum). Genes that show fold change >2 are highlighted. 3_ List of genes that have differential methylation (δβ ≥ 0.3) and differential gene expression (fold change >2).

Additional file 18: Table S17
Ingenuity Pathway Analysis (IPA) of genes associated with differentially methylated sites in infratentorial pilocytic astrocytomas and normal cerebellum IPA analysis of the top 20 significant biological and disease categories including their functional annotation (pvalue <0.05) identified for 161 genes that had ≤2 fold change in gene expression and δβ ≥ 0.3, sorted by pvalue.

Additional file 19: Table S18
Differentially methylated CpG sites identified between supratentorial pilocytic astrocytomas and cerebral cortex controls 1_ MethLAB analysis identified 90,249 CpG sites that are significantly differentially methylated in supratentorial pilocytic astrocytomas, diffuse astrocytomas, glial component of the cerebral cortex, neuronal component of the cerebral cortex, adult cerebral cortex and foetal cerebral cortex (Benjamini-Hochberg corrected p-value <0.05). The adult cerebral cortex controls (including the glial and neuronal components) were taken from [Guintivano et al, Epigenetics 8:290-302, 2013]. The number of differentially methylated sites (averaged beta values with a difference of δβ ≥ 0.3, Benjamini-Hochberg corrected p-value <0.05) identified in tumour and all controls was 382 CpG sites for supratentorial pilocytic astrocytomas. 2_ CpG sites hypermethylated and 3_ hypomethylated in pilocytic astrocytomas compared to controls. 4_ Expression analysis for the genes that show differentially methylated CpG sites. Highlighted genes show ≤ 2 fold change in expression compared to controls.
Additional file 20: Table S19 Differentially methylated CpG sites identified between diffuse astrocytomas and cerebral cortex controls 1_ MethLAB analysis identified 90,249 CpG sites that are significantly differentially methylated in supratentorial pilocytic astrocytomas, diffuse astrocytomas, glial component of the cerebral cortex, neuronal component of the cerebral cortex, adult cerebral cortex and foetal cerebral cortex (Benjamini-Hochberg corrected p-value <0.05). The adult cerebral cortex controls (including the glial and neuronal components) were taken from [Guintivano et al, Epigenetics 8:290-302, 2013]. The number of differentially methylated sites (averaged beta values with a difference of δβ ≥ 0.3, Benjamini-Hochberg corrected p-value <0.05) identified in tumour and all controls was 58 CpG sites for the diffuse astrocytomas. 2_ CpG sites hypermethylated and 3_ hypomethylated in diffuse astrocytomas compared to controls. 4_ Expression analysis for the genes that show differentially methylated CpG sites. Highlighted genes show ≤ 2 fold change in expression compared to controls.

Supplementary Figure 1. Correlation between Illumina 450K BeadChip and pyrosequencing analysis of methylation of CpG sites within 5 genes for the test tumour set
Correlation coefficients were calculated for the matching 450K (beta values) and pyrosequencing (% methylation) values at selected CpG sites (gene names and Illumina probe identifiers shown). R-values range from 0-1 with strong correlation having values close to 1. Tumours and controls tested are: pilocytic astrocytomas (PA, n=11), diffuse astrocytomas (DA, n=9) and normal brain controls (AB, Adult brain; FFL, foetal frontal lobe; FB, foetal brain; FCB, foetal cerebellum, n=4).

astrocytomas with methylation in an adult low-grade astrocytoma cohort and a high-grade glioma cohort
Heatmaps showing the hypomethylation signature in (a) Adult low-grade astrocytoma cohort (TCGA) and (b)

astrocytomas with a foetal brain cohort
Heatmap showing the hypomethylation signature in foetal brain cohort (n=179) representing different stages of brain development from 23dpc-184dpc (GSE58885 [9]. Columns represent individual samples; rows represent 450K CpG probes. Methylation beta values are represented in shades of yellow = 0 (unmethylated) to blue =1 (methylated). Study test tumours are shown, pilocytic (n=17) and diffuse (n=10) astrocytomas left of red line.
As values were missing from the cohort, 287/315 CpG sites are shown (Supplementary Table 8).