Methylator phenotype of malignant germ cell tumours in children identifies strong candidates for chemotherapy resistance

Abstract

Background: Yolk sac tumours (YSTs) and germinomas are the two major pure histological subtypes of germ cell tumours. To date, the role of DNA methylation in the aetiology of this class of tumour has only been analysed in adult testicular forms and with respect to only a few genes.

Methods: A bank of paediatric tumours was analysed for global methylation of LINE-1 repeat elements and global methylation of regulatory elements using GoldenGate methylation arrays.

Results: Both germinomas and YSTs exhibited significant global hypomethylation of LINE-1 elements. However, in germinomas, methylation of gene regulatory regions differed little from control samples, whereas YSTs exhibited increased methylation at a large proportion of the loci tested, showing a 'methylator' phenotype, including silencing of genes associated with Caspase-8-dependent apoptosis. Furthermore, we found that the methylator phenotype of YSTs was coincident with higher levels of expression of the DNA methyltransferase, DNA (cytosine-5)-methyltransferase 3B, suggesting a mechanism underlying the phenotype.

Conclusion: Epigenetic silencing of a large number of potential tumour suppressor genes in YSTs might explain why they exhibit a more aggressive natural history than germinomas and silencing of genes associated with Caspase-8-dependent cell death might explain the relative resistance of YSTs to conventional therapy.

Figure 1

Combination of bisulphite and restriction analysis (COBRA) of LINE-1 element methylation. (A) Diagram showing predicted digestion products. (B) Illustration of undigested (U) and digested (D) PCR products from representative tumour and control samples. (C) Example of GeneScan data from a representative control and tumour samples. (D) Percentage methylation levels of LINE-1 elements from individual samples as determined by GeneScan quantification. (E–G) Comparisons between tumour and control groups of samples showing that LINE-1 elements are significantly hypomethylated in all tumour groups as compared with controls, except where testicular tumours were compared with normal testes samples. However, the latter appears to reflect one of the three testicular control samples having a very low methylation value (x in D). ^*P<0.05, ^**P<0.01.

Figure 2

Cluster analysis of tumours according to methylation status. (A) Heat map showing that germinomas (G) cluster together with controls (C), whereas YSTs (Y) cluster separately, showing higher levels of methylation (red). (B) Bootstrapped hierarchical clustering using the R package pvclust (Suzuki and Shimodaira, 2006). Subgroups with an approximate unbiased P-values of <0.05 were significant. (C) The observed clustering patterns were assessed using principal component analysis and k means analysis. Yolk sac tumours are shown in blue, germinomas in red and controls in green. (D) Plot shows subgroup members selected by k means analysis plotted against the first two principal components. The optimal number of clusters for k means analysis was assessed using Scree plots. Differentially methylated loci between subgroups were identified using Mann–Whitney U-tests, with an adjusted P-value <0.05 after Benjamini–Hochberg false discovery rate correction for multiple testing. The colour reproduction of this figure is available at the British Journal of Cancer online.

Figure 3

Graphical representation of the 25 most differentially methylated genes in YSTs. Bars represent average difference of methylation to controls for each gene in either YSTs (dark bars) or germinomas (pale bars) in the first (A) and second (B) cohorts of tumours analysed. (C) The percentage of tumours across both cohorts analysed in which each gene was hypermethylated in either YSTs (dark bars) or germinomas (pale bars).

Figure 4

Pyrosequence of the CpG island of the PYCARD gene from selected tumour samples. (A) Graph showing percentage methylation of two YST and three germinoma samples at the six CpG positions included in the region of pyrosequencing for the PYCARD gene (100 bp between positions −234 and −135 proximal to the start of transcription). The YST samples show clear hypermethylation when compared with the germinoma samples. (B) Comparison between the percentage methylation at the sixth CpG (−151) in the five tumours shown in (A), as determined from either the pyrosequencing (dark bars) or from the methylation array (pale bars). This shows strong correlation between the array and pyrosequencing results and also suggests that the pyrosequencing gives more quantitative values at lower levels of methylation where these are below the level detectable in the array (arrow heads). Bars show the range for the two samples.

Figure 5

Quantitative RT–PCR analysis of DNMT3B expression. Graph showing the expression level of DNMT3B in five germinomas (pale bars) and five YST samples (dark bars). Error bars show standard deviation.