Klin Padiatr 2011; 223 - A16
DOI: 10.1055/s-0031-1292597

Identification of a CpG Island Methylator Phenotype (PA-CIMP) in a Subgroup of Pilocytic Astrocytoma

H Witt 1, 2, DTW Jones 1, M Zapatka 1, H Cin 1, M Ryzhova 4, M Remke 1, 2, AE Kulozik 2, O Witt 2, 3, P Lichter 1, A Korshunov 5, SM Pfister 1, 2
  • 1Division Molecular Genetics, German Cancer Research Center, Heidelberg, Germany
  • 2Department of Pediatric Oncology, Hematology and Immunology, University Hospital Heidelberg, Germany
  • 3Clinical Cooperation Unit Pediatric Oncology, German Cancer Research Center, Heidelberg, Germany
  • 4NN Burdenko Neurosurgical Institute, Moscow, Russia
  • 5Department of Neuropathology, University Hospital Heidelberg, Germany

The most common pediatric brain tumors are pilocytic astrocytomas (PA), classified as WHO°I. These tumors arise most often within cerebellum, and here complete resection is typically possible, leading to a favorable prognosis. By contrast, tumors in other locations, however, e.g. diencephalic, brain stem, or optic tract tumors, where complete resection is often challenging are associated with much worse prognosis.

PA are discussed to be a “single pathway disease” because MAPK pathway activation (e.g. by KIAA1549:BRAF fusion, BRAF V600E mutation, etc.) has been demonstrated in approximately 80% of cases. So far no specific genetic aberration of the remaining 20% of cases is causative for MAPK activation could be identified.

We performed an integrative study combining array-based genome-wide methylation and transcriptome analyses of in total 71 primary pilocytic astrocytomas. Both approaches were investigated on Illumina arrays, HumanWG-6 (mRNA profiling, >47,000 probes) and HumanMethylation27 (DNA methylation, >27,500 probes covering ˜14,000 promoters). To identify molecular subgroups, we performed unsupervised consensus/hierarchical clustering of the 1000 most variant probes on each platform. A nonparametric approach (Wilcoxon rank-sum test) was used to discover differenzially methylated or expressed CpG-sites/genes between groups. Two independent validation cohorts with detailed clinical follow-up data were also collected, including DNA of 90 PA samples and a tissue microarray containing 100 PAs.

We could identify two subgroups with distinct DNA methylation patterns. One cluster highlighted a particular subgroup of PA with a highly characteristic DNA methylation pattern, including multiple hypermethylated loci, previously referred to as CpG island methylator phenotype (CIMP). Interestingly, tumors in this PA-CIMP group showed a significantly lower frequency of MAPK alterations (P=0.0003), especially the KIAA1549:BRAF fusion (P<0.0001), and these tumors were associated with favorable progression free survival compared to the remaining subgroup. Comparison with transcriptome data from the same samples was used to identify concordantly hypermethylated and downregulated genes, identifying the potentially most interesting candidate genes.

We could identify a candidate tumor suppressor gene that is epigentically silenced in this subgroup, which normally functions as a MAPK antagonist. When treating low-grade glioma cell lines with 5-aza-2'-deoxycytidine we could re-induce expression of this candidate gene. These findings suggest an alternative mechanism of MAPK activation based on epigenetic silencing in a subgroup of PA. This hypothesis is currently being validated in an independent series of 190 PAs.