Improved risk-stratification for posterior fossa ependymoma of childhood considering clinical, histological and genetic features - a retrospective analysis of the HIT ependymoma trial cohort.

INTRODUCTION: Risk stratification of children with ependymomas of the posterior fossa in current therapeutic protocols is mainly based on clinical criteria. We aimed to identify independent outcome predictors for this disease entity by a systematic integrated analysis of clinical, histological and genetic information in a defined cohort of patients treated according to the German HIT protocols. METHODS: Tumor samples of 134 patients aged 0.2-15.9 years treated between 1999 and 2010 according to HIT protocols were analyzed for histological features including mitotic activity, necrosis and vascular proliferation and genomic alterations by SNP and molecular inversion probe analysis. Survival analysis was performed by Kaplan-Meier method with log rank test and multivariate Cox regression analysis. RESULTS: Residual tumor after surgery, chromosome 1q gain and structural genomic alterations were identified as predictors of significantly shorter event-free (EFS) and overall survival (OS). Furthermore, specific histological features including vascular proliferation, necrosis and high mitotic activity were predictive for shorter OS. Multivariate Cox regression revealed residual tumor, chromosome 1q gain and mitotic activity as independent predictors of both EFS and OS. Using these independent predictors of outcome, we were able to build a 3-tiered risk stratification model that separates patients with standard, intermediate and high risk, and which outperforms current stratification procedures. CONCLUSION: The integration of defined clinical, histological and genetic parameters led to an improved risk-stratification model for posterior fossa ependymoma of childhood. After validation in independent cohorts this model may provide the basis for risk-adapted treatment of children with ependymomas of the posterior fossa.

H3.3 G34R mutations in pediatric primitive neuroectodermal tumors of central nervous system (CNS-PNET) and pediatric glioblastomas: possible diagnostic and therapeutic implications?

Pediatric glioblastomas recently have been exon sequenced with evidence that approximately 30 % of cases harbour mutations of the histone H3.3 gene. Although studies to determinate their role in risk stratification are on-going, it remains to be determined whether H3.3 mutations could be found in other tumors such as pediatric primitive neuroectodermal tumors of the central nervous system (CNS-PNETs) and whether the presence of H3.3 mutations in glioblastomas could be used as diagnostic tool in their differential diagnosis with CNS-PNETs. We performed a large mutational pyrosequencing-based screening on 123 pediatric glioblastomas and 33 CNS-PNET. The analysis revealed that 39/123 (31.7 %) glioblastomas carry H3.3 mutations. The K27M (AAG → ATG, lysine → methionine) mutation was found in 33 glioblastomas (26 %); the G34R (GGG → AGG, glycine → arginine) was observed in 6 glioblastomas (5.5 %). However, we also identified 4 of 33 cases (11 %) of CNS-PNETs harbouring a H3.3 G34R mutation. Multiplex ligation-dependent probe amplification analysis revealed PDGFR-alpha amplification and EGFR gain in two cases and N-Myc amplification in one case of H3.3 G34R mutated CNS-PNET. None of H3.3 mutated tumors presented a CDKN2A loss. In conclusion, because pediatric patients with glioblastoma and CNS-PNET are treated according to different therapeutic protocols, these findings may raise further concerns about the reliability of the histological diagnosis in the case of an undifferentiated brain tumor harbouring G34R H3.3 mutation. In this view, additional studies are needed to determine whether H3.3 G34 mutated CNS-PNET/glioblastomas may represent a defined tumor subtype.