Genome-wide profiling using single-nucleotide polymorphism arrays identifies novel chromosomal imbalances in pediatric glioblastomas

Abstract

Available data on genetic events in pediatric grade IV astrocytomas (glioblastoma [pGBM]) are scarce. This has traditionally been a major impediment in understanding the pathogenesis of this tumor and in developing ways for more effective management. Our aim is to chart DNA copy number aberrations (CNAs) and get insight into genetic pathways involved in pGBM. Using the Illumina Infinium Human-1 bead-chip-array (100K single-nucleotide polymorphisms [SNPs]), we genotyped 18 pediatric and 6 adult GBMs. Results were compared to BAC-array profiles harvested on 16 of the same pGBM, to an independent data set of 9 pediatric high-grade astrocytomas (HGAs) analyzed on Affymetrix 250K-SNP arrays, and to existing data sets on HGAs. CNAs were additionally validated by real-time qPCR in a set of genes in pGBM. Our results identify with nonrandom clustering of CNAs in several novel, previously not reported, genomic regions, suggesting that alterations in tumor suppressors and genes involved in the regulation of RNA processing and the cell cycle are major events in the pathogenesis of pGBM. Most regions were distinct from CNAs in aGBMs and show an unexpectedly low frequency of genetic amplification and homozygous deletions and a high frequency of loss of heterozygosity for a high-grade I rapidly dividing tumor. This first, complete, high-resolution profiling of the tumor cell genome fills an important gap in studies on pGBM. It ultimately guides the mapping of oncogenic networks unique to pGBM, identification of the related therapeutic predictors and targets, and development of more effective therapies. It further shows that, despite commonalities in a few CNAs, pGBM and aGBMs are two different diseases.

Fig. 1.

Unsupervised hierarchical clustering of 24 GBMs (18 pediatric and 6 adult) and 1 control normal brain (CB). We used statistically significant CNAs present in at least one sample as data input (Welsch t-test, p < 0.001) and separated our data set accordingly. Pediatric (P) samples clustered separately from adult samples (A) and from the CB. DNA from one pediatric sample P1 was extracted from two separate parts of the same tissue sample. DNA from both extractions was independently subjected to whole-genome amplification prior to analysis on the 100K SNP array platform. P1a and P1b clustered similarly showing the reproducibility of the analysis of genome profiling. Primary adult glioblastoma (Ap) clustered separately from secondary adult glioblastoma (As).