Purpose: Traditional genetic approaches to identify gene mutations in cancer are expensive and laborious. Nonetheless, if we are to avoid rejecting effective molecular targeted therapies, we must test these drugs in patients whose tumors harbor mutations in the drug target. We hypothesized that gene expression profiling might be a more rapid and cost-effective method of identifying tumors that contain specific genetic abnormalities.
Materials and methods: Gene expression profiles of 46 samples of medulloblastoma were generated using the U133av2 Affymetrix oligonucleotide array and validated using real-time reverse transcriptase polymerase chain reaction (RT-PCR) and immunohistochemistry. Genetic abnormalities were confirmed using fluorescence in situ hybridization (FISH) and direct sequencing.
Results: Unsupervised analysis of gene expression profiles partitioned medulloblastomas into five distinct subgroups (subgroups A to E). Gene expression signatures that distinguished these subgroups predicted the presence of key molecular alterations that we subsequently confirmed by gene sequence analysis and FISH. Subgroup-specific abnormalities included mutations in the Wingless (WNT) pathway and deletion of chromosome 6 (subgroup B) and mutations in the Sonic Hedgehog (SHH) pathway (subgroup D). Real-time RT-PCR analysis of gene expression profiles was then used to predict accurately the presence of mutations in the WNT and SHH pathways in a separate group of 31 medulloblastomas.
Conclusion: Genome-wide expression profiles can partition large tumor cohorts into subgroups that are enriched for specific genetic alterations. This approach may assist ultimately in the selection of patients for future clinical trials of molecular targeted therapies.