Transcriptome Analysis of Pediatric cALL Versus Normal Fetal B Cells Reveals a Novel Signature of the Malignant Phenotype.

Differential gene expression profiling of pediatric common acute lymphoblastic leukemia (cALL) versus non-malignant tissues enables identification of aberrantly expressed genes in malignant cells, facilitating discrimination of leukemic from normal cells and possibly revealing specific disease mechanisms. Expression patterns of 29 pediatric cALL samples were analyzed by use of high-density DNA microarrays HG-U133A. Leukemic patients’ bone marrow samples were compared to sorted B cells from cord blood of healthy donors expressing CD19 and CD10 surface antigens. Principal component analysis clearly distinguished leukemia samples from normal controls. Significance analysis of microarrays revealed 723 genes significantly up-regulated, and 617 down-regulated genes in leukemic cells. Independent validation of deregulated genes by RT-PCR was chosen to address enrichment limitations of controls. A comparison to previous publications investigating genetically defined subsets of cALL revealed only 5 – 22% match with our differentially expressed genes. Furthermore, class prediction with only 14 differentially expressed genes correctly classified tumors and controls not included in the training set and hitherto not investigated samples including 3 leukemic tumor lines (NALM-6, CALL-2, 697) as cALL. Interestingly, terminal deoxynucleotidyl-transferase (DNTT) as well as in the context of cALL unknown genes, were found to be the strongest predictive genes for the malignant phenotype signifying the diagnostic value of our approach.