Purpose By parallel assessment of multiple apoptosis-related transcripts, we aimed to refine the current concept of apoptosis resistance in acute myeloid leukemia (AML) and identify the combination of genes best predicting overall survival (OS). Patients and Methods The reverse transcriptase multiplex ligation–dependent probe amplification technique was used for simultaneous quantification of 31 apoptosis-related transcripts in viable (7AAD–/AnnexinV–) blasts (CD45dim) from bone marrow aspirates of 120 newly diagnosed AML patients. By forward selection, a prognosis-predicting gene expression profile was constructed. The predictive validity of this profile was assessed by cross validation. Results High transcript levels were associated with poor OS for seven of 31 genes, three of which were proapoptotic. The average expression of all 12 antiapoptotic genes was associated with poor OS (P = .029). A similar association with poor OS was found for the average expression of all 19 proapoptotic genes (P = .009). Forward selection and cross validation revealed the antiapoptotic gene BIRC3 and the proapoptotic genes BAX-(l) and BMF to optimally predict OS. Three equally sized patient groups, constructed by ranking the cross-validated prognoses of the patients, were clearly distinct (median OS times were 8.2, 16.7, and 85.6 months). Conclusion High expression of both pro- and antiapoptotic genes predicted poor OS, which postulates a mechanism of activation of the apoptosis pathway as a whole. This mechanism, which culminates in a three-gene expression signature, allows accurate clinical outcome prediction in AML and puts efforts to target single antiapoptosis genes in a new perspective.
Epigenetic drug combination induces remission in mouse xenograft models of pediatric acute myeloid leukemia