Drug-Response Signature Predicts Outcome in Adult Acute Myeloid Leukemia and Associates Poor Response with Molecular Characteristics of Hematopoietic Stem Cells.
Abstract Resistance to induction chemotherapy is of independent prognostic value in acute myeloid leukemia (AML). DNA microarrays were used to determine the gene-expression profile of AML blasts in 38 patients with good or poor response to induction chemotherapy. We selected an 11-sample training set, applying prediction analysis of microarrays (PAM) to devise a drug-response predictor which was tested on the remaining 27 samples and an independent set of samples recently published (Bullinger et al. 2004). Our drug-response predictor with 46 clones divided the 27 samples into two prognostic subgroups, the poor response group having a significantly shorter overall survival (P= .021). A subset of these 46 clones was sufficient to divide the published 62-sample test-set with intermediate risk cytogenetics into prognostically relevant subgroups (P= .028), adding prognostic information to that of known risk factors in multivariate analysis (hazard ratio, 2.8; 95 percent confidence interval, 1.4 to 5.8; P= .005). Thirteen of 39 drug resistance-enriched genes are known to be upregulated in hematopoietic stem/progenitor cells, and the expression pattern in normal CD34+ cells is highly correlated to the drug-resistance signature. This suggests that drug resistant AMLs show molecular features of hematopoietic stem/progenitor cells and can be identified by a characteristic gene-expression profile.