signature=db62d16eb6fad19f830104fcda6fab51,Gene expression profiles associated with drug resistance ...

Abstract

1338

20-50 % of acute myeloid leukemia (AML) cases show primary resistance to induction chemotherapy, which has been shown to be an independent adverse predictive factor for patient outcome. The purpose of this study was to associate gene expression profiles in AML blasts of newly diagnosed patients with their response to Cytarabin-based induction chemotherapy. Response was assessed cytomorphologically by the initial cytoreduction, whereas good response is no blasts in peripheral blood, <5% blasts in bone marrow and no extramedullary manifestation, poor response is residual blasts in peripheral blood or 5% or more blasts in bone marrow or extramedullary AML manifestation on treatment day 15, respectively. We compared expression signatures of seven AML samples with good response and nine AML samples with poor response. Inclusion criteria were age above 16 years, FAB type M1-2 or M4-5 and >50% blasts in bone marrow or peripheral blood. The median percentage of blasts was 80%. The majority of patients showed intermediate risk cytogenetics (according to CALGB 2002). Patients with favorable risk cytogenetics were present in both groups, whereas 2 patients with adverse cytogenetics were present only among patients with poor response. cDNA microarrays containing ∼ 41,000 cDNA clones (SFGF, Stanford CA, USA) were used. Filtering out poorly hybridised clones greatly reduced clone number. Hierarchical clustering (Pearson correlation) identified two clusters, one consisting of five good response samples and four poor response samples and one consisting of two good response samples and five poor response samples. T-statistics identified 38 genes differentially expressed (p<0.01) between good and poor responders. In addition, Prediction Analysis of Microarrays (PAM) was performed using 10 randomly selected samples as a training set and the remaining six samples as a test set. Considering only those clones that had a four-fold difference in expression from the mean on at least one array, this analysis identified 39 genes that correctly predicted the treatment response of five of the six test patients. Interestingly, among the 39 discriminating genes, alpha 9 integrin and CD34 antigen were highly predictive for the chemoresistant phenotype, which might suggest that an immature phenotype and blast cell - stroma interactions play an important role in AML resistance. Confirmation of our findings in a larger patient cohort might lead to further risk stratification in AML therapy and new clues for elucidation of resistance mechanisms in AML.