ABSTRACT: In this series we described the gene expression profiles of 18 APL patients using a high density DNA-oligonucleotide microarray (Agilent) representing about 20.000 genes demonstrating the presence of a very uniform expression pattern of APL cases among which two mainly types of gene profiles were recognized by unsupervised analysis. Relationship between these two subgroups and the different clinical, haematological and molecular APL features, comprising FLT3 status, were investigated. To identify biologically meaningful subsets of APL, fluorescently-labelled cRNA was generated by in vitro transcription using Low RNA Input Fluorescent Linear Amplification Kit (Agilent Technologies) according to the manufacturer?s instructions. Amplified cRNA of each patient was labelled with cyanine 5-CTP and Human Universal Reference (Stratagene) with 3-CTP in each experiment. Samples were hybridized on Agilent Human 1A Oligo Microarray (V2), ink-jet printed microarray, comprising 20,173 (60-mer) experimentally validated oligonucleotide probes (features). Fluorescence data were analysed with Feature Extraction Software v.7.5 (Agilent Technologies). Log10 ratio of the dye-normalized Cy3 and Cy5 channel signals were calculated. Raw signal intensities from each scan were imported into the gene expression analysis software Luminator (Rosetta Biosoftware). Two-dimensional clustering analysis was performed with Rosetta Luminator using an agglomerative algorithm with average link heuristics and correlation with mean subtraction. Supervised classification of APL samples into categories based on gene expression profiles was performed using Bayesian classifier implemented by Rosetta Biosoftware. Resultant gene expression profiles were first analyzed using an unsupervised, agglomerative hierarchical clustering. The matrix view clearly evidenced the high grade of uniformity in gene expression pattern among APL patients being the majority of genes expressed at high (red colour) or low (green colour) levels relative to reference in all APL cases analyzed. This appearance of unsupervised matrix reflects the homogeneous nature of acute promyelocytic leukemia whose gene expression pattern definitely distinguishes it from all other types of acute myeloid leukemia. Beside the homogeneous pattern, subtle differences in gene expression had the strength to distinguish three clusters of APL patients (designed I, II, III). Comparisons between cluster I and cluster II patients clearly revealed a preferential distribution of FLT3 gene mutated cases in cluster I. Again, cluster I was related to microgranular morphology (M3v) (p=0.035) and short-type PML-RAR isoform (bcr3) (p=0.044). Furthermore cluster I was significantly associated with higher presentation circulant blast cell percentage (p=0.030) as well as hyperleucocitosis (p=0.009). To identify those genes distinguishing the FLT3-ITD leukemic cells from FLT3-WT, we applied a Bayesian classifier. 147 genes, significantly different among classes based on t-tests, were selected and included into the classifier. Among them, 92 genes were up-regulated in FLT3-ITD class and 55 were down-regulated. FLT3 internal tandem duplication gene expression signature consisted of several high expressed genes, among them we found genes that are involved in cytoskeleton organization, cell adhesion and migration, in proliferation and coagulation/inflammation pathways as well as down-regulated genes of myeloid granules and differentiation suggesting a role of FLT3 mutations in the pathogenesis and clinical manifestation of an APL subtype. SUPPLEMENTARY DATA:1) Anova gene list ITD, 2)Anova gene list M3 morphology, 3)Anova gene list PML/RARalpha isoforms (bcr), 4)Intersection gene list ITD_M3, 5) Intersection gene list ITD_bcr, 6)Intersection gene list M3_bcr, 7)Bayesian classifier gene list for FLT3-ITD vs FLT3-wt, 8)Bayesian classifier gene list for bcr1 vs bcr3
