Microarray Guided Flow Cytometry: CD44 and CD27 Are Extremely Powerful Discriminators of ALL Subtypes.

Abstract

Expression profiling (EP) studies in acute lymphoblastic leukemia (ALL) investigate thousands of genes per sample to predict principal ALL subtypes and/or prognosis. So far, insufficient amount of information is available on transferring the knowledge from EP to other methods. We have screened publicly available data of two EP studies (Yeoh et al., Cancer Cell.1(2), 2002; Ross et al., Blood.102(8), 2003) for genes that can be taken into flow cytometry (FC) diagnostics. The genes, which were identified as the best correlating with the childhood ALL subgroups (E2A/PBX1, MLL, TEL/AML1, BCR/ABL and hyperdiploid genotypes; relapse and therapy-induced AML) were individually screened. After recalculation of the data just for the B-cell precursor (BCP) ALL cases, we selected the genes in which the difference in expression was likely to be observed on protein level. Next, we selected molecules with available monoclonal antibody (mAb). Reactivity and specificity of mAbs was tested in healthy peripheral blood cells and/or in cell lines. Next, selected molecules (CD44, CD27, CD247, CD49f, CD103) were investigated by 4-color FC in diagnostic BM samples. A total of 62 children with BCP ALL and 13 children with T lineage ALL were investigated. CD44 expression (identified as predictor of poor prognosis T-ALL by EP) correlated with the risk group categories of T-ALL (p=0.0321). Strong positive correlation of CD27 (p<0.0001) and negative correlation of CD44 (p<0.0001) with TEL/AML1 genotype was found as predicted by some EP studies. This prompted a simultaneous analysis of the expression of CD44 and CD27. [Display omitted] Clearly, the expression of CD27 and CD44 is mutually exclusive in most cases. Furthermore, we investigated CD27 and CD44 expression in non-malignant BM by 7color-FC together with other key molecules of B cell differentiation. We identified physiological subpopulations with patterns of CD44 and CD27 expression corresponding to distinct ALL genotypes. EP discovered that CD247 (CD3zeta chain) is overexpressed in poor prognosis hyperdiploid ALL. We prove that this T-cell specific molecule is present in BCP ALL but we could not analyze its prognostic impact yet. EP data showed positive correlation of CD49f with BCR/ABL genotype. FC showed similar results with lower statistical significance (p=0.0432). CD103 expression was undetectable, although it should be higher in BCR/ABLneg cases. Among the 5 molecules tested in diagnostic BM, CD44 and CD27 appear extremely useful in discriminating genetic and/prognostic subsets, whereas CD247 and CD49f are potentially useful. Microarray-guided FC, which investigates molecules in mutual context, has comparable power to EP despite using much fewer molecules. Although EP describes almost only empirical correlations, it was claimed to be a candidate for replacement the existing lab techniques. Since major progress in ALL treatment efficacy is mostly awaited from therapies against specific targets, it seems reasonable to investigate these molecules directly (e.g, BCR/ABL by PCR or CD33 by FC).