A Low Density DNA Microarray for Comparing Gene Expression Profiles in Hematological Malignancies.

High density microarrays (HDM) are powerful tools for simultaneously profiling the expression levels of thousands of genes. The application of this technology to study of neoplastic hematological disorders.has identified new sub groups of disease not related previously and new prognosis markers. However there is a limited experience in the gene expression studies using low density microarrays (LDM) in neoplastic hematological disorders.

A gene expression analysis system based on a LDM containing 538 oligonucleotides has been developed. Whole technical process was optimized to improve the analysis of differential expression. We have analyzed mRNA from cell line cultures (Jurkat, U937), whole blood samples from healthy subjects and different hematological malignancies (HM) using this chip. A hierarchical clustering procedure applying Welch t-statistics with Bonferroni correction was used to analysis gene expression data The LDM generated a linear response of 2 magnitude orders and a CV values less than 20% for hybridization and label replicates. This procedure detects 0,2 fmols of mRNA. We have found genes with statistically significant differences between Jurkatt and U937 cells cultures, and blood samples from 15 healthy donors, 59 lymphocyte leukemia and 13 myeloid leukemia and myelodisplasia syndrome patients. A classification system based on gene expression data was constructed with an accuracy of 97%.to predict healthy or lymphocyte leukemia status. To identify different subsets of patients in the B-CLL group, whole blood samples from 12 B-CLL patients were collected and defined as stables, according to clinical and analytical criteria at the time of diagnosis, “stable” (n=6) if disease stability was maintained for more than five years after the diagnosis and “progressive” (n=6) if the disease progressed less than one year after the diagnosis. Applying Welch statistical test without correction and a p<0.05 yielded two lists of 29 and 19 probes differentially hybridized from VSN and quantile-robust normalized data, respectively. The supervised hierarchical clustering of B-CLL samples with 29 statistically significant probes shown that samples grouped together based on their stable or progressive behavior. Eighteen probes were statistically significant in both normalized data. In order to confirm the data expression of POU2F2, PSMB4, FCER2, LCP1, and ABCC5 genes represented by 5 of the 18 statistically significant probes, real-time RT-PCR was performed. Three out of 5 genes -POU2F2, PSMB2, and FCER2- were over-expressed in B-CLL stable patients. Differences were statistically significant (P<0.05) and, therefore, results obtained from the chip for POU2F2, PSMB2, and FCER2 genes were confirmed.

In conclusion, a viable LDM for gene expression analysis and a simple procedure has been developed useful for analysis of whole blood samples, without any cellular or sample manipulation prior to RNA extraction with variability and reproducibility similar to others commercial HDM. The application to different samples is capable to establish significant differences in gene clusters and could be useful for clinical application in HM