Abstract

B-cell prolymphocytic leukemia (B-PLL) is a rare disease, originally described in the early seventies and now recognized as a specific entity by the WHO classification. Diagnosis is based on clinical features and lymphocyte morphology (>55% circulating prolymphocytes), as no specific immunophenotypic or cytogenetic marker is available. According to WHO, cases of chronic lymphocytic leukemia (CLL) with increased prolymphocytes (CLL/PL) (>10% and <55%) should not be considered as B-PLL, because they have different genetic features. However, the existence of B-PLL as a separate entity from CLL has been questioned. We investigated the gene expression profiles of B-PLL and CLL to identify key genetic differences potentially useful for the diagnosis or involved in their different natural history. We retrospectively selected cryopreserved samples from 10 de-novo B-PLL and 10 untreated CLL. Diagnosis was well-defined by clinical features, lymphocyte morphology and immunophenotype. Matutes immunophenotypic score was 4–5 in all CLL; 3 cases showed CLL/PL morphology. B-PLL scored 0–3, with 3 CD5+ cases. Diagnosis of B-PLL was corroborated by excluding a leukemic form of MCL; t(11;14) assessed by fluorescent in situ hybridization (FISH) was absent in all but one case which, originally diagnosed as B-PLL, was reclassified as leukemic MCL. Five B-PLL and 2 CLL showed del(p53) by FISH. Total RNA was extracted from frozen blood mononuclear cells containing ≥95% purity of malignant cells, determined by flow cytometry. cDNA synthesis followed by biotin-labelled cRNA synthesis was carried out as per Affymetrix protocols. Microarray experiments were performed by MRC geneservice (UK HGMP Resource Centre), using the Affymetrix Human U133PLUS2 GeneChip array (54K probes). Hierarchical clustering was performed on samples using a filtered set of 9878 genes with >4 different algorithms. Prediction analysis for microarray (PAM) and significance analysis of microarray data (SAM) were used to evaluate class performance, and to partition genes using a priori defined labels of morphology, immunophenotype and cytogenetics. Unsupervised analysis reproducibly partitioned samples into two homogeneous distinct groups, corresponding to the diagnoses of B-PLL and CLL. SAM analysis identified 3957 differentially expressed transcripts (false discovery rates <1%), >77% of which showed an over 2-fold difference in expression between the groups. PAM analysis refined a sub-group of 46 genes which most efficiently differentiated the two diseases. Differentially expressed genes included those encoding surface antigens, oncogenes, transcription factors, adhesion molecules or involved in cell cycle/cell proliferation, lipidic metabolism and catalytic protein activity. Comparison of CD5 positive (13) versus CD5 negative (7) cases and cases with (7) or without (13) del(p53) showed no reliable class prediction. Our study formally demonstrates that B-PLL and CLL are two distinct diseases, each showing a specific gene expression. B-PLL has a homogeneous genomic profile irrespective of its heterogeneity in laboratory features. Validation of a model based on the expression of few genes predictive of diagnosis is on going. Further analysis of these data may also identify specific genes involved in B-PLL pathogenesis and drug resistance.

Author notes

Corresponding author