Positive selection of CD34+ cells is an attractive approach to reduce tumour cell contamination in bone marrow (BM) and peripheral blood progenitor cell (PBPC) autografts in malignancies not expressing CD34. All current selection methods use monoclonal antibodies (MoAbs) specific for the class I or class II CD34 epitopes, while for detection most investigators use class III MoAbs. Since the distribution of the different CD34 epitopes on haematopoietic progenitors differs, we studied their significance in CD34+ selection procedures. Testing MoAbs against class I, II and III CD34 epitopes on normal BM we observed that +/- 23% of class III positive cells was class I negative. A higher expression of the class III epitope compared with classes I or II was observed on the KG1 cell line, whereas no differences in binding capability were found. The class III epitope anti-CD34, 561, was compared with the class I epitope anti-CD34, BI-3C5, both coupled to M450 Dynabeads. The yield of CD34+ cells obtained with the 561 beads was 1.7% of the mononuclear cells versus 0.95% using the class I epitope, a 1.95-fold increase (1.3-2.7), whereas the purity was similar (96% in both cases). The absolute number of CD34+ cells was therefore twofold higher after 561 selection, including cells with a more mature phenotype. In single cell assay comparable numbers of highly proliferative progenitors but higher numbers of differentiated colonies per phenotypic subfraction were measured. In conclusion, M450 beads coated with the 561 anti-class III CD34 epitope are more efficient in isolating CD34+ cells from bone marrow, probably due to a broader distribution of the class III epitope.