Gene expression data from sorted primary human cord blood samples

Experiments using xenografts show that some solid tumours and leukemias are organized as cellular hierarchies sustained by cancer stem cells (CSC). Despite promise, the relevance of the CSC model to human disease remains uncertain. Here we show that acute myeloid leukemia (AML) follows a CSC model based on sorting multiple populations from each of 16 primary human AML samples and identifying which contain leukemia stem cells (LSC) using a sensitive xenograft assay. Analysis of gene expression from all functionally validated populations yielded an LSC-specific signature. Similarly, a hematopoietic stem cell (HSC) gene signature was established. Bioinformatic analysis identified a core transcriptional program shared by LSC and HSC, revealing the molecular machinery underlying stemness properties. Both stem cell programs were highly significant independent predictors of patient survival and also found in existing prognostic signatures. Thus, determinants of stemness influence clinical outcome of AML establishing that LSC are clinically relevant and not mere artifacts of xenotransplantation.