The cancer stem-cell hypothesis proposes that tumours are populated by a rare fraction of cells with stem-cell properties. Although cancer stem cells have been identified and characterized in leukaemias, there is only a small amount of evidence for their existence in solid tumours. Peter Dirks and colleagues report the isolation and characterization of a cell subpopulation from human brain tumours that have stem-cell properties in vivo.

Cancer stem cells are identified based on their ability to undergo self-renewal, as well as the ability of small numbers of these cells to form tumours with the same phenotype as the tumour they were derived from. Dirks and colleagues had previously identified a population of human brain tumour cells, isolated either from aggressive glioblastomas or medulloblastomas, with in vitro stem-cell-like properties. Cells from both tumour types where characterized by expression of the cell-surface glycoprotein CD133, which is also expressed by normal neural and haematopoietic stem and progenitor cells. In a study recently published in Nature, they describe the in vivo capabilities of these cells following injection into non-obese diabetic/severe combined immunodeficienct (NOD/SCID) mice. Injection of as few as 100 CD133+ cells produced a tumour that had the same characteristics as the patient's original tumour. Injection of up to 105 CD133− cells, however, did not cause tumour formation in these mice, indicating that CD133+ cells are the brain tumour stem cells.

What are the characteristics of these stem cells? In addition to their high rates of proliferation, both medulloblastoma and glioblastoma CD133+ cells expressed neural precursor markers, which CD133− cells did not. Instead, CD133− cells express differentiated neural lineage markers. This indicates that these cancer stem cells have taken on neural precursor phenotypes. CD133+ cells also had a key property of cancer stem cells — self-renewal capacity in serial re-transplantation assays. When as few as 1,000 CD133+ cells from original xenografts of both brain tumour types were re-injected into a second set of mice, these mice formed brain tumours with the same phenotype of the primary xenograft.

Both CD133+ and CD133− tumour cell types had the same cytogenetic alterations, indicating that they were derived from the same original clone. The CD133+ cells only make up about 20–30% of the overall tumour population, however, and the authors propose that these cells have somehow acquired patterns of abnormal differentiation. Further experiments will determine whether this stem-cell-initiating event occurs in a normal stem cell, or in normal progenitor/differentiated cells that have re-acquired stem-cell properties. Also, since only a small population of cancer cells appears to be capable of regenerating an entire tumour, treatment relapse is likely to be the result of failure to target the tumour-initiating cell.