�� 2005 Nature Publishing Group PERSPECTIVES only 1���4% of transplanted murine lym- phoma cells formed colonies in spleens of recipient animals8,9. This low clonogenic potential was also observed in human leukaemia cells by Jim Griffin and colleagues in 1985, who reported that acute myeloge- nous leukaemia (AML) blasts (a blast is a haematopoietic tumour cell with very primi- tive morphology, indicative of an acute leukaemia or lymphoma) formed colonies at low frequency in methylcellulose10,11. Also at this time, it was found that solid- organ cancer cells vary in their ability to prolif- erate in similar assays. In standard soft-agar assays,where colony formation provides a sur- rogate for transformation,Anne Hamburger and Sydney Salmon found that only 1 in 1,000 to 1 in 5,000 cells isolated from solid tumours (such as lung, ovarian and brain tumours) were capable of forming colonies12. In addi- tion, in 1961, experiments (which now could not be justified ethically) carried out by Chester Southam and Alexander Brunschwig showed that tumour cells that had been har- vested from patients with disseminated malig- nancy and then injected subcutaneously into the same patients led to a low frequency of tumour formation, and that tumours were only initiated when over 1,000,000 cells were injected13.These observations led the investiga- tors to speculate that the entire population of a tumour���s cells might arise from a few so-called ���cancer stem cells���12. A normal stem cell is defined by its dual properties of self-renewal and multilineage differentiation potential, and continuously repopulates the mature cells of the organ system that it serves. Although homeostatic pressures can dictate that a stem cell under- goes symmetric division to produce two daughter cells that are either both stem cells or both progenitor cells (this outcome depends on need) stem cells are defined by their ability to divide asymmetrically. Through this process, the division of a stem cell results in the formation of two daughter cells ��� one of which is another stem cell, and the other of which is a committed prog- enitor that is capable of further differentia- tion and proliferation but lacks the ability to self-renew.A cancer stem cell would function in a similar way to sustain the growth and spread of tumours while repopulating the distinct cell types represented within the tumour. However, a cancer stem cell would not be subject to the same intrinsic and extrinsic controls as normal stem cells. This ���cancer-stem-cell hypothesis��� repre- sents a modern-day interpretation of the pro- posal made by pathologists such as Rudolph Virchow and Julius Cohnheim ~150 years ago (TIMELINE) that cancer results from the activation of dormant embryonic-tissue rem- nants14,15. This ���embryonal-rest hypothesis���of cancer was based on the histological similari- ties between the developing fetus and certain types of cancer, such as teratocarcinomas14, and the observation that both tissues have an enormous capacity for both proliferation and differentiation, albeit aberrant differentiation in the case of tumours. There is substantial evidence in the literature to support this mechanism in paediatric cancers nephro- genic rests (abnormally retained embryonic kidney precursor cells arranged in clusters) that antedate the development of Wilm���s tumour can be detected by ultrasound during gestation16, and certain mutations can be detected at birth that are associated with a predisposition to paediatric leukaemias. These mutations include the somatic TEL���AML1 (REF. 17), MLL���AF4 (REF. 18), AML���ETO (REF. 19) and OTT���MAL translo- cations, as well as constitutional trisomy 21 (REF. 20). Van R. Potter and Barry Pierce Abstract | Many cancers seem to depend on a small population of ���cancer stem cells��� for their continued growth and propagation. The leukaemia stem cell (LSC) was the first such cell to be described. The origins of these cells are controversial, and their biology ��� like that of their normal-tissue counterpart, the haematopoietic stem cell (HSC) ��� is still not fully elucidated. However, the LSC is likely to be the most crucial target in the treatment of leukaemias, and a thorough understanding of its biology ��� particularly of how the LSC differs from the HSC ��� might allow it to be selectively targeted, improving therapeutic outcome. It is now half a century since bone-marrow reconstitution experiments, following lethal irradiation in mice, first indicated the exis- tence of the haematopoietic stem cell (HSC)1,2 ��� a cell first postulated to exist by Artur Pappenheim as early as 1917 (REF. 3). Although this cell population is still not fully characterized, its discovery awakened the field of stem-cell biology. The recent discov- ery of many more tissue-specific stem cells4���7 has kept this field of study at the forefront of biological research. At around the same time that the existence of the HSC was postulated, observations were reported of the heterogeneous potential of tumour cells to self-renew both in vitro and in vivo. For example, in 1973, Ernest McCulloch and colleagues observed that only 1 in 100 to 1 in 10,000 murine myeloma cells had the ability to form colonies in vitro. In 1963, Robert Bruce and colleagues showed that NATURE REVIEWS | CANCER VOLUME 5 | APRIL 2005 | 311 Leukaemia stem cells and the evolution of cancer-stem-cell research Brian J. P. Huntly and D. Gary Gilliland T I M E L I N E

�� 2005 Nature Publishing Group 312 | APRIL 2005 | VOLUME 5 www.nature.com/reviews/cancer P E R S P E C T I V E S Based on functional and immunopheno- typic analysis of subpopulations of cells with modern technologies, cancer has become viewed increasingly as a stem-cell disorder, in which the continued growth and propaga- tion of the whole tumour depends on a small subpopulation of self-renewing cancer stem cells. The HSC was the first adult somatic stem cell to be described. The exis- tence of cancer stem cells was also first described in the haematopoietic system. The demonstration of a leukaemia-initiating cell, now commonly referred to as the leukaemia stem cell (LSC), in AML and other leukaemias, along with the subsequent iden- tification of cancer stem cells in breast and central-nervous-system (CNS) tumours, provides evidence for the broad applicability of the cancer-stem-cell model. It is plausible that malignant stem cells share a number of biological features that are different from their normal-tissue counterparts and that these might be exploited for therapeutic benefits. As malig- nant haematopoieisis is well characterized and amenable to laboratory and clinical investigation, further investigation of the biology of the LSC could provide a para- digm for all cancer stem cells and could improve the therapy of both leukaemias and other solid-organ cancers. In addition, The haematopoietic system is amenable to ex vivo analysis, and, in the late 1980s and 1990s, developments in antibody technology allowed investigators ��� such as Irving Weissman and colleagues ��� to define the ontogeny of this system, based on surface immunophenotype in humans and mice (for a comprehensive review see REF. 25). These studies identified a large number of pheno- typic markers that are associated with defined lineages and developmental stages of haematopoietic cells, and concomitantly allowed the annotation of malignant haematopoietic ontogeny.Another technical advance that greatly facilitated the characteri- zation of leukaemia and other cancer stem cells was the development of high-speed multi- parameter flow cytometry. This is an auto- mated technique that is used to identify and/or purify distinct cell populations, based on the ability of fluorescently labelled antibod- ies to bind cell-surface antigens. Following excitation with a laser, the fluorochrome- bound antibody emits light of a specific wavelength, allowing the detection and enu- meration of cell populations that express the antigen. Furthermore, by applying an electric charge to deflect populations that are positive for a predefined fluorochrome, and therefore antigen,these populations can be collected for functional analysis26,27. updated these ideas in the 1970s and 1980s, describing cancer as the maturation arrest of tissue-determined stem cells21���23. However, formal proof of these hypotheses would have to await further technological advances. So, it seems that tumours are composed of a heterogeneous population of cells, within which resides a small population of cancer stem cells that are exclusively respon- sible for the growth and propagation poten- tial of the whole tumour. However, an alternative hypothesis, the stochastic model, could also explain the heterogeneous poten- tial of tumour cells to self-renew. This model predicts that all tumour cells have the poten- tial to self-renew and recapitulate the tumour, but that the probability that any particular tumour cell enters the cell cycle and finds an environment permissive for growth in an assay of tumorigenesis is low24. To differentiate between these two models it is necessary to define distinct populations of cells within tumours, based on surface immunophenotypic or functional character- istics, to purify these populations to homo- geneity and to develop long-term assays of their functional ability. The functional assessment of a cancer stem cell requires not only the ability to form a new tumour, but to recapitulate precisely the phenotype of the initial disease (BOX 1). Rudolph Virchow proposes the ���embryonal-rest hypothesis��� of tumour formation, based on histological similarities between tumours and embryonic tissues14. This theory, later expanded by other pathologists such as Julius Cohnheim, suggested that tumours develop from residual embryonic tissues15. Low in vivo clonogenic efficiency of mouse lymphoma cells demonstrated by Robert Bruce and Hugo Van der Gaag9. 1967���1981: Philip Fialkow and co- workers use patterns of X-linked gene inactivation in patient samples to demonstrate that leukaemias and myeloproliferative disorders are clonal in origin. These studies also indicate the involvement of an early stem or progenitor cell in chronic myelogenous leukaemia and acute myelogenous leukaemia (AML)28���31. 1968���1973: development of fluorescence- activated cell sorting by Leonard Herzenberg and co- workers121 and the production of the first commercially available machines27. 1989���1994: refinement of the SCID and non-obese diabetic (NOD)/SCID mouse assays for assessment of normal and leukaemic stem-cell function by John Dick and colleagues, as well as other investigators. Van R. Potter, Barry Pierce and other investigators extend the embryonal-rest hypothesis to suggest that tumours arise from maturation arrest in tissue-specific stem cells. Advances in cell sorting with the commercial manufacture of multiparameter high-speed machines. 1980���1990s: creation of monoclonal antibodies against numerous haematopoietic cell- surface antigens. Timeline | The evolution of cancer-stem-cell research Characterization of the NOD/SCID immunocompro- mised mouse123. Irving Weissman and colleagues are the first to use SCID mice as an assay system for human haematopoietic cells33. 1855 1961 1963 1967 1968 1970s 1985 1988 1989 1980 1990s 1992 Autotransplantation assays in humans demonstrate a low frequency of tumour-initiating cells in various solid-organ malignancies13. James Till and Ernest McCulloch suggest the existence of the haematopoietic stem cells, based on experiments demonstrating the ability of bone marrow to provide protection to lethally irradiated mice2. Characterization of the severe combined immunodeficiency (SCID) mutation in mice122. Low clonogenic efficiency of AML blasts in vitro demonstrated by Jim Griffin, Bob Lowenberg and co-workers10,11.