Review | Published:

Self-renewal and solid tumor stem cells



Solid tumors arise in organs that contain stem cell populations. The tumors in these tissues consist of heterogeneous populations of cancer cells that differ markedly in their ability to proliferate and form new tumors. In both breast cancers and central nervous system tumors, cancer cells differ in their ability to form tumors. While the majority of the cancer cells have a limited ability to divide, a population of cancer stem cells that has the exclusive ability to extensively proliferate and form new tumors can be identified based on marker expression. Growing evidence suggests that pathways that regulate the self-renewal of normal stem cells are deregulated in cancer stem cells resulting in the continuous expansion of self-renewing cancer cells and tumor formation. This suggests that agents that target the defective self-renewal pathways in cancer cells might lead to improved outcomes in the treatment of these diseases.


Since the discovery of oncogenes (McBride et al., 1982; Parada et al., 1982; Varmus and Bishop, 1986), our knowledge of the genetic events that lead to cancer has advanced rapidly. However, our understanding of the cellular consequences of these mutations in the organs in which cancers arise has lagged. Unraveling the underlying cellular biology of these organs coupled with a detailed analysis of the effects of oncogenic mutations on development should help us better determine how the resultant alterations in both the stem and progenitor cells' molecular signaling pathways leads to tumor formation.

Epithelial cancers such as cancer of the colon, breast, lung and prostate are the most common cancers in adults. In each of these tissues, the mature cells of the tissue are thought to be constantly replenished by a minority population of tissue stem cells (Spangrude et al., 1988; Reya et al., 2001; Terskikh et al., 2001). These stem cells give rise to a rapidly dividing population of progenitor cells, sometimes called transiently amplifying cells, which finally give rise to the mature epithelial cells in the tissue. In most tissues, the only long-lived cells are the stem cells, whereas other cells typically have a lifespan measured in days or weeks.

Recent evidence has demonstrated that cancers can be viewed as an abnormal organ in which tumor growth is driven by a population of cancer stem cells (CSCs), which can give rise to both more CSCs as well as nontumorigenic cancer cells (Lapidot et al., 1994; Akashi et al., 2003; Al-Hajj et al., 2003; Hemmati et al., 2003; Singh et al., 2003; Kondo et al., 2004; Matsui et al., 2004; Setoguchi et al., 2004). In marked contrast to the CSCs, these latter cells have either no or a markedly diminished capacity to form new tumors (Lapidot et al., 1994; Al-Hajj et al., 2003; Hemmati et al., 2003; Singh et al., 2003). This observation has implications for the biology of tumor formation as well as the diagnosis and treatment of cancer. To treat cancer effectively, the CSCs must be eliminated. Otherwise, the tumor will rapidly reform if the therapy eliminates nontumorigenic cancer cells but spares a significant population of the CSCs.

Normal stem cells

The stem cells in different tissues share two common properties, the ability to self-renew, for example, to divide and form at least one new stem cell, as well as to differentiate into the mature cells of the organ in which it resides. Although some studies suggested that plasticity allowed stem cells from different tissues such as the brain or blood system to transdifferentiate and form mature cells of many different tissues, it is now clear that such plasticity is frequently the result of a rare fusion of the stem cell or its progeny with a cell of another organ (Wagers et al., 2002; Vassilopoulos et al., 2003; Wang et al., 2003). The ability of stem cells to expand in number is under tight genetic constraints (Phillips et al., 1992; Phillips et al., 2000; Morrison et al., 2002). This is not surprising since unlimited stem cell expansion, coupled with the ability of the stem cells to enter the circulation (essentially metastasize) (Wagers et al., 2002), would result in a cell with a phenotype similar to that of a cancer cell. All that would be lacking would be the property of tissue invasion.

Cancer stem cells

Solid tumors consist of a mixed populations of cancer cells that often form abnormal structures reminiscent of their normal counterparts (Heppner, 1984; Nowell, 1986). The different cell populations could arise, in part, from sequential mutations occurring due to either genetic instability and/or environmental factors. An alternative explanation for this heterogeneity is that a tumor is essentially an aberrant organ containing a tumorigenic (stem cell) population that drives tumor growth and forms both the CSCs as well as their nontumorigenic progeny (Southam and Brunschwig, 1961; Wodinsky et al., 1967; Bergsagel and Valeriote, 1968; Hamburger and Salmon, 1977; Ozols et al., 1980; Weisenthal and Lippman, 1985). Although both the stem cells and the nontumorigenic populations of cells contain the oncogenic mutations that result in cancer, the latter population would lack the ability to self-renew. Several clinical observations suggest that the stem model accounts for much of the cellular heterogeneity seen in tumors, although genetic instability and environmental factors undoubtedly also contribute to the variability in phenotypes (Aubele and Werner, 1999). Many types of cancer contain heterogeneous populations of cells that variably express differentiation markers that reflect the tissue or origin, as well as cancer cells that have an immature morphology that do not express these markers (Fidler and Kripke, 1977; Fidler and Hart, 1982; Heppner, 1984). For example, X-linked inactivation studies suggested that the mature erythrocytes, platelets and granulocytes are often clonally derived from leukemia cells in patients with acute myeloid leukemia (AML) and chronic myeloid leukemia (CML) (Fialkow, 1976a, 1976b, 1990; Douer et al., 1981). Frequently in teratocarcinomas, only a minority of the cancer cells express immature cell markers such as β-human chorionic gonadotropin (β-HCG), whereas some of the cancer cells form teeth and hair. In leukemia, host red blood cells and granulocytes are often derived from the malignant clone. As these terminally differentiated cancer cells are unlikely to be able to proliferate and form new tumors, it is likely that the minority population of immature leukemia cells or α-fetoprotein expressing germ cell cancer cells have the exclusive ability to form new tumors consisting of more tumorigenic cells as well as the phenotypically diverse populations of abnormally differentiated cells lacking the ability to self-renew. If one viewed the cancer as an organ, these immature cells could be considered CSCs.

Identification of CSCs

The findings that only a minority of the tumor cells found in either hematopoietic malignancies or solid cancers are clonogenic when placed in tissue culture or injected into immunodeficient mice led investigators to postulate the existence of tumor stem cells (Southam and Brunschwig, 1961; Wodinsky et al., 1967; Bergsagel and Valeriote, 1968; Park et al., 1971; Salsbury, 1975; Heppner, 1984; Henrique et al., 1997). However, an alternative explanation for these observations was that all cancer cells have the intrinsic ability to proliferate extensively, but only some do so in a particular assay (Weisenthal and Lippman, 1985). The techniques of stem cell biology can be used to distinguish between these possibilities (Spangrude et al., 1988; Uchida et al., 2000; Akashi and Weissman, 2001). To prove that a phenotypically distinct population of cancer cells was solely responsible for perpetuating the disease requires isolating different populations of cancer cells and demonstrating that one or more groups were enriched for the ability to initiate the disease while other populations lacked this ability. This was first shown in AML when a leukemia tumor-initiating subpopulation of cells was prospectively identified and purified from multiple patients' bone marrows (Lapidot et al., 1994; Bonnet and Dick, 1997; see also Warner et al., in this issue). Recent evidence suggests that multiple myeloma may also have a stem cell population. The majority of the myeloma cells express CD138, a mature plasma cell antigen. However, a minority population of CD138- myeloma cells preferentially engrafted immunodeficient mice and were able to generate CD138+ cells the same light chain restriction as the original tumor cells present in the patient's tumor (Matsui et al., 2004).

Recently, tumorigenic and nontumorigenic subsets of cancer cells have been isolated from human breast cancer tumors, providing the first direct evidence for CSCs in solid tumors. To assay the tumorigenic cancer cells, a xenograft model for human breast cancer was developed that allowed breast cancer tumors isolated directly from patients to be passaged reliably in vivo. In this model, only a subset of the breast cancer cells had the ability to form new tumors (Al-Hajj et al., 2003). In cancer cells isolated from most patients' tumors, tumorigenic cells could be distinguished from nontumorigenic cancer cells based upon surface marker expression. In eight out of nine patients, tumorigenic cells could be prospectively identified and isolated by flow cytometry as CD44+CD24−/lowLineage cells (Al-Hajj et al., 2003). Limiting-dilution assays demonstrated that as few as one hundred tumorigenic cancer cells were able to form tumors, while tens of thousands of the other populations of cancer cells failed to form tumors in NOD/SCID mice. These tumorigenic cells have been serially passaged, and each time cells within this population generated new tumors containing additional CD44+CD24−/lowLineage tumorigenic cells as well as the phenotypically mixed populations of nontumorigenic cancer cells. Importantly, the phenotypic distribution of cells closely resembled that of the original tumor. These data demonstrate the presence of a hierarchy of cells within a breast cancer tumor in which only a fraction of the cells have the ability to generate a new tumor, which contains similar populations of tumorigenic and nontumorigenic cancer cells, suggesting that the tumorigenic cells can both generate both populations of cells. Thus, tumorigenic breast cancer cells from most tumors appear to exhibit the properties of CSCs. However, before these cells can definitively be called CSCs, new assays such as retroviral marking will be necessary to demonstrate that a single transplanted cell gives rise to all of the diverse populations of cancer cells within a tumor. For example, there might be populations of self-renewing cancer cells that are restricted with respect to the phenotypic populations of nontumorigenic cancer cells that they can produce.

Using culture conditions similar to that used to culture normal neuronal stem cells as ‘neurospheres’, two groups have now shown that pediatric tumors of neuronal origin contain a minority population of cancer cells that are clonogenic in vitro (Hemmati et al., 2003; Singh et al., 2003). The number of clonogenic cells in the tumors able to form spheres ranged from 1 to 25% of the cancer cells. Individual cells from dissociated spheres were able to form new spheres during serial passage in tissue culture, demonstrating that the cells could self-renew. When cultured cells were isolated based on the expression of CD133, a marker expressed by normal CNS stem cells (Uchida et al., 2000), only the CD133+ fraction of cells was capable of forming spheres. These studies suggest that CNS tumors of neural origin contain a stem cell population, but these conclusions need to be confirmed with studies demonstrating that the CD133+ cells isolated directly from patients' tumors are tumorigenic in an animal model. More details on this issue are provided by Singh et al. (in this issue).

Molecular regulation of self-renewal

Proliferation and self-renewal are not synonomous. Self-renewal is a unique cell division in which the capacity of one or both progeny to proliferate and differentiate is similar to those of the parental cell. Although a committed progenitor cell might have an extensive ability to proliferate, it is destined to eventually become terminally differentiated and stop dividing. For example, committed multipotent hematopoietic progenitor cells can give rise to mature blood elements for up to 2 months time (Morrison and Weissman, 1994). However, with each cell division, the progenitor cell's progeny become progressively more differentiated and their proliferative capacity diminishes. On the other hand, a self-renewing cell division of a hematopoietic stem cell (HSC) results in a cell that maintains its proliferative capacity and can reconstitute the blood system for the life of an animal (Spangrude et al., 1988; Morrison and Weissman, 1994). Indeed, a single HSC or a progeny that arose from a self-renewing cell division can be serially transplanted several times and restore blood production in lethally irradiated animals.

Most tumors develop over a period of months to years and like normal tissues consist of heterogeneous populations of cells. In previous models of cancer, the unregulated growth of tumors was attributed to the serial acquisition of genetic events that resulted in: turning on genes promoting proliferation, silencing genes involved in inhibiting proliferation and circumventing genes involved in programmed cell death. In the stem cell model for cancer, another key event in tumorigenesis is the disruption of genes involved in the regulation of stem cell self-renewal. Thus, some of the cancer cells within a tumor share with normal stem cells the ability to replicate without losing the capacity to proliferate.

It is not surprising then that a number of genes initially identified as oncogenes have been implicated in normal stem cell self-renewal decisions. Genes that have been demonstrated to be involved in regulation of self-renewal in normal stem cells from many tissues include Bmi-1, Notch, Wnt and Shh (Austin et al., 1997; Spink et al., 2000; Bhardwaj et al., 2001; Taipale and Beachy, 2001; Calvi et al., 2003; Lemischka and Moore, 2003; Lessard and Sauvageau, 2003; Reya et al., 2003; Zhang et al., 2003). All of these genes were initially identified for their roles in tumor formation. Bmi-1 is a member of the polycomb family that functions to repress the transcription of its target genes via an epigenetic mechanism (Hanson et al., 1999) and in a mouse model of leukemia, Lessard and Sauvageau (2003) showed that Bmi-1 is also needed for the self-renewal of the leukemia initiating cell (LIC).

Since both CSCs and their nontumorigenic progeny share the same mutations that drive tumor formation, it is likely that epigenetic events are responsible for the generation of at least some of the nontumorigenic cancer cells. Supporting this notion are studies that have examined the ability of oocytes to reprogram the nuclei of cancer cells. When nuclei obtained from medulloblastoma tumor cells arising in Ptc1 heterozygous mice were transfered into enucleated oocytes, cells from the resulting blastocysts were unable to form tumors when injected into mice suggesting that the oocytes were able to reprogram cancer cell nuclei via an epigenetic mechanism and suppress their tumorigenicity (Li et al., 2003).

Implications of the stem cell model for cancer

There are major implications for the way we study, diagnosis and treat cancer. If the same populations of cancer cells that are tumorigenic the xenograft model are also tumorigenic in humans, then this must be taken into account when diagnosing and treating cancer. Current therapeutic strategies fail to account for potential differences in drug sensitivity or target expression between the CSCs and their nontumorigenic progeny. Differential sensitivity to treatments between the CSCs and nontumorigenic populations may account for the inability of our current armamentarium of drugs to consistently eradicate solid tumors other than testicular cancer.

Classically, treatments for cancer have relied on the ability to shrink tumors. Since in many cases the CSCs represent a minority cell population of the tumor (Lapidot et al., 1994; Al-Hajj et al., 2003; Singh et al., 2003; Matsui et al., 2004), agents selectively killing the CSCs are likely overlooked in our current screening methods, which rely on rapid reduction of tumor size. If an agent spares a significant number of the CSCs, then the remaining cells could rapidly reform the tumor. Supporting this hypothesis, recent studies demonstrated that CD34+CD38− leukemic cells were significantly less sensitive to daunorubicin or ctyarabine than the bulk population of leukemia blast cells (Costello et al., 2000; Guzman et al., 2002).

In addition to its impact on our understanding of the efficacy of our current therapies, the stem cell model for cancer is likely to have an impact on the identification of future therapeutic targets. The gene expression patterns of normal stem cells and their more differentiated progeny can differ significantly (Akashi et al., 2003). DNA and tissue microarrays of tumors to date have focused on either the entire tumor or isolated cancer cells that will make it difficult to de-convolute the gene expression pattern of the CSCs. The identification of novel diagnostic markers and novel therapeutic targets should be simplified by focusing expression studies on the CSCs. Such a strategy successfully identified the role of Bmi-1 in the self-renewal of normal stem cells (Park et al., 2003).

The ability to prospectively identify CSCs has implications for the design of future studies aimed at improving our ability to identify individuals at risk for relapse. In some patients with early stage breast cancer, disseminated cytokeratin-positive breast cancer cells can be detected in the bone marrow of patients that never relapse (DiStefano et al., 1979; Braun and Pantel, 1998). It has been thought that in these patients, the cancer cells lie dormant until some unknown event triggers them to renew proliferation. Alternatively, it is possible that the disseminated cancer cells in this group of patients arose from the spread of nontumorigenic cells, and only when CSCs disseminate and subsequently self-renew will patients relapse with macroscopic metastases. Diagnostic reagents that easily identify CSCs in the blood or the bone marrow could allow us to predict which patients will develop metastatic disease. This could allow clinicians to treat only those patients most likely to benefit from adjuvant therapy and thus spare many patients the severe systemic toxicities often associated with cytotoxic drugs.

Future questions

Do CSCs arise from normal stem cells or normal progenitor cells?

Self-renewal is essential for maintenance of both CSCs and normal stem cells. If CSCs are derived from normal stem cells, then the cancer cells can utilize the already functional self-renewal pathways active in these cells. On the other hand, if the CSCs are derived from progenitor cells, then oncogenic mutations must include reactivation of self-renewal pathways in the malignant cells. In the majority of cases of AML, the phenotype of the leukemic stem cells is CD34+CD38. The phenotype of the normal HSC is Thy1+ CD34+CD38 (Uchida and Weissman, 1992; Lapidot et al., 1994; Bhatia et al., 1997; Bonnet and Dick, 1997; Miyamoto et al., 2000; George et al., 2001). This led to the speculation that the leukemic stem cells arose directly from their normal counterparts. However, since LIC lack the Thy 1 expression characteristic of normal HSCs, it is possible that while the early mutations occurred in the HSC, the final transforming mutations occurred in early downstream progenitors lacking Thy 1 expression. Alternatively, the Thy 1 expression could have been lost as a consequence of neoplastic tansformation (Miyamoto et al., 2000). In human AML, X-linked gene inactivation studies suggested that in some cases of AML the LIC appeared to involve a multipotent cell, whereas in others the differentiation potential of the LIC was restricted to the granulocyte/macrophage lineage (Fialkow, 1990). Thus, it appeared that the leukemic stem cells might be derived from a multipotent cell in some patients, and committed progenitor cell in others. Recently, it was shown in a mouse model of leukemia that the MLL/ENL oncogene, which causes human leukemia, was able to transform both normal HSCs and committed progenitor cells (Cozzio et al., 2003). Interestingly, the phenotype of the resultant leukemia was identical in each case. Taken together, these results suggest in leukemia, the leukemic stem cells can be derived either from transformed HSCs, or progenitor cells that have gained the ability to self-renew (Figure 1).

Figure 1

Cancer stem cells origin and their ability to self-renew. Unlike cancer stem cells derived from progenitor cells that require oncogenic mutations to re-activate self-renewal pathways, a cancer stem cell derived from a transformed normal stem cell can make use of the already active self-renewal machinery

What is the developmental hierarchy of breast epithelial tissue?

The ability to prospectively identify HSCs and the different progenitor cells in the developmental hierarchy enables the isolation of specific lineages to determine the functional properties of each type of cell. Similarly, the isolation of CNS neuronal stem cells and neural crest stem cells has enabled the identification of molecular pathways involved in their self-renewal and differentiation (Morrison et al., 2000; Uchida et al., 2000; Hitoshi et al., 2002; Ivanova et al., 2002; Hemmati et al., 2003; Molofsky et al., 2003). Thus, the prospective identification of the breast epithelial stem and progenitor cell populations should result in a much clearer understanding of processes such as pathways that regulate the self-renewal of breast epithelial cells or the role of estrogen/progesterone receptors in the biology of normal breast development and breast cancer.

There are two possible hierarchies of breast development (Figure 2). In the first model, there are two populations of self-renewing cells in the breast, myopethelial stem cells and epithelial stem cells. In the second model, the self-renewing stem cell is the precursor to all three populations. Identifying the correct model is important because prior to frank cancer formation, transforming mutations are likely to accumulate in any self-renewing cell population. These cells should be the focus of prevention interventions.

Figure 2

Models of breast development hierarchy. Identification and purification of the different mammary ductal epithelium stem and progenitor populations will result in a better understanding of breast cancer development. Model 1 depicts two different stem cells, both possessing the ability to self-renew, that can give rise to cells of the epithelial and myoepithelial lineages, respectively. In the second model, a single self-renewing stem cell gives rise to all three populations of cells within the duct

Most evidence to date suggests that the second model is correct. Two groups have reported that ‘side population’ cells that exclude Hoechst 33342 can reconstitute the breast in the mouse (Welm et al., 2002; Alvi et al., 2003). Retroviral marking studies have demonstrated that a single cell could contribute to the breast epithelium after serial transplantation (Chepko and Smith, 1999; Smith and Boulanger, 2003). However, whether a single marked cell can give rise to all three lineages awaits analysis that conclusively proves all three lineages arose from the marked cells. Tissue culture studies suggest that is likely to be the case. Colonies arising from human MUC-1-to +/−/CALLA+/− to +/ESA+ breast cells gave rise to both myopethelial cells and epithelial cells in tissue culture and an ESA+MUC1− cell line established by immortalizing cultured mammary cells with the human papilloma virus E6/E7 gene could also produce both cell types (Stingl et al., 1998; Gudjonsson et al., 2002). When breast cells are placed in tissue culture, spheres are formed that contain cells of all three ductal lineages (Dontu et al., 2003). Although these studies strongly suggest that there is a breast stem/progenitor cell with tri-lineage differentiation potential, it is possible that in the organism, the cells do not have the same differentiation capacities seen in tissue culture. Again, definitive proof awaits demonstration via a single-cell analysis that a single normal, uncultured cell can self-renew and give rise to each of the differentiated populations.

Does cancer progression result in an expansion of self-renewing cancer cells?

The CSC model could help to explain the apparent increased virulence of some cancers over time. Cancer arises after the accumulation of multiple mutations that eventually leads to malignant transformation of a cell (Knudson et al., 1973; Land et al., 1983; Fearon and Vogelstein, 1990) and many different combinations of mutations can lead to cancer. During the progression of a cancer it appears that additional mutations in some of the cancer cells can lead to an increased tempo of the progression of the disease. Since the CSCs likely play a central role in tumor growth and spread, it follows that at least some mutations that lead to rapidly growing, quickly spreading tumors might cause a significant expansion of self-renewing cells when compared to their less aggressive counterparts (Figure 3a–c). There are multiple ways in which this could occur. For example, a particular mutation could favor cell divisions that produce more stem cells over those that produce nontumorigenic cancer cells (Figure 3b). Another possibility is that a mutation could confer self-renewal capacity to a non-self-renewing cancer progenitor cell population that originally lacked this capability (Figure 3c). In each of these latter two cases, the expansion of the self-renewing cell populations would undoubtedly lead to a more aggressive cancer.

Figure 3

Any event that results in the expansion of the pool of self-renewing cells in a tumor will likely make the tumor more aggressive. (a) An oncogenic mutation results in a limited expansion of the stem and progenitor cell pools. Some aberrantly differentiated cells are derived from the cancer stem cell. (b) A second mutation that favors symmetric self-renewing stem cell division will result in a faster growing tumor. If a third mutation occurs that confers self-renewal ability on cancer progenitor cells in the tumor that formerly lacked this ability, then there would be a substantial increase in the self-renewing cancer cells within the tumor, and, hence, the most aggressive cancer (c)

Although more research is needed to conclusively prove that the CSC content of a tumor has prognostic significance, there is evidence that this may be the case. In eight of nine patients examined, the phenotype of the breast CSCs was CD44+CD24−/low and the CSCs represented a minority population (Al-Hajj et al., 2003). However, in one patient with a Comedo-type adenocarcinoma of the breast, a very aggressive form of breast cancer, tumorigenic cancer cell populations were found in both the CD44+CD24−/low and the CD44+CD24+ fractions (Al-Hajj et al., 2003). In this cancer, more than 66% of the cells were contained in the tumorigenic cell fraction (Al-Hajj et al., 2003). If the CD44+CD24+ and CD44+CD24−/low cells represent different stages of differentiation, then these data imply that in this comedo carcinoma one or more cell populations have gained the ability to self-renew leading to a more aggressive cancer with an expanded stem cell population. Furthermore, the tumor with the smallest CSC population is also one of the slowest growing in the immunodeficient mouse model (M Al-Hajj and M Clarke, unpublished data). The number of clonogenic cells found in pediatric tumors varied markedly, ranging from 1% of the cancer cells in pilocystic astrocytomas to 25% of the cancer cells in meduloblastomas (Singh et al., 2003). Since meduloblastomas are generally more aggressive tumors than astrocytomas, this again suggests that increased CSC content in a tumor has poor prognostic significance. Finally, in teratocarcinoma, patients' tumors occasionally contain not only germ cell cancer cells but also nests of sarcoma cancer cells (Ulbright et al., 1984, 1985; Kesler et al., 1999). The germ cell elements are often cured by platinum-based chemotherapy, but the sarcoma cells remain after treatment (Kesler et al., 1999). It is not unusual for patients to relapse years later with distant metastases that contain only the sarcoma cells (Kesler et al., 1999). This suggests that the original tumor contained two different CSC compartments, the germ cell CSCs and the sarcoma stem cells, as well as terminally differentiated teratoma cells without metastatic potential.

Do oncogenic mutations affect CSCs, nontumorigenic progeny, or both?

Currently, there is an intense effort to develop therapies that target molecular pathways involved in cancer development. However, the effects of mutations can be manifested in the CSCs, their nontumorigenic progeny, or both (Figure 4a and b). For example, a particular mutation could be present in the CSCs but its ability to cause cancer could be secondary to an expansion of the non-self-renewing progeny of the CSCs (Figure 4b). In such a situation, therapies that target this pathway would not eliminate the CSC population and would not be curative. There is evidence in CML that this scenario occurs in some cancers. The drug Gleevec® targets the ATP-binding domain of the Abl kinase present in all cases of CML. Most patients treated with Gleevec® respond to the drug and have a virtually complete cytogenetic responses (Druker et al., 1996; Mauro and Druker, 2001). However, In the majority of patients RT–PCR detects the fusion transcript in the patients' bone marrow cells suggesting that this therapy may not be curative (Paterson et al., 2003). In some patients in remission and negative for the translocation by fluorescent in situ hybridization (FISH) on unfractionated bone marrow, Bhatia et al. (2003) demonstrated the presence of the BCR/ABL fusion by FISH analysis in the between 6.5 and 13% of their CD34+ progenitor cells. Since isolation of CD34+ cells was likely to have enriched for the CML stem cells, this suggests that Gleevec® eliminates CML progenitor cells while sparing the CML stem cells. This could be as a result of the absence of target mRNA expression by the abnormal stem cell population. Alternatively, the BCR-ABL/ATP binding domain signaling pathway might be required for the survival of progenitor cells, but not the CML stem cells. These studies highlight the importance of understanding the effect of the particular pathway being targeted on crucial CSC functions such as survival and self-renewal. Such knowledge should help in the effort to discover more effective therapies.

Figure 4

Oncogenic mutation could affect either the cancer stem cells or the nontumorigenic progeny. (a) An oncogenic mutation results in a block in differentiation of the cancer stem cells and an expansion of the cancer stem cell pool. Thus, the mutation's effects are in the cancer stem cell population. (b) A mutation occurs in the cancer stem cells, but it causes increased tumor size by expanding a non-self-renewing cancer cell population. (c) Effective cancer treatments are only possible when therapies target the cancer stem cells. Top panel: If a particular mutation is required for an essential function in cancer stem cells, then drugs that target the mutation would eliminate the cancer stem cells and thus have curative potential. Bottom panel: If a particular mutation is results in expansion of non-self-renewing cancer cells but does not have an essential function in cancer stem cells, then drugs that target the mutation would eliminate the non-self-renewing cells, but not the cancer stem cells. In this case, the tumor would shrink, but it would return when the therapy is stopped

Can we develop therapeutics that target CSCs while sparing normal stem cells?

There has been steady but slow progress in developing curative cancer therapies. Notably, many patients with malignancies derived from hematopoietic cells such as large cell lymphoma and childhood acute lymphocytic leukemia (ALL) can be cured with chemotherapy. Unfortunately, with the exception of testicular cancer, patients with solid tumors are rarely cured after the cancer has metastasized. In most solid tumors, chemotherapy can often initially shrink a tumor, but these effects are usually short-lived and the tumors are destined to return. Thus, new agents are desperately needed to treat these diseases. If the same cells are responsible for tumor growth in the patients and the immunodeficient mice, then our present drugs must initially shrink a tumor but residual CSCs rapidly reform the tumor (Figure 4c). However, if an agent were to effectively eliminate the CSCs, then even if some nontumorigenic cells were spared, patients would likely be cured (Figure 4c). Testicular cancer again provides clinical evidence that this is the case. When patients with testicular cancer are treated with platinum-based chemotherapy, they are often left with residual masses. When these masses are resected, if there are any immature cancer cells in the tumor, patients must be treated with more chemotherapy or they have a substantial chance of relapsing (Williams et al., 1987; Donohue et al., 1994). On the other hand, if the tumor contains only mature teratoma, such patients do not require further therapy and they are often cured (Donohue et al., 1994).

The observation that a self-renewing cancer cell population drives tumor formation suggests that identification of agents that specifically inhibit CSC self-renewal or survival while sparing normal stem cells and other essential normal tissues would result in more effective and less toxic treatments. Indeed, several of our current therapies probably function in this manner. Retinoic acid compounds such as ATRA are quite active in the treatment of acute promyelocytic leukemia (APL) (Kogan et al., 2001; Ohno et al., 2003; Ravandi et al., 2004). These agents probably function by inhibiting the self-renewal of APL stem cells and inducing them to differentiate into terminally differentiated progeny (Breitman and Gallo, 1981). Sufficient normal HSCs are not affected by these agents and are able to repopulate the bone marrow and to restore normal hematopoiesis (Ohno et al., 2003). Similarly, Platinum-based chemotherapy eliminates the germ cell CSCs while sparing sufficient normal stem cells, even sufficient gonadal germ cells to allow maintenance of fertility in many patients (Drasga et al., 1983). In multiple myeloma, the minority clonogenic CD138- blood cells, but not CD138+ cells, express CD20. When myeloma cells were exposed to rituximab, a therapeutic antibody that targets CD20, clongenicity was markedly reduced (Matsui et al., 2004). Together, these observations suggest that understanding the molecular pathways that regulate self-renewal of normal and CSCs is critical. It may be possible to target the differences in self-renewal pathways in CSCs in tumors that are presently hard to treat so that the self-renewal of these cells might be selectively inhibited.

Conflict of interest statement

I am aware that the University of Michigan has applied for a patent for the isolation and use of cancer stem cells. This patent has been licensed to a company in which the University and Dr Clarke have a financial interest.



acute myeloid leukemia


hematopoietic stem cell


cancer stem cell


leukemia initiating cell


chronic myeloid leukemia






fluorescent in situ hybridization


  1. Akashi K, He X, Chen J, Iwasaki H, Niu C, Steenhard B, Zhang J, Haug J and Li L . (2003). Blood, 101, 383–389.

  2. Akashi K and Weissman IL . (2001). Developmental Biology of Hematopoiesis. Oxford University Press: New York.

  3. Al-Hajj M, Wicha MS, Benito-Hernandez A, Morrison SJ and Clarke MF . (2003). Proc. Natl. Acad. Sci. USA, 100, 3983–3988.

  4. Alvi AJ, Clayton H, Joshi C, Enver T, Ashworth A, Vivanco MM, Dale TC and Smalley MJ . (2003). Breast Cancer Res., 5, R1–R8.

  5. Aubele M and Werner M . (1999). Anal. Cell Pathol., 19, 53–58.

  6. Austin TW, Solar GP, Ziegler FC, Liem L and Matthews W . (1997). Blood, 89, 3624–3635.

  7. Bergsagel DE and Valeriote FA . (1968). Cancer Res., 28, 2187–2196.

  8. Bhardwaj G, Murdoch B, Wu D, Baker DP, Williams KP, Chadwick K, Ling LE, Karanu FN and Bhatia M . (2001). Nat. Immunol., 2, 172–180.

  9. Bhatia M, Wang JC, Kapp U, Bonnet D and Dick JE . (1997). Proc. Natl. Acad. Sci. USA, 94, 5320–5325.

  10. Bhatia R, Holtz M, Niu N, Gray R, Snyder DS, Sawyers CL, Arber DA, Slovak ML and Forman SJ . (2003). Blood, 101, 4701–4707.

  11. Bonnet D and Dick J . (1997). Nat. Med., 3, 730–737.

  12. Braun S and Pantel K . (1998). Breast Cancer Res. Treat., 52, 201–216.

  13. Breitman TR and Gallo RC . (1981). Blood Cells, 7, 79–89.

  14. Calvi LM, Adams GB, Weibrecht KW, Weber JM, Olson DP, Knight MC, Martin RP, Schipani E, Divieti P, Bringhurst FR, Milner LA, Kronenberg HM and Scadden DT . (2003). Nature, 425, 841–846.

  15. Chepko G and Smith GH . (1999). J. Mammary Gland Biol. Neoplasia, 4, 35–52.

  16. Costello RT, Mallet F, Gaugler B, Sainty D, Arnoulet C, Gastaut JA and Olive D . (2000). Cancer Res., 60, 4403–4411.

  17. Cozzio A, Passegue E, Ayton PM, Karsunky H, Cleary ML and Weissman IL . (2003). Genes Dev., 17, 3029–3035.

  18. DiStefano A, Tashima CK, Yap HY and Hortobagyi GN . (1979). Cancer, 44, 196–198.

  19. Donohue JP, Fox EP, Williams SD, Loehrer PJ, Ulbright TM, Einhorn LH and Weathers TD . (1994). World J. Urol., 12, 190–195.

  20. Dontu G, Abdallah WM, Foley JM, Jackson KW, Clarke MF, Kawamura MJ and Wicha MS . (2003). Genes Dev., 17, 1253–1270.

  21. Douer D, Levin AM, Sparkes RS, Fabian I, Sparkes M, Cline MJ and Koeffler HP . (1981). Br. J. Haematol., 49, 615–619.

  22. Drasga RE, Einhorn LH, Williams SD, Patel DN and Stevens EE . (1983). J. Clin. Oncol., 1, 179–183.

  23. Druker BJ, Tamura S, Buchdunger E, Ohno S, Segal GM, Fanning S, Zimmermann J and Lydon NB . (1996). Nat. Med., 2, 561–566.

  24. Fearon ER and Vogelstein B . (1990). Cell, 61, 759–767.

  25. Fialkow PJ . (1976a). Biochim. Biophys. Acta., 458, 283–321.

  26. Fialkow PJ . (1976b). Birth Defects Orig. Artic. Ser., 12, 123–132.

  27. Fialkow PJ . (1990). Verh. Dtsch. Ges. Pathol., 74, 43–47.

  28. Fidler IJ and Hart IR . (1982). Science, 217, 998–1003.

  29. Fidler IJ and Kripke ML . (1977). Science, 197, 893–895.

  30. George AA, Franklin J, Kerkof K, Shah AJ, Price M, Tsark E, Bockstoce D, Yao D, Hart N, Carcich S, Parkman R, Crooks GM and Weinberg K . (2001). Blood, 97, 3925–3930.

  31. Gudjonsson T, Villadsen R, Nielsen HL, Ronnov-Jessen L, Bissell MJ and Petersen OW . (2002). Genes Dev., 16, 693–706.

  32. Guzman ML, Swiderski CF, Howard DS, Grimes BA, Rossi RM, Szilvassy SJ and Jordan CT . (2002). Proc. Natl. Acad. Sci. USA, 99, 16220–16225.

  33. Hamburger AW and Salmon SE . (1977). Science, 197, 461–463.

  34. Hanson RD, Hess JL, Yu BD, Ernst P, van Lohuizen M, Berns A, van der Lugt NM, Shashikant CS, Ruddle FH, Seto M and Korsmeyer SJ . (1999). Proc. Natl. Acad. Sci. USA, 96, 14372–14377.

  35. Hemmati HD, Nakano I, Lazareff JA, Masterman-Smith M, Geschwind DH, Bronner-Fraser M and Kornblum HI . (2003). Proc. Natl. Acad. Sci. USA, 100, 15178–15183.

  36. Henrique D, Hirsinger E, Adam J, Le Roux I, Pourquie O, Ish-Horowicz D and Lewis J . (1997). Curr. Biol., 7, 661–670.

  37. Heppner GH . (1984). Cancer Res., 44, 2259–2265.

  38. Hitoshi S, Alexson T, Tropepe V, Donoviel D, Elia AJ, Nye JS, Conlon RA, Mak TW, Bernstein A and van der Kooy D . (2002). Genes Dev., 16, 846–858.

  39. Ivanova NB, Dimos JT, Schaniel C, Hackney JA, Moore KA and Lemischka IR . (2002). Science, 298, 601–604.

  40. Kesler KA, Rieger KM, Ganjoo KN, Sharma M, Fineberg NS, Einhorn LH and Brown JW . (1999). J. Thorac. Cardiovasc. Surg., 118, 692–700.

  41. Knudson Jr AG, Strong LC and Anderson DE . (1973). Prog. Med. Genet., 9, 113–158.

  42. Kogan SC, Brown DE, Shultz DB, Truong BT, Lallemand-Breitenbach V, Guillemin MC, Lagasse E, Weissman IL and Bishop JM . (2001). J. Exp. Med., 193, 531–543.

  43. Kondo T, Setoguchi T and Taga T . (2004). Proc. Natl. Acad. Sci. USA, 101, 781–786.

  44. Land H, Parada LF and Weinberg RA . (1983). Nature, 304, 596–602.

  45. Lapidot T, Sirard C, Vormoor J, Murdoch B, Hoang T, Caceres-Cortes J, Minden M, Paterson B, Caligiuri M and Dick J . (1994). Nature, 17, 645–648.

  46. Lemischka IR and Moore KA . (2003). Nature, 425, 778–779.

  47. Lessard J and Sauvageau G . (2003). Nature, 423, 255–260.

  48. Li L, Connelly MC, Wetmore C, Curran T and Morgan JI . (2003). Cancer Res., 63, 2733–2736.

  49. Matsui W, Huff CA, Wang Q, Malehorn MT, Barber J, Tanhehco Y, Smith BD, Civin CI and Jones RJ . (2004). Blood, 103, 2332–2336.

  50. Mauro MJ and Druker BJ . (2001). Curr. Opin. Oncol., 13, 3–7.

  51. McBride OW, Swan DC, Santos E, Barbacid M, Tronick SR and Aaronson SA . (1982). Nature, 300, 773–774.

  52. Miyamoto T, Weissman IL and Akashi K . (2000). Proc. Natl. Acad. Sci. USA, 97, 7521–7526.

  53. Molofsky AV, Pardal R, Iwashita T, Park IK, Clarke MF and Morrison SJ . (2003). Nature, 425, 962–967.

  54. Morrison S, Perez S, Qiao Z, Verdi J, Hicks C, Weinmaster G and Anderson D . (2000). Cell, 101, 499–510.

  55. Morrison SJ, Qian D, Jerebek L, Thiel BA, Park I-K, Ford PS, Kiel MJ, Schork NJ, Weissman IL and Clarke MF . (2002). J. Immunol., 168, 635–642.

  56. Morrison SJ and Weissman IL . (1994). Immunity, 1, 661–673.

  57. Nowell PC . (1986). Cancer Res., 46, 2203–2207.

  58. Ohno R, Asou N and Ohnishi K . (2003). Leukemia, 17, 1454–1463.

  59. Ozols RF, Willson JK, Weltz MD, Grotzinger KR, Myers CE and Young RC . (1980). Cancer Res., 40, 4109–4112.

  60. Parada LF, Tabin CJ, Shih C and Weinberg RA . (1982). Nature, 297, 474–478.

  61. Park CH, Bergsagel DE and McCulloch EA . (1971). J. Natl. Cancer Inst., 46, 411–422.

  62. Park IK, Qian D, Kiel M, Becker MW, Pihalja M, Weissman IL, Morrison SJ and Clarke MF . (2003). Nature, 423, 302–305.

  63. Paterson SC, Smith KD, Holyoake TL and Jorgensen HG . (2003). Br. J. Cancer, 88, 983–987.

  64. Phillips R, Ernst R, Brunk B, Ivanova N, Mahan M, Deanehan J, Moore K, Overton G and Lemischka I . (2000). Science, 288, 1635–1640.

  65. Phillips RL, Reinhart AJ and Van Zant G . (1992). Proc. Natl. Acad. Sci. USA, 89, 11607–11611.

  66. Ravandi F, Kantarjian H, Giles F and Cortes J . (2004). Cancer, 100, 441–454.

  67. Reya T, Duncan AW, Ailles L, Domen J, Scherer DC, Willert K, Hintz L, Nusse R and Weissman IL . (2003). Nature, 423, 409–414.

  68. Reya T, Morrison SJ, Clarke MF and Weissman IL . (2001). Nature, 414, 105–111.

  69. Salsbury AJ . (1975). Cancer Treat. Rev., 2, 55–72.

  70. Setoguchi T, Taga T and Kondo T . (2004). Cell Cycle, 3, 414–415.

  71. Singh SK, Clarke ID, Terasaki M, Bonn VE, Hawkins C, Squire J and Dirks PB . (2003). Cancer Res., 63, 5821–5828.

  72. Smith GH and Boulanger CA . (2003). Cell Prolif., 36 (Suppl 1), 3–15.

  73. Southam C and Brunschwig A . (1961). Cancer, 14, 971–978.

  74. Spangrude GJ, Heimfeld S and Weissman IL . (1988). Science, 241, 58–62.

  75. Spink KE, Polakis P and Weis WI . (2000). EMBO J., 19, 2270–2279.

  76. Stingl J, Eaves C, Kuusk U and Emerman J . (1998). Differentiation, 63, 201–213.

  77. Taipale J and Beachy PA . (2001). Nature, 411, 349–354.

  78. Terskikh AV, Easterday MC, Li L, Hood L, Kornblum HI, Geschwind DH and Weissman IL . (2001). Proc. Natl. Acad. Sci. USA, 98, 7934–7939.

  79. Uchida N, Buck D, He D, Reitsma M, Masek M, Phan T, Tsukamoto A, Gage F and Weissman I . (2000). Proc. Natl Acad Sci USA, 97, 14720–14725.

  80. Uchida N and Weissman IL . (1992). J. Exp. Med., 175, 175–184.

  81. Ulbright TM, Clark SA and Einhorn LH . (1985). Hum. Pathol., 16, 268–272.

  82. Ulbright TM, Loehrer PJ, Roth LM, Einhorn LH, Williams SD and Clark SA . (1984). Cancer, 54, 1824–1833.

  83. Varmus H and Bishop JM . (1986). Cancer Surv., 5, 153–158.

  84. Vassilopoulos G, Wang PR and Russell DW . (2003). Nature, 422, 901–904.

  85. Wagers AJ, Sherwood RI, Christensen JL and Weissman IL . (2002). Science, 297, 2256–2259.

  86. Wang X, Willenbring H, Akkari Y, Torimaru Y, Foster M, Al-Dhalimy M, Lagasse E, Finegold M, Olson S and Grompe M . (2003). Nature, 422, 897–901.

  87. Weisenthal L and Lippman ME . (1985). Cancer Treat. Rep., 69, 615–632.

  88. Welm BE, Tepera SB, Venezia T, Graubert TA, Rosen JM and Goodell MA . (2002). Dev. Biol., 245, 42–56.

  89. Williams SD, Birch R, Einhorn LH, Irwin L, Greco FA and Loehrer PJ . (1987). N. Engl. J. Med., 316, 1435–1440.

  90. Wodinsky I, Swiniarski J and Kensler CJ . (1967). Cancer Chemother. Rep., 51, 415–421.

  91. Zhang J, Niu C, Ye L, Huang H, He X, Tong WG, Ross J, Haug J, Johnson T, Feng JQ, Harris S, Wiedemann LM, Mishina Y and Li L . (2003). Nature, 425, 836–841.

Download references

Author information

Correspondence to Michael F Clarke.

Rights and permissions

To obtain permission to re-use content from this article visit RightsLink.

About this article

Publication history

  • Published

  • Issue Date


Further reading