IJC International Journal of Cancer

Complexity of cancer stem cells Eiji Sugihara1,2 and Hideyuki Saya1,2 2

Division of Gene Regulation, Institute for Advanced Medical Research, School of Medicine, Keio University, Tokyo, Japan Japan Science and Technology Agency, CREST, Tokyo, Japan

Heterogeneity of tumor tissue has been accounted for in recent years by a hierarchy-based model in which cancer stem cells (CSCs) have the ability both to self-renew and to give rise to differentiated tumor cells and are responsible for the overall organization of a tumor. Research into CSCs has progressed rapidly and concomitantly with recent advances in the biology of normal tissue stem cells, resulting in the identification of CSCs in a wide range of human tumors. Studies of mouse models of human cancer have provided further insight into the characteristics of CSCs as well as a basis for the development of novel therapies targeted to these cells. However, recent studies have revealed complexities, such as plasticity of stem cell properties and clonal diversity of CSCs, in certain tumor types that have led to revision of the original CSC model. In this review, we summarize the history of the discovery and characterization of CSCs, as well as address recent advances that have revealed the complexity of these cells and their therapeutic implications.

Identification of Cancer Stem Cells Many tumors consist of phenotypically and functionally heterogeneous cancer cells. For many years, such heterogeneity was considered to be explained by the stochastic (clonal evolution) model1 (Fig. 1a), which is based on the notion that all cancer cells possess tumorigenic potential and can develop tumor dependent on genetic and/or epigenetic changes. However, more recent studies have suggested a new paradigm: that tumors show hierarchy, with a subpopulation of cancer cells having a tumorigenic potential much greater than that of other cancer cells.2 This subpopulation of cells at the top of the hierarchy comprises cancer stem cells (CSCs), and tumors are thus thought to manifest a hierarchical organization—consisting of stem cells, progenitors and differentiated cells—similar to that of normal tissues (Fig. 1a). A CSC has been defined3 as ‘‘a cell within a tumor that possess[es] the capacity to selfrenew and to cause the heterogeneous lineages of cancer cells Key words: cancer stem cell, cell of origin, plasticity, clonal diversity, CSC-targeted therapy Abbreviations: ALDH: aldehyde dehydrogenase; AML: acute myeloid leukemia; Ara-C: arabinosylcytosine; CML: chronic myeloid leukemia; CSC: cancer stem cell; EpCAM: epithelial cell adhesion molecule; GFP: green fluorescent protein; HSC: hematopoietic stem cell; iCSC: induced cancer stem cell; iPSC: induced pluripotent stem cell; pre-B-ALL: pre-B acute lymphoblastic leukemia Grant sponsor: Ministry of Education, Culture, Sports, Science, and Technology of Japan DOI: 10.1002/ijc.27961 History: Received 8 Jul 2012; Accepted 15 Nov 2012; Online 26 Nov 2012 Correspondence to: Hideyuki Saya, Division of Gene Regulation, Institute for Advanced Medical Research, School of Medicine, Keio University, 35 Shinano-machi, Shinjuku-ku, Tokyo 160-8582, Japan, Tel.: þ81-3-5363-3982, Fax: þ81-3-5363-3982, E-mail: [email protected]

C 2012 UICC Int. J. Cancer: 132, 1249–1259 (2013) V

that comprise the tumor.’’ Similarly, the term ‘‘tumor-initiating cell’’ has been used to describe a cell with the potential to initiate a tumor. If this term is used to refer to the subpopulation of cells within an established tumor that gives rise to a new tumor when transplanted, then tumor-initiating cells are essentially functionally equivalent to CSCs. In1963, Bruce and Van Der Gaag first documented that there is a small number of mouse lymphoma cells capable to form colonies in spleen using same transplantation method of normal hematopoietic cells.4 Although the existence of CSCs had been subsequently suggested in mouse myeloma and acute myeloid leukemia (AML) in earlier studies,5–7 human CSCs were first prospectively identified for AML in 1997 as the CD34þCD38– cell subpopulation by xenotransplantation into immunocompromised (NOD/SCID) mice.8 Since this initial discovery, CSCs have been identified by flow cytometry-based prospective analyses in a wide variety of human cancers including those of the breast (CD44þCD24–/ low cells),9 brain (CD133þ cells),10 prostate (CD44þCD24– cells),11,12 colon (CD133þ and EpCAMhiCD44þ cells)13 and pancreas (CD44þCD24þESAþ cells).14 These cell subpopulations are relatively rare and form secondary tumors that recapitulate the heterogeneity and pathology of the original tumor when transplanted into experimental animals. Although CSC-markers such as CD44 and CD133 are detected in CSCs for a broad range of tumors, tissue-specific CSC markers have been also reported. For instance, ABCB5, an ABC-transporter protein, has been shown as a CSC-marker for human melanoma and is assumed to be involved in transport of melanin as well as chemo-therapeutic agents.15,16 A cell adhesion receptor, Integrin a2b1, was reported as a marker for prostate CSCs in addition to the above markers and implicates prostate cancer progression.17,18 Some recent studies, however, have revealed that cell surface markers of CSCs are not restricted to those identified

Special Section Paper

1

Complexity of cancer stem cells

Special Section Paper

1250

Figure 1. Two models for cancer heterogeneity and methods for isolation of CSCs. (a) The stochastic model and cancer stem cell (CSC) model are the two major models to account for heterogeneity of tumors. In the stochastic model (left), all tumor cells possess tumorigenic activity and are the product of clonal evolution through the acquisition of genetic mutations and epigenetic changes. In the CSC model (right), only a subpopulation of tumor cells possesses high tumorigenic activity, with these cells representing the top of a hierarchical organization similar to that of normal tissue. CSCs (red) both undergo self-renewal and generate differentiated progeny cells (pink and yellow) that are no longer able to form tumors. (b) Four methods for the isolation and enrichment of CSCs. CSCs can be prospectively isolated by cell sorting based on a combination of cell surface markers (I). Given that they possess self-renewal activity, CSCs for leukemia form colonies in methylcellulose medium whereas those for solid tumors form spheroids in serum-free medium under nonadherent conditions (II). When tumor cells are exposed to Hoechst 33342 and subjected to ultraviolet irradiation followed by cell sorting, CSCs are enriched in a cell subpopulation corresponding to a side area (red box) of a flow cytometric plot of Hoechst blue versus red fluorescence (III). CSCs have high ALDH activity and therefore can be isolated by cell sorting on the basis of their green fluorescence signal when treated with Bodipy-conjugated aminoacetaldehyde (IV).

C 2012 UICC Int. J. Cancer: 132, 1249–1259 (2013) V

prospectively. For instance, CD133– cells in brain tumors as well as CD34þCD38þ cells in AML were also shown to possess high tumorigenic activity.19–21 These studies indicate that CSCs might not be definable simply on the basis of the expression of cell surface markers, and that methods based on such markers have limitations with regard to the identification and purification of CSCs.

Methods for Isolation of CSCs In addition to prospective cell sorting (Fig. 1bI), several methods have been developed to distinguish CSCs on the basis of their functional properties. Colony or spheroid formation assays have thus been used to enrich CSCs on the basis of their self-renewal activity (Fig. 1bII). Colony formation assays with methylcellulose medium were originally applied to analyze the self-renewal and multilineage differentiation of immature hematopoietic cells. Consecutive repetition of such assays (replating) can result in the isolation of subpopulations of hematopoietic cancer cells with self-renewal ability and also allows determination of whether genes of interest have the potential to confer selfrenewal activity in non-transformed hematopoietic cells.22,23 Spheroid formation assays were performed originally to isolate neural stem or progenitor cells and are now often used to evaluate stem-like characteristics of tumor cells. Suspension culture in medium not containing serum but supplemented with basic fibroblast growth factor and epidermal growth factor allows the isolation of tumor cells with self-renewal capacity.24 The side-population assay is often used for isolation of normal or tumor cells with the ability to exclude certain drugs (Fig. 1bIII). Similar to normal tissue stem cells, a subpopulation of tumor cells expresses drug transporters, such as ABCG2, and is thereby thought to be rendered resistant to anticancer drugs.25 In this assay, tumor cells are treated with the DNA binding dye Hoechst 33342, which is then subjected to fluorescence excitation by ultraviolet irradiation. A small cell subpopulation with no fluorescence emission can be detected in a side position on the flow cytometric plot. This ‘‘side population’’ has been shown to possess tumorigenic activity greater than that of other cells after cell sorting and transplantation.26,27 The side-population assay may therefore be useful for the identification of CSCs in some tumor types for which CSC markers have not been discovered. A method based on the enzymatic activity of aldehyde dehydrogenase (ALDH) is also widely used to identify and evaluate CSCs (Fig. 1bIV).28–30 ALDH catalyzes the irreversible conversion of retinaldehyde to retinoic acid and thereby contributes to cell proliferation and differentiation.31 High ALDH activity in CSCs is shown to be related to their resistance to chemotherapy.32 CSCs isolated from breast and ovarian cancers on the basis of both ALDH activity and previously identified CSC markers manifested higher growth and tumorigenic potential than did those isolated on the basis of cell surface markers alone.33,34 C 2012 UICC Int. J. Cancer: 132, 1249–1259 (2013) V

Although these methods have been developed based on the established methods for isolating normal stem cells, some problems are still argued. For instance, because the sidepopulation assay uses cytotoxic dye (Hoechst 33342), this method potentially causes problems to judge tumorigenicity and cell viability between cells in side population and the other population that imports this dye.35 In addition, each method cannot cover whole CSCs.36 The combination of CSC markers and functional assays might thus provide the basis for more robust and powerful strategies for the identification of bona fide CSCs.

Cells of Origin for CSCs The cells of origin for CSCs are assumed to be normal stem cells in most instances, given that the biological characteristics of normal stem cells, including the abilities to undergo self-renewal and to generate differentiated progenies, are similar to those of CSCs. In some cancer types, CSCs are enriched in cell subpopulations expressing normal stem cell markers. Furthermore, stem cells live much longer than differentiated cells, which can provide time for accumulation of mutations sufficient for transforming them. In solid tumors, the origin of CSCs has been debated because of difficulty to monitor the process from the origin to CSCs. Recently, Clevers and coworkers developed the ingenious system that allows tracing adenoma initiated from intestinal stem cell marker-positive (Lgr5þ) cells and further re-traces Lgr5þ cells in established adenoma.37 They found that low population of Lgr5þ cells in adenoma generates Lgr5þ cells and all other adenoma cell types.37 This finding experimentally proved that intestinal stem cells are the cell of origin for CSCs in intestinal adenoma. In hematopoietic tumors, studies on the origin for CSCs have been more extensively performed than those in solid tumors due to sufficient markers to isolate a variety of cell types and well-established transplantation methods. In the case of human AML, hematopoietic stem cells (HSCs) have been also considered to be the cells of origin for AML stem cells because the markers identified for the latter are identical to those for the former.8,38 On the other hand, some studies have suggested that the cells of origin for mouse AML stem cells are not only restricted to HSCs but also include progenitor cells committed to the myeloid lineage using bone marrow transplantation (BMT) model.22,39 In the case of B acute lymphoblastic leukemia (B-ALL), although CSCs were found in cell subpopulations expressing HSC markers or B cell markers,40–44 the cells of origin for B-ALL stem cells remain unclear. We recently established a Myc-induced mouse model of pre-B-ALL based on BMT model.45 We found that CSCs in this model were restricted to cells positive for B cell markers (B220þ and CD19þ). We further found that Myc-transduced HSCs showed tumorigenicity higher than that of corresponding progenitor or committed cells, suggesting that the cells of origin for our mouse model of pre-B-ALL are HSCs. Thus, despite the cells of

Special Section Paper

1251

Sugihara and Saya

1252

Complexity of cancer stem cells

Special Section Paper

cells induced from Ink4a/Arf-null progenitor B cells manifested extensive apoptosis after treatment with Nutlin-3, an inhibitor of the p53-targeting ubiquitin ligase Mdm2, whereas the tumor cells induced from wild-type HSCs did not undergo Nutlin-3-induced apoptosis, likely because of p53 mutation in these cells (Fig. 2b). These results suggested that CSCs derived from different cells of origin are differentially sensitive to Ara-C and Nutlin-3. The human gene locus of Ink4a and Arf is frequently deleted in malignant types of B-ALL such as Philadelphia chromosome-positive and recurrent ALL.47 Analysis of Ink4a and Arf expression in such types of ALL cells might thus in the future provide information relevant to drug selection at the time of diagnosis. The Mdm2 inhibitor is currently under evaluation for leukemia treatment in clinical trials. Thus, although it is often difficult to identify the origin of human CSCs, mouse cancer models have the potential to provide important insight into such cells of origin.

Plasticity in the CSC Model

Figure 2. Different cells of origin result in distinct phenotypes for pre-B-ALL stem cells. (a) HSCs and committed progenitor B cells are potential cells of origin for a Myc-induced model of pre-B-ALL. HSC-derived ALL stem cells are sensitive to Ara-C but resistant to Nutlin-3 (an Mdm2 inhibitor), whereas progenitor B cell-derived ALL stem cells are sensitive to Nutlin-3 but resistant to Ara-C. (b) Molecular mechanisms underlying the differential sensitivity of the two types of ALL stem cells to Nutlin-3. HSC-derived Myc-induced ALL stem cells constantly acquire p53 mutation, which is unable to induce apoptosis even if Nutlin-3 inhibits Mdm2. In contrast, progenitor B cell-derived ALL stem cells are null for Ink4a/Arf but express wild-type p53, with the result that stabilization of p53 by Nutlin-3 leads to the induction of apoptosis.

origin being HSCs, B cell marker-positive cells become preB-ALL stem cells during the course of the tumor development, indicating that differentiation status may differ between cells of origin and CSCs. Moreover, committed progenitor B cells were also able to serve effectively as cells of origin in this model in the absence of Ink4a and Arf (Fig. 2a), the expression of which is maintained at low levels in HSCs. It is important to identify the cells of origin for tumors in order to understand the context in which tumor cells develop, which in turn may provide useful information for preventive and therapeutic strategies in the clinical setting.46 In our mouse pre-B-ALL model, we compared the therapeutic sensitivity of CSC-enriched cell preparations derived from tumors with different cells of origin. We found that the tumor cells induced from Ink4a/Arf-null progenitor B cells were more resistant to Ara-C treatment compared with those induced from wild-type HSCs (Fig. 2a). In contrast, tumor

Similar to those for other cancers, CSCs for human melanoma are considered to constitute a rare subpopulation and have been identified by prospective isolation based on cell surface markers (including CD20, CD133, ABCG2, ABCB5 and CD271).16,48–50 However, Morrison and coworkers have questioned whether implantation of human melanoma cells into NOD/SCID mice might lead to underestimation of the frequency and tumorigenicity of isolated CSCs with defined markers, given that these animals still possess natural killer T cells even though they lack B and T cells of the adaptive immune system. These researchers therefore used NOD/ SCID/IL-2Rc-null (NSG or NOG) mice, which lack natural killer T cells and in which both the innate and adaptive immune systems are therefore compromised,51 to reevaluate the frequency of cells with tumor-initiating ability and that express various cell surface markers including previously identified CSC markers.52,53 They found that 28% of individual melanoma cells tested was able to give rise to melanoma reminiscent of the original tumor, and that cell subpopulations with any cell surface markers examined could initiate melanoma irrespective of the patient from which they were derived. These results suggested that a large proportion of melanoma cells possess tumorigenic activity and that melanoma does not therefore conform to the hierarchical CSC model.53 Around the same time, Herlyn and coworkers proposed a new concept related to the CSC model for melanoma. They focused on the histone H3 lysine-4 demethylase JARID1B, which had been found to be expressed in a small subpopulation of human malignant melanoma.54 To monitor the dynamics of melanoma cells expressing JARID1B, they manipulated the cells to express the gene for green fluorescent protein (GFP) under the control of the JARID1B promoter and then assessed the proliferation and tumorigenic activities of GFPþ and GFP– cells. They found that GFPþ C 2012 UICC Int. J. Cancer: 132, 1249–1259 (2013) V

1253

Special Section Paper

Sugihara and Saya

Figure 3. New concepts of plasticity and clonal diversity in the CSC model. (a) The dynamic stemness model is based on the conventional CSC model but introduces the concept of dynamic stemness, whereby a non-CSC can acquire stemness and become a CSC. CSCs and nonCSCs thus exhibit plasticity (reversibility) of stemness as indicated by red arrows. (b) The CSC niche is required for maintenance of stem cell properties and is considered to consist of many factors including stromal cells, blood vessels, extracellular matrix (ECM), growth factors, cytokines and hypoxia. Exposure of non-CSCs to niche factors may result in their acquisition of stem cell properties. The EMT may be one process by which non-CSCs acquire stemness. (c) A combination of the CSC model and the stochastic (clonal evolution) model has been proposed to account for clonal diversity of CSCs. Each CSC clone is thought to evolve through the acquisition of genetic mutations. Phenotypically and functionally distinct major clones and minor clones may exist in a tumor. Each clone is organized into a hierarchical structure.

(JARID1Bþ) cells cycled only slowly, possessed a high selfrenewal potential, generated GFP– (JARID1B–) cells and were enriched in the side population after staining with Hoechst 33342.55 Furthermore, knockdown of JARID1B expression in melanoma cells by RNA interference resulted in a marked reduction in tumorigenic activity, suggesting that JARID1Bþ cells have characteristics as CSCs. Of note, the researchers also found that JARID1B– cells gave rise to JARID1Bþ cells during long-term culture and that JARID1B– cells showed high tumorigenic activity when injected into immunocompromised mice.55 The JARID1B– cells were therefore able to gain JARID1B expression and to acquire stem cell properties under certain conditions. This dynamic transition of stem C 2012 UICC Int. J. Cancer: 132, 1249–1259 (2013) V

cell properties, termed ‘‘dynamic stemness’’, might account for the concern as to whether melanoma follows the CSC model. Whereas the CSC model incorporates a stringent hierarchy in which CSCs generate non-CSCs in a unidirectional manner, the notion of dynamic stemness allows non-CSCs to gain stemness and to become CSCs, indicative of phenotypic plasticity between CSCs and non-CSCs (Fig. 3a). Such phenotypic plasticity may result from dynamic epigenetic changes regulated by signals from the stem cell niche, a microenvironment that is crucial for the maintenance of stem cell properties (self-renewal and an undifferentiated state) and comprises many elements including stromal cells, blood vessels, extracellular matrix, growth factors and

Special Section Paper

1254

cytokines as well as hypoxia (Fig. 3b).56 The epithelial to mesenchymal transition (EMT) may be one process by which non-CSCs acquire stemness and may confer phenotypic plasticity on tumor organization in breast cancer (Fig. 3b).57 In melanoma, given that JARID1B is shown to be a target of hypoxia-inducible factor 1 (HIF1),58 hypoxia may up-regulate JARID1B expression and thereby give rise to a pattern of histone modification that underlies a change in gene transcription required for stem cell function. It has yet to be clarified how tumors can acquire plasticity between CSCs and non-CSCs unlike rigid hierarchy between normal stem cells and differentiated cells. Our speculation is that plasticity can be acquired during the course of therapies. Many chemical therapeutic agents and irradiations drastically destroy environmental status in the tumor tissues and often cause stress conditions such as reactive oxygen species and/or hypoxia, leading to formation of de novo microenvironment (niche) to evoke stem cell properties to non-CSCs similar to EMT (Fig. 3b). Once non-CSCs acquire the adaptive ability to become CSCs, tumors can exploit this plasticity in order to escape and survive from further therapies. Thus, it might be important to create new strategies to prevent the formation of de novo niche in addition to conventional treatments.

Induced Cancer Stem Cells Induced pluripotent stem cells (iPSCs) were first established in 2006 by the ectopic expression of defined factors in mouse embryonic or tail-tip fibroblasts.59 Research on iPSCs has had a large impact on various biological fields including cancer research, given that the process underlying the reprogramming of somatic cells is similar to that responsible for cancer initiation. The establishment of iPSCs inspired us to generate induced cancer stem cells (iCSCs) from somatic cells of various tissues through expression of a set of defined factors. iCSCs were originally termed as the somatic cells that have embryonic stem cell-like genes expression with high tumorigenicity and were induced by three factors (Ras, IjB, c-Myc).60 We made some minor modifications to the original concept of iCSCs and proposed that iCSCs are tumor-initiating cells derived from normal somatic cells by induction of various combinations of defined factors (not only above three factors) (Figs. 4a and 4b). To date, iCSCs have been established from normal somatic cells such as primary tissue stem, progenitor and differentiated cells through retroviral transduction with driver genes such as Myc or Ras (Fig. 4a). Similar to iPSCs, such iCSCs have the potential to undergo self-renewal and to generate differentiated progenies, and they form tumors when transplanted into recipient mice. By using the combinations of defined factors that reflect oncogenic events found in human tumors, we have established several types of iCSCs capable of forming tumors in mice, such as osteosarcoma,61 brain tumors,62 ovarian tumors,63 choriocarcinoma64 and leukemia-lymphoma.45 Those tumors have been found to share phenotypic and functional characteristics with the cor-

Complexity of cancer stem cells

responding human cancers (Fig. 4b). For instance, the iCSCs that initiate osteosarcoma were established from Ink4a- and Arf-null bone marrow stromal cells ex vivo by retroviral transduction of c-Myc.61 These osteosarcoma-initiating iCSCs have the capacity to undergo trilineage differentiation into adipocytes, osteocytes and chondrocytes and are highly tumorigenic, giving rise to metastases in multiple organs including the lung and bone, when transplanted into syngeneic mouse recipients.61 There are several advantages to the study of iCSCs: (i) The entire process of tumor development can be monitored comprehensively from the cell of origin to malignant transformation. We found that tissue invasion by and migration of brain tumor-initiating iCSCs preceded tumor mass formation, which was apparent only a few days after transplantation of the cells into the brain of recipient mice.62 This cell migration occurred along fiber tracts and was followed by perivascular infiltration. The cells then proliferated and finally formed tumors with pathological characteristics similar to those of human glioblastoma within a 4-week period.62 We were thus able to recapitulate the entire process of gliomagenesis with the use of iCSCs. (ii) Given that iCSCderived tumors have a short latency period and tumor penetrance is 100% for all types of iCSCs examined,45,61–64 CSCs can be readily isolated from the iCSC-derived tumors for analysis of CSC characteristics such as tumorigenicity, metastasis and drug resistance. For example, with the use of ovarian cancer-initiating iCSCs, we found that prospectively purified tumor cells positive for epithelial cell adhesion molecule (EpCAM) had a tumorigenic potential higher than that of EpCAM– cells,63 suggesting that mouse ovarian CSCs are present within the EpCAMþ subpopulation of tumor cells in this model. (iii) Large numbers of CSCs can be readily prepared from iCSC-derived tumors for examination of the potential efficacy of anticancer drugs. In contrast, it is difficult to perform such screening assays with CSCs of human tumors both because of the limited number of such cells isolatable from clinical specimens and because of related ethical considerations. (iv) Mice with tumors formed by transplanted iCSCs can be subjected to preclinical studies of potential new drugs. We have thus examined and confirmed the efficacy of arabinosylcytosine (Ara-C or cytarabine), a standard treatment for hematopoietic tumors, in a mouse model of pre-B-ALL; the effective doses and each amount of Ara-C administered were similar to those for actual clinical protocols.45

Therapeutic Implications of CSCs The resistance of CSCs to conventional cancer treatments such as chemotherapeutic agents and radiation is considered a formidable problem because remaining CSCs presumably trigger relapse after treatment termination. Development of new therapeutic strategies based on the CSC model has therefore become a key goal in the challenge to achieve complete eradication of cancer. To date, four strategies have been considered (Fig. 5). C 2012 UICC Int. J. Cancer: 132, 1249–1259 (2013) V

1255

Special Section Paper

Sugihara and Saya

Figure 4. Generation and study of iCSCs. (a) iCSCs are established by the introduction of defined genes such as Myc and Ras into somatic cells including stem, progenitor and differentiated cells of a variety of tissues. The iCSCs are able to undergo both self-renewal and differentiation, and they form tumors when transplanted into recipient mice. The generated tumors contain CSCs, which can be purified and subjected to further experiments. (b) Types of iCSCs that have been established to date.45,61–64 Abbreviations not defined in text: NSC, neural stem cell; siRNA, small interfering RNA; SV40, simian virus 40; LBL, lymphoblastic lymphoma.

The first strategy is to attack CSCs directly. CSCs can be targeted by agents that kill them specifically or that promote their differentiation into non-CSCs, which in turn will undergo apoptosis, senescence or terminal differentiation (Fig. 5a). A transgenic system was used to target only CSC marker-positive cells for cell death in a mouse model of AML.65 This approach resulted in the eradication of all AML cells, suggesting that CSCs are responsible for the overall cellular organization of AML. A cytotoxic antibody specific for the cell surface molecule TIM-3, which is expressed in human AML stem cells but not in normal HSCs, was used to selectively attack AML stem cells and thereby to eradicate reconstituted human AML cells in xenografted mice.66 Furthermore, a monoclonal antibody to CD47, a cell surface C 2012 UICC Int. J. Cancer: 132, 1249–1259 (2013) V

protein that protects cells from attack by phagocytes, was found to induce phagocytosis of AML stem cells by macrophages.67,68 Such studies of antibodies that preferentially target AML stem cells may provide more general insight into CSC-targeted therapy based on antibodies with minimal adverse effects. We recently showed that a splice variant form of CD44, a well-known CSC marker, interacts with the cystine transporter subunit xCT and effectively eliminates reactive oxygen species through increased cellular synthesis of reduced glutathione.69 This CD44 variant may play an important role in homeostasis of CSCs by limiting oxidative stress. Indeed, ablation of CD44 or treatment with an xCT inhibitor attenuated development of gastric cancer in mice.69 Redox regulation is therefore a potential target for anti-CSC therapy.

Special Section Paper

1256

Figure 5. CSC model-based strategies for cancer therapy. (a) The direct targeting of CSCs can result in their eradication, leaving only non-CSCs, which eventually will undergo senescence, apoptosis or terminal differentiation. Remaining non-CSCs are thought to be sensitive to the conventional chemo- and radio-therapies. (b) Promotion of the transition of CSCs from the quiescent to the proliferative state renders them sensitive to the induction of apoptosis by conventional anticancer drugs. (c) Targeting of the niche for CSCs can lead to CSC differentiation and eventual senescence, apoptosis or terminal differentiation of the resulting non-CSCs. (d) Inhibition of the transitions from CSCs to non-CSCs and from non-CSCs to CSCs may result in long-term dormancy of CSCs and block the generation of new CSCs from non-CSCs, respectively.

The second strategy for CSC eradication is to promote the exit of CSCs from the quiescent state and their entry into the cycling state (Fig. 5b). Quiescent AML stem cells were found to reside around endosteal regions of bone marrow after transplantation of human AML cells into NOG mice.70 Stimulation with granulocyte colony-stimulating factor triggered the entry of these quiescent AML stem cells into the cell cycle and thereby increased their sensitivity to treatment with Ara-C. The tumor suppressor proteins PML and FOXO were found to be critical for maintenance of the dormancy of chronic myeloid leukemia (CML) stem cells in a mouse model of CML.71,72 The targeting of either PML or FOXO (or the upstream transforming growth factor-b signaling pathway) forced quiescent CML stem cells to proliferate,

Complexity of cancer stem cells

resulting in an increased sensitivity of these cells to treatment with imatinib,71,72 a selective tyrosine kinase inhibitor of the BCR-ABL fusion oncoprotein. The combination of treatment to promote the proliferation of CSCs and conventional anticancer drugs that target proliferating cancer cells is thus a promising approach to CSC-targeted therapy. The third strategy is to attack the CSC niche (Fig. 5c). As mentioned in the previous section, the niche is thought to provide a supportive environment for maintenance of CSCs (Fig. 3b).56 Tumor blood vessels are implicated as a major component of the vascular niche for glioblastoma CSCs.73 The vascular niche is assumed to be formed via angiogenesis stimulated by vascular endothelial growth factor (VEGF), which is presumably released from CSCs. Bevacizumab, a neutralizing antibody to this growth factor, was found to block endothelial cell migration and tube formation, resulting in suppression of glioma formation, in a xenograft model.74 Inhibition of angiogenesis is thus a potential approach to therapy that targets the CSC niche. The chemokine CXCL12 (SDF1) and its receptor CXCR4 have been implicated in metastasis of breast cancer cells and in homing of leukemic cells.75 Given that CXCR4 is expressed on both breast CSCs and leukemic stem cells,75,76 CXCR4 antagonists such as plerixafor might also prove effective for disruption of the interaction between CSCs and their niche and therefore lead to a loss of stem cell properties. Further characterization of the components and molecular framework of the CSC niche should facilitate the development of targeted therapies. The fourth strategy is to inhibit the transitions between CSCs and non-CSCs (Fig. 5d). Blockade of the differentiation of CSCs into non-CSCs is a conceptual approach to maintain CSC dormancy and thereby to increase the effectiveness of direct CSC-targeted therapy (Fig. 5a). The notion of blocking the de-differentiation of non-CSCs into CSCs derives from the concept of plasticity in the hierarchical organization of tumors such as that incorporated into the dynamic stemness model (Fig. 3a). Similar to normal stem cell differentiation, the differentiation of CSCs is associated with dynamic epigenetic changes.77 Recent studies using global transcriptome analysis and next-generation sequencing have revealed aberrant expression or mutation of genes related to epigenetic regulation in many tumors.78,79 The formation and maintenance of CSCs are therefore thought to require epigenetic abnormalities resulting from such aberrant expression of mutation of epigenetic modifiers such as those responsible for DNA methylation, histone acetylation and histone methylation.78,79 Plasticity in the CSC model might be driven by changes in the pattern of histone modification that give rise to changes in gene transcription, rather than by changes in DNA methylation, which tend to be less dynamic. Although development of inhibitors of histone-modifying enzymes is being widely pursued, further studies are required before novel therapeutic approaches that target the differentiation and plasticity of CSCs and non-CSCs can be realized. C 2012 UICC Int. J. Cancer: 132, 1249–1259 (2013) V

Clonal Diversity in the CSC Model Two independent groups recently showed that CSCs (leukemia-initiating cells) for human B-ALL with the fusion genes BCR-ABL or ETV6-RUX1 are genetically diverse.80,81 Profiling of DNA copy number alterations in individual leukemic cells with the use of global SNP (single nucleotide polymorphism) arrays or multiplex fluorescence in situ hybridization revealed the existence of multiple genetically distinct clones that could initiate leukemia when transferred to immunocompromised mice. These clones are assumed to evolve by acquiring mutations in a multiple-branching manner rather than in linear succession. Furthermore, the pattern of clones found at diagnosis was different from that at the stage of overt leukemia or relapse, suggesting that only certain clones present at diagnosis, presumably those resistant to drug treatment and with a high tumorigenic activity, were able to survive and evolve. Some cases of B-ALL are thus likely to be best explained by a combination of the CSC model and the stochastic (clonal evolution) model (Fig. 3c). With regard to solid tumors, CSCs for human colon cancer have also been found to consist of functionally distinct clones.82 The authors isolated CSCs based on the property to generate spheroid formation. Using molecular marking approach with lentiviral vectors, at least three major clones were found in CSCs: a clone that maintained self-renewal activity over the long-term in serial transplants, a clone with limited or no self-renewal activity that was able to form tumors only in primary mouse recipients and a clone with delayed tumorigenic activity that became apparent in secondary or tertiary recipients. Of note, the clone with long-term self-renewal activity disseminated to bone marrow and was alone responsible for metastasis.82 These findings thus suggest the existence of clonal diversity that conforms to a hierarchical organization originated from CSCs, with the clones being functionally and phenotypically distinct presumably as a result of differential acquisition of mutations (Fig. 3c).

Although it remains to be determined whether diversity of CSCs is common or restricted to a small number of tumor types, the possible existence of distinct CSC clones with differential sensitivities to antitumor therapies should be kept in mind and warrants further investigation. Characterization of individual CSC clones may shed light on common mechanisms underlying the maintenance of stem cell properties in all clones and thereby inform the development of new CSCtargeted therapies.

Conclusions Ever since the first experimental identification of CSCs, the CSC model has been a major topic of debate as a result of uncertainties concerning the properties of these cells such as their defining cell surface markers, frequency and plasticity.83 However, we believe that the introduction of the CSC concept has resulted in important advances in cancer research. For instance, the CSC-based hierarchical model has provided a better understanding of the tumor heterogeneity. In addition, the goal of development of anticancer drugs will change from reduction of tumor size to targeting of subpopulations of tumor cells with a high tumorigenic potential. Future studies of CSCs will need to expand beyond the xenograft approach, in which cells derived from human tumors are transplanted into immunocompromised mice, and include the development and characterization of mouse tumor models that recapitulate aspects of tumor heterogeneity and the microenvironment in order to provide further insight into the complexity of CSCs observed in actual human tumors.

Acknowledgements The work in the authors’ laboratory was supported by grants from the Ministry of Education, Culture, Sports, Science, and Technology of Japan (to E.S. and H.S.). The authors declare no potential conflicts of interest.

References 1. 2. 3.

4.

5.

6.

7.

Nowell PC. The clonal evolution of tumor cell populations. Science 1976;194:23–8. Dick JE. Stem cell concepts renew cancer research. Blood 2008;112:4793–807. Clarke MF, Dick JE, Dirks PB, et al. Cancer stem cells—perspectives on current status and future directions: AACR Workshop on cancer stem cells. Cancer Res 2006;66:9339–44. Bruce WR, Van Der Gaag H. A quantitative assay for the number of murine lymphoma cells capable of proliferation in vivo. Nature 1963;199:79–80. Park CH, Bergsagel DE, McCulloch EA. Mouse myeloma tumor stem cells: a primary cell culture assay. J Natl Cancer Inst 1971;46:411–22. Sabbath KD, Ball ED, Larcom P, et al. Heterogeneity of clonogenic cells in acute myeloblastic leukemia. J Clin Invest 1985;75: 746–53. Griffin JD, Lowenberg B. Clonogenic cells in acute myeloblastic leukemia. Blood 1986;68: 1185–95.

8.

Bonnet D, Dick JE. Human acute myeloid leukemia is organized as a hierarchy that originates from a primitive hematopoietic cell. Nat Med 1997;3:730–7. 9. Al-Hajj M, Wicha MS, Benito-Hernandez A, et al. Prospective identification of tumorigenic breast cancer cells. Proc Natl Acad Sci USA 2003; 100:3983–8. 10. Singh SK, Clarke ID, Terasaki M, et al. Identification of a cancer stem cell in human brain tumors. Cancer Res 2003;63:5821–8. 11. Patrawala L, Calhoun T, Schneider-Broussard R, et al. Highly purified CD44þ prostate cancer cells from xenograft human tumors are enriched in tumorigenic and metastatic progenitor cells. Oncogene 2006;25:1696–708. 12. Hurt EM, Kawasaki BT, Klarmann GJ, et al. CD44þ CD24(
) prostate cells are early cancer progenitor/stem cells that provide a model for patients with poor prognosis. Br J Cancer 2008; 98:756–65.

C 2012 UICC Int. J. Cancer: 132, 1249–1259 (2013) V

13. Ricci-Vitiani L, Lombardi DG, Pilozzi E, et al. Identification and expansion of human colon-cancer-initiating cells. Nature 2007;445: 111–15. 14. Li C, Heidt DG, Dalerba P, et al. Identification of pancreatic cancer stem cells. Cancer Res 2007;67: 1030–7. 15. Frank NY, Margaryan A, Huang Y, et al. ABCB5mediated doxorubicin transport and chemoresistance in human malignant melanoma. Cancer Res 2005;65:4320–33. 16. Schatton T, Murphy GF, Frank NY, et al. Identification of cells initiating human melanomas. Nature 2008;451:345–9. 17. Collins AT, Berry PA, Hyde C, et al. Prospective identification of tumorigenic prostate cancer stem cells. Cancer Res 2005;65:10946–51. 18. Patrawala L, Calhoun-Davis T, SchneiderBroussard R, et al. Hierarchical organization of prostate cancer cells in xenograft tumors: the CD44þalpha2beta1þ cell population is enriched

Special Section Paper

1257

Sugihara and Saya

1258

19.

Special Section Paper

20.

21.

22.

23.

24.

25.

26.

27.

28.

29.

30.

31.

32.

33.

34.

in tumor-initiating cells. Cancer Res 2007;67: 6796–805. Beier D, Hau P, Proescholdt M, et al. CD133(þ) and CD133(
) glioblastoma-derived cancer stem cells show differential growth characteristics and molecular profiles. Cancer Res 2007;67:4010–15. Joo KM, Kim SY, Jin X, et al. Clinical and biological implications of CD133-positive and CD133-negative cells in glioblastomas. Lab Invest 2008;88:808–15. Eppert K, Takenaka K, Lechman ER, et al. Stem cell gene expression programs influence clinical outcome in human leukemia. Nat Med 2011;17: 1086–93. Huntly BJ, Shigematsu H, Deguchi K, et al. MOZ-TIF2, but not BCR-ABL, confers properties of leukemic stem cells to committed murine hematopoietic progenitors. Cancer Cell 2004;6: 587–96. Somervaille TC, Cleary ML. Identification and characterization of leukemia stem cells in murine MLL-AF9 acute myeloid leukemia. Cancer Cell 2006;10:257–68. Lee J, Kotliarova S, Kotliarov Y, et al. Tumor stem cells derived from glioblastomas cultured in bFGF and EGF more closely mirror the phenotype and genotype of primary tumors than do serum-cultured cell lines. Cancer Cell 2006;9: 391–403. Zhou S, Schuetz JD, Bunting KD, et al. The ABC transporter Bcrp1/ABCG2 is expressed in a wide variety of stem cells and is a molecular determinant of the side-population phenotype. Nat Med 2001;7:1028–34. Hirschmann-Jax C, Foster AE, Wulf GG, et al. A distinct ‘‘side population’’ of cells with high drug efflux capacity in human tumor cells. Proc Natl Acad Sci USA 2004;101:14228–33. Kondo T, Setoguchi T, Taga T. Persistence of a small subpopulation of cancer stem-like cells in the C6 glioma cell line. Proc Natl Acad Sci USA 2004;101:781–6. Pearce DJ, Taussig D, Simpson C, et al. Characterization of cells with a high aldehyde dehydrogenase activity from cord blood and acute myeloid leukemia samples. Stem Cells 2005; 23:752–60. Cheung AM, Wan TS, Leung JC, et al. Aldehyde dehydrogenase activity in leukemic blasts defines a subgroup of acute myeloid leukemia with adverse prognosis and superior NOD/SCID engrafting potential. Leukemia 2007;21:1423–30. Ginestier C, Hur MH, Charafe-Jauffret E, et al. ALDH1 is a marker of normal and malignant human mammary stem cells and a predictor of poor clinical outcome. Cell Stem Cell 2007;1: 555–67. Ma I, Allan AL. The role of human aldehyde dehydrogenase in normal and cancer stem cells. Stem Cell Rev 2011;7:292–306. Tanei T, Morimoto K, Shimazu K, et al. Association of breast cancer stem cells identified by aldehyde dehydrogenase 1 expression with resistance to sequential Paclitaxel and epirubicinbased chemotherapy for breast cancers. Clin Cancer Res 2009;15:4234–41. Croker AK, Goodale D, Chu J, et al. High aldehyde dehydrogenase and expression of cancer stem cell markers selects for breast cancer cells with enhanced malignant and metastatic ability. J Cell Mol Med 2009;13:2236–52. Silva IA, Bai S, McLean K, et al. Aldehyde dehydrogenase in combination with CD133

Complexity of cancer stem cells

35.

36.

37.

38.

39.

40.

41.

42.

43.

44.

45.

46. 47.

48.

49.

50.

51.

52.

defines angiogenic ovarian cancer stem cells that portend poor patient survival. Cancer Res 2011; 71:3991–4001. Welte Y, Adjaye J, Lehrach HR, et al. Cancer stem cells in solid tumors: elusive or illusive? Cell Commun Signal 2010;8:6. Brescia P, Richichi C, Pelicci G. Current strategies for identification of glioma stem cells: adequate or unsatisfactory? J Oncol 2012;2012: 376894. Schepers AG, Snippert HJ, Stange DE, et al. Lineage tracing reveals Lgr5þ stem cell activity in mouse intestinal adenomas. Science 2012;337: 730–5. Huntly BJ, Gilliland DG. Leukaemia stem cells and the evolution of cancer-stem-cell research. Nat Rev Cancer 2005;5:311–21. Krivtsov AV, Twomey D, Feng Z, et al. Transformation from committed progenitor to leukaemia stem cell initiated by MLL-AF9. Nature 2006;442:818–22. Cobaleda C, Gutierrez-Cianca N, Perez-Losada J, et al. A primitive hematopoietic cell is the target for the leukemic transformation in human philadelphia-positive acute lymphoblastic leukemia. Blood 2000;95:1007–13. Cox CV, Evely RS, Oakhill A, et al. Characterization of acute lymphoblastic leukemia progenitor cells. Blood 2004;104:2919–25. Castor A, Nilsson L, Astrand-Grundstrom I, et al. Distinct patterns of hematopoietic stem cell involvement in acute lymphoblastic leukemia. Nat Med 2005;11:630–7. Hong D, Gupta R, Ancliff P, et al. Initiating and cancer-propagating cells in TEL-AML1-associated childhood leukemia. Science 2008;319:336–9. le Viseur C, Hotfilder M, Bomken S, et al. In childhood acute lymphoblastic leukemia, blasts at different stages of immunophenotypic maturation have stem cell properties. Cancer Cell 2008;14: 47–58. Sugihara E, Shimizu T, Kojima K, et al. Ink4a and Arf are crucial factors in the determination of the cell of origin and the therapeutic sensitivity of Myc-induced mouse lymphoid tumor. Oncogene 2012;31:2849–61. Visvader JE. Cells of origin in cancer. Nature 2011;469:314–22. Mullighan CG, Williams RT, Downing JR, et al. Failure of CDKN2A/B (INK4A/B-ARF)mediated tumor suppression and resistance to targeted therapy in acute lymphoblastic leukemia induced by BCR-ABL. Genes Dev 2008;22: 1411–15. Fang D. A tumorigenic subpopulation with stem cell properties in melanomas. Cancer Res 2005;65: 9328–37. Monzani E, Facchetti F, Galmozzi E, et al. Melanoma contains CD133 and ABCG2 positive cells with enhanced tumourigenic potential. Eur J Cancer 2007;43:935–46. Boiko AD, Razorenova OV, van de Rijn M, et al. Human melanoma-initiating cells express neural crest nerve growth factor receptor CD271. Nature 2010;466:133–7. Shultz LD, Lyons BL, Burzenski LM, et al. Human lymphoid and myeloid cell development in NOD/LtSz-scid IL2R gamma null mice engrafted with mobilized human hemopoietic stem cells. J Immunol 2005;174:6477–89. Quintana E, Shackleton M, Sabel MS, et al. Efficient tumour formation by single human melanoma cells. Nature 2008;456:593–8.

53. Quintana E, Shackleton M, Foster HR, et al. Phenotypic heterogeneity among tumorigenic melanoma cells from patients that is reversible and not hierarchically organized. Cancer Cell 2010;18:510–23. 54. Roesch A, Becker B, Meyer S, et al. Retinoblastoma-binding protein 2-homolog 1: a retinoblastoma-binding protein downregulated in malignant melanomas. Mod Pathol 2005;18: 1249–57. 55. Roesch A, Fukunaga-Kalabis M, Schmidt EC, et al. A temporarily distinct subpopulation of slow-cycling melanoma cells is required for continuous tumor growth. Cell 2010; 141:583–94. 56. Cabarcas SM, Mathews LA, Farrar WL. The cancer stem cell niche—there goes the neighborhood? Int J Cancer 2011;129:2315–27. 57. Mani SA, Guo W, Liao M-J, et al. The epithelial– mesenchymal transition generates cells with properties of stem cells. Cell 2008;133:704–15. 58. Xia X, Lemieux ME, Li W, et al. Integrative analysis of HIF binding and transactivation reveals its role in maintaining histone methylation homeostasis. Proc Natl Acad Sci USA 2009;106:4260–5. 59. Takahashi K, Yamanaka S. Induction of pluripotent stem cells from mouse embryonic and adult fibroblast cultures by defined factors. Cell 2006;126:663–76. 60. Wong DJ, Segal E, Chang HY. Stemness, cancer and cancer stem cells. Cell Cycle 2008;7:3622–4. 61. Shimizu T, Ishikawa T, Sugihara E, et al. c-MYC overexpression with loss of Ink4a/Arf transforms bone marrow stromal cells into osteosarcoma accompanied by loss of adipogenesis. Oncogene 2010;29:5687–99. 62. Sampetrean O, Saga I, Nakanishi M, et al. Invasion precedes tumor mass formation in a malignant brain tumor model of genetically modified neural stem cells. Neoplasia 2011;13: 784–91. 63. Motohara T, Masuko S, Ishimoto T, et al. Transient depletion of p53 followed by transduction of c-Myc and K-Ras converts ovarian stem-like cells into tumor-initiating cells. Carcinogenesis 2011;32:1597–606. 64. Kobayashi Y, Shimizu T, Naoe H, et al. Establishment of a choriocarcinoma model from immortalized normal extravillous trophoblast cells transduced with HRASV12. Am J Pathol 2011;179:1471–82. 65. Aikawa Y, Katsumoto T, Zhang P, et al. PU.1mediated upregulation of CSF1R is crucial for leukemia stem cell potential induced by MOZTIF2. Nat Med 2010;16:580–5, 1p following 5. 66. Kikushige Y, Shima T, Takayanagi S, et al. TIM-3 is a promising target to selectively kill acute myeloid leukemia stem cells. Cell Stem Cell 2010; 7:708–17. 67. Jaiswal S, Jamieson CH, Pang WW, et al. CD47 is upregulated on circulating hematopoietic stem cells and leukemia cells to avoid phagocytosis. Cell 2009;138:271–85. 68. Majeti R, Chao MP, Alizadeh AA, et al. CD47 is an adverse prognostic factor and therapeutic antibody target on human acute myeloid leukemia stem cells. Cell 2009;138:286–99. 69. Ishimoto T, Nagano O, Yae T, et al. CD44 variant regulates redox status in cancer cells by stabilizing the xCT subunit of system xc(
) and thereby promotes tumor growth. Cancer Cell 2011;19:387–400.

C 2012 UICC Int. J. Cancer: 132, 1249–1259 (2013) V

70. Saito Y, Uchida N, Tanaka S, et al. Induction of cell cycle entry eliminates human leukemia stem cells in a mouse model of AML. Nat Biotechnol 2010;28:275–80. 71. Ito K, Bernardi R, Morotti A, et al. PML targeting eradicates quiescent leukaemia-initiating cells. Nature 2008;453:1072–8. 72. Naka K, Hoshii T, Muraguchi T, et al. TGF-betaFOXO signalling maintains leukaemia-initiating cells in chronic myeloid leukaemia. Nature 2010; 463:676–80. 73. Gilbertson RJ, Rich JN. Making a tumour’s bed: glioblastoma stem cells and the vascular niche. Nat Rev Cancer 2007;7:733–6. 74. Bao S. Stem cell-like glioma cells promote tumor angiogenesis through vascular

75.

76.

77.

78.

endothelial growth factor. Cancer Res 2006;66: 7843–8. Burger JA, Peled A. CXCR4 antagonists: targeting the microenvironment in leukemia and other cancers. Leukemia 2009;23:43–52. Hwang-Verslues WW, Kuo WH, Chang PH, et al. Multiple lineages of human breast cancer stem/progenitor cells identified by profiling with stem cell markers. PLoS One 2009;4:e8377. Lotem J, Sachs L. Epigenetics and the plasticity of differentiation in normal and cancer stem cells. Oncogene 2006;25:7663–72. Baylin SB, Jones PA. A decade of exploring the cancer epigenome—biological and translational implications. Nat Rev Cancer 2011;11:726–34.

C 2012 UICC Int. J. Cancer: 132, 1249–1259 (2013) V

79. Rodriguez-Paredes M, Esteller M. Cancer epigenetics reaches mainstream oncology. Nat Med 2011;17:330–9. 80. Anderson K, Lutz C, van Delft FW, et al. Genetic variegation of clonal architecture and propagating cells in leukaemia. Nature 2011;469:356–61. 81. Notta F, Mullighan CG, Wang JCY, et al. Evolution of human BCR–ABL1 lymphoblastic leukaemia-initiating cells. Nature 2011;469:362–7. 82. Dieter SM, Ball CR, Hoffmann CM, et al. Distinct types of tumor-initiating cells form human colon cancer tumors and metastases. Cell Stem Cell 2011;9:357–65. 83. Magee JA, Piskounova E, Morrison SJ. Cancer stem cells: impact, heterogeneity, and uncertainty. Cancer Cell 2012;21:283–96.

Special Section Paper

1259

Sugihara and Saya