Neoplastic stem cells: Current concepts and clinical perspectives

1. Introduction 

The principle concept of cancer stem cells (CSC) has gained acceptance in recent years [1], [2], [3], [4]. The CSC concept postulates the existence of subfractions of “tumor stem cells” within each neoplasm. These CSC exhibit the capacity for self-renewal and unlimited growth, and in this regard differ from more mature neoplastic cells (progeny) that have only a limited capacity to divide and to survive. The concept of tumor stem cells may provide explanations for the failure of certain treatments to induce long-term remission. In fact, in many instances, conventional chemotherapy may act only on more mature cells, whereas immature neoplastic stem cells exhibit resistance, so that these drugs fail to target and eliminate CSC [1], [2], [3], [4]. Thereby, the CSC concept points to the need to develop new treatment strategies through which CSC can be eliminated. A prerequisite for the evaluation of CSC as “target-cell” in oncology is their identification and knowledge about target expression profiles. Therefore, substantial efforts have recently been made to identify CSC in various types of cancer and to identify molecular targets and expression profiles in these cells.

A number of different monoclonal antibodies directed against various cell surface antigens have been used to identify CSC-enriched cell populations in various neoplasms, and to purify these cells for molecular and functional studies [5], [6]. These experiments focus on the identification and characterization of molecular targets, and effects of natural ligands, response modifiers, and targeted drugs on these cells.

The current article provides a summary of our current knowledge about cancer- and leukemia-initiating cells, with special focus on clinical implications and perspectives.

2. Definition of CSC 

Cancer/leukemia stem cells (CSC) are undifferentiated cells and are defined by three key features [1], [21]: first, these cells can differentiate into most or all types of cells that are produced by the original tumor. Second, CSC have the ability to self-renew. Finally, CSC maintain the stem cell pool and most (or even all) mature elements of the tumor/leukemia for unlimited time periods by balancing between self-renewal (proliferation without maturation) and proliferation plus differentiation and maturation by asymmetrical cell division(s) [1], [4]. The process(es) of cell division, of self-renewal, and of differentiation of CSC are considered to be regulated by a network of cytokines and by the microenvironment, similar to normal stem cells [1], [4]. In many instances, the same cytokine that regulates growth of normal (stem) cells in a certain organ will also promote the growth and self-renewal of neoplastic stem cells [1], [4].

3. Limitations of in vivo CSC assays 

Despite the obvious value of an in vivo model that is sufficient to demonstrate the tumor-initiating potential of distinct subpopulations of neoplastic cells, a number of limitations of the xenotransplant assay have to be considered. First, neoplasms with a low growth rate (e.g. indolent neoplasms, low-grade malignancies, preneoplasm) are difficult to analyze in a mouse xenotransplant model as in most instances, the development phase of the neoplasm exceeds the lifetime of the mouse. Second, the microenvironment is usually species-specific and often tumor-specific. In fact, the microenvironment of the normal mouse may differ in several aspects from the tumor microenvironment that supports the growth of neoplastic (stem) cells in the natural (human) host. Likewise, microenvironment receptors and cytokines in the mouse may not cross-interact in all cases with the respective receptors expressed on human tumor/leukemia (stem) cells. To overcome this problem, NOD/SCID mice have been treated with human cytokines or are cotransplanted with microenvironmental cells in order to facilitate better growth of neoplastic (stem) cells [9]. Another important limitation is that most mouse models that have been used in the past, including NOD/SCID mice, harbour a residual immune system through which these mice can eliminate subfractions of injected cancer/leukemia cells, especially when these cells display numerous immunogenic antigens (more mature cells) or when cells are antibody-laden (antibody-stained) cells. The same cells may, however, grow and form tumors in mice with a more severely impaired immune system [22], [23]. Finally, most neoplasms may grow in permanently established subclones with subclone-specific stem cells. In other words, the stem cell pool in most neoplasms is composed of several (many) different subsets of stem cells with varying biological properties and diverse growth characteristics, and it remains unknown whether all (relevant) subclones can be grown in one xenotransplant model.

4. In vitro assays 

As stem cell research using NOD/SCID mice is expensive and time-consuming and may have several limitations, in vitro long-term growth assays are often used in order to screen for stem cell fractions or CSC-regulating compounds. Such in vitro long-term growth assays have been established for myeloid and lymphoid neoplasms as well as various solid tumors [24], [25], [26], [27], [28]. Interestingly, in most instances, a stroma cell layer supports the long-term growth of immature neoplastic cells in these assays, which is in line with the concept that the (tumor) microenvironment essentially contributes to growth and survival of CSC and thus the development and biology of these neoplasms (stem cell niche). In myeloid neoplasms, the Dexter type long-term culture system, originally established for normal pluripotent hematopoietic progenitor cells [29] has been employed and found to support the long-term growth of leukemias [30]. However, interestingly, some of these leukemias may even have a “growth-disadvantage” compared to normal stem cells (long-term culture initiating cells) in these cultures [31]. In lymphoid neoplasms, including CLL and ALL, it is well known that stromal cells can inhibit apoptosis and support the growth of leukemic cells in vitro [32], [33]. In various solid tumors and melanomas, isolated cells are found to form three-dimensional spheres that presumably are composed of immature neoplastic cells and supporting/nutritive stromal cells, and thus may resemble an in vitro model of the so-called stem cell niche [34]. Such sphere formation has been described for neural cancer (stem) cells, especially glioblastomas, colon cancer, breast cancer, and melanoma cells [24], [25], [26], [27], [28]. The culture system is based on the ability of cancer cells to form three dimensional spheres in vitro, and the ability of sphere-derived cells to induce long-term growth of tumors [24], [25], [26], [27], [28]. In fact, at least some of the tumor cells within these spheres are considered to have self-renewal capacity and tumor-initiating (CSC) potential.

5. Antigens commonly expressed on CSC 

Neoplastic stem cells are considered to express a similar antigen pattern, to display similar functional properties, and to be regulated by similar receptor ligands when compared to normal stem cells (derived from the same organ system). Therefore, many stem cell/progenitor cell markers are also markers of neoplastic stem cells. These antigens include cytokine receptors, homing receptors, and various drug transporters (Table 1).

Table 1. Cell surface molecules detectable on stem and progenitor cells.

	Marker/antigen	Background and function of marker/antigen
	CD133	CD133 antigen is a transmembrane molecule expressed on hematopoietic stem and progenitor cells, on circulating endothelial progenitor cells, neural stem cells, renal and prostate stem cells [215].
	CD34	CD34 is an adhesion molecule expressed on human hematopoietic stem and progenitor cells, endothelial progenitor cells and vascular endothelial cells [142].
	CD44	CD44 is a cell adhesion receptor, and its ligands are hyaluronate and the cytokine osteopontin. CD44 is expressed in lymphoid, mieloid and erythroid cells, mesenchymal stem cells and may be useful predictor of lymph node metastasis [1].
	CD29	CD29 is a beta1 integrin involved in cell adhesion embryogenesis, tissue repair, immune response and metatastatic diffusion of tumor cells reacting with thrombocytes, monocytes and a T and B lymphocytes and is also expressed on mesenchymal stem cells [216].
	CD24	CD24 antigen is a glycosylphosphatidylinositol-linked membrane sialoglycoprotein. CD24 is present on B cells, from the stage pre-B to the mature B cell stage, but not on plasma cells. It is expressed on mature granulocytes and on a variety of epithelial cell types [1].
	CD166	CD166 molecule, a mesenchymal stem cell marker that displayed heterogeneous expression patterns in CRC epithelial cells (15, 16) and whose increased expression levels were previously associated with poor clinical outcome in CRC patients [45].
	CD326	EpCAM (CD326) is a pan-epithelial differentiation antigen expressed on the basolateral surface of various carcinomas to varying degrees. As a homotypic cell adhesion molecule, it is intimately integrated within the Cadherin–Catenin and WNT pathways. It has recently been shown to modulate the expression of proto-oncogenes such as c-myc [217].
	CD90	Thy-1 (CD90) is expressed on many cell types, including T cells, thymocytes, neurons, endothelial cells, and fibroblasts. Activation of Thy-1 can promote T cell activation, and this role of Thy-1 is reviewed elsewhere. Thy-1 also affects numerous nonimmunologic biological processes, including cellular adhesion, neurite outgrowth, tumor growth, migration, and cell death [218].
	CD123	The specific [alpha] subunit of the interleukin-3 receptor (IL-3R-alpha, CD123) is expressed on hematopoietic cells, including monocytes, neutrophils, basophils, and megakaryocytes but not on peripheral T cells, natural killer cells, platelets, and red blood cells [219].
	CD9	CD9 belongs to a tetraspanin superfamily and is expressed in a variety of blood cells including pre-B lymphocytes but not in HSCs. It is also expressed in many types of solid tumors, and is involved in a various kinds of cell processes, such as cell adhesion, motility, and signalling events through an association with integrin family proteins [220].
	CD20	CD20 (human B-lymphocyte-restricted differentiation antigen, Bp35), is a hydrophobic transmembrane protein with a molecular weight of approximately 35kDa located on pre-B and mature B lymphocytes. The antigen is expressed on most B-cell non-Hodgkin's lymphomas but is not found on stem cells, pro-B cells, normal plasma cells or other normal tissues. Plasma blasts and stimulated plasma cells may express CD20. CD20 regulates an early step(s) in the activation process for cell cycle initiation and differentiation, and possibly functions as a calcium ion channel. CD20 is not shed from the cell surface and does not internalize upon antibody binding. Free CD20 antigen is not found in the circulation; thus a drug that reacts with CD20, such as an antibody, would not be neutralized before binding to its target cell [221].
	ABCB1 (MDR1) and ABCG2	ABCB1 and ABCG2 are recognized as belonging to a family of at least 48 human ABC transporters involved in a variety of essential cellular transport processes. They are the products of MDR genes and confer multidrug resistance by pumping out chemotherapy. They also pump out Hoechst dye and rhodamine. Reversal of MDR in vitro was easily attained with a variety of inhibitors like verapamil [222].
	ABCB5	ABCB5 [subfamily B (MDR/TAP)] is a novel human ABC transporter encoded on chromosome 7p15.3 [223]. ABCB5, like ABCB1, acts as an energy-dependent drug efflux transporter for the fluorescent probe rhodamine-123 [224].
	CLL-1	Human CLL-1 (also known as MICL or CLEC12A), is a type II transmembrane glycoprotein and member of the large family of C-type lectin-like receptors involved in immune regulation. The intracellular domain of CLL-1 contains both an immunotyrosine-based inhibition motif as well as a YXXM motif, suggesting a role for CLL-1 as a signalling receptor [86], [225].
	Erythropoietin receptor	The erythropoietin receptor (EpoR) consists of two peptide chains and is a member of the cytokine receptor family. The interaction of erythropoietin with its cell surface receptor induces a conformational change of receptor homodimers leading to the activation of intracellular signal transduction that mediates the ability of erythropoietin to support the proliferation, terminal differentiation and survival [226].

To identify potential surface markers of CSC, it is helpful to look for interactions of these cells with the surrounding microenvironment. Based on the behaviour of normal stem cells, CSC should interact with the supporting microenvironment via several biologically relevant surface receptors mediating specific functions. One specific stem cell function related to the microenvironment is “stem cell homing” [35], [36], [37], [38]. Major homing receptors discussed as being expressed on CSC are integrins, selectin-ligands, chemokine receptors, other cytoadhesion molecules (CAMs), and ligands of matrix molecules such as L1 or the hyaluronic acid receptor CD44 [39], [40]. L1 is also involved in the epithelial–mesenchymal transition (EMT) and is found on the edge of invasive colon cancer and its metastases [41], [42]. Together with L1, CD133 appears to be necessary for tumor growth of gliomas [43]. CD133 and CD44 show overlapping expression in various tumors and CSC [26], [44], [45]. CD133, a glycoprotein also known as Prominin 1 (PROM1), is a member of the pentaspan family of transmembrane glycoproteins (5-transmembrane, 5-TM) which specifically bind to cellular protrusions [46], [47], [48]. CD133 is expressed in hematopoietic stem cells, endothelial progenitors, as well as on glioblastoma, colon CSC, and other solid tumor CSC [46], [47], [48]. The CD44 protein is a cell-surface glycoprotein involved in cell–cell interactions, adhesion, migration, and homing [49]. It is a receptor for hyaluronic acid and can also interact with other ligands, such as osteopontin, collagen species, and matrix metalloproteinases (MMPs) [49]. A specialized sialofucosylated glycoform of CD44, called HCELL, is usually expressed on human hematopoietic stem cells, and is a highly potent E-selectin and L-selectin ligand [50]. HCELL functions as a “bone marrow homing receptor” directing migration of human hematopoietic stem cells and mesenchymal stem cells to the bone marrow [50]. HCELL is expressed on breast-, prostate-, pancreatic-, and colon CSC, and is regulated by the WNT pathway [13], [16], [45], [51]. The frequently observed expression of CD44 and CD133 on CSC suggests a functional role for these receptors in CSC biology. However, it remains unknown whether these receptors are indeed necessary for stem cell functions or are just expressed on CSC without a specific function. Whatever the answer to this question is, these markers are commonly found on immature neoplastic cells (CSC) and may thereby help in the identification of CSC in various organs and tumors. More recently, it has been reported that in a colon adenoma strain, LT97, the CD44+ subpopulation exhibits a faster growth rate and expresses higher levels of other stem cell antigens (such as Musashi-1 and EphB2) compared to the CD44-negative fraction of adenoma cells [52]. Similarly, the CD133+ fraction of the HCT116 colon cancer cell line has recently been described to display higher levels of stem cell antigens and a faster growth rate in vitro and in vivo compared to the CD133− fraction [53].

Another basic requirement for CSC is to protect themselves against various external toxic stimuli and drugs, which makes them often drug-resistant. One example for such a drug transporter protein is ABCG2 which is able to pump out not only cell-specific substances but also exogenous toxins and cytotoxic drugs, and thus is responsible (in part) for resistance of CSC against various drugs. Other drug transporters that have been discussed as indicating the long-term repopulation-potential of malignant cells include MDR-1 and ABCB5 [54], [55]. Various specific dyes like the Hoechst 33342 dye, are also transported from the cell into the extracellular space via specific transporters, which allows the definition of the so-called “side population” which seems closely related to the stem cell fraction in many tumors [56]. A similar distribution and association with stem cells has been described for the enzyme aldehyde dehydrogenase (aldolase) [57], [58], [59], [60], [61]. This enzyme has been described to be involved in the metabolization of several cytostatic drugs including cyclophosphamide and thereby may be associated with chemoresistance [60], [62], [63].

Lastly, CSC are considered to respond to various external physiologic (sometimes paracrine or autocrine) stimuli including various cytokines. In line with this assumption, CSC express certain cytokine and chemokine receptors on their cell surface. For example the IL-3 receptor (CD123/CD131), SCF receptor (KIT) and the G-CSF-receptor are usually expressed on leukemic CSC in AML [64], [65]. It has also been described that EGF receptor family members including HER2 are expressed on epithelial CSC including mammary CSC [66], [67]. There is also evidence that IGF- and FGF-receptors play an important role in solid tumors and may be expressed on solid tumor CSC [68], [69]. More recently, it has been described that pancreatic CSC express CXCR4 [12]. A clinically important question is whether solid tumor CSC or leukemic CSC express receptors for erythropoietin (EPO), G-CSF, or GM-CSF, as these cytokines are often applied in cancer/leukemia patients [70]. At least for AML and some tumor cell types it has been described that CSC indeed express receptors for SCF, G-CSF, and sometimes also for GM-CSF [65].

6. General problems with the so-called ‘stem cell markers’ 

One general problem is that the so-called stem cell markers are by far not specific for stem cells or progenitor cells. Rather, most of these antigens are broadly expressed on various mesenchymal cells. Likewise, CD44 is expressed not only on hematopoietic and non-hematopoietic stem cells but also on most mature cells, including monocytes, lymphocytes, granulocytes, epithelial cells, and melanocytes [50], [71]. Similarly, in most leukemias and solid neoplasms, CD44 is expressed on mature cells. Prominin (CD133) is expressed on myeloid progenitor cells and on various (immature and mature) mesenchymal cells, including endothelial cells [46], [47], [48]. Even the CD34 antigen (hematopoietic precursor cell antigen-1, HPCA1) is not only expressed on hematopoietic stem cells but also on more mature progenitors and also on endothelial cells [72]. All in all, no stem cell specific antigen has been identified so far. Therefore, it is essential to apply combinations of markers and antibodies and to define stem cell-enriched fractions on the basis of typical antigen combinations. Usually, one or two markers are employed by investigators to gate for or to exclude a germ layer or an organ system (e.g. CD45 as pan-hematopoietic cell marker). In certain instances a cocktail of antibodies is used to exclude more mature cells in a certain organ system, e.g. more mature hematopoietic cells by lineage-specific markers (“Lin-cocktail” defining Lin-negative progenitor cells) [19], [73]. Here, one problem may be that in certain leukemias, CSC may aberrantly express lineage-related marker antigens. In these leukemias, application of the “Lin-cocktail” would lead to a loss of CSC (sub)fractions.

Finally, markers that are typically expressed on immature cells of a given neoplasm or organ, are applied, e.g. CD34 for lymphohematopoietic cells and most myeloid leukemias. In these defined progenitor fractions, further subpopulations in which CSC reside, are defined. This is performed by antibodies that positively or negatively identify these subfractions. Based on antibody-binding patterns and identification of subfractions enriched in CSC, these cells can be isolated by multi-color flow cytometry and cell sorting. The read out that is then used to confirm the presence of CSC is either an in vitro long-term culture system or – preferably – the immunodeficient mouse model. In most instances, the NOD/SCID mouse has been employed in these assays. However, there are a number of caveats that have to be considered when using sorted cells in these mouse models and bioassays. The most important caveat may be that the antibody itself may stimulate or may interfere with in vitro growth, engraftment, or/and long-term repopulation of progenitor cells in NOD/SCID mice [22]. Therefore, it has to be excluded by appropriate control experiments, that the antibodies used to positively or negatively define CSC would per se induce or inhibit engraftment of CSC in NOD/SCID mice, or would interfere with in vitro CSC growth in the bioassay applied or with in vivo engraftment [22]. Recently, Taussig et al. revealed a procedure-related problem in the description of AML CSC defined by expression of CD34 and lack of CD38 [22]. In their paper, they were able to show that the original description of leukemic stem cells as CD38-negative cells in the mouse model may be due to the clearance of leukemic CSC together with the CD38 antibody by the residual immune system of NOD/SCID mice. Clearance was not observed when mice were further immunosuppressed by drugs or when the antibody was degraded into fab fragments which allowed CD38+ leukemic cells to repopulate these mice [22]. Moreover, CD34+/CD38+ grew well in more severely immunodeficient IL2rgamma(null) mice (NSG mice) [22]. All in all, it appears that NOD/SCID mice may not be the most suitable strain to study CSC growth in solid tumors and leukemias. Rather, more severely immunodeficient mice (NOG or NSG mice) may be required for optimal engraftment of CSC [10], [11]. It can therefore be expected that stem cell research will employ NOG or NSG mice in various CSC models in the future.

Recently, Kim et al. found that a polyclonal anti-human CD24 rabbit antibody initiates cross-linking of CD24 on tumor cells, and that this cross-link induces apoptosis in breast cancer cells in a cell culture system using MCF-7 cells [74]. This may have implications for the characterization of breast cancer stem cells, since these cells supposedly reside within the CD44+CD24low (CD24-negative) fraction of tumor cells. If CD24+ tumor cells would just be eliminated (faster than CD24low cells) by the antibody-induced cross-linking of CD24, the CD24-negativity could no longer be regarded as a stem cell-related feature. Whether indeed CD24-induced stem cell depletion occurs after antibody-binding remains to be shown.

Another major limitation may be that CSC receptors and their ligands and homing receptors in the tissue environment are sometimes species-specific. Therefore, CSC are often injected directly into a certain target organ (e.g. bone marrow, pancreas, brain, others) in these mice [12], [13], [14], [15], [16], [17], [44], [75], [76]. Nevertheless, it is difficult to know whether some of the CSC will not (cannot) grow in a xenotransplant model simply because essential mouse-derived ligands are not cross-reacting with human receptors expressed on CSC. A possible solution to this problem may be to co-transplant human stroma cells, to humanize mice (organ-specific microenvironment), or to inject mice with all important (human) ligands in order to guarantee proper engraftment of CSC.

Probably the most important problem with CSC markers and CSC-reactive antibodies is stem cell plasticity. In fact, it has been described that leukemic and solid tumor CSC fractions are usually composed of several different subclones defined by varying profiles of molecular markers, point mutations in critical target genes (often causing drug resistance in subclones), and expression of cell surface receptors including stem cell markers. This means, that CSC in a given tumor may display varying combinations of cell surface antigens which makes it difficult to define all CSC subfractions (subclones) by monoclonal antibodies. In some instances, the phenotypic heterogeneity may also be associated with functional heterogeneity. However, so far very little is known about cell surface antigens defining subfractions of CSC in various solid tumors and leukemias. Likewise, in AML, leukemic CD34+/Lin− (stem) cells are often composed of a CD133+ and a CD133− subfraction [77].

In the following paragraphs, we review human neoplasms where CSC have recently been identified.

7. Myeloid neoplasms 

It is generally assumed that most if not all myeloid neoplasms derive from a clonal immature hematopoietic progenitor (stem) cell. Therefore, myeloid neoplasms are optimal models to study the CSC hypothesis. Normal hematopoietic stem cells are considered to reside within the Lin-negative and CD38-negative portion of CD34-positive progenitor cells. Leukemic stem cells have first been described in AML, and later in CML [19], [73], [78]. In other myeloid neoplasms, neoplastic stem cells are less well defined. The phenotype of leukemic stem cells is considered to be similar to that of normal stem cells (Lin−, CD34+, CD38− and in part CD38+) [8], [19], [79] (Table 2). Several attempts have been made to identify markers that distinguish between normal and neoplastic meyloid stem cells. This would enable therapeutic approaches that might be able to spare normal stem cells [80]. Indeed, several surface molecules may be expressed abundantly on CD34+/Lin− AML stem cells, whereas normal hematopoietic stem cells lack or express only low levels of these antigens. Examples for antigens preferentially expressed on leukemic CSC in AML are the alpha chain of the IL-3 receptor (CD123), the Mylotarg-receptor Siglec-3 (CD33), CD96, CXCR4, and CLL-1 [64], [81], [82], [83], [84], [85], [86]. CD33 is of special interest as CD33-targeting drugs are available and are used to treat patients with (refractory) AML. Fig. 1 shows the effects of Mylotarg on survival of CD34+/CD38+ and CD34+/CD38− AML cells. Neoplastic stem cells have been defined in several but not all variants of AML [8], [19]. Likewise, in a group of AML patients, leukemic blast cells and immature progenitors are CD34-negative. In these patients, it is very difficult to define stem cell compartments. Other examples are promyelocytic leukemia, NPM-mutated AML variants, and monoblastic leukemias where most of the neoplastic cells may be CD34-negative cells [87]. In many of these leukemias it even remains to be demonstrated that AML stem cells reside within the (small) CD34+ subpopulation of clonal leukemic cells. It is also important to note that AML types in which the stem cell origin of the leukemia can be demonstrated using a NOD/SCID mouse xenotransplant model are those variants where blast cells exhibit a high proliferative potential and resistance against chemotherapy [88].

Table 2. Published cell surface phenotype of neoplastic stem cells in various malignancies.

	Disease	Surface marker	Refs.
	AML	CD34+, CD38−, CD44, CD123+	Bonnet and Dick [19]
	ALL Ph+	CD34+, CD38−	Cobaleda et al. [105]
	TEL-AML1–positive c-ALL	CD34+/CD38−/low/CD19+	Hong et al. [106]
		CD34+/CD19+	Kong et al. [107]
	Childhood B-ALL	CD133(+)/CD19(−) and CD38(−)	Cox et al. [108]
	B lineage ALL	CD34+/CD9+	Nishida et al. [109]
	CML	CD34+, CD38−, CD123+	Holyoake et al. [73]
	Breast	CD44+, CD24−/low, ESA+	Al-Hajj et al. [13]
	Pancreas	CD44+	Li et al. [16]
		CD133+	Hermann et al. [12]
	Liver	CD133+	Ma et al. [131]
		CD90+	Yang et al. [17]
	Colon	CD133	Ricci-Vitian et al. [26], O’Brien et al. [44]
		CD44	Dalerba et al. [45]
	Prostate	CD133+/alpha 2 beta 1 integrin/CD44+	Collins et al. [142]
		CD44+/CD24−	Hurt et al. [51]
	Sarcoma	CD133+	Suva et al. [151]
	Melanoma	CD20+	Fang et al. [24]
		CD133+	Monzani et al. [162]
	CNS	CD133+	Singh et al. [15]
			Hemm et al. [27]

View Large Image Download to PowerPoint

Fig. 1. Apoptosis in AML stem cells induced by Mylotarg (GO). Blast cells obtained from a patient with AML (FAB M0) were incubated with control medium (upper graphs) or with gemtuzumab/ozogamicin (GO=Mylotarg), 1μg/ml, at 37°C for 48h (lower graphs). Then, apoptosis was analyzed by combined flow cytometry using antibodies against CD34 and CD38, and AnnexinV-FITC. The percentage of apoptotic cells was analyzed in CD34+/CD38+ cells (right panel) and in the CD34+/CD38− fraction of the clone (left panel) after gating for vital (7-AAD negative) cells.

In CML, most clonal cells are CD34-negative cells. In these patients, a complex multi-step enrichment technique for the isolation of putative CD34+/Lin− stem cells has been proposed [73]. Indeed, when these cells were injected into NOD/SCID or NOD/SCID-beta2microglobulin−/− mice, they were found to repopulate these mice with a CML-like disease. By contrast, more mature CD34+ CML cells only produced an early transient leukemic repopulation that was no longer detectable after 6 weeks [89]. Similar to the AML stem cell, the CML stem cell was found to express the IL-3 receptor [90]. Other target antigens like CD33, CD44, or CD117 were also found to be expressed on CD34+/CD38− stem cells in CML patients [64]. CML stem cells also express ABCB1 and ABCG2, whereas the levels of OCT-1 expressed on CML stem cells are rather low [55], [91]. Both ABCG molecules are known to mediate the efflux of imatinib and other drugs and thus confer resistance. OCT-1, on the other hand, mediates the uptake of imatinib, and low expression of these transporter molecules is associated with low drug-uptake and thus low intracellular levels of imatinib. All these mechanisms may contribute to the intrinsic resistance of CML CSC against imatinib [55], [92]. More recently, imatinib-resistance of CML stem cells has been employed to screen for stem cell-related markers in BCR/ABL transformed (imatinib-exposed) murine cells [93]. In these studies, several antigens potentially involved in leukemic stem cell growth and survival have been identified. One of these CML stem cell-related genes is the Alox5 gene that encodes the 5-lipoxygenase (5-LO) [87]. In the absence of Alox5/5-LO, BCR/ABL failed to induce a CML-like disease in mice, whereas Alox5-deficiency showed no effects on growth of normal stem cells [93]. These data suggest that growth and survival of leukemic stem cells in CML are regulated by specific gene products, and the same may hold true for AML. Also, recent data suggest that neoplastic stem cells in various myeloid neoplasms may use similar if not identical signalling pathways for growth and survival in vivo. These pathways include the PI3 kinase-mTOR pathway, WNT-β-catenin pathway, and Notch signalling-pathway [94]. However, all these pathways may be shared by normal and leukemic stem cells, and therefore may not be optimal targets, at least when treatment should spare normal (stem) cells. Also, Hedgehog signalling seems to be important for the development of myeloid leukemias and may be a promising target pathway [95]. Other pathways found in leukemic stem cells are the JAK2-STAT5 pathway and NF-kappa-B pathway. In most instances, signalling is initiated by specific oncoproteins, like BCR-ABL in CML, or PML-RAR-alpha in acute promyelocytic leukemia [96], [97], [98], [99], [100].

In other myeloid neoplasms, very little is known about the biology and phenotype of neoplastic stem cells. Florian et al. examined the phenotype of CD34+/CD38− cells in various myeloid neoplasms including myelodysplastic syndromes, CML, and systemic mastocytosis [64]. In all these myeloid neoplasms, the phenotype, determined by antibody-staining, appeared to be very similar, and included major surface targets such as CD123 and CD33 [64]. In classical JAK2 V617F-mutated myeloproliferative disorders, the JAK2 mutant is usually detectable in the CD34+/CD38− fraction of clonal cells [101], [102]. However, JAK2 V617F may not be detectable in all neoplastic stem cell subclones in these patients [103], and leukemic progression (secondary AML) is often accompanied by a loss of the JAK2 mutant, which may be explained by expansion of a more immature JAK2 V617F-negative subclone during disease progression [103]. Otherwise, very little is known about the biology, target expression profiles, and phenotypes of neoplastic stem cells in JAK2+ MPN. Recently, a defective stem cell niche has been discussed as an important factor contributing to the pathogenesis of JAK2-mutated neoplasms [104].

8. Lymphoid neoplasms 

9. Solid tumors 

9.2. Colon CSC 

Since the gastrointestinal tract is an organ-system with a high turnover of cells where proliferation and cell self-renewal take place, it was of great interest to learn about the location, biology, and phenotype of colon CSC. Several lines of evidence suggest that immature colon cells (and presumably also CSC) are located in colon crypt bottoms. With regard to the phenotype, first reports pointed to CD133 as a potential CSC marker antigen colon cancer [26], [44] (Table 2). When CD133-positive cells were injected into the kidney capsule or subcutaneously into NOD/SCID mice, colon tumors developed after several weeks [26], [44]. Later, Dalerba et al. found that CD44 is another potential marker for CSC in colon cancer patients [45]. When CD44+/EpCAM+ cells were injected subcutaneously into NOD/SCID mice tumors developed after 20 weeks in these mice [45]. Furthermore, they found that the CSC population is further characterized by coexpression of CD166 [45]. Recently, it was reported that CD133+ cells derived from colon spheres are clonogenic cells and can form adenocarcinomas in a mouse xenotransplant model [28]. These CD133+ sphere-derived cells in part co-expressed CD24, CD29, CD44, and CD166. Interestingly, of all markers tested, only CD24 was found to enrich for cells with even higher clonogenic activity. All in all, several surface markers, including CD44 and CD133 have been discussed as potential markers of colon CSC (Table 2). With regard to CD44, this may also hold true for colon adenomas. In fact, it has been shown that the colon adenoma cell strain LT97 consists of a CD44+ and a CD44− portion, and that both subfractions differ in their growth kinetics and expression of stem cell markers [52]. In particular, CD44-positive LT97 cells attach and grow for unlimited time periods, whereas CD44− cells are slowly growing cells (Fig. 2) [52]. Moreover, in contrast to CD44− LT97 cells, CD44+ LT97 cells display nuclear beta-catenin and express beta-catenin target genes, such as ephrin B receptor (ephB2) and the musashi1 antigen (msi1) [52]. These stem cell antigens are also detectable in colon CSC [124]. All in all, CD44 may not only be a CSC marker for established colon adenocarcinomas, but also for CSC in preneoplastic lesions, i.e. colon adenomas which may have implications for the biology of tumor development. In fact, colon carcinoma CSC may develop from adenoma CSC by subclone selection after accumulation of further hits [125], [126].

View Large Image Download to PowerPoint

Fig. 2. Correlation between expression of CD44 and proliferative potential of LT97 cells. The adenoma cell strain LT97 was cultured in complete medium and passaged once every 10–14 days. Expression of CD44 on LT97 cells was determined by flow cytometry in early (passage up to 20), intermediate (passage 20–30), and late (passage above 30) cultures. The figure shows the percentage of CD44+ cells (black bars) and the plating efficiency (hatched bars) over time (early, intermediate, late passage).

10. Melanoma stem cells (MSC) 

So far, little is known about tumor-initiating cells in skin cancer patients. In fact, CSC have only been investigated and partly characterized in malignant melanomas. Fang et al. described that melanoma spheres can be grown from primary melanoma cells, and that melanoma sphere-derived cells exhibit long-term growth, multilineage potential, and tumor-initiating potential in SCID mice [24]. They found that a subpopulation of cells in these spheres is CD20+/CD45− cells that co-express melanoma antigens. This CD20+ fraction of melanoma cells, when cultured separately, was found to form larger spheres than CD20− cells, and showed long-term proliferation in vitro. Based on these data, it is tempting to speculate that melanoma stem cells reside within the CD20+ fraction of cells, although this has not formally been proven by mouse experiments using primary melanoma cells sorted for CD20+ and CD20− cells. In addition, the expression of CD20 on melanoma cells has not been confirmed in other studies. Another marker that has been discussed as a potential stem cell marker for human melanoma is CD133 [162]. It has been described that only 0.2–0.8% of primary metastatic melanoma cells express CD133 [162]. Monzani et al. observed that 40–50 days after subcutaneous injection of 3.5×105 CD133-positive cells into NOD/SCID mice, these mice develop detectable melanoma lesions, whereas the CD133-negative fraction did not form tumors in NOD/SCID mice [162]. These data suggest that melanoma stem cells may reside within a CD133+ fraction of the clone, at least in metastatic melanomas [163]. In line with this hypothesis, several melanoma cells lines display CD133 [164], [165]. Other studies have shown that certain drug-transporters that (when expressed) are indicative of chemoresistance, may be expressed on melanoma stem cells (and stem cells in other tumors). Among these transporters are MDR1, ABCG2 and ABCB5 [166]. Schatton et al. described that the ABCB5+ fraction of primary (freshly isolated) melanoma cells is enriched for melanoma-repopulating stem cells, whereas ABCB5-negative cells are less capable of initiating melanomas in NOD/SCID mice [54]. ABCB5+ melanoma cells were found to co-express other potential stem cell markers including ABCB1, TIE1, nestin, and CD44 [54], [167].

However, the melanoma-initiating potential of single (stem) cells may greatly depend on the microenvironment and the mouse strain employed to demonstrate repopulation. Quintana et al. have recently shown that up to 30% of all melanoma cells are melanoma-repopulating cells (independent of their phenotype) when interleukin-2 receptor gamma null NOD/SCID mice are used, whereas tumorigenicity of the same cells is much lower when NOD/SCID mice with an intact interleukin-2 receptor are employed [23]. These data suggest that the frequency of melanoma-initiating cells may be relatively high, at least in a severely immunocompromised host. Whether the same cells (all these cells) are also able to repopulate melanomas in an immunocompetent host or in patients remains unclear. In fact, the stem cell function of a given tumor/melanoma cell may not only be predetermined by intrinsic factors (genetic and epigenetic stem cell programs) but also be microenvironmental factors (factors defining the stem cell niche) and also the immune system (immunosurveillance, apoptosis-induction by killer cells, tumor cell phagocytosis).

The observation that up to 30% of melanoma cells may have the potential to repopulate melanomas in NOG mice and thus have stem cell function, has changed our views on what types and subpopulations of melanoma cells indeed are repopulating and non-repopulating cells. Unfortunately, only a few markers can clearly discriminate larger subpopulations of melanoma cells with different potential to metastasize and to form new tumor lesions in patients. One such marker is the erythropoietin (EPO) receptor. In fact, the EPO receptor is only expressed in trace amounts in normal melanocytes, but is expressed in melanoma cells [168], [169], [170], [171]. Whereas in primary melanomas, only a small subpopulation of cells display the EPO receptor, in metastatic melanomas up to 30% of all cells express this antigen [168]. Moreover, it has recently been described that EPO receptor expression is associated with disease progression and the metastasizing potential of melanomas [169], [172]. Finally, EPO has been described to initiate signalling and promote survival in human melanomas [171]. Whether indeed the EPO receptor is a functionally revelant antigen expressed specifically on melanoma initiating cells is currently under investigation.

11. CSC plasticity—the Hydra Model of CSC development 

A number of previous and more recent data suggest that neoplastic stem cell clones display substantial plasticity and are often composed of several different subclones. In many cases, subclone formation may precede the development of a frank neoplasm, and only a few (or even only one) of these subclones may progress to an overt malignancy. This assumption is consistent with the multi-hit theory of cancer development [173], [174], [175], [176] and would predict that neoplastic stem cells detectable in the NOD/SCID or NOG (NSG) mouse assay would not all be able to repopulate all components (cell subclones) in a given neoplasm. Rather, this model predicts early formation of subclones with different transformation potential, and each subclone must be expected to contain subclone-specific stem cells. A good example for the presence of multiple subclones in one neoplasm is the coexistence of two histologically different hematopoietic disorders in one patient. Another example is CML, where during treatment with imatinib, one or more different subclones exhibiting imatinib-resistant BCR/ABL mutants may be selected (by treatment), and each of these subclones may progress into a clinically relevant leukemia (subclone-specific progress) [91], [177], [178]. Whereas many of these subclones bearing (drug-resistant) BCR/ABL mutations may be present (as small clones) before imatinib is started, some of these subclones may progress to leukemia during imatinib therapy which may point to clonal instability and a potential effect of the drug on subclone formation. Similar observations of stem cell plasticity have been made in other hematopoietic malignancies and may also apply to non-hematopoietic neoplasms [179].

Recent data suggest the stem cell plasticity may even involve the supportive stroma or tumor/leukemia-associated angiogenesis. In particular, it has been described that endothelial cells in lymphomas and other neoplasms are of monoclonal origin [180]. Other studies suggest that cultured bone marrow stroma cells and mesenchymal progenitors in patients with Ph+ CML are of clonal origin [181]. However, other studies suggest that stromal cells and mesenchymal progenitor cells in patients with CML are non-clonal cells and not derived from the Ph+ clone [182], [183].

The above described heterogeneity and plasticity of CSC populations in various neoplasms may be one reason for the complexity of CSC evolution and function, and may also explain why it is difficult to design effective treatment approaches sufficient to eliminate all these cells and subclones in cancer/leukemia patients [184], [185], [186].

12. Strategies for the successful elimination of CSC 

Several different strategies have been considered to inhibit growth and/or survival of CSC, with the ultimate goal to eliminate all CSC in these malignancies. These concepts include relevant surface targets, signal transduction molecules, and certain survival molecules expressed in CSC (Fig. 3). Many concepts still relate to myeloid or lymphoid neoplasms, whereas so far, only a few treatment strategies have been presented for solid tumors and melanomas. Moreover, whereas much is known about the expression of various surface and cytoplasmic drug targets in solid tumors and leukemias, only very little is known about expression of the same targets in neoplastic stem cells (Fig. 3).

View Large Image Download to PowerPoint

Fig. 3. Expression of molecular targets in and on neoplastic cells and their progenitors. Whereas much is known about target expression profiles in more mature neoplastic cells (left part of panel), little is known about expression of molecular targets in leukemic (neoplastic) stem cells (right panel part “?”). ST, signal transduction; TK, tyrosine kinase(s); TF, transcription factor(s); TSG, tumor suppressor gene(s); DR, death receptors.

In myeloid neoplasms, potential surface targets include the IL-3R alpha chain CD123, the Mylotarg-receptor Siglec-3 (CD33), CD44, and the tyrosine kinase receptor KIT (CD117). Target expression can be exploited by selecting blocking antibodies or by constructing antibody–toxin conjugates, cytokine–ligand–toxin conjugates, or antibody–drug conjugates. A good example for a ligand–toxin conjugate is the IL-3-diphteria toxin fusion protein [187]. An example for an antibody–drug conjugate is the CD33-targeting fusion molecule gemtuzumab–ozogamcin (Mylotarg) [81], [82], [83] (Fig. 1). It has also been reported that an unconjugated blocking CD123 antibody can effectivity reduce leukemia cell growth in mice [188] and even in a few patients with AML [189]. So far, only a few studies have attempted to demonstrate functional significance of target expression in leukemic stem cells. In one study, Mylotarg was found to induce apoptosis in CD34+/CD38−/CD33+ stem cells in AML [65] (Fig. 1). A novel and promising marker for stem cell targeting in AML may be CLL-1 that is able to discriminate between normal and leukemic stem cells in the bone marrow [86]. Another promising marker is CD96 since Hosen et al. showed that CD96 is specific for AML stem cells. He showed that CD96+ AML cells can engraft irradiated Rag2(−/−) gamma(c)(−/−) mice [85]. However, so far no targeted drugs specific for CD96 or CLL-1 have been developed. Several different CD antigens are employed as targets of therapy in B-cell Non-Hodgkin's lymphomas, including CD20, CD22, CD23, CD25, or CD52. Clinical trials have shown that antibodies directed against some of these antigens (especially CD20 and CD52) can improve treatment and prognosis in these patients. It is therefore tempting to speculate that some of these antigens are also expressed on immature lymphoma-initiating tumor cells. Studies are ongoing to define the CD antigen profile in lymphoid stem cells in various NHLs. An interesting aspect is that some of the lymphoid markers, like CD20, are also expressed on non-hematopoietic CSC. In solid tumors, cell surface targets include members of the ErbB receptor family, IGF receptors, TGFβ receptors, and FGF receptors [190], [191], [192], [193]. A general problem with surface targets is that CSC subclones often exhibit or develop resistance against antibody-based or other drugs. Several different mechanisms of drug resistance have been discussed, including expression of multi-drug resistance gene products like MDR1 or reduced antibody-binding. In addition, subclones that do not express the surface CD antigen may be selected by antibody-based therapy. Therefore, antibody-therapy is usually combined with conventional or with other targeted drugs.

Apart from surface markers, also intracellular targets have been identified in CSC. Among those, signal transduction molecules and survival molecules may represent most promising target antigens. These drugs include tyrosine kinase inhibitors, farnesyl transferase inhibitors, and other kinase inhibitors including PI3 kinase and mTOR blockers [194], [195], [196]. In some myeloid neoplasms like CML, BCR/ABL tyrosine kinase inhibitors have been used with considerable success [197]. More recently tyrosine kinase inhibitors have also been employed in clinical trials in solid tumors and melanomas [98], [196], [198], [199], [200], [201], [202], [203], [204], [205], [206], [207], [208], [209]. Among survival-related targets, most promising molecules may be members of the BCL-2 family (BCL-2, MCL-1, BCL-xL, others) and various Heat shock proteins (HSP32, HSP70, HSP90, others) [210], [211], [212]. Studies are ongoing to define whether these targets are expressed in CSC in various hematopoietic or solid tumors. Likewise, it has been described that the HSP32, also known as heme oxygenase 1, is expressed in CSC in various leukemias and solid tumor models [213], [214].

Despite intensive research and some progress in targeted therapy in solid tumors, only a few targeted drugs have produced convincing results in clinical trials, which may be due to resistance and the complexity of the signalling cascades in cancer cells and CSC. This is mainly due to (the various forms of) resistance as well as CSC plasticity with target-negative subclone formation. Therefore, it seems reasonable to combine targeted drugs (antibodies, small molecules) with each other or with conventional therapy (chemotherapy) in order to overcome drug resistance in CSC.

Finally, provided that specific antigens of CSC are identified, an attractive approach would be to induce specific immune responses and to establish vaccination therapy approaches in these malignancies.

13. Concluding remarks and future directions 

The emerging concept of neoplastic stem cells and CSC-related targets may offer new insights into the biology and the pathogenesis of various malignant disorders and new possibilities for the design of targeted drug therapies. Since CSC display considerable heterogeneity and plasticity, eradication of these cells and thus cure may only be reached when combinations of anticancer drugs (therapies) are applied. Therefore, treatment designs aiming at cancer/leukemia stem cell elimination need to take all potential targets and all relevant stem cell subclones into account. There is hope for the future that such novel treatment approaches will improve therapy in cancer patients.

Conflict of interest 

All the authors declare that they have no proprietary, financial, professional or other personal interest of any nature or kind in any product, service and/or company that could be construed as influencing the position presented in the review entitled “Neoplastic stem cells: Current concepts and clinical perspectives”.

Reviewers 

Dr. Dominique Bonnet, Cancer Research UK, London Research Institute, Haematopoietic Stem Cell Laboratory, 44 Lincoln's Inn Fields, London WC2A 3PX, United Kingdom.