Introduction

Metastasis, the spread of cancer cells from primary tumor sites to distant organs and tissues, accounts for over 90% of lethality in cancer patients 1. In spite of its clinical importance, the underlying cellular and molecular mechanisms of cancer metastasis are still poorly understood. In recent years, stem cells have been the focus of a tremendous amount of biomedical research because of their apparent potential for regenerative medicine. The discovery of cancer stem cells (CSCs) has stimulated great excitement, as well as heated debates, for both stem cell and cancer biologists. How the CSC theory fits into the general scheme of cancer progression, particularly with respect to metastasis, has not been well-defined. This review begins by comparing some characteristic features of normal and malignant stem cells that are highly relevant to cancer metastasis. These attributes lead us to propose that the ability of a tumor to metastasize is an inherent property of a subset of CSCs, coined here as metastatic CSCs (mCSCs). This ability is modulated through the interactions of the mCSCs with the local microenvironment or “niche”. Guided by current knowledge about normal stem cells, future characterizations of the origin, as well as the cellular and molecular mechanisms governing the in vivo behaviors of mCSCs, should shed light on potential therapeutic applications related to metastasis. Before proceeding, however, it is important to keep in mind that various aspects of the CSC theory remain to be unequivocally confirmed in different cancer types, particularly in solid tumors 2. Therefore, this review should be considered as a discussion of the current status of the field and as a guide for future research rather than summaries of proven hypotheses.

Normal stem cells

Two golden standards for defining normal stem cells were first established from studies of hematopoiesis 3. In order to sustain a lifelong supply of blood cells, hematopoietic stem cells (HSCs) must have the ability to (1) self-renew and (2) generate differentiated blood cells. In addition to the well-characterized HSCs, adult stem cells have also been found in a variety of tissues and organs including muscle, intestine, brain, skin/hair follicles, heart, lung and, more recently, mammary glands 4, 5, 6, 7, 8. The physiological function of normal stem cells is to maintain tissue homeostasis as well as to regenerate tissue after damage or injury. Prospective isolation of normal adult stem cells such as HSCs, neural stem cells and mammary epithelial stem cells (MESCs) has been accomplished by using combinations of cellular phenotypic markers (summarized in Table 1). Additionally, isolating a “side population” based on Hoechst33342 dye efflux (due to high activity of cell membrane multi-drug transporters) has been used as an alternative way of identifying stem cells 9.

Table 1 Summary of reported human (i) or mouse (ii) cell surface makers for normal or malignant tissue stem cells

As exemplified by HSCs, under normal physiological conditions, adult stem cells can live in a prolonged state of quiescence. Cell cycle regulators p21CIP1 and p18INK4C have been shown to regulate the quiescence of HSCs 10, 11. Once they have exited from the quiescent state, stem cells either self-renew or differentiate to generate progenies depending on the nature of both intrinsic and extrinsic stimulatory signals 3, 12, 13. These signals exist as a delicate “Yin & Yang” balance between positive and negative regulators. Using genetic models, key factors of the Wnt signaling pathway have been shown to be essential for promoting stem cell self-renewal in various systems 14, 15, 16, 17, 18, 19, 20. Notch and hedgehog signaling pathways have also been demonstrated to play crucial regulatory roles for self-renewal regulation of HSCs 21, 22, 23, 24. On the other hand, signals derived from BMP and TGF-β pathways negatively regulate stem cell proliferation 13, 25, 26, 27, 28. Depending on the cellular context, cross-talk between BMP and Wnt signaling is critical for the proper cell fate decision of normal stem cells 13. Tipping the delicate balance between positive and negative regulators of stem cell self-renewal in either direction can be problematic in vivo. For example, genetic changes leading to aberrant Wnt signaling in either stem cells or progenitors have been implicated as early events in the development of leukemia and other cancers 20.

The differentiation capacity of adult stem cells is normally restricted to specific cell lineages of the respective tissue type. However, adult stem cells have been observed to exhibit a certain level of cellular flexibility 29. For instance, muscle derived stem cells have been shown to give rise to blood cells 30. Cell fusion, albeit at a low frequency, has been suggested as one of the possible mechanisms underlying this phenomenon (reviewed in 8). The limited plasticity of adult stem cells and other potential complications may largely restrict the clinical applications of adult stem cells in regenerative medicine. Nonetheless, these observations potentially have very important implications for cancer metastasis 12. For example, cellular plasticity in stem cells may facilitate the epithelial to mesenchymal transition, which has been postulated as a key event during the early phase of cancer metastasis 31.

Discovery and properties of CSCs

With the landmark work accomplished by John Dick and his co-workers in human leukemia studies, recent years have witnessed the exciting discoveries of CSCs in solid tumors, including those of the breast and the brain 32, 33, 34, 35, 36. Several excellent reviews have covered the supporting evidence and implications of CSCs for carcinogenesis 37, 38, 39, 40. Applying principles established from stem cell research, human CSCs are functionally defined by their enriched capacity to regenerate cancers using xenograft mouse models. Similar to normal stem cells, CSCs can reproduce themselves through the process of self-renewal, which can be studied in serial transplantation assays. Additionally, cancers derived from purified CSCs recapitulate the heterogeneous phenotypes of the parental cancer from which they were derived, reflecting the differentiation capacity of CSCs 32, 33. These observations suggest that CSCs contain the complete genetic programs necessary to initiate and sustain tumor growth. The demonstration of CSCs in solid tumors, however, has not been reproduced in a large number of independent research groups and still awaits further validation.

As shown in Table 1, surface markers have been used to describe CSCs from different cancers. Interestingly, cell surface phenotypes for leukemia stem cells (CD34+CD38) are virtually identical to normal HSCs. Similarly, CD133, which is a marker for normal neural stem cells, has also been used to enrich for brain tumor stem cells 35, 41. Although they share many molecular pathways with normal stem cells, it is not surprising that CSCs have their own distinguishing molecular profiles. For example, the Bmi-1 proto-oncogene plays an essential role in the regulation of self-renewal for both leukemia stem cells as well as HSCs 42, 43. Bmi-1 is also required for neural stem cell self-renewal and is highly expressed in brain tumor CSCs 35, 44, 45. In contrast, loss of the PTEN tumor suppressor functionally distinguishes leukemia stem cells from normal HSCs 46. In the future, identifying and characterizing unique features of CSCs that discriminate them from normal stem cells will be pivotal for devising specific therapies that would spare normal stem cells. Additionally, an understanding of the origin of CSCs may provide stronger diagnostic and therapeutic power to clinicians.

Origins of CSCs

The discovery of CSCs begs the question regarding the origin of these cells. Are they derived from normal stem cells with a cancerous phenotype? Or do previously differentiated progenitor cells with oncogenic mutations regain the ability to self-renew? A third theory hypothesizes that CSCs may come from a rare fusion event between stem cells and other cells. Normal stem cells may be ideal target cells for accumulating mutations that are necessary for stepwise malignant transformation due to their inherent self-renewal capacity. Since multiple pathways are involved in self-renewal of stem cells, it seems conceptually more difficult for a differentiated cell to regain this ability through mutations. But the rareness of stem cells in tissues may counter this theory because of the low probability that they could be targeted by mutations. The relative abundance of transient amplifying immediate progenitor cells, derived from stem cells retaining partial self-renewal capacity, makes them likely candidates for initial transforming events. As documented below, evidence exists for both the stem cell and the committed progenitor as the origin of CSCs.

Similarity in cell surface markers suggests that normal tissue stem cells may be the targets of oncogenic transformation and give rise to CSCs (see Table 1). Expansion of mammary stem cells in mouse breast cancer models prior to cancer development is also indicative of a potential connection between normal tissue stem cells and CSCs 6. Similar observations were made in other cancer models 5, 47. However, the mere expansion of a normal stem cell in a tumor is not sufficient to justify the conclusion that it is the cell of origin for tumors. A more direct proof was shown by the fact that the ectopic expression of both Met and Myc oncogenes in MESCs/progenitors was sufficient to drive breast cancer development, although the identity of the initial cell population needs to be characterized better 48. Moreover, bone marrow derived cells (BMDCs) have been shown to initiate a gastric intraepithelial neoplasia that proceeds to gastric adenocarcinoma after chronic Helicobacter infection 49. In this study, mesenchymal stem cells from bone marrow have been proposed as candidate cells that give rise to gastric cancer. However, definitive experimental evidence showing normal tissue stem cells as origins for CSCs remains to be established.

Several stronger lines of evidence support the notion that a committed progenitor can be the cancer-initiating cell as a result of oncogenic transformation. Co-expression of Bcl-2 and the BCR/ABL protein (the fusion protein found in 90% of CML patients) in committed myeloid progenitors is sufficient to drive leukemia development in mice 50. Leukemic granulocyte-macrophage progenitors have been shown to be able to self-renew through the activation of the Wnt/β-catenin signaling pathway 51. More recently, the MLL-AF9 fusion protein has been shown to transform committed granulocyte-macrophage progenitors into leukemia stem cells. Reactivation of a subset of signature genes expressed in HSCs correlates with regaining of an enhanced self-renewal capacity 52. Researches in brain tumor development also indicate that more committed neural progenitor cells are likely to be the targets of tumorigenic mutations 53.

Although no direct experimental evidence is currently available for the cell fusion origin of CSCs, as mentioned above, cell fusion has been shown to be one of the mechanisms for the apparent cellular plasticity associated with tissue stem cells 8. However, it is not clear from previous studies whether stem cells themselves are fused with other cell types in different tissues in vivo. Conceptually, cell fusion between stem cells and mutated cells might lead to regaining of self-renewal capacity to allow further accumulation of transforming mutations. A recent study has shown that BMDCs were able to fuse to neoplastic epithelium 54. Additionally, the fusogenic factor CD44 is used as a positive surface marker for CSCs in breast cancer implying that these cells may have the capacity to fuse with other cell types. If cell fusion is an origin of CSCs, it could easily explain the detection of both fusogenic proteins and aneuploid cells commonly associated with neoplastic malignances 55.

While future studies are certainly needed to provide definitive proof for identifying the origin of CSCs, it is important to remember that demonstrating one model for the formation of CSCs in a given system does not necessarily exclude other mechanisms. The possible origins for CSCs are not mutually exclusive. For example, it has been shown that loss of Ink4a and Arf expression combined with EGFR activation can cause de-differentiation of astrocytes as well as the transformation of neural stem cells, both of which result in a similar high grade malignant glioma 56. Regardless of their origin, CSCs may play a critical role in metastasis because of their potential to migrate into different tissues. Additionally, CSCs from different origins may possess different metastasis abilities. Although comparatively less is known about the governing of mCSC migration, well-characterized evidence has elucidated the role that the niche plays in regulating normal stem cell migration.

Stem cell niche and tumor migration

In 1978, Schofield suggested that stem cells live in a niche, i.e. a physiologically defined supportive microenvironment, as demonstrated through early co-culture and transplantation studies 57. Recently, significant progress has been made in characterizing the in vivo architecture and functions of stem cell niches in different model systems including Caenorhabditis elegans, Drosophila and mammals 13, 58, 59. Niches for germ line stem cells (GSCs) in the Drosophila ovary or testis have been well characterized with great anatomic detail using genetic and cell biological approaches 60, 61. Critical signaling molecules that govern the self-renewal and differentiation of Drosophila GSCs, including activators of the BMP pathway and Jak-Stat signaling, are derived from the niche 62, 63, 64, 65, 66. One mechanism through which niche factors can modulate stem cell fate decision is the control of symmetric (producing two identical daughter cells) versus asymmetric (producing one identical and one differentiated cell) division 67.

In mammals, niches for adult stem cells have been characterized in the bone marrow, skin/hair follicle, intestine, neural system and testis 59. Cell types and architectures of niches for specific stem cells are variable from tissue to tissue. Nonetheless, many key players in stem cell niches have evolutionally conserved functions in both normal and malignant tissues, and thus may play key roles in tumorigenesis and metastasis to different target organs. For example, BMPs and their antagonists are known to play a crucial role in stem and progenitor cell biology as regulators of the balance between expansion and differentiation. BMPs promote differentiation of stem cells, thus promoting exit from the stem cell compartment. Gremlin 1, a secreted antagonist of the BMP pathway, was found to be overexpressed in tumor stromal cells derived from basal cell carcinoma (BCC), but not in those derived from nontumor skin 111. BMP inhibits and Gremlin 1 promotes proliferation of cultured BCC cells. Thus, factors secreted by tumor stroma may influence the stem cell niche of the tumor microenvironment, providing a suitable milieu for cancer development.

Potentially related to cancer metastasis, one critical function of the niche is to serve as an anchoring site for stem cells. For instance, HSCs are physically attached to their osteoblastic niche cell (N-cadherin+/CD45) in the bone through a membrane bound N-cadherin/β-catenin complex 68, 69. It is noteworthy that β-catenin is a key downstream mediator of the canonical Wnt signaling pathway, which is essential for self-renewal of stem cells. Retention of β-catenin at the stem cell membrane may prevent precocious activation of the Wnt signaling pathway 20. Osteopontin (Opn), a glycoprotein that negatively regulates the pool size of HSCs in bone marrow, is also critical for breast cancer bone metastasis 70, 71, 72, 73. Other critical adhesion molecules required for stem cell localization include integrins 74, 75, 76, 77, 78, 79.

With respect to HSCs, two distinct niches have been defined. The first is the endosteal bone surface niche, which is composed primarily of osteoblasts and correlates positively with the pool size of HSCs 68, 69, 80, 81. An alternative vascular niche for HSCs composed of endothelial cells has recently been identified in both the bone marrow and the spleen 82, 83. Similar to the loss of the osteoblastic niche, depletion of endothelial cells in vivo also leads to diminished HSC activity and resulting hematopoietic failure 84. Why then are there two distinct niches for the same HSC cells? Supported by several experimental observations, an attractive hypothesis proposes that the osteoblastic niche functions primarily in maintaining HSC quiescence, whereas the vascular niche promotes HSC proliferation, differentiation and migration 85.

By far the most detailed knowledge concerning normal stem cell migration comes from the hematopoietic system. Intriguingly, many of the factors known to govern HSC migration are also critical mediators of cancer metastasis. Under normal physiological conditions, small numbers of HSCs are found in the bloodstream. In response to mobilizing agents such as G-CSF and signaling through the laminin receptor, increased numbers of HSCs circulate out of the bone marrow 86. The laminin receptor has also been shown to play a key role in cancer metastasis 87. Similarly, stromal cell derived factor and its receptor CXCR4 form a critical regulatory axis for HSC migration, engraftment and homing 88, 89, 90, 91, 92, and also function in the metastasis of breast, prostate and other types of cancer 73, 93, 94, 95, 96, 97, 98, 99, 100. Matrix metalloproteinase-9 (MMP-9), belongs to a family of MMPs that plays a critical role during cancer cell invasion 101, 102, 103, 104, and it is also involved in HSC homing and migration 105, 106, 107. In addition to providing niches for HSCs, skeletal bones are also the most common sites for cancer metastasis 108. It has been shown that HSCs lacking a calcium sensing receptor (CaR) are unable to localize to the endosteal niche in the bone 109. Elevated CaR expression in primary breast cancer samples correlates positively with bone metastasis 110. Therefore, bone-specific factors such as the level of calcium ions may serve as chemo-attractants for guiding the migrating cancer cells into the bone. These striking similarities that connect normal stem cells to metastatic cancer cells raise the question of what role CSCs may play in cancer metastasis.

CSCs and metastasis

Metastasis is a complex, multi-step process. Tissue tropisms associated with cancer metastasis indicate that specific and distinct cellular and molecular mechanisms are involved. The prevailing clonal selection model of metastasis contends that genetic mutations attained late in tumorigenesis provide a selective advantage for cells to metastasize. However, recent studies now lend their support to the notion that metastasis capacity is pre-determined by genetic changes acquired at the initial stages of tumor development 112. Applying genomic approaches, molecular signatures have been defined to successfully predict poor prognosis for patients due to the metastatic potential of solid tumors. This suggests that the metastasis gene program is shared by the majority of cancer cells found in primary tumors 113, 114, 115, 116. Presently, the functional relevance of those signature genes to the underlying mechanisms for cancer progression and metastasis is not immediately obvious, partially due to the lack of significant overlap among reported signatures. Furthermore, these studies did not address the lingering questions regarding what factors govern the tissue tropism for a given cancer, which has been observed for over one hundred years 117. Fortunately, functional genomic studies have begun to shed light on the cellular and molecular mechanisms of tissue-specific metastasis 73, 118, 119, 120. Results from these studies have shown that a minor population of cancer cells within a heterogeneous breast tumor is already programmed to preferentially metastasize to specific organs. The molecular signatures of these tissue-specific metastatic cells can be distinguished from the general poor prognosis signatures. Nevertheless, it remains to be determined if the defining signature of tissue specific metastasis overlaps with the expression profile of CSCs. If an overlap exists, is there a subset of the gene profile responsible for determining the different tissue tropisms during metastasis? Part of the answer to this question likely involves the interactions between CSCs and their microenvironments.

Major clues as to the relation of niche formation and metastasis came from a recent study by Kaplan et al. 121, which characterized the initiation of the pre-metastasis niche by BMDCs after implantation of lung cancer or melanoma cells. In this study, it was shown that BMDCs are directed to the future sites of metastasis prior to cancer cell arrival by secreted factors in the cancer cell conditioned media. In addition, blocking the availability of the pre-metastasis niche components largely eliminated cancer metastasis, potentially indicating the functional importance of the pre-metastasis niche. The recruitment of BMDCs to the pre-metastasis niche could involve the extracellular matrix protein Opn. As mentioned above, Opn is a secreted protein known to play a role in metastasis of many tumors 72, 114. It has been shown to be overexpressed in subpopulations of highly bone-metastatic breast cancer cells from the MDA-MB-231 cell line. When co-overexpressed with IL-11, it induces normally non-metastatic cells to metastasize 73. One of the natural ligands for Opn is the α4β1 integrin complex, a heterodimer expressed on the surface of the recruited BMDCs in the Kaplan study. Putting these observations together, one can envision a scenario whereby tumor cells secrete factors such as Opn, which aid in recruiting BMDCs to the future sites of metastasis where they restructure the local microenvironment, making it amenable to the growth of metastatic cells. Although it remains an open question whether only CSCs are responsible for orchestrating the formation of the pre-metastasis niche, it is clear that the tumor cells are able to secrete as yet undetermined factors that can localize and initiate the pre-metastasis niche through unknown mechanisms. This study represents an important conceptual advance; it shows that incoming metastatic cells are capable of remodeling the microenvironment at preferred sites into a more permissive/supportive location.

Several characteristics of CSCs make them likely candidates to occupy and thrive in these foreign sites. First, it is theoretically possible that only CSCs within tumors have the ability to initiate and sustain cancer growth. It has been known for years that just one cell can initiate a metastatic lesion 122. Therefore, even if non-CSCs migrate (which is likely, given the number of cancer cells that can be detected in the blood), they would not be able to propagate into heterogeneously diverse metastatic lesions. Furthermore, the inherent plasticity of stem cells makes them more adept to survive in a foreign environment (albeit primed by the pre-metastasis niche) where growth factors and other signaling molecules are different than in the primary tumor site. Increased genetic instability in CSCs is also likely to provide a selective advantage in adapting to foreign sites. However, it remains to be seen whether all CSCs are equally capable of forming metastasis at different sites. Formation of organ-specific metastases may require both the CSCs property and the ability of either the mCSCs or their progenies to adapt to a particular target tissue microenvironment. Tumor-initiating capacity is required at any metastasis site along with a niche for it. However, the immediate progeny of mCSCs may soon succumb unless they have an ability to exploit that environment. In other words, CSCs may be necessary to re-initiate the tumor in the strange new environment at metastasis sites, but they will be insufficient to maintain metastatic growth if their progeny cannot survive owing to a lack of organ-specific adaptability.

In light of significant advances in metastasis and stem cell research, we propose a CSC-based model for both tumorigenesis and metastasis (depicted in Figure 1). Initial transforming events could occur either in adult stem cells, their more differentiated progenies or fused cells to give rise to CSCs. Gain of self-renewal capacity at early phases of cancer is essential for further accumulation of oncogenic transformations and eventual development of cancer. During the establishment of the CSC pool, CSCs inherit a unique set of genetic and/or epigenetic changes that determine the cancer malignancy, metastatic potential and the tissue tropism. The initial origin of CSCs may influence the phenotypes of developing cancer, including the metastatic property. Molecular crosstalk between the primary tumor and the pre-metastasis niche through secreted stimulatory signals helps govern the homing of mCSCs. Trafficking towards preferred tissues and organs of mCSCs is guided by cues such as oxygen gradients or other chemo-attractants derived from niche sites 94, 123, 124, 125.

Figure 1
figure 1

A model for tissue-specific metastasis mediated by mCSCs. Irrespective of the cell of origin, the first step of this model is a transformation event (a) after which the self-renewal capacity leads to an expansion of the CSC pool. This pool of tumor-initiating cells has the capacity to expand into a fully heterogenous primary tumor mass (b). Secretion of pre-metastasis niche forming factors (c) plays a critical role in determining the tissue tropism of the future metastatic lesion. Once the mCSCs begin to migrate through the blood (d), they are guided by homing and anchorage factors produced by the niche (e). After seeding, the local microenviroment in the niche helps determine if the mCSCs will either proliferate into a metastatic lesion directly (f), or will enter a quiescent period (g), which can be cut short by reactivation signals (h) that promote expansion into a full blown metastatic lesion. These key steps of metastasis present several potential targets for therapeutic interventions. Ø = Potential therapeutic intervention

CSCs and mCSCs at primary and secondary sites can either hijack the niches for normal stem cells or recruit new components to form a permissive niche. The availability of niche components and the influence of other factors such as immune surveillance also contribute to primary tumor and metastasis development. mCSCs may then either actively proliferate at the new site or stay dormant, similar to the quiescent state of normal stem cells. Stimulatory factors from the niche can lead to reactivation of the CSCs and formation of a metastatic lesion, which may partly explain temporal patterns of primary tumors versus secondary tumors, although this remains to be determined. Since only a few mCSCs would need to leave the primary tumor site to initiate metastasis, this could help reconcile the observation that cancer cells can be detected in distant sites long before any detectable dissemination occurs at the primary tumors. mCSCs at metastatic sites maintain most of the genetic programs acquired at the primary tumor site through self-renewal, which explains the phenotypic similarities between primary and metastatic cancers. However, mCSCs in secondary sites are able to evolve independently by accumulating additional genetic alterations that render them resistant to treatments that are effective against primary tumors.

Implication of CSCs for metastasis therapy

If mCSCs are proven to be critical for metastasis, they will have highly significant implications in the realm of cancer treatment. Our model predicts that preventative treatment for metastasis should be applied significantly earlier than current practices. The fact that stem cells rarely divide and have other cellular properties distinct from the rest of tumor population, coupled with the observation that they may have high levels of drug transporters to pump chemotherapy agents out of the cell, has led many to believe that traditional chemo- or radiation therapies are not sufficient to clear these tumor initiating cells from the body. A recent study in glioma has provided concrete evidence for the ability of CSCs to contribute to radioresistance through preferential activation of the DNA damage checkpoint response and an increase in DNA repair capacity 126. Therefore, targeted therapies to eliminate CSCs and mCSCs could lead to a revolution in the way cancer is treated. A few possible mechanisms to target these cells are presented below and highlighted in Figure 1.

Self-renewal and differentiation

If proliferation of CSCs is important for the growth of the primary tumor as well as progression to metastatic disease, one obvious target for eliminating these cells is to hamper their self-renewal capacity. Previously developed therapies such as cyclopamine (targeting the hedgehog signaling) and exisulind, bromoindirubin-3′-oxime, and imatinib (targeting the Wnt/β-catenin pathways) have looked to inhibit self-renewal pathways likely critical to CSCs with varying levels of success (for a more detailed review, see 127). By forcing CSCs to differentiate, their ability to self-renew would indirectly be eliminated. One of the most successful (and general) differentiation inducers to be used in clinical practice has been all trans-retinoic acid for patients with acute premyelocytic leukemia. Other general inducers such as TPA, DMSO, butyric acid and vitamin D3 have also been tried for solid tumors; however, using targeted therapies such as nerve growth factors, PPARα activators or compounds such as vesnarinone may be more successful (see 128 for a review). Better functional characterization of CSC self-renewal and differentiation will translate into more targeted therapies compared to this first round of general differentiation inducers.

Drug transporters

Another clinically relevant inherent property of stem cells is their ability to pump drugs out of the cell through the use of the ABC family of drug transporters. If these efflux pumps could be inhibited, CSCs could be more susceptible to current or newly designed chemotherapeutic agents used in conjunction. Previous attempts to inhibit the ABCB1 transporter resulted in limited clinical success, but a new wave of ABCG2 specific inhibitors may lead to better results 129.

Homing/seeding

If mCSCs are highly mobile and able to generate metastases, then targeting the homing or seeding of these cells in the potential metastatic niches at the initial time of tumor presentation could significantly inhibit disease progression. In fact, blocking the homing factor CXCR4 has been shown to effectively prevent both primary tumor formation as well as metastasis in animal models 97. Identification and characterization of niches for mCSCs will be helpful in identifying new targets for blocking the seeding of mCSCs. Furthermore, characterizing the secreted factors that set up the pre-metastasis niche could provide both diagnostic as well as therapeutic benefits.

Reactivation

Targeting homing and seeding of mCSCs may be technically difficult because by the time of tumor presentation, mCSCs or secreted factors may have already migrated to set up the pre-metastasis niche. Therefore, another potential clinical target would be to block the reactivation of dormant mCSCs at the metastatic sites. Therapies based on this idea are far off since the dormancy model needs to be validated and characterized before picking targets for drug therapy. Insights into the organization of the mCSC niche will provide clues as to the pathways associated with reactivation.

Cautions

With all of the previously mentioned therapeutic targets, there are two important caveats that must be considered. First, many of the molecular mechanisms that govern normal stem cell function also govern CSC function. Therefore, agents to target these pathways could have harmful effects on the homeostatic function of normal stem cells. As yet, it is unknown whether or not the increased proliferation of CSCs could provide a therapeutic window whereby a drug would more likely target CSCs over normal stem cells. Better discrimination between normal stem cells and CSCs will likely provide for targeted therapeutics. For instance, differential dependency on PTEN function of leukemia initiating cells versus HSCs will be useful in discovering specific therapeutic targets 46. The precise degree of discrimination that is needed between normal and CSCs is likely to vary from system to system as targeting mammary or prostate stem cells, in addition to their cancerous counterparts, would not be detrimental to patients willing to undergo a mastectomy or prostatectomy. The second caveat to consider is that the current standard of measuring the therapeutic benefit of a potential therapy (i.e. the amount the drug can shrink a tumor) is potentially problematic for CSC specific therapies. As the CSC population is a minority population within the whole tumor, administration of agents to eliminate them alone, or in combination with conventional chemotherapies, will most likely not shrink the primary tumor to a significant degree. Studies to evaluate the effectiveness of these treatments should instead look for decreased reoccurrence or metastasis formation as a measure of drug effectiveness. While a number of drugs that may have unintentionally targeted CSCs have had limited success, future experiments to better understand the origin and function of CSCs in primary tumorigenesis and, especially, metastasis may lead to a new wave of therapeutic agents with the potential to go beyond the traditional way cancer is viewed and treated.