Concise Review: Cancer Stem Cells
In vitro models †‡§
Author contributions: JPM did research and writing.
Disclosure of potential conflicts of interest is found at the end of this article.
First published online in STEM CELLSEXPRESS November 10, 2011.
The current resurgence of interest in the cancer stem cell (CSC) hypothesis as possibly providing a unifying theory of cancer biology is fueled by the growing body of work on normal adult tissue stem cells and the promise that CSC may hold the key to one of the central problems of clinical oncology: tumor recurrence. Many studies suggest that the microenvironment plays a role, perhaps a seminal one, in cancer development and progression. In addition, the possibility that the stem cell-like component of tumors is capable of rapid and reversible changes of phenotype raises questions concerning studies with these populations and the application of what we learn to the clinical situation. These types of questions are extremely difficult to study using in vivo models or freshly isolated cells. Established cell lines grown in defined conditions provide important model systems for these studies. There are three types of in vitro models for CSCs: (a) selected subpopulations of existing tumor lines (derived from serum-containing medium; (b) creation of lines from tumor or normal cells by genetic manipulation; or (c) direct in vitro selection of CSC from tumors or sorted tumor cells using defined serum-free conditions. We review the problems associated with creating and maintaining in vitro cultures of CSCs and the progress to date on the establishment of these important models. STEM CELLS 2012; 30:95–99.
The concept of cancer stem cells (CSC) has been discussed for almost a century (for reviews see [1–3]). The current resurgence of interest in this hypothesis as possibly providing a unifying theory of cancer biology is fueled by the growing body of work on normal adult tissue stem cells and the promise that CSC may hold the key to one of the central problems of clinical oncology: tumor recurrence . Typically, the development of an in vivo model allows rapid advances in understanding the factors that regulate cell and organ homeostasis and differentiation: autocrine and paracrine growth factors, extracellular matrix, and cell–cell interactions. Many studies suggest that the microenvironment plays a role, perhaps a seminal one, in cancer development and progression [5–8]. Thus, in vitro model systems would be expected to play an important role in understanding these factors. There is recognition that in vitro systems would be useful in understanding CSC. However, the progress in developing such models has been slow . Many problems involved in developing in vitro CSC models are shared with the field as whole: defining a “cancer stem cell”; assays for “stemness”; and obtaining unique markers allowing the rapid identification, analysis and/or isolation of cells meeting the above criteria. We will briefly review these issues as they pertain to in vitro model systems and then turn out attention to the state of development of such systems for the study of CSC.
The CSC hypothesis states that there exists, within a tumor, a minority cell type that has the characteristics of stem cells; that is, these cells can self-renew and can differentiate to form all of the cell types that constitute the original tumor from a small number of cells . It seems clear that, at least in hematopoetic  and some solid tumors including pancreatic , prostate , colon [1, 12, 13], breast [8, 14], lung , and brain [16, 17], a small subset of cells can be isolated that can self-renew and form well-differentiated tumors similar to that of the patient's tumor from which they arise. However, the CSC hypothesis may not be true for all tumor types or for all of these cells all of the time [18–21]. In fact, it has been suggested that stemness may be a dynamic state, which is a function of the cell's interaction with the environment .
Most CSCs are operationally defined by the enrichment of a subpopulation of tumor cells using prospectively defined cell surface “markers” and sorting for these cells using immunomagnetic bead or cytometric-based technologies. These populations are demonstrated to be capable of self-renewal and forming differentiated tumors in immune-deficient mice. Cell surface markers for hematopoetic stem cells (HSCs) have been elucidated, and the separation technologies have been applied for isolating CSC for hematopoetic cancers [2, 10]. However, the markers used to enrich for solid tumor CSC have, in most cases, been drawn from this same set of HSC markers, especially when little is known of the actual adult stem cell markers in the tissues from which the tumors arise . Thus, there has been significant disagreement regarding the utility of specific markers, for example, CD44 or CD133, for defining the CSC for solid tumors [18, 19, 24].
Certain other investigators have turned to markers found in embryonic or adult tissue stem cells such as: aldehyde dehydrogenase (ALDH1A1) [25, 26], Lgr5 [5, 27], and ephB2 . Another approach has been to use stem cell properties such as dye exclusion or quiescence as selection criteria . However, as normal tissue stem cell division rates vary widely , quiescence is obviously not a uniformly useful criterion. Added confusion arises from a widespread conflation of the selection marker with the biologically defined CSC.
There is general agreement that CSC should be defined based on biologic criteria. That is, a CSC is a tumor cell that is uniquely capable of self-renewal and of regenerating the original tumor phenotype . Each of these aspects is both qualitatively and quantitatively dependent on the assay used and the state/environment of the cells being assayed.
The assays for self-renewal are straightforward but time consuming. Two assays are: the ability to continue to form tumors after multiple (three to four) animal-to-animal transplants and the ability to expand the desired population in vitro for more than 60 population doublings. Defining self-renewal for CSC can be complicated by the fact that all cell lines have been naturally selected, or experimentally induced, to exhibit immortality. Asymmetric division, inherent in the ability of the CSC to form differentiated progeny as well as continue to proliferate, is a more difficult concept to prove but is not necessary for stem cell identity per se [29, 30]. As complete forced differentiation of CSC is a theoretically useful cancer treatment [31, 32], deriving assay conditions that allow study of this crucial step would be worthwhile.
Assaying tumor cells, especially human cells, for tumorigenicity is complicated by both technical and biological problems. The ability of a cell to form a tumor is highly dependent on the genetic background of the host, with more immune-deficient animals allowing for qualitative and quantitative increases in tumorigenicity index (TI) [10, 18]. One inescapable difference in the human and experimental tumor models is that the xenograft tumors grown in mice contain murine stroma and vasculature and human tumor cells. There is a growing understanding of the importance of the tumor stroma in initiating and regulating tumor growth  and the importance of the microenvironment in regulating stem cells in general and tumor stem cells as well [5, 7]. The site of implantation (subcutaneous, subrenal, or orthotopic), cell carrier (e.g., collagen, Matrigel, other cells), extent of cell damage, state of replication, and number of input cells, and time in vivo before assessment all play a significant role in determining TI. The real goal here should be to find conditions that most closely replicate the in situ tumor environment not just fastest growth. Comparison of the TI of putative CSC with other control populations has been the strength of the marker-sorted populations, which allow the immediate comparison of marker plus versus marker minus populations. However, the differences in these two populations are almost always quantitative rather than qualitative. Culture systems allow manipulations of the cells and environment but not the immediate comparisons to the nondividing tumor cell populations as these are rapidly lost in vitro.
Defining “unique subset”
The uniqueness of the presumed CSC and the quantitative number of such cells present in tumors has been perhaps the most contentious, and hardest to quantify, aspect of the definition [19, 33]. The main issues are dependence of the quantitative outcome on assay variables and differences in markers and methods used to isolate CSC. More recent data calls into question the idea that the CSC is a “minority” cell type in some tumor types such as melanoma or even that some of the marker-defined subclasses of cells are more tumorigenic than others . It would, in any case, benefit the field if there were different, marker-independent, means of isolating the CSC-like populations of cells from the original solid tumor. There is growing evidence that many solid tumors may be composed of several distinct subtypes of tumors [34, 17], which may have distinct CSC. Finally, CSC may be characterized by a unique plasticity allowing epigenetic or microenvironmentally driven reversible changes to occur thus complicating definition or selection based on changeable characteristics [7, 20, 21, 35]. Defined in vitro models should be used to study this plasticity, its regulation, and its implications for cancer treatment .
The biologically based assays for CSC will necessarily use living cells and results will depend on the makeup and quality of the input cells as well as the assay conditions. Many studies with human tumor CSC use cells directly isolated from the patient tumor or xenografts. This approach has the advantage that there is a close temporal relationship between tumor and cells being studied, with the concomitant disadvantage that the cells isolated must always represent a snapshot of a subpopulation of the cells in the original tumor. Solid tumors must be extensively treated mechanically and/or enzymatically to disperse the cells. The method and harshness of the treatment will depend on tumor variables such as the amount of connective tissue, necrotic tissue, and vasculature, which will vary with tumor type, from patient to patient, and even within a single tumor. Generally, digested tumor pieces are filtered and the single-cell suspension is enriched using various sorting methods and assayed directly, or cultured for further enrichment before assay (e.g., [12, 37]. The longer the enzymatic digestion process, the higher the yield of single cells and the greater the damage to the cells obtained, including depletion of cell surface proteins. A shorter digestion process results in more undigested tissue being discarded with no measure of how representative the easily dissociated cells may be of the entire tumor. The use of cultured cells has both advantages and disadvantages compared to the use of directly dispersed and sorted cells. Culture, particularly in defined conditions, allows more control over the viability and environment of the cells.
IN VITRO MODELS FOR CSCS
There are three types of in vitro models for CSCs that have been described in the literature: (a) selected subpopulations of existing tumor lines (derived from serum-containing medium) [14, 38, 39]; (b) creation of lines from tumor or normal cells by genetic manipulation [21, 40, 41]; or (c) direct in vitro selection of CSC from tumors [17, 42] or sorted tumor cells [12, 13, 43] using defined serum-free conditions .
Existing Tumor-Derived Cell Lines
Multiple human tumor-derived cell lines have been established over the last 60 years using serum-containing media and plastic substrates. These lines, however, may be much changed from the original tumor because of selection of mutations or conditions that allow for faster growth or the inappropriateness of the medium for maintaining the original characteristics of the cells. While these lines are capable of self-renewal and are tumorigenic to varying degrees, many of the tumors that arise from these lines are uniformly poorly differentiated and therefore not thought to reflect the original patient tumor morphology. It is not surprising, however, that cells can be selected from such lines which self-renew and even form differentiated tumors. Recently Yeung et al.  have shown that a subset of cells can be isolated from several long-established colon cancer-derived cell lines, which have the properties of CSC, including the ability to form well-differentiated tumors in mice. Similar studies have shown that selected clones of an established breast or prostate cancer line have “CSC-like” properties [14, 39]. Van Steveren et al. have reviewed the pros and cons of using long-established cancer cell lines and conclude:“….the available human cancer cell lines are not exactly representative of the in vivo cell population of cancers. However their representativity can be much improved by using conditions, not designed for their maximal proliferation as classically, but closer to those of the in vivo situation.”
However, interpretation of results from subfractions of established cell lines as broadly representing the properties of the hypothetical CSC in the tumor is subject to misinterpretation. We cannot know how these lines have been altered by their initial selection and growth over many years in serum-supplemented medium  which can, itself, lead to genetic instability [45, 46]. This is of particular importance as freshly isolated cancer stem-like cells have been shown, in several systems, to be altered, differentiated, or lost in serum-containing medium [12, 42, 43].
Creation of Lines Using Genetic Modification
Another approach for creating in vitro models of cancer cells has been to transform normal cells with a distinct set of genes that immortalize them [40, 41, 47]. Interestingly, the phenotype of the transformed cells was dependent on the conditions originally used to culture the normal cells before transformation . Mani et al. have shown that ectopic expression of genes known to induce epithelial–mesenchymal transition (EMT) (TWIST and SNAI1) in these immortal cells can give them CSC-like characteristics including the ability to form well-differentiated tumors in vivo . The EMT may well play a role in the process of metastasis [48, 49]. These lines have been particularly useful in helping to identify the interplay of genetics and environment in creating stem-like properties and in identifying the EMT as potentially important in CSC biology. However, although they show what biological characteristics can be acquired by the overexpression of specific genes, they do not tell us what actually happens in vivo to make a cancer undergo uncontrolled growth.
In Vitro-Defined Conditions to Select and/or Expand Enriched CSC Populations
A third approach has been to determine defined in vitro conditions that allow the predictable selection and expansion of unselected or marker-enriched CSC-like populations from tumors. The use of defined serum-free culture conditions is a powerful method to select one particular cell type from mixed populations and has been particularly important in creating in vitro models for stem and progenitor cell expansion and differentiation [45, 50, 51]. These conditions always require both elimination of serum and the provision of growth factors (e.g., epidermal growth factor) and extracellular matrix  or three-dimensional cell configurations such as “tumorspheres” [12, 13, 15, 37, 43]. A strength of this approach is that it allows/requires the definition of the correct environment (i.e., the niche characteristics) required to maintain the stem-like cells. This approach has been used to select and expand pure cell populations with CSC characteristics from multiple patients with colon and lung cancers  and gliomas . Tumorspheres containing mixed populations can also be expanded [12, 37] or induced to differentiate in vitro  or in xenograft models. Pollard et al.  reported that defined serum-free culture on matrix protein (laminin) gave an advantage over tumorsphere cultures as a method of expanding pure populations of glioma initiating cells. Our work with defined adherent cultures of ovarian, lung, and colon tumor initiating cells (unpublished data) supports this conclusion. Culture in serum-free medium has been shown to lead to greater chromosomal stability in vitro compared with culture in serum-containing medium . More studies need to be performed to look at the long-term stability of CSC expanded in defined media. However, use of cell banks, frozen at early passage, can minimize this problem.
There is strong evidence that the microenvironment plays a strong role in the initiation and progression of cancer [5–8]. Indeed, it has been suggested that stemness might be a state definable only in a biological context rather than a biochemically distinct entity . While this may seem an esoteric distinction, it has important consequences for our thinking about drug development and patient treatment. By their nature, in vitro models are especially useful in understanding the role of the environment in the control of stem cell plasticity and differentiation. Defined media also allow significant expansion, or derivation of permanent lines, of CSC which can then be banked as frozen cells so that experiments can be readily compared over time, and between lines derived from patients with different tumor phenotypes. Changing the environment leads to controlled differentiation in vitro and allows generation of fully differentiated tumors, making them ideal for generating more relevant in vivo models . Additionally, the expanded cell number provides screening tools for research and drug discovery [53, 54].
STEM CELL PLASTICITY
One important idea, which has been associated with normal stem cells, is the concept of stem cell plasticity. Several different approaches suggest that plasticity may be exhibited by CSCs or a subset of these cells [7, 20, 36, 55]. Melanomas contain a set of “primitive” stem cells that can create “vasculature-mimics” as well as the tumor tissue . Epigenetic changes have also proven to be relevant to oncology research and in the clinic . As a number of studies have shown that specific marker expression and even drug sensitivity  can be a reversible phenomenon in some cell populations, the widespread occurrence of plasticity as a property of CSC must be seriously considered. The possibility that the stem cell-like component of tumors is capable of rapid and reversible changes of phenotype raises several questions concerning studies with these populations and the application of what we learn to the clinical situation. These types of questions are extremely difficult to study using in vivo models or freshly isolated cells. Established cell lines grown in defined conditions provide important model systems for these studies.
While an overarching agreement on the validity of the CSC hypothesis is yet to be achieved, some aspects of this work will remain a part of our understanding of cancer for the foreseeable future. It also seems probable that the CSC hypothesis will permanently change oncology drug discovery and clinical development. In vitro model systems of CSC are likely to play an increasing role in this process [17, 53, 54]. Many important questions are unanswered and remain open to investigation using in vitro approaches: (a) How do CSCs relate to adult tissue stem and/or progenitor cells, metastatic cells [14, 56, 57], circulating tumor cells, treatment-resistant cells? (b) What roles do mutation and the stem cell microenvironment play in these processes? (c) How does the existence of CSC relate to disease treatment ? (d) Would eradicating the tumor CSC be sufficient to effect a cure or would CSC plasticity allow these cells to escape targeted killing and repopulate the tumor [36, 55]? (e) Could forcing CSC differentiation provide possibilities for treatment or for increasing sensitivity to current therapies [31, 32]? It seems clear that the excitement generated by the CSC hypothesis and the expenditure of time, energy, and resources in an exploration of the CSC hypothesis over the last few years will provide important insights, even if the biology proves to be more complex than originally thought [55, 59]. The use and further development of in vitro model systems for cancer stem like cells is likely to play an increasingly important role in understanding this fascinating and complex component of the disease.
I thank my colleagues at Raven Biotechnologies and MacroGenics, Inc. with whom it has been my pleasure to work with on the creation of in vitro models for the study of cancer stem cells and their use in developing drugs for the treatment of cancer. I thank Penelope Roberts, Kathleen King, Ezio Bonvini, and Paul Moore for stimulating discussions. MacroGenics, Inc. provided financial support for the research.
DISCLOSURE OF POTENTIAL CONFLICTS OF INTEREST
J.P.M. is an employee of MacroGenics and potentially owns shares or options in the company.