Abstract
In recent years cancer stem cells (CSCs) have been hypothesized to comprise only a minor subpopulation in solid tumors that drives tumor initiation, progression, and metastasis; the so-called “cancer stem cell hypothesis.” While a seemingly trivial statement about numbers, much is put at stake. If true, the conclusions of many studies of cancer cell populations could be challenged, as the bulk assay methods upon which they depend have, by, and large, taken for granted the notion that a “typical” cell of the population possesses the attributes of a cell capable of perpetuating the cancer, i.e., a CSC. In support of the CSC hypothesis, populations enriched for so-called “tumor-initiating” cells have demonstrated a corresponding increase in tumorigenicity as measured by dilution assay, although estimates have varied widely as to what the fractional contribution of tumor-initiating cells is in any given population. Some have taken this variability to suggest the CSC fraction may be nearly 100% after all, countering the CSC hypothesis, and that there are simply assay-dependent error rates in our ability to “reconfirm” CSC status at the cell level. To explore this controversy more quantitatively, we developed a simple cellular automaton model of CSC-driven tumor growth dynamics. Assuming CSC and non-stem cancer cells (CC) subpopulations coexist to some degree, we evaluated the impact of an environmentally dependent CSC symmetric division probability and a CC proliferation capacity on tumor progression and morphology. Our model predicts, as expected, that the frequency of CSC divisions that are symmetric highly influences the frequency of CSCs in the population, but goes on to predict the two frequencies can be widely divergent, and that spatial constraints will tend to increase the CSC fraction over time. Further, tumor progression times show a marked dependence on both the frequency of CSC divisions that are symmetric and on the proliferation capacities of CC. Together, these findings can explain, within the CSC hypothesis, the widely varying measures of stem cell fractions observed. In particular, although the CSC fraction is influenced by the (environmentally modifiable) CSC symmetric division probability, with the former converging to unity as the latter nears 100%, the CSC fraction becomes quite small even for symmetric division probabilities modestly lower than 100%. In the latter case, the tumor exhibits a clustered morphology and the CSC fraction steadily increases with time; more so on both counts when the death rate of CCs is higher. Such variations in CSC fraction and morphology are not only consistent with the CSC hypothesis, but lend support to it as one expected byproduct of the dynamical interactions that are predicted to take place among a relatively small CSC population, its CC counterpart, and the host compartment over time.
Introduction
Normal tissues undergo constant turnover, with cells dying due to age, injury, or shedding, and being replaced by new healthy cells. Homeostasis is accomplished by a potent subpopulation of stem cells. In recent years, a potent subpopulation of stemlike cells has also been proposed to exist as a minority population in cancers. First in leukemia and later in solid tumors, distinct cell populations were isolated that were either capable or not capable of initiating and sustaining, and re-initiating tumor growth (Furth and Kahn, 1937; Al-Hajj et al., 2003; Singh et al., 2003; Visvader and Lindeman, 2008; Vlashi et al., 2009). The picture to emerge – that of a potent cancer stem cell (CSC) that initiates and progresses the tumor, with the bulk of the growing tumor being composed of replication-limited cancer cells (CCs) – stands in marked contrast to the long-established paradigm that cancer cells typically are long-lived, escape cell death, and have limitless replicative potential (Hanahan and Weinberg, 2000, 2011), and argues against one recent study (Quintana et al., 2008), which has been interpreted to suggest we may simply be “under-assaying” the preponderant stem cell population (Baker, 2008). Summing up this alternative paradigm is the CSC hypothesis, perhaps better described as a cancer “non-stem cell” hypothesis, which posits that in fact only a few cells in the tumor population exhibit the immortal, stemlike trait. Confusion sometimes accompanying the CSC terminology regards the cell of origin of the disease. While the term “CSC” clearly suggests these cells possess stemlike qualities, this should not be taken to suggest they originate from normal stem cells. Indeed, the literature is split on the matter of origin (Haeno et al., 2009; Rahman et al., 2009; Shibata and Shen, 2013), with debates occurring occasionally even within the same tumor type, as in the case in glioblastoma (Stiles and Rowitch, 2008; Hambardzumyan et al., 2011).
Populations enriched for stemness can be isolated using different surface markers. The number of cells from such enriched populations that is necessary to form tumors gives an indication of the fraction of cells that are CSCs in the primary tumor. We calculated these ratios from data reported in the literature (Visvader and Lindeman, 2008) (Table 1), and found these tend to support numerous other reports that CSCs are indeed a rare population within a tumor (Reya et al., 2001; Pardal et al., 2003). In addition to identification of CSCs through surface proteins in vitro and in vivo mouse xenograft transplantation assays, novel approaches emerge that trace tumor hierarchy and help estimate CSC kinetics and frequency in spontaneous tumors or orthotropic models. One approach to monitor the division kinetics of stem and progenitor cells in normal epithelial tissues, skin papilloma, and invasive squamous cell carcinoma during unperturbed growth emerged from clonal analysis using genetic lineage tracing in mice (Driessens et al., 2012). Gao et al. (2013) used an integrated experimental and cellular Potts model approach to simulate glioblastoma population growth and response to irradiation, which identified the (a)symmetric division kinetics of glioblastoma stem cells necessary to reproduce the observed ratio of 2–3% of such cells. Another integrated approach of single-molecule genomic data, spatial agent-based modeling, and statistical inference was recently introduced to derive tumor ancestral trees in patient-specific colorectal cancer samples that lead to the observation of a CSC fraction of 0.5–4% (Sottoriva et al., 2013).
Table 1
| Tumor type | Cells expressing CSC marker (%) | Minimal number of cells expressing marker for tumor formation | Calculated cancer stem cell ratio |
|---|---|---|---|
| Breast | 11–35 | 200 | 1.1 × 10−3 |
| ND | 2000 | ||
| 3–10 | 500 | ||
| Brain | 19–29 | 100 | 3.3 × 10−4 |
| 6–21 | 100 | ||
| Colon | 1.8–25 | 200 | 5.4 × 10−4 |
| 0.7–6 | 3000 | ||
| 0.03–38 | 200 | ||
| Head and neck | 0.1–42 | 5000 | 4.2 × 10−5 |
| Pancreas | 0.2–0.8 | 100 | 3.5 × 10−5 |
| 1–3 | 500 | ||
| Lung | 0.32–22 | 10,000 | 1.1 × 10−4 |
| Liver | 0.03–6 | 5000 | 6 × 10−6 |
Cancer stem cells in solid tumors.
Adapted from Visvader and Lindeman (2008).
One mechanism responsible for establishing the CSC fraction within a tumor is the relative frequency with which CSCs either create another CSC (by symmetric division) or a non-CSCs (by asymmetric division) (Caussinus and Hirth, 2007; Dingli et al., 2007b). Mechanisms known to directly affect the symmetric division probability, in turn, include availability of certain host growth factors such as EGF, and growth-factor-rich niches, which can skew division modes in favor of symmetric production of CSC up to 85% (Lathia et al., 2011). Another mechanism responsible for the observed CSC fraction in tumors is factor-independent, and may be traced to the aggregate population-level action of cell proliferation, migration, and apoptosis; a process we have previously described as “self-metastatic” growth (Norton, 2005; Enderling et al., 2009b). Underlying this notion, each CSC can only form a cluster of limited size (Prehn, 1991), until such time as it can opportunistically migrate out of its current cluster to seed a new cluster nearby.
To show how these influences comprising the CSC hypothesis can give rise to realistic tumor growth dynamics and morphologies, we used an agent-based cellular automaton model of tumor population dynamics that considers the kinetics and interactions of CSC and CCs. Consistent with observation, we show that host-dependent variations in (a)symmetric CSC division ratios can yield tumors with substantially different CSC pool sizes and overall tumor morphologies. Furthermore, we show, by virtue of the properties of CSCs and their progeny, that the CSC fraction in a tumor grows over time, regardless of how quickly the tumor as a whole grows. As will be argued, these findings can explain the large variation in CSC fractions within and among tumors reported throughout the literature (Reya et al., 2001; Visvader and Lindeman, 2008), and offer a new paradigm for cancer development within the CSC hypothesis.
Materials and Methods
We use an agent-based cellular automaton model to describe the behavior of individual tumor cells dependent on intrinsic mechanisms of proliferation, migration, and cell death. By tracking the fate of multiple cells over time we simulated the emergence of interacting tumor cell populations that compete for the same environment (see Deutsch and Dormann, 2005 for an overview of similar approaches). Such theoretical frameworks are increasingly utilized to investigate different aspects of the CSC hypothesis (Deasy et al., 2003; Dingli and Michor, 2006; Ganguly and Puri, 2006; Ashkenazi et al., 2007; Bankhead et al., 2007; Dingli et al., 2007b; Michor, 2008; Piotrowska et al., 2008; Galle et al., 2009; Glauche et al., 2009; Johnston et al., 2010; Sottoriva et al., 2010). Ganguly and Puri (2006) developed a compartment model of normal stem cells, early and late progenitors, and mature cells in the neural lineage as well as their abnormal counterparts. Through numerical simulation of physiologic homeostasis in their deterministic ordinary differential equation model, the authors explored the impact of mutations in the stem cell and early progenitor populations. As stem cells have a larger proliferation potential, they found that mutations in stem cells lead to a larger tumor population growth rate. Considering normal and abnormal stem and differentiated cell populations of the hematopoietic system, Dingli and Michor (2006) showed in a simple ordinary differential equation model that successful therapy must eradicate the CSCs. Therapy that targets mature CCs or partially induces differentiation of stem cells is unable to provide tumor control. Recent mathematical investigations into the fraction of CSCs in solid tumors either assumed a fixed proportion of these cells and explored the cell kinetic parameters in a hierarchical ordinary differential equation model that lead to proportional stability in the cancer lineage (Molina-Peña and Álvarez, 2012), or simulate the exponential phase of tumor growth and observe a constant minor proportion of CSCs in an ordinary differential equation model (Johnston et al., 2010) or in a cellular automaton approach (Morton et al., 2011). Recently, Hillen et al. (2013) used reaction-birth processes and developed a mean-field integro-differential equation system to describe spatio-temporal tumor growth under the CSC hypothesis. Analyzing the simplified ordinary differential equation system of their model, the authors were able to show that tumor growth accelerates with increased cell death and that the tumor population monotonically evolves to a pure CSC state. Using a partial differential equation approach to simulate the spatio-temporal dynamics of cell lineage in solid tumors, Youssefpour et al. (2012) were able to observe complex pattering with CSCs being predominantly located in individual clusters at the outer rim of the total population in response to a variety of cellular feedback mechanisms and oxygen tension. Similar to our agent-based model investigations into CSC-driven solid tumor growth (Enderling et al., 2009a,b), Sottoriva et al. (2010) recently developed a hybrid cellular automaton model to study tumor morphology and phenotypical heterogeneity in the classical cancer model where all cells can be considered CSCs, and in the CSC model with populations heterogeneous for proliferation potential. In addition to increasing invasiveness with decreased CSC fraction, the model revealed a constant CSC fraction during exponential growth phases. While Sottoriva and co-workers assumed intratumoral proliferation through pushing adjacent cells toward the tumor periphery, we set out to explore tumor growth dynamics and morphology evolution when proliferation is restricted to the tumor outer rim as a result of competition for space in the tumor interior (Brú et al., 2003; Drasdo and Höhme, 2005; Galle et al., 2009). We will explore global cancer features in an agent-based model as a function of environmentally modulated CSC symmetric division rates as well as the proliferative potential of the CC population, and discuss evolution of the tumor population and CSC fraction during local tissue invasion.
A detailed description of the agent-based model assumptions and biological motivation can be found elsewhere (Enderling et al., 2009a,b; Enderling and Hahnfeldt, 2011). To summarize, at time t = 0, we initiate a single cell in the center of our computational lattice (domain) of 10 mm × 10 mm, subdivided into 1000 × 1000 equal-sized lattice points of (10 μm)2 that can hold at the most one cell at any time. By simulating (a)symmetric cell proliferation, migration, and cell death kinetics at discrete time intervals Δt, a population of cancer cells emerges, which we track until it reaches 100,000 cells. In this procedure, we assume CCs are able to proliferate a certain number of times, ρmax, before inevitable cell death and ultimate removal from the simulation. For CSCs, we assume ρmax = ∞. CSCs divide symmetrically with fixed probability ps, and asymmetrically with probability 1 − ps. Cells need to mature through the cell cycle before division can occur, which takes a cell-type-dependent time τ. With available adjacent space, a cell can migrate with rate μ. Otherwise, the cell is forced into quiescence until space becomes available. With probability α, CCs undergo spontaneous cell death and vacate the space they occupy. CSCs are assumed to be immortal (α = 0). A flowchart of the simulation process and decisions at the cell level is shown in Figure 1.
Figure 1
Results
Symmetric stem cell division probability ps and tumor growth
Tumor growth is simulated and analyzed for various stem cell division probabilities ps = (0.1, 0.25, 0.5, 0.75, 0.99) and various progeny proliferation capacities ρmax = (10, 15, 20), which are in line with reported progeny division potentials for different tissues (Bernard et al., 2003; Ashkenazi et al., 2007). As previously shown, tumors in which CCs have high values of ρmax grow significantly slower than tumors with CCs with limited proliferation capacity (low ρmax) (Enderling et al., 2009a). With proliferation being dependent on available space, tumor growth is restricted to cells on the outer rim of cell clusters, which is frequently observed in vitro and in vivo (Brú et al., 2003; Drasdo and Höhme, 2005; Galle et al., 2009). CSCs become “trapped” in the tumor core and are forced into quiescence until space becomes available again – either after adjacent CCs have migrated away or died. These kinetics have been shown to be inversely dependent on CC proliferation capacity and death rate (Enderling and Hahnfeldt, 2011; Morton et al., 2011).
We now relate the effect of symmetric CSC division probability ps on tumor growth. A high symmetric CSC division probability (ps = 0.99; i.e., 99%) is found to yield tumors with a very large CSC population (93%). Tumors of 100,000 cells are formed within 49 days regardless of the CC proliferation capacity. Lower symmetric stem cell division probabilities of ps = 75% and ps = 50% result in both a smaller stem cell compartment (15.5 and 5.4%, respectively) and more progeny CCs that act to encapsulate and slow the expansion of the available CSCs, prolonging tumor growth to 56 and 72 days, respectively, for ρmax = 10, and 56(63) and 226(586) days for ρmax = 15(20) (Figure 2). CSC fractions are reduced further to 1.8 and 0.5% for ps = 25% and 10%, respectively. In the latter case, tumor growth takes as long as 3351(1451 and 282) days for ρmax = 20(15 and 10), respectively.
Figure 2
The intrinsic and/or extrinsic mechanisms that regulate (a)symmetric CSC division are not yet understood, but modeling predicts that if the symmetric division probability is even modestly less than 100% – which it must be given the observed cell fate heterogeneity within a tumor – the CSC compartment rapidly becomes a minor subpopulation within tumors (Figure 2) and remains that way at least for a long while.
Symmetric stem cell division probability ps and tumor morphology
Within the CSC hypothesis tumor growth follows an interesting pattern. CSCs can only form tumor clusters of limited size before self-inhibition (Prehn, 1991). Tumor growth and progression is achieved by shedding of CSCs from the tumor clusters and seeding of new clusters nearby – a process termed self-metastatic growth (Norton, 2005; Enderling et al., 2009b, 2010). Figure 3 shows the spatio-temporal evolution of tumor morphology for various symmetric stem cell division probabilities and thus various stem cell pool sizes within the tumors. As expected, and in line with the literature, homogeneous tumor populations consisting of only stem cells (or a sufficiently large stem cell compartment) grow in a radially symmetric manner (Drasdo and Höhme, 2005; Enderling and Hahnfeldt, 2011). Small symmetric stem cell division probabilities, by contrast, result in CSCs being scattered throughout the tumors through self-metastatic progression. As this continues, the tumor clusters become bigger and intermingle, giving rise to clusters of stem cells, likened to stem cell niches in normal tissues (Figure 3).
Figure 3
Tumor morphology is often used as a measure of tumor invasiveness. A compact, circular morphology is associated with less invasiveness than an irregular, fingering morphology (Anderson, 2005; Anderson et al., 2006). To assess this, we measured tumor compactness, or “circularity,” i.e., tumor density as a function of radial distance from the tumor’s center of mass, for various symmetric division probabilities ps (Figure 4). It is observed that tumor morphology can be described by three distinct regions: (i) a non-linear (saturated) region without significant changes over time in the bulk tumor core close to the center of mass, (ii) a deterministic region dominated by cell proliferation and diffusion, and (iii) a highly fluctuating outer rim zone distant from the center of mass with a low cell density that is dominated by random cell migration (Hatzikirou et al., 2010). Tumors comprised primarily of CSCs form a very dense, saturated tumor without a largely fluctuating outer rim, as the effect of random migration at the tumor boundary is nullified by continuous stem cell proliferation. A smaller dense core and a larger deterministic and highly fluctuating self-metastatic region characterize tumors with lower stem cell fractions. Of note, in tumors with larger progeny proliferation capacities [ρmax = (15, 20)] there is a non-linear, U-shaped dependence of tumor compactness on stem cell fraction. Smaller and larger symmetric stem cell division probabilities ps yield a more compact tumor morphology, whereas a symmetric division probability of ps = 50% features the most invasive morphology with the smallest dense core and the most widespread proliferative and motile regions.
Figure 4
Stem cell fraction as a function of time
Beyond the dependence of tumor growth dynamics on cell kinetic parameters, environmental constraints may also modulate tumor progression and architecture. Tumors growing in vivo compete with the environment not only for oxygen and nutrients (which are not very limiting in the early phases of tumor growth), but also for space to grow and proliferate. Different host tissues, such as epithelial membranes or dense muscle structures (Figure 5) retard tumor progression. Single cells will eventually be able to infiltrate the tissue and thus form a path for local invasion. We introduce a “host tissue” in our computational domain by discretizing a hematoxylin-eosin (H&E)-stained tissue sample of a murine Lewis lung carcinoma – mouse muscle tissue interface. For the purpose of our study, we did not model anatomically precise cell structures but limited the tissue structure to average local cell densities (Figure 5). We introduce a single CSC in the top right corner of the domain and simulate tumor growth with either ps = 100%, representing a pure CSC tumor, or ps = 10%. Tumors consisting solely of stem cells quickly populate the empty space representative of tissue with low density, and with a reduced growth rate into the dense muscle structure and less dense tissue areas beyond. Heterogeneous tumor populations (ps = 10%) populate the less dense tissue via formation of self-metastases, as previously described (Enderling et al., 2009b), and the denser structures with a further reduced growth rate. Snapshots of the spatial distribution of 13,000 tumor cells for both tumors (ps = 100 and 10%) reveal no difference in morphology (Figure 5). The circular and self-metastatic tumor morphologies evident in the respective tumors without host constraints (c.f., Figure 3) are no longer present. Instead, in vivo tumor morphology is dictated by host tissue architecture. What is seen, however, is a significant difference in tumor growth and growth rate history. While stem cell tumors fill the available space and form a mass of 13,000 cells within about 100 days, the heterogeneous tumor takes 20 times longer (72 months) to reach a comparable size and morphology (Figure 5). In the presented sample simulation, the empty space to the right of the muscle tissue is populated by the heterogeneous tumor within 12 months, but the invasion of the muscle structure takes significantly longer (another 60 months), during which time the overall CSC fraction goes from 60/9500 (0.6%) to 2050/13,000 (15.8%). Simulation snapshots at different time points clarify the heterogeneous tumor growth dynamics and stem cell fraction evolution (Figure 5). When the tumor population reaches the dense muscle structure, single cells try to invade the narrow gaps between muscle cells. As more than 99% of the tumor is comprised of CCs at the time invasion commences, the infiltrating and invading cells can only form small clusters of cells that eventually die out, blocking invasion routes, and inhibiting tumor invasion. Only the eventual opportunistic infiltration of a CSC results in successful metastatic seeding of these interstices within the muscle architecture. Figure 5 shows how these spaces fill with microscopic cancer nodules that over time disappear again, awaiting the chance entry of a stem cell. The necessity of rare CSCs, which are predominantly migration- and proliferation-inhibited by their own progeny as well as by host tissue, to infiltrate and invade host tissue structures can, at least in part, explain the frequently observed poor efficiency of metastization (Luzzi et al., 1998; Wyckoff et al., 2000; Dingli et al., 2007a). As remarked, the overall CSC frequency increases over time as dying CCs get opportunistically replaced by CSCs. Such a time-dependent tumor stem cell fraction offers one plausible explanation for the wide spread of stem cell fractions reported in the literature even for tumors of the same tissue of origin (Quintana et al., 2008; Visvader and Lindeman, 2008) (c.f., Table 1). These results are amplified when the CC death rate is higher. With competition for space being a pivotal tumor-inhibiting factor, the introduction of a 10-fold higher spontaneous death rate α in CCs results in accelerated tumor growth and host tissue invasion, along with a more rapid increase in CSC number and percentage within the tumor (Figure 6).
Figure 5
Figure 6
Discussion
The existence of a minor subpopulation of CSCs within a tumor that drives tumor initiation, growth, and progression is an attractive hypothesis to explain primary tumor dynamics and transplantation experiments. The existence of a minor fraction of CSCs in leukemia has long been appreciated (Furth and Kahn, 1937), but over the last 10 years or so, CSC-like populations have also been reported in solid tumors of, for example, the breast, brain, prostate, and colon (Al-Hajj et al., 2003; Singh et al., 2003; Cammareri et al., 2008; Hurt et al., 2008). The reported frequencies with which these cells occur in a tumor varies by many orders of magnitude; dependent, for example, on the chosen experimental setup and purification methods (Quintana et al., 2008; Visvader and Lindeman, 2008). Furthermore, the size of the CSC pool can be modulated by availability of certain host growth factors like EGF (Lathia et al., 2011), Sonic hedgehog (Takezaki et al., 2011), Wnt (Vermeulen et al., 2010), or Notch (Wang et al., 2009), supporting the idea of a CSC niche (He et al., 2009; Borovski et al., 2011). Here we present a cellular automaton model of tumor growth and invasion of heterogeneous cancer populations comprised of CSCs and their progeny. Simulations of the model reveal multiple indications on the fraction of CSCs in solid tumors: (i) intrinsic stem cell symmetric division probabilities results in tumors with different stem cell ratios and morphologies, (ii) CSCs are intrinsically a minor subpopulation within a tumor, and (iii) the stem cell ratio within a tumor is variable over time with intratumoral and environmental competition for limited resources – space in the case of muscle invasion – selecting for and thus enriching in CSC fraction. In fact, competition such as for space and the resulting selection for CSCs has recently been shown to yield a pure CSC population over time (Hillen et al., 2013).
The frequency of symmetric division events in CSCs has previously been identified as a pivotal determinant of stem cell proportion experimentally (Cicalese et al., 2009) as well as in a variety of theoretical approaches including differential equations (Johnston et al., 2010), agent-based approaches (Enderling et al., 2009a), and hybrid models (Sottoriva et al., 2010).
CC proliferation capacity and the space- and time-dependent evolution of the CSC fraction in solid tumors offers a novel augmentation to the ongoing discussion about the frequencies at which CSCs are observed (Pardal et al., 2003; Quintana et al., 2008; Visvader and Lindeman, 2008). Furthermore, the CSC fraction, as determined by host environmental factors that control the symmetric division probability, determines whether the tumor exhibits an invasive or compact morphology, although the relationship is non-monotonic. Intrinsic tumor growth can be described as conglomerates of self-metastases (Enderling et al., 2009b, 2010), but over time a solid tumor core forms with more or less invasive boundary clustering. While lower and higher stem cell fractions yield more compact morphologies, intermediate fractions result in the most invasive tumor morphologies. These results augment findings of a monotonic increase of invasiveness with decreased stem cell fraction (Sottoriva et al., 2010). When additional host spatial constraints are imposed on the growing tumor, intrinsic morphological features disappear. Conceivably, the indistinguishable pathological morphologies that result could, in a clinical setting based on empirical observations, lead to the recommendation of comparable treatment protocols. However, due to their different CSC fractions, morphologically comparable tumors could in fact demonstrate response patterns ranging from complete regression [for low CSC content (Enderling et al., 2009c)] to resistance and accelerated re-growth (Gao et al., 2013).
Herein we limited our study to early avascular tumor growth where the total population is sufficiently small such that oxygen diffusion and tension can be neglected and global tumor growth dynamics be derived from different intrinsic CC kinetics. Simulations of larger tumor volumes will require physiological extension of the model to include nutrient delivery (Anderson and Chaplain, 1998; Ribba et al., 2004; Frieboes et al., 2007; Macklin et al., 2009) and vascular carrying capacities (Folkman, 1971; Folkman and Hochberg, 1973; Hahnfeldt et al., 1999). Furthermore, a translation from the cellular level model to a tissue-level continuous description might be more feasible to augment our understanding of the dynamics of larger populations (Hillen et al., 2013). For computational convenience we also limited this study to two spatial dimensions, but emphasize that extension to three spatial dimensions is algorithmically straightforward (Enderling et al., 2009b), with no qualitative change to be expected in the results here reported.
However, one transcendent feature expected to survive model simplifications is the possibility of widely varying stem cell compositions, highly dependent on host structural and biochemical context. This finding needs to be taken into account in both the clinical and research arenas, where heretofore, the threat has been presumed to come from the tumor bulk as a whole, not from a limited subpopulation within it.
Statements
Acknowledgments
The work was supported by the National Cancer Institute under Award Number U54CA149233 (to Lynn Hlatky).
Conflict of interest
The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
References
1
Al-HajjM.WichaM. S.Benito-HernandezA.MorrisonS. J.ClarkeM. F. (2003). Prospective identification of tumorigenic breast cancer cells. Proc. Natl. Acad. Sci. U.S.A.100, 3983–3988.10.1073/pnas.0530291100
2
AndersonA. R. (2005). A hybrid mathematical model of solid tumour invasion: the importance of cell adhesion. Math. Med. Biol.22, 163–186.10.1093/imammb/dqi005
3
AndersonA. R.ChaplainM. A. (1998). Continuous and discrete mathematical models of tumor-induced angiogenesis. Bull. Math. Biol.60, 857–899.10.1006/bulm.1998.0042
4
AndersonA. R.WeaverA. M.CummingsP. T.QuarantaV. (2006). Tumor morphology and phenotypic evolution driven by selective pressure from the microenvironment. Cell127, 905–915.10.1016/j.cell.2006.09.042
5
AshkenaziR.JacksonT. L.DontuG.WichaM. S. (2007). Breast cancer stem cells-research opportunities utilizing mathematical modeling. Stem Cell Rev.3, 176–182.10.1007/s12015-007-0026-2
6
BakerM. (2008). Melanoma in mice casts doubt on scarcity of cancer stem cells. Nature456, 553.10.1038/456322a
7
BankheadA.IIIMagnusonN. S.HeckendornR. B. (2007). Cellular automaton simulation examining progenitor hierarchy structure effects on mammary ductal carcinoma in situ. J. Theor. Biol.246, 491–498.10.1016/j.jtbi.2007.01.011
8
BernardS.BélairJ.MackeyM. C. (2003). Oscillations in cyclical neutropenia: new evidence based on mathematical modeling. J. Theor. Biol.223, 283–298.10.1016/S0022-5193(03)00090-0
9
BorovskiT.de Sousa E MeloF.VermeulenL.MedemaJ. P. (2011). Cancer stem cell niche: the place to be. Cancer Res.71, 634–639.10.1158/0008-5472.CAN-10-3220
10
BrúA.AlbertosS.Luis SubizaJ.García-AsenjoJ. L.BrúI. (2003). The universal dynamics of tumor growth. Biophys. J.85, 2948–2961.10.1016/S0006-3495(03)74715-8
11
CammareriP.LombardoY.FrancipaneM. G.BonventreS.TodaroM.StassiG. (2008). Isolation and culture of colon cancer stem cells. Methods Cell Biol.86, 311–324.10.1016/S0091-679X(08)00014-9
12
CaussinusE.HirthF. (2007). Asymmetric stem cell division in development and cancer. Prog. Mol. Subcell. Biol.45, 205–225.10.1007/978-3-540-69161-7_9
13
CicaleseA.BonizziG.PasiC. E.FarettaM.RonzoniS.GiuliniB.et al (2009). The tumor suppressor p53 regulates polarity of self-renewing divisions in mammary stem cells. Cell138, 1083–1095.10.1016/j.cell.2009.06.048
14
DeasyB. M.JankowskiR. J.PayneT. R.CaoB.GoffJ. P.GreenbergerJ.et al (2003). Modeling stem cell population growth: incorporating terms for proliferative heterogeneity. Stem Cells21, 536–545.10.1634/stemcells.21-5-536
15
DeutschA.DormannS. (2005). “Cellular automaton modeling of biological pattern formation: characterization, applications, and analysis,” in Modeling and Simulation in Science, Engineering and Technology, eds DeutschA.DormannS. (Boston: Birkhäuser), 59–104.
16
DingliD.MichorF. (2006). Successful therapy must eradicate cancer stem cells. Stem Cells24, 2603–2610.10.1634/stemcells.2006-0136
17
DingliD.MichorF.AntalT.PachecoJ. M. (2007a). The emergence of tumor metastases. Cancer Biol. Ther.6, 383–390.10.4161/cbt.6.3.3720
18
DingliD.TraulsenA.MichorF. (2007b). (A)symmetric ion and cancer. PLoS Comput. Biol.3:e53.10.1371/journal.pcbi.0030053
19
DrasdoD.HöhmeS. (2005). A single-cell-based model of tumor growth in vitro: monolayers and spheroids. Phys. Biol.2, 133–147.10.1088/1478-3975/2/3/001
20
DriessensG.BeckB.CaauweA.SimonsB. D.BlanpainC. (2012). Defining the mode of tumour growth by clonal analysis. Nature488, 527–530.10.1038/nature11344
21
EnderlingH.AndersonA. R.ChaplainM. A.BeheshtiA.HlatkyL.HahnfeldtP. (2009a). Paradoxical dependencies of tumor dormancy and progression on basic cell kinetics. Cancer Res.69, 8814–8821.10.1158/0008-5472.CAN-09-2115
22
EnderlingH.HlatkyL.HahnfeldtP. (2009b). Migration rules: tumours are conglomerates of self-metastases. Br. J. Cancer100, 1917–1925.10.1038/sj.bjc.6605071
23
EnderlingH.ParkD.HlatkyL.HahnfeldtP. (2009c). The importance of spatial distribution of stemness and proliferation state in determining tumor radioresponse. Math. Model. Nat. Phenom.4, 117–133.10.1051/mmnp/20094305
24
EnderlingH.HahnfeldtP. (2011). Cancer stem cells in solid tumors: is “evading apoptosis” a hallmark of cancer?Prog. Biophys. Mol. Biol.106, 391–399.10.1016/j.pbiomolbio.2011.03.007
25
EnderlingH.HlatkyL.HahnfeldtP. (2010). Tumor morphological evolution: directed migration and gain and loss of the self-metastatic phenotype. Biol. Direct5, 23.10.1186/1745-6150-5-23
26
FolkmanJ. (1971). Tumor angiogenesis: therapeutic implications. N. Engl. J. Med.285, 1182–1186.10.1056/NEJM197111182852108
27
FolkmanJ.HochbergM. (1973). Self-regulation of growth in three dimensions. J. Exp. Med.138, 745–753.10.1084/jem.138.4.745
28
FrieboesH. B.LowengrubJ. S.WiseS.ZhengX.MacklinP.BearerE. L.et al (2007). Computer simulation of glioma growth and morphology. Neuroimage37(Suppl. 1), S59–S70.10.1016/j.neuroimage.2007.03.008
29
FurthJ.KahnM. C. (1937). The transmission of leukemia of mice with a single cell. Am. J. Cancer31, 276–282.
30
GalleJ.HoffmannM.AustG. (2009). From single cells to tissue architecture-a bottom-up approach to modelling the spatio-temporal organisation of complex multi-cellular systems. J. Math. Biol.58, 261–283.10.1007/s00285-008-0172-4
31
GangulyR.PuriI. K. (2006). Mathematical model for the cancer stem cell hypothesis. Cell Prolif.39, 3–14.10.1111/j.1365-2184.2006.00369.x
32
GaoX.McDonaldJ. T.HlatkyL.EnderlingH. (2013). Acute and fractionated irradiation differentially modulate glioma stem cell division kinetics. Cancer Res.73, 1481–1490.10.1158/0008-5472.CAN-12-3429
33
GlaucheI.MooreK.ThieleckeL.HornK.LoefflerM.RoederI. (2009). Stem cell proliferation and quiescence two sides of the same coin. PLoS Comput. Biol.5:e1000447.10.1371/journal.pcbi.1000447
34
HaenoH.LevineR. L.GillilandD. G.MichorF. (2009). A progenitor cell origin of myeloid malignancies. Proc. Natl. Acad. Sci. U.S.A.106, 16616–16621.10.1073/pnas.0908107106
35
HahnfeldtP.PanigrahyD.FolkmanJ.HlatkyL. (1999). Tumor development under angiogenic signaling: a dynamical theory of tumor growth, treatment response, and postvascular dormancy. Cancer Res.59, 4770–4775.
36
HambardzumyanD.ChengY.-K.HaenoH.HollandE. C.MichorF. (2011). The probable cell of origin of NF1- and PDGF-driven glioblastomas. PLoS ONE6:e24454.10.1371/journal.pone.0024454
37
HanahanD.WeinbergR. A. (2000). The hallmarks of cancer. Cell100, 57–70.10.1016/S0092-8674(00)81683-9
38
HanahanD.WeinbergR. A. (2011). Hallmarks of cancer: the next generation. Cell144, 646–674.10.1016/j.cell.2011.02.013
39
HatzikirouH.BruschL.SchallerC.SimonM.DeutschA. (2010). Prediction of traveling front behavior in a lattice-gas cellular automaton model for tumor invasion. Comput. Math. Appl.59, 2326–2339.10.1016/j.camwa.2009.08.041
40
HeS.NakadaD.MorrisonS. J. (2009). Mechanisms of stem cell self-renewal. Annu. Rev. Cell Dev. Biol.25, 377–406.10.1146/annurev.cellbio.042308.113248
41
HillenT.EnderlingH.HahnfeldtP. (2013). The tumor growth paradox and immune system-mediated selection for cancer stem cells. Bull. Math. Biol.75, 161–184.10.1007/s11538-012-9798-x
42
HurtE. M.KawasakiB. T.KlarmannG. J.ThomasS. B.FarrarW. L. (2008). CD44+ CD24(-) prostate cells are early cancer progenitor/stem cells that provide a model for patients with poor prognosis. Br. J. Cancer98, 756–765.10.1038/sj.bjc.6604242
43
JohnstonM. D.MainiP. K.Jonathan ChapmanS.EdwardsC. M.BodmerW. F. (2010). On the proportion of cancer stem cells in a tumour. J. Theor. Biol.266, 708–711.10.1016/j.jtbi.2010.07.031
44
LathiaJ. D.HitomiM.GallagherJ.GadaniS. P.AdkinsJ.VasanjiA.et al (2011). Distribution of CD133 reveals glioma stem cells self-renew through symmetric and asymmetric cell divisions. Cell Death Dis.2, e200.10.1038/cddis.2011.80
45
LuzziK. J.MacDonaldI. C.SchmidtE. E.KerkvlietN.MorrisV. L.ChambersA. F.et al (1998). Multistep nature of metastatic inefficiency: dormancy of solitary cells after successful extravasation and limited survival of early micrometastases. Am. J. Pathol.153, 865–873.10.1016/S0002-9440(10)65628-3
46
MacklinP.McDougallS.AndersonA. R.ChaplainM. A.CristiniV.LowengrubJ. (2009). Multiscale modelling and nonlinear simulation of vascular tumour growth. J. Math. Biol.58, 765–798.10.1007/s00285-008-0216-9
47
MichorF. (2008). Mathematical models of cancer stem cells. J. Clin. Oncol.26, 2854–2861.10.1200/JCO.2007.15.2421
48
Molina-PeñaR.ÁlvarezM. M. (2012). A simple mathematical model based on the cancer stem cell hypothesis suggests kinetic commonalities in solid tumor growth. PLoS ONE7:e26233.10.1371/journal.pone.0026233
49
MortonC. I.HlatkyL.HahnfeldtP.EnderlingH. (2011). Non-stem cancer cell kinetics modulate solid tumor progression. Theor. Biol. Med. Model.8, 48.10.1186/1742-4682-8-48
50
NortonL. (2005). Conceptual and practical implications of breast tissue geometry: toward a more effective, less toxic therapy. Oncologist10, 370–381.10.1634/theoncologist.10-6-370
51
PardalR.ClarkeM. F.MorrisonS. J. (2003). Applying the principles of stem-cell biology to cancer. Nat. Rev. Cancer3, 895–902.10.1038/nrc1232
52
PiotrowskaM. J.EnderlingH.van der HeidenU.MackeyM. C. (2008). “Mathematical modeling of stem cells related to cancer,” in Cancer and Stem Cells, eds DittmarT.ZaenkerK. S. (New York: Nova Science Publishers, Inc.).
53
PrehnR. T. (1991). The inhibition of tumor growth by tumor mass. Cancer Res.51, 2–4.
54
QuintanaE.ShackletonM.SabelM. S.FullenD. R.JohnsonT. M.MorrisonS. J. (2008). Efficient tumour formation by single human melanoma cells. Nature456, 593–598.10.1038/nature07567
55
RahmanR.HeathR.GrundyR. (2009). Cellular immortality in brain tumours: an integration of the cancer stem cell paradigm. Biochim. Biophys. Acta1792, 280–288.10.1016/j.bbadis.2009.01.011
56
ReyaT.MorrisonS. J.ClarkeM. F.WeissmanI. L. (2001). Stem cells, cancer, and cancer stem cells. Nature414, 105–111.10.1038/35102167
57
RibbaB.AlarcónT.MarronK.MainiP. K.AgurZ. (2004). “The use of hybrid cellular automaton models for improving cancer therapy,” in Proceedings, Cellular Automata: 6th International Conference on Cellular Automata for Research and Industry, ACRI, 2004, Lecture Notes in Computer Science 3305, eds SlootP. M. A.ChopardB.HoekstraA. G. (Berlin: Springer-Verlag), 444–453.
58
ShibataM.ShenM. M. (2013). The roots of cancer: stem cells and the basis for tumor heterogeneity. Bioessays35, 253–260.10.1002/bies.201200101
59
SinghS. K.ClarkeI. D.TerasakiM.BonnV. E.HawkinsC.SquireJ.et al (2003). Identification of a cancer stem cell in human brain tumors. Cancer Res.63, 5821–5828.
60
SottorivaA.SpiteriI.ShibataD.CurtisC.TavaréS. (2013). Single-molecule genomic data delineate patient-specific tumor profiles and cancer stem cell organization. Cancer Res.73, 41–49.10.1158/0008-5472.CAN-12-2273
61
SottorivaA.VerhoeffJ. J.BorovskiT.McWeeneyS. K.NaumovL.MedemaJ. P.et al (2010). Cancer stem cell tumor model reveals invasive morphology and increased phenotypical heterogeneity. Cancer Res.70, 46–56.10.1158/1538-7445.AM10-46
62
StilesC. D.RowitchD. H. (2008). Glioma stem cells: a midterm exam. Neuron58, 832–846.10.1016/j.neuron.2008.05.031
63
TakezakiT.HideT.TakanagaH.NakamuraH.KuratsuJ.KondoT. (2011). Essential role of the Hedgehog signaling pathway in human glioma-initiating cells. Cancer Sci.102, 1306–1312.10.1111/j.1349-7006.2011.01943.x
64
VermeulenL.de SousaE.MeloF.van der HeijdenM.CameronK.de JongJ. H.et al (2010). Wnt activity defines colon cancer stem cells and is regulated by the microenvironment. Nat. Cell Biol.12, 468–476.10.1038/ncb2048
65
VisvaderJ. E.LindemanG. J. (2008). Cancer stem cells in solid tumours: accumulating evidence and unresolved questions. Nat. Rev. Cancer8, 755–768.10.1038/nrc2499
66
VlashiE.KimK.LagadecC.DonnaL. D.McDonaldJ. T.EghbaliM.et al (2009). In vivo imaging, tracking, and targeting of cancer stem cells. J. Natl. Cancer Inst.101, 350–359.10.1093/jnci/djn509
67
WangZ.LiY.BanerjeeS.SarkarF. H. (2009). Emerging role of Notch in stem cells and cancer. Cancer Lett.279, 8–12.10.1016/j.canlet.2008.09.030
68
WyckoffJ. B.JonesJ. G.CondeelisJ. S.SegallJ. E. (2000). A critical step in metastasis: in vivo analysis of intravasation at the primary tumor. Cancer Res.60, 2504–2511.
69
YoussefpourH.LiX.LanderA. D.LowengrubJ. S. (2012). Multispecies model of cell lineages and feedback control in solid tumors. J. Theor. Biol.304, 39–59.10.1016/j.jtbi.2012.02.030
Summary
Keywords
mathematical model, cellular automaton, cancer stem cell, symmetric division, invasion, morphology
Citation
Enderling H, Hlatky L and Hahnfeldt P (2013) Cancer Stem Cells: A Minor Cancer Subpopulation that Redefines Global Cancer Features. Front. Oncol. 3:76. doi: 10.3389/fonc.2013.00076
Received
15 January 2013
Accepted
24 March 2013
Published
15 April 2013
Volume
3 - 2013
Edited by
Katarzyna A. Rejniak, H. Lee Moffitt Cancer Center & Research Institute, USA
Reviewed by
Alberto D’Onofrio, European Institute of Oncology, Italy; Andrea Sottoriva, USC Keck School of Medicine, USA
Copyright
© 2013 Enderling, Hlatky and Hahnfeldt.
This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits use, distribution and reproduction in other forums, provided the original authors and source are credited and subject to any copyright notices concerning any third-party graphics etc.
*Correspondence: Philip Hahnfeldt, Center of Cancer Systems Biology, St. Elizabeth’s Medical Center, Tufts University School of Medicine, 736 Cambridge Street, CBR-115, Boston, MA 02135, USA. e-mail: philip.hahnfeldt@tufts.edu
This article was submitted to Frontiers in Molecular and Cellular Oncology, a specialty of Frontiers in Oncology.
Disclaimer
All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article or claim that may be made by its manufacturer is not guaranteed or endorsed by the publisher.