Radiation Resistance of Cancer Stem Cells: The 4 R’s of Radiobiology Revisited

Abstract

There is compelling evidence that many solid cancers are organized hierarchically and contain a small population of cancer stem cells (CSCs). It seems reasonable to suggest that a cancer cure can be achieved only if this population is eliminated. Unfortunately, there is growing evidence that CSCs are inherently resistant to radiation, and perhaps other cancer therapies. In general, success or failure of standard clinical radiation treatment is determined by the 4 R’s of radiobiology: repair of DNA damage, redistribution of cells in the cell cycle, repopulation, and reoxygenation of hypoxic tumor areas. We relate recent findings on CSCs to these four phenomena and discuss possible consequences.

Introduction

It has been postulated for more than 4 decades that most if not all cancers are hierarchically organized and contain a subtle subpopulation of cancer stem cells (CSCs) [1–4] within a tumor that possess the capacity to self-renew and to cause the heterogeneous lineages of cancer cells that comprise the tumor [5]. The CSC concept has its origin in the leukemia literature [6], and although the term CSC does suggest normal tissue stem cells as the cell of origin in leukemia, there is evidence that leukemia stem cells may actually arise from progenitor cells (recently reviewed in [7] and [8]). However, since most radiotherapy is delivered locally to solid tumors, we will focus this review exclusively on CSCs in solid cancers and their response to radiation.

The existence of CSCs in solid tumors has been hotly debated [9–11]. It is an appealing hypothesis that has often been challenged theoretically but its opponents have not yet been able to disprove it experimentally [12]. A recent study by Quintana et al. [13] challenged the existence of CSCs in human melanoma on the basis of the ability of single cells to initiate tumors in severely immune-compromised mice. However, this study did not investigate CSC-defining characteristics of self-renewal and differentiation capacity and unlimited proliferative capacity [5] in serial marker-defined human to mouse xenotransplantation experiments. Although the CSC model may not fit perfectly the behavior of all solid tumors, it describes the way many epithelial and brain tumors respond and recur after radiation treatment more accurately than a stochastic model in which all cancer cells have the same tumorigenic potential [14]. Although no one would argue against the existence of cells with the potential to regrow a tumor, the key questions any model has to address are their frequency, do they possess “stemness”, and do they have a distinct response to therapy that might allow them to be responsible for a significant number of tumor recurrences.

It is very likely that the frequency of CSCs will depend on the type of tumor and the model system studied and the actual numbers may vary substantially [15–17], which makes any analysis that is numerically based difficult.

Numerous studies that prospectively identified subpopulations of cells with stem cell phenotypic markers in many solid cancers have been the major source of support for the existence of a hierarchical structure derived from CSCs. These cells can be enriched if grown as spheroids under serum-free conditions, express gene expression programs seen during normal tissue development, such as Wnt and Notch [18–20], and appear to be highly tumorigenic in transplantation assays [15, 21–26]. Additionally, breast CSCs share with normal mammary gland stem cells-specific molecular gene expression patterns like downregulation of microRNA clusters [27]. Their origin from normal tissue stem cells, or at least very early progenitor cells, is further supported by four recent murine studies demonstrating that oncogene expression or loss of tumor suppressor genes in the stem cell compartment, but not in committed progenitor or differentiated cells, is both required and sufficient for full malignant transformation [28–31]. Although these data are convincing, the exact definition of stemness is elusive and stemness may be more of a continuum or a property that may be regained in cancer, which would suggest that neither the hierarchical model nor the stochastic model are exclusively right.

The CSC concept has been elevated to a higher level of significance in cancer therapy by recent evidence in several cancers that they can resist conventional treatments including ionizing radiation [19, 20, 32–39] and chemotherapy [16, 33, 40]. This has been explained by a metabolic status that is associated with high free radical scavenger levels [19, 41], low proteasome activity [42], activated DNA checkpoints [32], and expression of the ABCB5 multi-drug resistance protein [16]. Their possible relative radioresistance indicates the need for re-evaluation of the mechanisms underlying the response of solid tumors to conventional and newer radiation treatments with a specific emphasis on CSCs. This has recently been discussed with reference to classic radiobiological end points [43] and will therefore not be addressed here. This review will instead summarize current data pertaining to the radiation responsiveness of CSCs within the framework of the “4 R’s of Radiobiology” that determine the outcome of a conventional fractionated course of radiation therapy for cancer, as originally described by Withers: repair of sublethal DNA damage, cell repopulation, redistribution of cells in the cell cycle, and reoxygenation of previously hypoxic tumor areas [44] (Fig. 1). It is obvious that tumor responses to radiation treatment are modulated by many additional factors, and in fact, Steel even suggested 2 decades ago that intrinsic radiosensitivity should be considered as the 5th R [45]. However, the intrinsic radiosensitivity of individual CSCs has not yet been investigated and it is not clear if this changes during a fractionated course, and therefore we will focus on the 4 R’s originally described by Withers [44] as they represent hubs on which many other mechanisms converge and because they provide a simple model for understanding the efficacy of fractionated radiotherapy for cancer.

Figure 1.

The 4 R’s of radiation biology. (A): Repair of sublethal DNA damage. DNA double-strand breaks (DSBs) after exposure to ionizing radiation are mainly repaired by NHEJ. NHEJ involves recognition of the DNA DSBs by Ku70/80, recruitment of the histone H2AX to the DNA leasion, phopshorylation of H2AX by ATM, DNA-PKcs, or ATR, and finally rejoining of the strand ends by XRCC4 and Ligase 4. (B): Redistribution. Mammalian cells exhibit different levels of radioresistance during the course of the cell cycle. Cells in the late S-phase are especially resistant and cells in the G2/M-phase are most sensitive to ionizing radiation. During fractionated radiation cells in the G2/M-phase are preferentially killed. The time between two fractions allows resistant cells from the S-phase of the cell cycle to redistribute into phases in which cells are more radiosensitive. (C): Repopulation. Normal and malignant stem cells have the ability to perform asymmetric cell division, which give rise to a daughter stem cell and a committed progenitor cell. In a symmetric cell division in contrast, stem cells divide into two committed progenitor cells or two daughter stem cells. If the latter happens only in 1% of the stem cell divisions, the number of stem cells after 20 cell doublings will be twice as high as the number of committed progenitor cells. This indicates that small changes in the way stem cells divide have huge impact on the organization of a tissue or tumor and are thought to be the mechanism behind accelerated repopulation. (D): Reoxygenation. Tumors contain regions of hypoxia in which cancer cells are thought to be resistant to radiation. During fractionated radiotherapy, these regions are reoxygenated by various mechanisms including reduction of intratumoral pressure and normalization of the vasculature. Reoxygenation between radiation fractions will lead to radiosensitization of previously hypoxic tumor areas and is thought to increase the efficiency of radiation treatment. Abbreviations: CSCs, cancer stem cells; NHEJ, nonhomologous end joining.

It should be noted that the vast majority of experimental studies that were used to originally define the 4 R’s were based on clonogen survival in hierarchical normal tissues or tumors, such as in vitro clonogenic cell survival assays, in vivo splenic colony-forming unit (CFU) or colonic/jejunal crypt cell assays, or tumor regrowth assays and evaluated responses in the short term (see Appendix). Recent data indicate that normal and also malignant stem cells, for example, in the normal colon [46], in bone marrow (long-term repopulating hematopoetic stem cells), or in melanoma [13], cycle very slowly and would not be evaluated by most of the standard radiobiological assays, such as those listed above. Although these assays are often described as measuring “stem” cells, in fact, they rather favor progenitor cells. Such progenitor cell responses are highly relevant to preserving normal tissue function because in hierarchical normal tissues they rapidly restore tissue integrity. However, in tumors, complete cell kill is required to prevent a recurrence and this will be determined by a composite of the radioresistance of different subpopulations and the number of cells with that level of radioresistance. The representation of CSCs may be particularly important for clinical consideration of the relative radiation sensitivities of cancers like melanomas in which the frequency of CSCs may vary [13, 16] and epithelial cancers where it may be low [15]. With our growing ability to study CSC populations directly, future studies should be able to take this heterogeneity into account.

Repair of Radiation-Induced DNA Damage

Cell kill by ionizing radiation is based on production of unrepairable lesions involving DNA double-strand breaks (DSBs). Most radiation-induced DNA damage is however sublethal. Although this is repaired at lower doses, at higher doses accumulation of sublethal lesions also contributes to lethality. Repair of sublethal damage between radiation fractions is exploited in radiation therapy because critical normal tissues and tumors often differ in their ability to repair radiation damage.

Highly reactive oxygen species (hROS), which radiation generates by ionization of water molecules, are short-lived and rapidly interact with various biomolecules in cells. Those that are generated within 2 nm of the DNA are more important in causing DNA damage than direct ionization of the DNA strands and, consequently, free radical scavengers, such as glutathione, within this location play a major role in determining the extent of initial radiation-induced DNA damage and cell survival [47].

CD24^−/low/CD44⁺ breast cancer cells, which are believed to be a clinically relevant CSC-containing subpopulation, when compared to the whole population were originally found to have increased tumorigenicity and to be relatively resistant to radiation at the DNA and cellular levels, which could be attributed to significantly lower levels of basal and radiation-induced ROS, indicating higher levels of free radical scavengers [19]. The low ROS levels before and after irradiation of murine and human breast CSCs from cell lines was recently confirmed by Diehn and coworkers using primary breast cancer cultures and who further described an anti-oxidant gene expression profile for breast CSCs by single-cell reverse transcription-polymerase chain reaction (RT-PCR) [41]. In this study, depletion of glutathione by buthionine sulfoximine reversed the radioresistant phenotype of breast CSCs. Low constitutive and radiation-inducible ROS levels may therefore be a useful marker for identification of CSCs [48] and perhaps even normal stem cells, and could also be crucial in determining the response of this subpopulation to radiation and other therapies. They may be why breast CSCs were found to be more radiation-resistant in clonogenic assays [19, 20, 41], in particular, in the lower dose range, and why CSCs are enriched by repeated fractions of radiation that preferentially kill the more radiosensitive, less tumorigenic cancer cells [19].

A hallmark of DNA DSB recognition and repair is phosphorylation of the histone H2AX by ATM or ATR (γ-H2AX) [49], which is thought to form one immunohistochemically detectable focus per DSB. Radiation induced few [19] or significantly less [50] γ-H2AX foci in human breast CSCs, and in murine breast CSCs they resolved faster than in non-CSC populations [20]. Also, in glioma, although CSCs showed a normal initial γ-H2AX response to irradiation, here also the DNA DSBs were repaired more efficiently and more rapidly [32]. This response may however depend on the experimental context as the ability of glioma stem cells to repair DNA damage more efficiently than their non-CSC counterparts was recently challenged by Ropolo and coworkers who reported no change in base excision repair, resolution of γ-H2AX foci, or single-strand DNA repair in cell lines enriched for CD133⁺ cells in vitro [51]. Additionally, McCord et al. reported [52] that CD133⁺ glioma cells were not always more radiation-resistant than CD133⁻ cells, although this study did not attempt to demonstrate aspects of the CSC phenotype other than CD133 positivity.

Currently, it is prudent to believe that there is no ideal single marker for CSCs in any tumor system [53]. Furthermore, sorting of cancer stem cells and how to report the isolation methods used still need to be standardized to allow comparison of data obtained from different laboratories [54]. For example, CD133 surface expression is commonly used to identify CSCs in glioma, but doubts have been expressed as to whether this marker may define progenitor cells rather than CSCs or has any specificity for glioma CSCs, and as to whether it reflects a state of bioenergetic stress rather than stemness [55–57]. Recent data on neural stem cells even suggest the possibility that expression of CD133 depends on the cell cycle and is specifically downregulated in the G₀/G₁ phase [58]. In spite of these caveats, CD133⁺ glioma cells were used to show increased ability to repair single-strand breaks by the alkaline Comet assay and preferential activation of the DNA damage checkpoint [59] in the response to radiation [32], as assessed by hyperphosphorylation of Chk1, and to a lesser extent Chk2. Inhibition of this response radiosensitized CD133⁺ glioma cells [32] (Fig. 2C). This was confirmed in CD133⁺ atypical teratoid/rhabdoid tumor cells, a rare and aggressive pediatric brain tumor of uncertain origin [34] and in glioma cultures enriched for CD133⁺ cells [51]. Chk1 phosphorylation protects cells from radiation cytotoxicity, although not through stimulating repair by nonhomologous end joining [60]. However, RAD51, a protein involved in the search for homology and strand pairing during homologous recombination of DNA double-strand breaks [61], is part of the CSCs signature in breast cancer [62], suggesting that Chk1-dependent homologous recombination may be more important in DNA repair in CSCs.

Figure 2.

CSCs and DNA repair. (A): CSCs exhibit less DNA double-strand breaks (DSBs) after exposure to ionizing radiation than nontumorigenic cells. GCL catalyzes the reaction of cysteine and glutamate to form γ-glutamylcysteine in an ATP-dependent step. In a second step, GSS condensates γ-glutamylcysteine and glycine to form glutathione. Breast CSCs were found to express high levels of GCL and GSS. Consequently, most radiation-induced free radicals were scavenged in breast CSCs and ionizing radiation caused only little DNA damage if compared to nontumorigenic cells. Inhibition of glutamate cysteine ligase by buthionine sulfoximine reversed the radioresistant phenotype. (B): CSCs repair DNA DSBs more efficiently than nontumorigenic cells. CSCs in breast and brain cancers hyperphosphorylated the DNA checkpoint kinases Chk1 and Chk2 constitutively and in response to ionizing radiation, thereby removing DNA DSBs more rapidly and more efficiently. Abbreviations: BSO, Buthionine Sulfoximine; CSCs, cancer stem cells; GCL, glutamate cysteine ligase; GSS, glutathione synthetase; GSSG, glutathione disulfide; ROS, reactive oxygen species.

It is sometimes difficult to dissociate DNA repair, or lack thereof, from induced cell death following radiation exposure. Indeed, induction of the apoptosis inhibitor survivin was proposed by Woodward et al. [20] as an additional cytoprotective mechanism of breast CSC following irradiation based on their observation that β-catenin and survivin expression was induced in normal murine Sca-1⁺ (stem cell antigen 1) but not Sca-1⁻ mammary epithelial cells. Survivin expression has been linked to radioresistance in other studies [63]. The mechanisms by which survivin affects DNA repair are incompletely understood but seem to involve changes in cell cycle distribution and direct effects on DNA DSB repair [64–66]. Survivin may be targeted by small-molecule inhibitors [63], but since radiation-induced survivin expression may be a conserved response of normal and malignant stem cells, the existence of a therapeutic window for these agents in combination with radiation therapy needs to be shown before these drugs can be considered as having potential radiotherapeutic benefit in targeting CSC.

An alternative radioprotective mechanism for CD133⁺ glioma stem cells was recently suggested by Lomonaco et al. to be induced autophagy. In their study, CD133⁺ cells expressed higher levels of the autophagy-related proteins LC3, ATG5, and ATG12 than CD133⁻ cells after irradiation and inhibition of autophagy preferentially sensitized CD133⁺ cells to radiation and decreased sphere-forming capacity [39].

Fractionation is believed to allow repair of slowly proliferating, late-responding tissue like the central nervous system at the expense of tumors that seem less able to repair sublethal damage. In fact, cells vary greatly in their intrinsic cellular radioresistance and many tumors seem quite adept at repair. Variation both within a tumor and between tumors of the same entity is therefore to be expected and it is difficult to draw hard and fast conclusions that apply in all circumstances. In part, the outcome of radiation exposure will depend upon the extent of DNA damage and repair, but the downstream DNA damage response that determines cell death and cell cycle arrest, in coordination with the signaling pathways that are active and activated, will play roles. It is important to remember that cancer-associated mutations influence DNA repair, cell cycle, and cell arrest and their influence on radiation response should be taken into account when comparing tumors from different individuals. The therapeutic resistance of tumors is therefore multifactorial and complex but the fact that CSCs may differ in the way they handle radiation-induced DNA damage should be considered as a potential parameter determining the outcome of fractionated radiotherapy. In essence, if the dose per fraction of radiation is insufficient to cause sufficient cell death in the CSC population, the frequency of CSCs in a tumor may even increase during treatment, although clinically the tumor may regress macroscopically. However, parameters other than DNA damage and repair must be considered as contributing to this equation.

Redistribution

Redistribution acknowledges the fact that cells exhibit differential radiation sensitivity while in the different phases of the cell cycle, with cells in mitosis being most sensitive to DNA damaging agents and cells in late S-phase being most resistant [67]. Because of cell cycle progression of surviving cells between radiation fractions, dose fractionation allows redistribution of radioresistant S-phase tumor cells into a more sensitive phase of the cell cycle, and a resultant therapeutic benefit for slowly cycling normal cells (Fig. 3A) [68]. Normal tissue stem cells and CSCs are believed to exist in the G0-phase of the cell cycle and to cycle slowly. They are thought to be maintained in this state by intrinsic genetic programs and extrinsic influences from the niche in which they reside. Although there is good experimental evidence for the quiescent state of hematopoetic stem cells [69, 70], evidence for the quiescent state of cancer stem cells comes currently rather from theoretical considerations and mathematical modeling than from experimental data. Several molecules like, for example, Notch ligands, have been implicated in the transition of stem cells into cycle and their emergence from the niche [71], which potentially offers novel opportunities for therapeutic intervention [72, 73]. Niches for most CSCs have yet to be convincingly demonstrated but glioma CSCs appear to reside in a perivascular location [42, 74]. Most putative CSCs in this site are negative for the proliferation marker Ki67 [42]. Multiple fractions of radiation promotes recruitment of CSCs from the niche and increases the proportion of cycling cells [42], with a concomitant increase in radiosensitivity. For normal tissue stem cells and CSCs, therefore, redistribution following irradiation may be tied to their mobilization and entry into the cell cycle, and thus regeneration. This raises the interesting possibility that mobilization of CSCs may be strategy for increasing their radiosensitivity. The effect of radiation on expression of developmental gene expression profiles is still in its infancy, but it is interesting to note that the effects of fractionated irradiation on Notch/Jagged expression by breast CSCs were far greater than those for single doses [19], suggesting that fractionated irradiation may specifically alter the kinetics of CSCs and perhaps normal stem cell regenerative behavior through specific activation of developmental pathways.

Figure 3.

Radiation-induced redistribution and accelerated repopulation employs the developmental Notch pathway. (A): In breast cancer, ionizing radiation induces the expression of Notch receptor ligands on the surface of nontumorigenic cells and possibly other nonmalignant stem cells niche cells like, for example, endothelial cells, which are finally depleted by radiation. Activation of Notch signaling in CSCs may than redistribute quiescent CSCs into the cell cycle in a symmetric type of stem cell division and finally cause repopulation of the tumor. (B): In this model system, TGF-β is produced by the mass of the nontumorigenic, radiosensitive cancer cells and activated by radiation. It antagonizes the proliferative effects of Notch, which is activated in CSCs through interaction with their niche. During the course of fractionated irradiation, most of the nontumorigenic cancer cells are killed. This causes TGF-β levels to drop while Notch is still being activated, resulting in increased regrowth rates and thus accelerated repopulation. Abbreviation: CSCs, cancer stem cells.

Redistribution of cells during fractionated irradiation has been interpreted as sparing dose-limiting tissues with a small content of rapidly cycling cells such as the central nervous system in comparison with normal tissues such as the bone marrow and intestine and tumors that are considered to have a high content of cells with rapid turnover. If, in fact, tumors have a high percent of slowly cycling CSCs, this logic may not apply.

Repopulation

Repopulation of tumors may be one of the most common reasons for the failure of conventional fractionated courses of radiation therapy [75, 76], as judged by the dramatic effects of treatment prolongation over the conventional 6 weeks on local control rates [75, 77, 78]. For decades, radiation oncologists hypothesized that depopulation of certain hierarchical normal tissues, such as the jejunum, by ionizing radiation caused “stem” cells to switch from an asymmetric type of cell division, which gives rise to a daughter stem cell and a lineage-committed progenitor cell, to a symmetric form of cell division that results in two proliferative daughter stem cells. Cell loss from the proliferative compartment in normal tissues seems to be decreased until regeneration is complete. A similar process was postulated for tumors, giving rise to accelerated repopulation, which describes the situation where the regrowth rate of a tumor after treatment with a sublethal radiation dose exceeds the growth rate of the untreated tumor.

Importantly, like normal tissue stem cells, CSCs employ developmental signaling pathways like the Notch, Wnt, and Sonic hedgehog pathways [18–20, 79, 80] that are able to perform the switch from an asymmetric to a symmetric type of cell division. Our initial report of activation of the Notch pathway by radiation [19] has now been confirmed by others in endothelial cells [81], indicating that this pathway might be part of the acute response to ionizing radiation. Activation of the Notch receptor relies on cell-cell contacts and binding of Notch receptors to ligands of the Delta or Jagged family. Upon ligand binding, the extracellular part of the Notch receptor is shed and internalized into the ligand-expressing cell, whereas the remaining part of the Notch receptor undergoes intramembranous cleavage by γ-secretase, which finally releases the intracellular domain of the receptor (Notch-ICD) for nuclear translocation. In the nucleus, Notch-ICD binds to CBF-1, which turns it from a transcriptional repressor into a transcriptional activator, thereby initiating the transcription of gene products that promote progression into the S-phase of the cell cycle [82]. Activation of Notch signaling can recruit quiescent stem cells into the cell cycle [83] and a sustained Notch signal maintains the stem cell phenotype whereas termination of this signal leads to differentiation [71]. An important issue here is the extent to which this takes place within and is dependent on cell-cell contact within the “niche” and the relative radiosensitivity of cells as they go through activation.

Another developmental pathway activated in response to radiation is the TGF-β pathway [84–86], which is thought to be an antiproliferative pathway that controls tissue homeostasis [87]. It mediates its effects through proteins of the smad family, which can compete with Notch-ICD for binding to CBF-1 [88, 89]. This draws an interesting model of tumor tissue homeostasis and accelerated repopulation during fractionated radiation therapy for epithelial cancers in which the bulk of the tumor is characterized by relative radiosensitivity but produces most of the TGF-β. If CSCs can respond to TGF-β, it could antagonize their division until most of the non-CSCs are eliminated and TGF-β levels drop, whereupon the Notch pathway would be activated, driving CSC self-renewal and leading to a rapid relapse (Fig. 3B) [90]. This mechanism could also apply in normal hierarchical tissues that are also dependent on developmental pathways like Notch, which supports the proposal that the normal stem cell/early progenitor compartment is the origin of CSCs [28–31], and at the same time offers novel targets for therapeutic intervention [73, 91–93] in combination with radiation therapy.

Accelerated repopulation is very difficult to demonstrate in vitro, and in vivo results can be influenced by many factors other than altered proliferative status. However, with marker profiles for CSCs available for a variety of different solid cancers, several groups have recently reported an increase in such phenotypes after repeated, clinically relevant doses of radiation in vitro and in vivo [19, 20, 32]. However, it is possible that if radiation recruits CSCs into the proliferating pool, their intrinsic radiosensitivity may alter, which would give a therapeutic advantage to fractionated as opposed to single-dose radiotherapy.

Clearly, repopulation of tumors and regeneration of normal tissue are critical elements in the success of fractionated radiotherapy. Again, drawing broad conclusions is difficult but the CSC concept provides markers for studying this process in vivo and targets for possible intervention.

Reoxygenation

Since the initial experiments of Schwarz in 1909, Holthusen in 1921, and Thomlinson and Gray in 1955, oxygen has been known as one of the most potent modifiers of radiation sensitivity and hypoxic cells have been repeatedly shown to be 2–3 times more resistant to radiation [94–96]. In addition, human tumors contain regions of acute and chronic hypoxia that have often been shown to be associated with poor prognosis because of local recurrence or systemic disease [97–101]. The tumor microenvironment is dynamic with ever-changing oxygen and pH gradients [102, 103]. Transient areas of acute hypoxia due to intermittent vessel closure may reoxygenate rapidly, whereas chronic hypoxia due to the limitation of oxygen diffusion may take longer. A major concept in clinical radiobiology is that tumor subpopulations in hypoxic areas are critical to target for increased therapeutic benefit but there is still discussion as to whether acute or chronic hypoxic cells are most important [103].

Reoxygenation between dose fractions is generally believed to improve the efficacy of radiation treatment by increasing tumor radiosensitivity. It should however be noted that many of these experiments were performed under conditions where oxygen levels were rapidly decreased and tumors were then reoxygenated after irradiation. This drastic and stressful change in the tumor microenvironment could affect CSCs, for example, by triggering rapid differentiation [48]. Such studies should be interpreted with caution as they may fail to recapitulate tumor microenvironmental conditions in vivo.

As discussed above, it seems that brain CSCs reside in a perivascular niche [42, 74]. Although one might consider such cells as being in the most perfectly oxygenated region in a tumor, it is also likely that these would be exposed to rapidly changing bouts of hypoxia-reperfusion. This can generate damaging free radicals. It seems likely that the low hROS metabolic profile [19, 41] and slow cycling of CSCs may selectively protect them from some of these free radical effects and, indeed, survival pathways may be induced that encourage radioresistancy at the expense of non-CSC cells (Fig. 4).

Figure 4.

CSCs and tumor hypoxia. Like long-term repopulating hematopoetic stem cells, quiescent CSCs may exist in a nonperivascular hypoxic niche, relatively protected from ionizing radiation. Activated, and thus cycling, CSCs are found in a perivascular niche with confers increased radiation sensitivity and the dependence of CSCs on that niche makes them vulnerable to anti-angiogenic strategies, which target endothelial cells, thereby destroying the CSC niche. Reoxygenation of the hypoxic CSC niche during radiation fractionation redistributes quiescent CSCs as increasing oxygen levels will modify the niche conditions to render those found in pervascular regions and may cause the transition from a quiescent into an activated, proliferative CSC state. Abbreviation: CSCs, cancer stem cells.

There is increasing evidence that the length of time that cells are under hypoxic conditions and the extent of the hypoxia are critical factors in terms of the biological and radiation response of these cells. Indeed, cells irradiated shortly after reoxygenation or after exposure to long-term chronic hypoxia are radiosensitive [104] compared to those irradiated after 4–24 hours of hypoxia. The underlying mechanisms are incompletely understood but the suggestion is that the radioprotective effect of hypoxia does not exclusively rely on the radiochemistry of oxygen but rather involves complex signaling events, which adjust the homeostatic rheostat if the hypoxia persists long enough, thereby losing its protective effects. Indeed, chronic hypoxia may make cells radiosensitive by decreasing DNA repair, in particular, RAD51-mediated homologous recombination [105]. In light of these observations, cells undergoing intermittent hypoxia might be the most relevant for therapy resistance, and because of their presence in a perivascular niche, it would be of great interest to know more about the response of CSCs to irradiation under varying hypoxic conditions. Indeed, if as seems likely intermittent hypoxia were to have an effect similar to that of irradiation in selecting for CSCs, recruiting them into cycle, and mobilizing them [19, 42], CSCs could be responsible for the relationship that has been found between hypoxia and metastasis [98, 106].

There are other possible consequences of the presence of CSCs in the perivascular niche. They may be more accessible to drugs delivered systemically and also more dependent on endothelial cell viability and function than cells more distant from the vasculature. It is of interest in this regard that an earlier report showed that the effect of ionizing radiation on tumor cells in vivo may be preceded by radiation-induced death of endothelial cells [107]. Furthermore, synergy between anti-angiogenic therapies and radiation has been observed [108], which is counterintuitive from a classic view of tumor hypoxia as one could argue that anti-angiogenesis should increase tumor hypoxia and thus radiation resistance. Normalization of blood flow [109–111] may partly explain this synergy but the proximity of CSCs to endothelial changes offers an alternative explanation for this synergy. It should also be noted that single and fractionated radiation regimens can “prune” tumor and normal tissue vasculature, converting acute hypoxia to chronic hypoxia [112], and it will be important to determine the fate of CSCs as the microenvironment changes with time after exposure.

The observation of a perivascular CSC niche may explain the variable results [113] of clinical trials aimed at improving tumor oxygenation. One other possible conclusion would be that, like long-term repopulating bone marrow cells [48], CSCs exist in two different stem cell niches, one hypoxic for quiescent CSCs and one adjacent to endothelial cells for more activated CSCs. Transition between both states could be bidirectional, as it is in bone marrow [48]. Better marker profiles for quiescent CSCs will be needed to elucidate whether or not this is the case.

Another example of the potential relevance of CSCs to clinical practice is the use of erythropoiesis-stimulating agents (ESAs) in anemic cancer patients. These aim to decrease hypoxia and increase therapeutic benefit, but have been found to cause an unexpected decrease in local control rates after radiation treatment of epithelial cancers [114–116]. Although pre-clinical models investigating the effects of ESAs on unselected cell populations led to conflicting results [117–122], breast CSCs were clearly stimulated by ESAs, explaining the unexpected clinical findings [18].

Related to the question of how relevant hypoxia is for clinical radiation responses of tumors is the metabolic state of CSCs. Rapidly proliferating cancer cells, which form the bulk of the tumor, rely mainly on glycolysis [123], a less efficient way of energy production that requires drastically increased glucose uptake, a phenomenon utilized in ¹⁸FDG-PET imaging of tumors. It is a general assumption that CSCs, in contrast, are mostly quiescent and as such metabolize glucose by oxidative phosphorylation rather than by glycolysis [124]. Consequently, areas with high numbers of quiescent CSCs with moderate or low glucose uptake will not be detectable by ¹⁸FDG-PET and should therefore not be excluded from the defined clinical radiation target volume.

Concluding Remarks

Despite enormous research efforts, systemic therapies still fail to cure most solid cancers, irrespective of the stage and spread of the disease. Even so-called targeted therapies are mainly cytostatic and as such aim to turn cancer into a chronic and eventually controllable rather than curable disease. Surgery and radiation therapy are currently the only treatment options leading to cancer cure for patients suffering from localized cancers, but fail against systemic disease. More importantly, most current strategies do not target CSCs but rather more differentiated tumor cells [125]. Existing paradigms for cancer treatment need to be re-evaluated for relevance based upon recent insights into the response of CSCs to conventional treatments, such as radiation therapy. Conventional fractionated radiation therapy has evolved in over a century as a means of delivering radiation dose over a period of time and is a compromise between sparing normal tissues at the expense of tumors while avoiding loss of tumor control. The principles involved are enshrined in the 4 R’s of radiobiology. At this point in time, these are not tailored for individual treatment and predictive markers are needed to allow this to be part of the treatment planning strategy. Obviously, CSC markers should be part of this effort as they may change the way radiation is delivered.

CSCs also offer novel targets to enhance the efficacy of radiation therapy [32, 36, 50] and future targeted therapies should have this aim. Also, because their metabolic status may differ from the majority of cells in a tumor, the effect of this heterogeneity on functional imaging of tumors should be taken into account. With the advent of novel imaging technologies for CSCs [42], biology-guided radiation treatment planning may offer ways for specifically delivering high radiation doses to areas with high CSCs numbers. Finally, it should be noted that partial tumor responses to therapy mean little if CSCs are the major cells determining outcome. Radiation therapy will kill CSCs and their number and relative radioresistance will determine if they survive a fractionated course of radiation, or whether another delivery strategy or even therapy would be superior.

Appendix

Classic Radiobiology Assays

Clonogenic Survival Assay: The In Vitro Gold Standard Assay To Assess the Effect of Radiation on Tumor Cells

A defined low number of cells is plated into culture dishes and exposed to increasing doses of radiation. The low number of cells treated allows outgrowth of colonies from single surviving cells (clonogens). In general, colonies of more than 50 cells are considered survivors. This is equivalent to 5–6 cell divisions if none of the cells are lost. The surviving fraction of cells for each radiation dose is normalized to the surviving fraction of the corresponding control to compensate for acute toxicity of the assay itself. Results are presented on a log-linear scale and measure the number of clonogens, which reflects the survival of committed progenitor cells and cancer stem cells.

Splenic CFU Assay: In Vivo Assay To Assess the Number of Clonogens of Normal Hematopoetic Cells, Leukemia, and Lymphoma Cells

A defined number of cells is injected into immunodeficient animals. The number of macroscopic colonies on the spleen surface is counted and CFU numbers after radiation treatment are normalized to numbers obtained from nontreated controls. The radiation can be applied in vivo or ex vivo to the cells injected. The assay suffers from the same problems described for in vitro clonogenic survival assays and results may also depend on the ability of the injected cells to home to the microenvironment in the spleen.

Colonic/Jejunal Crypt Assay

The jejunum is one of the most radiosensitive organ systems and loss of its integrity defines the gastrointestinal syndrome in the response of mammalian organisms to radiation. Radiation causes denudation of the jejunum of its crypts and recovery of the crypts can be measured to estimate the fraction of surviving stem cells/progenitor cells.

Tumor Assays

In limiting dilution assays, a decreasing number of cells are injected into immunodeficient animals, either subcutaneously or orthotopically into the corresponding organ site. True cancer stem cell populations can be serially transplanted from one animal to the next.