|Home | About | Journals | Submit | Contact Us | Français|
There is increasing evidence that glioblastoma possess ‘stem‐like’ cells, low concentrations of which can initiate a tumour. It has been proposed that these cells are radioresistant, and that this property contributes to the poor treatment outcomes of these tumours. In this paper we propose that radioresistance is not simply an intrinsic characteristic of glioma stem cells but a result of interactions between these cells and microenvironmental factors, i.e. the ‘microenvironment – stem cell unit’. The critical role of the microenvironment, along with glioma stem cells, is supported directly or indirectly by the following observations: glioma stem cells have been shown to reside preferentially in specific niches, the characteristics of which are known to influence cellular responses to radiation; radiation modifies environmental factors; and, contrarily to the consistency of clinical data, in vitro experiments have reported a wide variety in the radiation response of these cells.
The paper, therefore, focuses on the interaction between tumour stem cells and the microenvironment, analyzing how its various elements (endothelial cells, extracellular matrix, cytokines, nitric oxide, oxygen levels) are affected by radiation and how these might influence the response of tumour stem cells to radiation.
Finally, we summarize the ongoing debate on the optimal culture conditions for glioma stem cells and the difficulties in designing assays that reliably characterize their radiation response.
There is increasing evidence that human glioblastoma possess ‘stem‐like’ cells, small numbers of which are capable of initiating a tumour that closely resembles the original cancer. It has been proposed that these tumour initiating cells are resistant to radiation therapy, and that this property contributes to the poor treatment outcomes associated with these tumours. After analyzing the conflicting data published on this subject over the past seven years, this article proposes that radioresistance is not simply an intrinsic characteristic of glioma stem cells but a result of the interaction between these cells and microenvironmental factors, i.e. a property of the ‘microenvironment‐stem cell unit’. The latter term is used to define a functional entity which includes both tumour stem cells and microenvironmental factors. The critical role of interactions between the microenvironment and glioma stem cells is supported directly or indirectly by the following observations: (1) glioma stem cells have been shown to reside preferentially in specific niches, the characteristics of which are known to influence cellular responses to radiation; (2) radiation modifies environmental factors; and (3) contrarily to the relative consistency of clinical data, in vitro experiments have reported a wide variation in the radiation responses of glioma stem cells.
The paper, therefore, focuses on the interaction between tumour stem cells and the microenvironment, analyzing how its various elements (endothelial cells, extracellular matrix, cytokines, nitric oxide, oxygen levels) are affected by radiation and how these factors might influence the response of tumour stem cells to radiation.
Finally, we will summarize the ongoing debate on how best to culture glioma stem cells to facilitate their study in vitro, and consider the difficulties encountered when designing assays to characterize their radiation responses.
The first experimental evidence for the existence of “stem‐like” cells in glial brain tumours was reported in 2002, when these malignancies were shown to contain cells that were capable of clone‐formation under culture conditions used in the study of ‘normal’ neural stem cells, and could also be induced to undergo differentiation along astrocytic and neural lineages (Ignatova et al., 2002). Two years later, a study using an intracranial xenograft model provided evidence for the ability of a subpopulation of cells isolated from glioblastoma specimens to induce tumours in vivo at a very high frequency. Injection of as few as 100 selected cells into the brains of NOD‐SCID (non‐obese diabetic, severe combined immunodeficient) mice was sufficient for the formation of human brain tumours, which phenotypically resembled the patient's original specimen and could be serially transplanted. In contrast, injection of up to 105 negatively‐selected cells did not generate tumours (Singh et al., 2004). Thereafter, several studies have analysed the behaviour of primary glioblastoma cell lines sorted by expression of candidate stem cell markers (CD133, A2B5, SEEA‐1) or by growth pattern (sphere or adherent) in serum free medium (SFM) supplemented with growth factors. These cells were characterised in terms of self‐renewal (tested with the neurosphere formation assay, NFA), differentiation potential and in vivo tumorigenicity (tested in intracranial xenograft models). On one hand, the reported data strongly support the existence of a subpopulation of highly tumorigenic cells, as indicated by the tumour induction rates observed for putative stem cells in individual studies (see Table 1). On the other hand, the inconsistencies between the various studies highlight the limitations of present cell sorting methods and the need for standardisation of assays that test for ‘stemness’ (Singh et al., 2004; Hemmati et al., 2003; Galli et al., 2004; Yuan et al., 2004; Bao et al., 2006a,; Beier et al., 2007; Wang et al., 2008; Gunther et al., 2008; Ogden et al., 2008; Son et al., 2009; Liu et al., 2009; Kondo et al., 2004; Chang et al., 2009; Clement et al., 2010). A detailed analysis of the validity of individual markers or techniques is beyond the scope of this paper but an excellent overview of the topic is provided in a recent review (Campos and Herold‐Mende, 2011).
Studies characterizing glioma cells in terms of self‐renewal (tested with the neurosphere formation assay) and in vivo tumorigenicity (tested in xenograft models).
The accumulating evidence that only a small subpopulation of cells is capable of giving rise to a tumour has led to the theory that the almost inevitable recurrence of glioblastoma is due to persistence of these cells despite multimodality treatment. Several clinical studies have investigated the prognostic value of putative stem cell markers and related cellular features in tumour specimens. Although these studies have generally involved relatively small numbers of patients and have tended to focus on only one or two markers, the results have consistently supported the implication of glioma stem cells in treatment resistance. In a retrospective study of 95 glioma specimens of different grades (47 glioblastomas), multivariate analysis including histological grade, patient age and extent of resection showed significant associations of CD133 expression (>1% vs ≤1% positive cells) and organization of these cells in clusters (cluster vs single cells) with shorter overall survival (Zeppernick et al., 2008). Similarly, a prospective multivariate analysis of specimens from 44 patients with glioblastoma indicated that CD133 expression (>2% vs ≤2% positive cells) and in vitro neurosphere formation ability (present vs absent) were prognostic factors for a higher risk of death. This association was independent of symptom duration, extent of surgical resection, patient age, MGMT status and p53 status (Pallini et al., 2008). Consistent with these results are two additional retrospective analyses of glioma specimens. One reported significantly lower survival in patients with gliomas co‐expressing CD133 and nestin (Zhang et al., 2008), the other showed a correlation between decreasing overall survival and increasing levels of nestin expression in tumour cells (Strojnik et al., 2007).
The greatest limitation of these data is the retrospective design of three out of the four studies. Also, although the results of the latter two studies were generated by multivariate analysis, their value is lessened because adjustments were made only for tumour grade in the former study and for patient age and sex in the latter, without incorporating the other established prognostic factors. Despite these weaknesses, the consistency of the association between prognosis and expression of the putative stem cell marker CD133 in all four studies is striking.
Evidence for a role of glioma stem cells in determining treatment resistance also comes from a recent publication that analysed expression of CD133 in glioma specimens from ten patients that had undergone surgical resection before and after high‐dose irradiation delivered by stereotactic radiosurgery (Gamma Knife) followed by external beam radiation (Tamura et al., 2010). The percentage of CD133+ cells was significantly higher in the post‐treatment tumour material than in the original specimens. Although these data do not establish a causative relationship, they are consistent with the hypothesis that glioma stem cells are capable of surviving high doses of radiation.
Because radiation therapy is the primary treatment modality for patients with glioblastoma, the proposed radioresistance of glioma stem cells is a corollary of the theory that holds tumour stem cells responsible for the very low cure rates observed. However, laboratory studies comparing radiation responses of glioma stem cells with those of differentiated or ‘non‐stem’ glioma cells have until now produced conflicting data. To our knowledge there are only two studies (Bao et al., 2006a; McCord et al., 2009) that have performed a comparative analysis of clonogenic survival assays in the two cellular subpopulations. Clonogenic survival is traditionally recognized as the ‘gold standard’ endpoint for testing intrinsic radiosensitivity in a clinically relevant manner. In the first published paper (Bao et al., 2006a,), colony formation assays were performed using CD133+ and CD133− cells derived either from a biopsy specimen or from glioma xenografts. Cells were either irradiated (5 Gray (Gy)) or left untreated. Colony formation rates were not quantified, but representative images of surviving colonies were presented to support a relative radioresistance of CD133+ cells. As a measure of post‐irradiation clonogenicity, the study also reported tumour formation rates of CD133+ and CD133− cells derived from tumour specimens or glioma xenografts. The authors demonstrated that in vitro irradiation with 2 Gy did not significantly affect tumourigenicity of CD133+ cells, while a higher dose of 5 Gy resulted in a 5–10 fold increase in the minimum number of cells required for tumour initiation compared to unirradiated cells (evaluated after 8 weeks). As CD133− cells did not generate tumours, comparison of radiation response curves, and therefore radiosensitivity, of CD133+ and CD133− cells by this method was not feasible. In the more recent study (McCord et al., 2009) reporting clonogenic analysis of primary cell lines obtained from glioblastoma specimens, the surviving clonogenic fraction of CD133+ cells was significantly higher than that of the CD133− population in one of the two cell lines tested. Interestingly, when clonogenic assays were performed on a panel of CD133+ primary cell lines and three established glioma cell lines, substantial variability was observed in their survival curves, with all six CD133+ cell lines being more radiosensitive than the established cell lines. These findings highlight the heterogeneity of in vitro radiosensitivity that exists amongst primary cell lines, despite prospective selection for putative stem cell markers, and illustrate the inherent problems associated with comparing radiosensitivity parameters across cell lines of different origins.
To further investigate the intrinsic radiosensitivity of glioma stem cells relative to non‐stem cells, the authors of the landmark Bao study (Bao et al., 2006a) analysed additional radiation responses of the two subpopulations in vitro and in vivo. An in vitro cell mixing and repopulation experiment, in which CD133+ and CD133− cells from a tumour specimen were differentially labelled, mixed in definite ratios and left untreated or irradiated (5 Gy), showed that, after 8 days of culture, the percentage of CD133+ cells increased from 5% to ~80% in the irradiated population and to ~25% in the untreated one. In the corresponding in vivo study, in which CD133+ and CD133− cells were labelled, mixed and implanted into the brains of athymic nude mice that were subsequently irradiated (5 Gy) or left untreated, the ratio of CD133+:CD133− cells increased from 0.2 to more than 4 in the irradiated mice and to ~2 in the control group. These data are consistent with enhanced survival of glioma stem cells after radiation but also indicate that, under these conditions, a component of the survival advantage of CD133+ cells over CD133− cells is independent of radiation.
To identify possible mechanisms underlying an intrinsic radioresistance of CD133+ cells, proteins involved in apoptosis and early DNA damage checkpoint responses were analysed in cells derived from tumour specimens and xenografts (Bao et al., 2006a). Western blot for cleaved caspase‐3 (before and 24 h after 2 or 5 Gy) and immunofluorescence staining for Annexin‐V‐FITC (before and 20 h after 3 Gy) both showed lower rates of apoptosis in CD133+ cells compared to CD133− cells. Furthermore, analysis of phosphorylated ATM, Rad17, Chk1 and Chk2 (before and 1 h after 3 Gy) demonstrated higher activation of these cell cycle checkpoint proteins in CD133+ cells than in CD133− cells. Alkaline comet assays and quantification of cells staining positively for phosphorylated histone 2AX performed before and after irradiation (3 Gy) were consistent with a potential mechanism involving faster resolution of both single and double strand breaks in CD133+ cells than in CD133− cells.
A more recent study also compared CD133+ and CD133− populations from primary glioma cell lines with regard to early DNA damage checkpoint responses and DNA repair capacity. While increased activation of Chk 1 and Chk2 was confirmed in both irradiated (1 h post 3 Gy) and unirradiated CD133+ cells, no significant difference in DNA damage induction or repair was observed between the two cellular subpopulations, as measured by comet assay and phosphorylated histone 2AX positivity (Ropolo et al., 2009).
Two further studies have analysed in vitro viability of glioma stem and non‐stem cells (defined by expression of CD133) after irradiation and demonstrated a higher percentage of surviving cells in the former population (Chang et al., 2009; Lomonaco et al., 2009). These observations are of limited value, however, as viability at early time points is not a measure of clonogenicity and may not correlate with a clinically relevant response to radiation.
Finally, four studies have shown that radiosensitivity of glioma stem cells can be increased by inhibiting specific proteins (SirT1 (Chang et al., 2009), Notch (Wang et al., 2010b), Chk1/Chk2 (Bao et al., 2006a), autophagy‐related proteins Beclin and ATG5 (Lomonaco et al., 2009)), two of which (Chang et al., 2009; Wang et al., 2010b) have been associated with stemness or tumorigenicity. Although these findings suggest that there are intrinsic cellular features of glioma stem cells that can be targeted to modify radiation responses, the specificity of these mechanisms to this cellular subpopulation needs to be investigated further.
More detailed discussion of relevant aspects of the DNA damage response in carcinogenesis and cancer treatment, including glioblastoma, is provided in three articles that are also published in this edition of Molecular Oncology. These reviews cover chromosomal instability (Krämer et al.) and therapeutic approaches targeting defects in DNA repair (Evers and Helleday) and replication stress (Toledo et al.).
In summary, although a number of findings indicate that there are differences between the radiation responses of glioma stem cells and their differentiated counterparts, it is difficult to evaluate their significance in terms of radiosensitivity for the following reasons: 1) the importance of apoptosis in determining clonogenicity after radiation has not been established and probably varies between cell types (Steel, 2001); 2) the apoptosis data refer to single time points whereas the time‐course of apoptosis after radiation varies considerably between cell lines; 3) DNA damage responses need to be interpreted in the light of clonogenic survival data as their validity in terms of predicting for radiosensitivity is still under debate (Lobrich et al., 2010). Also, the inconsistency of the DNA repair data suggest that some mechanisms may be cell line specific. Considering the extensive heterogeneity of glioblastoma, the conclusion that intrinsic radioresistance of CD133+ cells is responsible for treatment resistance should probably be applied with caution at this stage.
The variability and scarcity of existing data regarding radiosensitivity of glioma stem cells and non‐stem cells can, on one hand, be viewed as a reflection of the inherent difficulties in designing clonogenic survival assays for cellular populations that present distinct growth patterns and require different culture conditions. On the other hand, since existing experimental models testing radiation responses do not take into account factors present in the tumour microenvironment, they can also be considered as an indication of the need for more representative assays for determining cellular radiosensitivity. This consideration together with the observation that glioma stem cells reside preferentially in specific niches (Zeppernick et al., 2008; Calabrese et al., 2007; Li et al., 2009; Christensen et al., 2008) leads to our proposal that radioresistance is more likely to be a property of the ‘microenvironment‐stem cell unit’, a functional entity within which glioma stem cells are able to maintain or enhance intrinsic cellular features that contribute to radiation resistance.
Indirect evidence for the relevance of the ‘microenvironment‐stem cell unit’ in determining radiosensitivity comes from a recent study comparing induction and repair of radiation‐induced DNA damage in CD133+ cells grown in vitro and as intracranial xenografts. Levels of phosphorylated histone 2AX foci decreased more rapidly in the in vivo setting, suggesting a more efficient DNA repair ability of these cells in the presence of microenvironmental factors (Jamal et al., 2010). Also consistent with the proposed theory are the results of a study that used a 3‐dimensional organotypic ‘explant’ system of surgical glioblastoma specimens to test the effect of radiation and/or Notch inhibition on glioma stem cells (Hovinga et al., 2010). Radiation alone (10 Gy) had a dramatically lower impact on self‐renewal capacity of tumour cells in the explants than in neurosphere cultures. The effect of Notch inhibition also varied between the two models, causing a further decrease in neurosphere formation rate only when tumour cells were treated in the explant setting. The importance of the ‘microenvironment‐stem cell unit’ is also indicated by a study that used mouse models of medulloblastoma to investigate mechanisms responsible for regional differences in radiation responses (Hambardzumyan et al., 2008). After demonstrating that Akt signalling exerts different effects in nestin‐expressing cells residing in the perivascular niche compared with cells forming the tumour bulk, the authors conclude that in order to detect cell‐type‐specific responses it is critical to use models that recapitulate the various cell types within an appropriate environment.
Despite the limitations of these studies (for example, the uncertain validity of radiation‐induced foci and apoptosis levels as predictive indices of cellular radiosensitivity and the possibility that neurosphere culture conditions select for specific subpopulations), they all highlight the importance of interrogating the ‘microenvironment‐stem cell unit’ when investigating the biology of therapeutic radiation responses.
In order to understand how the tumour microenvironment influences radiation responses of glioma stem cells, it is essential to characterize the niche in which they are thought to reside. Studies investigating whether glioma stem cells exist in specific niches have analyzed the distribution in tumour specimens of cells positive for CD133 or other markers associated with a stem cell phenotype. The first published series (Calabrese et al., 2007) evaluated the distance of nestin+ and nestin− cells from the nearest CD34+ endothelial cell in frozen sections from ten glioblastoma specimens. Nestin+ cells, three quarters of which co‐expressed CD133, were significantly closer to capillaries than were nestin− cells. Also, 3D reconstruction of serial images from four glioma specimens, using multiphoton laser‐scanning microscopy, showed that nestin+ tumour cells were often in direct contact with tumour capillaries. A close relationship between CD133+ cells and vascular structures was also found in a study (Christensen et al., 2008) analyzing paraffin sections from seventy‐two glioblastoma specimens: 54% of tumours exhibited CD133+ niches, which were defined as “a limited entity of cells corresponding to a minimum of 5–10 cells identifiable at low magnification”. Many of these entities were perivascular or associated with necrotic regions. The latter association was indirectly confirmed by immunofluorescence studies on frozen tumour samples, which demonstrated co‐expression of CD133 and hypoxia inducible factor 2α (HIF2α). Finally, a study that was mentioned in the previous section reported that CD133+ cells were found in clusters in 41 of 47 glioblastoma specimens (Zeppernick et al., 2008).
The interaction of glioma stem cells with blood vessels, and more specifically with endothelial cells and vascular endothelial growth factor (VEGF), is one of the main areas of interest in this field (Knizetova et al., 2008). New insights were recently offered by two studies, which demonstrated that CD133+ cells are capable of undergoing differentiation along the endothelial lineage and that a subpopulation of endothelial cells within glioblastomas are directly derived from tumour cells (Ricci‐Vitiani et al., 2010; Wang et al., 2010a). Regardless of the function and/or fate of glioma stem cells in the perivascular area, their presence in this niche prompts the question of whether there are specific elements in such regions that contribute to radiation resistance. Apart from the obvious interaction with endothelial cells, glioma stem cells residing in perivascular regions are in contact with extracellular matrix (ECM) components that are preferentially expressed within and around blood vessels. The vascular basement membrane is composed of collagen, fibronectin, laminin, heparan sulphate, entactin and vitronectin, while the tumour ECM surrounding blood vessels is rich in tenascin C, secreted protein acidic and rich in cystein (SPARC) and thrombospondin. Interestingly, all components but vitronectin are scarcely present in the remaining tumour ECM. A clear and schematic representation of the heterogeneous distribution of ECM proteins in glioblastoma is provided in a review discussing their role in modulating glioma cell invasion (Bellail et al., 2004). An additional component of the perivascular niche is nitric oxide (NO), synthesized by endothelial nitric oxide synthase (eNOS) which is known to be highly expressed in vessels within glioblastomas (Iwata et al., 1999).
Expanding on these observations, the following paragraphs will analyze evidence relating to the theory that elements of the perivascular niche (endothelial cells, ECM proteins, NO and relevant cytokines) and hypoxic areas modulate cellular responses to radiation. This discussion will include data derived from studies of glioblastoma along with information from other cancers. A schematic representation of these interactions is given in Figure 1.
Schematic representation of interactions (proven or suggested by data on other cell types) between glioma stem cells and components of the microenvironment. Blue arrows indicate a positive regulation, the brown arrow a negative one, in terms of proliferation ...
Several studies analyzing the association between endothelial cells and glioma stem cells support the hypothesis that interactions between these two cellular populations are reciprocal. On the one hand, glioma stem cells have been shown to exert a pro‐angiogenic effect that is mediated by stimulation of endothelial cells through production of VEGF (Bao et al., 2006a; Folkins et al., 2009; Salmaggi et al., 2006). A recent study investigating the interaction between glioma and endothelial cells in both a 3D co‐culture system and an in vivo model demonstrated that glioblastoma cells became incorporated into the tumour vasculature (Shaifer et al., 2010). Interestingly, vascular networks formed in co‐culture remained stable for weeks, unlike those arising from endothelial cells alone, which regressed soon after formation. On the other hand, endothelial cells have been proven to play a role in glioma stem cell maintenance. The evidence for this comes from the previously mentioned study by Hovinga (Hovinga et al., 2010) and colleagues that used a 3‐dimensional organotypic ‘explant’ system of surgical glioblastoma specimens. Selective elimination of endothelial cells from the model casued a >50% reduction in the neurosphere formation capacity of single cells obtained by dissociation of the explants. Data from other central nervous system tumour types are consistent with these results. A study comparing the ability of endothelial and other cells (CD133− tumour cells, fibroblasts or astrocytes) to maintain the viability of CD133+/nestin+ tumour cells derived from medulloblastoma and ependymoma demonstrated that a larger percentage of tumour spheres survived if they were co‐cultured with endothelial cells for 5 days. After 2 weeks, these spheres were up to five times larger than those grown with control cells (Calabrese et al., 2007). Co‐culture with endothelial cells was also more likely to maintain self‐renewal of CD133+/nestin+ brain tumour cells, demonstrated by the ability of these cells to generate spheres after two weeks of co‐culture with the various cell types (Calabrese et al., 2007).
Evidence for reciprocal regulation of radiation responses between endothelial and tumour cells is accumulating. The above mentioned study investigating the interaction between glioma and endothelial cells in a 3‐dimensional co‐culture system demonstrated a survival advantage after irradiation (6 Gy) for mosaic vasculature compared to blood vessels formed by endothelial cells alone. Similarly, a study comparing radiation‐induced apoptosis in endothelial cells either in mono‐culture or in co‐culture with glioma cells showed that after 3 Gy there were significantly higher levels of apoptosis in the mono‐cultured cells (Brown et al., 2004). Consistent with these findings are the results of a study analyzing the effect of endothelial susceptibility to apoptosis on radiation‐induced apoptosis in fibrosarcoma and melanoma in vivo models. After a large single dose of 15 Gy, tumour cell apoptosis increased in tumours with an apoptosis‐sensitive endothelium (grown in asmase +/+ or Bax+/+ mice), but not in those with apoptosis resistant endothelial cells derived from asmase −/− or Bax−/− mice (Garcia‐Barros et al., 2003).
Although suggestive of a radioprotective effect of glioma cells on endothelial cells, the correlation between radiosensitivity of gliomas and the endpoints used in these in vitro studies (regression of blood vessels and apoptosis of endothelial cells) is not known. Furthermore, the use of large single doses of radiation in these experiments limits extrapolation of the findings to the clinical setting. Further evidence including clonogenic survival data is required to confirm the existence of an enhancing feedback loop between endothelial and glioma stem cells that might modulate radiation responses and thence survival. In the light of the two recent studies reporting the existence within glioblastomas of a subpopulation of endothelial cells derived directly from glioma stem cells (Ricci‐Vitiani et al., 2010; Wang et al., 2010a), the underlying mechanisms of such a relationship are likely to be complex and multifaceted and their elucidation will require detailed analysis in robust model systems.
The spatial association of glioma stem cells with specific ECM components expressed within and around blood vessels is particularly intriguing given that evidence is accumulating across cancer types for the importance of cell‐ECM interactions in influencing the response of tumour cells to radiation. Review of the relevant literature identifies three main mechanisms that might influence glioma stem cell survival and radiosensitivity. ECM components have been proposed to 1) serve as a deposit for proteins which modulate radiation responses, 2) operate as a substratum for the activation of pro‐survival integrin‐mediated signalling cascades in tumour cells following radiation, and 3) create a more favourable niche for proliferation of cells that survive irradiation.
The first property is suggested by the ability of heparan sulphate, a component of the basement membrane, to bind basic Fibroblast Growth Factor (bFGF) (Folkman et al., 1988), which has been shown to stimulate growth (Loilome et al., 2009) and inhibit radiation‐induced apoptosis (Bao et al., 2006a) of glioma stem cells in vitro. More evidence for a modulatory role of this growth factor in radiation responses comes from studies on other cell types. bFGF has been found to increase radioresistance of endothelial cells both in clonogenic survival assays and an in vivo model where it protected against development of lethal radiation pneumonitis (Fuks et al., 1994). A protective effect was also demonstrated in HeLa cells, using clonogenic survival assays (CohenJonathan et al., 1997). Studies analysing the effect of bFGF on clonogenicity of glioma stem cells post‐irradiation are needed to test more rigorously the potential radioprotective role of bFGF in these cells. Heparan sulphate acts as a binding site for many additional proteins (Lindahl and Li, 2009); these should also be interrogated with regards to their effect on glioma stem cell survival.
The second mechanism finds its basis in the literature correlating increased radioresistance of glioma cell cultures with expression and activation of integrins β1 (Cordes et al., 2006), αvβ3 and αvβ5 (Monferran et al., 2008). The role of integrins in modulating radiation responses has been demonstrated in various cancer types (Sandfort et al., 2007). A study comparing clonogenic survival of glioma cells irradiated 24 h after plating on fibronectin, Matrigel, bovine serum albumin (BSA) or polystyrene showed that the first two substrata significantly increased survival, but only in one of the four cell lines tested (Cordes et al., 2003). However, substratum‐dependent survival has also been demonstrated in a lung carcinoma cell line and a lung fibroblast stem cell line (Cordes and van Beuningeni, 2003). Characterization of the effects of glioma stem cell attachment to the various ECM components present in the tumour is essential in order to clarify whether any of these factors play a role in determining the fate of glioma stem cells after irradiation.
The third mechanism is supported by three observations related to tenascin C. This molecule stimulates tumour cell proliferation (Midwood and Orend, 2009); its expression in glioma specimens correlates inversely with the degree of cell differentiation (Higuchi et al., 1993); and a study analysing radiation‐induced connective tissue changes in women who had undergone radiation therapy for breast cancer demonstrated an increased expression of tenascin C in irradiated areas(Riekki et al., 2001). Taken together, these data suggest that a post‐irradiation increase in tenascin C, which is already overexpressed in the perivascular niche where glioma stem cells preferentially reside, might have a protective role. The likely mechanism involves tenascin C stimulating proliferation of surviving cells thus counterbalancing radiation cell killing. The role of this molecule in treatment resistance is also suggested by an immunohistochemical study, in which a significant reduction in survival was observed in glioblastoma patients whose tumours strongly expressed tenascin C (87 patients), compared to those with reduced expression (12 patients) (Leins et al., 2003). Preliminary but encouraging evidence supporting tenascin C as a therapeutic target in glioblastoma has been reported in radioimmunotherapy studies (Reardon et al., 2002; Reardon et al., 2008) and a clinical study in which tenascin C was downregulated by RNA interference (Rolle et al., 2010).
Modulation of glioma stem cell radiation responses by other ECM components such as SPARC and thrombospondin is less well supported, although data from other cancer sites suggest they might exert a radiosensitizing effect (Tai et al., 2005; Maxhimer et al., 2009). These factors merit further investigation in the context of glioblastoma, especially with regards to SPARC, expression of which has been reported to correlate inversely with survival in astrocytic tumours (Rich et al., 2005; Capper et al., 2010).
Taken together these data highlight the importance of studies analysing the complex interactions between various ECM components and tumour cells, in order to clarify how the vicinity of glioma stem cells to specific elements influences their survival post‐irradiation. As mentioned previously, however, the heterogeneity observed amongst glioblastomas indicates that the individual mechanisms outlined above are likely to exert different levels of influence on treatment responses in different patients.
A role for NO in promoting radioresistance of glioma stem cells is suggested by two recent publications. One, using a genetically engineered mouse model of platelet derived growth factor (PDGF)‐induced glioma, demonstrated that endothelial nitric oxide synthase (eNOS) co‐localized with endothelial cells that were surrounded by tumour cells which co‐expressed nestin, Notch and the NO receptor soluble guanylyl cyclase. Furthermore, NO activated Notch signaling and promoted stemness in primary cultured mouse glioma cells (Charles et al., 2010). The other study showed that inhibition of the Notch pathway with γ‐secretase inhibitors (GSI) increased radiosensitivity of glioma stem cells (defined as CD133+ cells). Expression of the constitutively active intracellular domains of Notch1 or Notch2, which function downstream of γ‐secretase, rendered glioma stem cells more resistant to radiation than control cells, regardless of treatment with GSI (Wang et al., 2010b). Given this evidence, it is plausible to speculate that, in the perivascular niche, NO synthesized by highly expressed eNOS activates the Notch pathway in adjacent glioma stem cells and increases their radioresistance.
The role of hypoxia in radioresistance of tumours was first observed more than a century ago and is now well established (Overgaard, 2007). In general, the reduced cellular radiosensitivity observed in a hypoxic environment is thought to be due to the different fate of free radicals produced in the cell by radiation. The reaction of these radicals with oxygen changes and stabilizes their chemical composition so that DNA damage is more likely to occur. This is often described as chemical ‘fixation’. In the absence of oxygen, free radicals are more likely to react with H+ ions, returning to their original form and reducing the level of DNA damage (Horsman and Overgaard, 2002). Tumour cells that survive hypoxic stress generate a range of pro‐survival responses that are induced by transcription of various genes in response to the hypoxia inducible factor (HIF) family of transcription factors. Evidence for a differential response of glioma stem cells to hypoxia was found in a study that reported higher levels of HIF2α and various HIF‐regulated genes in these cells compared to non‐stem cells (Li et al., 2009). Furthermore, knockdown of HIFs in glioma stem cells resulted in reduced stemness in vitro and in vivo. A recent study conducted in orthotopic glioblastoma xenografts showed that increasing intratumoral oxygenation through normalization of the vasculature by treatment with interferon‐beta or bevacizumab enhanced radiosensitivity of these tumours (McGee et al., 2010).
The hypoxia‐stem cell interaction may therefore be viewed as a typical example of the role of the ‘microenvironment‐stem cell unit’, in the sense that hypoxic conditions maintain and enhance intrinsic cellular features of glioma stem cells rendering them more resistant to treatment.
Ever since the isolation of glioma stem cells, using cell culture techniques developed for the study of neural stem cells, characterization of the radiation responses of this cellular subpopulation has proved to be very challenging. If, on one hand, culturing glioma stem cells as non‐adherent spheres in serum‐free medium supplemented with growth factors has the advantage of exploiting a recognized feature of these cells and has been shown to preserve tumour genotype and phenotype (Lee et al., 2006; Hamer et al., 2008), on the other it introduces several uncertainties.
Fundamentally, comparisons of radiation sensitivity in stem and non‐stem cell populations are limited by the requirement for different culture conditions to maintain their respective phenotypic features. Furthermore, conventional clonogenic survival assays require some degree of cellular adhesion to the substratrum on which the cells are cultured. Three main strategies have been adopted to try to circumvent these issues: 1) exposure of both populations to serum‐free medium without growth factors for 24 h post‐irradiation followed by transfer to serum‐containing medium (Bao et al., 2006a); 2) plating of both populations on poly‐l‐lysine coated wells in serum‐free medium (McCord et al., 2009); 3) maintaining each cellular population in the appropriate medium, and quantifying clonogenicity of differentiated cells by colony formation and stem cell clonogenicity by neurosphere formation (Wang et al., 2010b). The first two approaches have the advantage of exposing the two populations to the same conditions during the experiment and of comparing the same type of colony, but it is very likely that the pro‐attachment methods (growth factor withdrawal, serum‐containing medium or poly‐l‐lysine coating) induce some degree of differentiation among the stem cells. Also, it is not possible to control for a potential differential effect of these changes on cellular responses to radiation damage. The third approach has the advantage of studying the radiosensitivity of the two cellular populations without compromising their culture conditions, but it requires comparison of two distinct types of colonies, formation rates of which might be influenced in different and unpredictable ways by the different methodologies.
In order to address some of these problems, and other issues related to characterization of cells growing as non‐adherent sphere cultures, alternative techniques for culturing glioma stem cells are being investigated. To date two studies have been published proposing protocols for adherent cultures of glioma stem cells in serum‐free medium, using ECM (Al‐Mayhani et al., 2009) or laminin (Pollard et al., 2009b) as attachment factors. Both studies claim that 2D (adherent monolayer) cultures are superior to 3D (non‐adherent sphere) cultures because they provide uniformity of exposure to growth factors, oxygen and nutrients, which would result in a more homogeneous cellular population. Further evidence is needed in order to evaluate whether these techniques offer a more efficient method for culturing glioma stem cells and whether they can be implemented in clonogenic survival assays to improve the reliability of radiosensitivity studies (Reynolds and Vescovi, 2009; Pollard et al., 2009a).
As well as exploring such innovations, it is essential that researchers in this field consider, and control for wherever possible, biases related to stem cell culture conditions. These include potential effects of growth factors on cell cycle progression, DNA repair pathways and apoptosis. There is increasing evidence that epidermal growth factor (EGF) and fibroblast growth factor (FGF) interact with processes involved in the radiation response. Studies analyzing the effect of these growth factors in maintenance of glioma and neural stem cells suggest that EGF and FGF activate pathways that modulate proteins involved in cell cycle progression and apoptosis (Loilome et al., 2009; Sato et al., 2010). An impact of EGF and FGF on apoptosis (measured as percentage of Annexin‐V positive cells) has been clearly shown in both CD133+ and CD133− glioblastoma cells exposed to serum‐free medium, with or without irradiation (Bao et al., 2006a). Evidence for a possible link between DNA repair pathways and EGF and FGF is derived from investigations on other cancer cell lines. A study in bronchial carcinoma cells analyzed the relationship between EGFR localization and DNA repair after irradiation and showed that nuclear translocation of the receptor resulted in increased activity of DNA‐dependent protein kinase (DNA‐PK, a key component of the non‐homologous end‐joining pathway) (Dittmann et al., 2005). Although the study found a correlation of nuclear import of EGFR with irradiation but not with exposure to EGF, an effect of long‐term stimulation with EGF cannot be excluded as the cells were treated with the growth factor at a single concentration and were analyzed after a maximum of 20 min. The possible relevance of these factors is supported by data from a model of wound repair, which showed that long‐term exposure to EGF was required in order to observe its stimulatory effect (Buckley et al., 1985). Evidence for a link between FGF and DNA‐PK activity comes from a study on HeLa cells, in which up‐regulation of the DNA repair enzyme correlated with the radioprotective effect of the growth factor (Ader et al., 2002). Given the cell specific nature of many of the biological effects of growth factors, the relevance of these interactions to glioma stem cells should be investigated before any conclusions are drawn.
It is hoped that the design of future radiation response studies will take into consideration these factors, both to increase the accuracy of comparative studies and to enhance our understanding of the role of these proteins in regulating radiosensitivity.
If radioresistance is a property of the ‘microenvironment‐stem cell unit’, as this paper proposes, it is essential that experimental models are developed that allow investigation of radiation responses of glioma stem cells in conditions that resemble as closely as possible their tumoral niche. Exciting advances in 3D tissue culture systems have been accomplished and to date have been applied predominantly to embryonic or adult stem cells used in studies of developmental biology and regenerative medicine. Available models vary greatly in composition and complexity, ranging from simple 3D scaffolds consisting of synthetic self‐assembling peptide hydrogels (Thonhoff et al., 2008), biodegradable polymers (Levenberg et al., 2003) or polystyrene (Bokhari et al., 2007), to in vitro systems that couple 3D matrices with microfluidic devices that enable multi‐parameter manipulation(Vickerman et al., 2008). Studies comparing the phenotypes of a variety of cell types grown in 2D and 3D models have clearly demonstrated the functional superiority of the latter. These systems have the capacity to allow exploration of the interactions that occur between the various elements discussed in this paper. A promising approach in this context is also the three‐dimensional explant system used in the above mentioned Hovinga study (Hovinga et al., 2010). It is hoped that these models will prove to be efficient tools to accurately identify targets for the modulation of radiation response and enable the development of new therapeutic agents to improve outcomes for patients with glioblastoma.
While data demonstrating the existence of glioma stem cells continue to accumulate, the published evidence supporting the hypothesis that radioresistance of glioblastoma is caused simply by an intrinsic property of glioma stem cells is less convincing. Data from gliomas and other cancer models strongly support a complementary theory that radiation responses are determined by the ‘microenvironment‐stem cell unit’. Based on published data, a variety of mechanisms that might contribute to the radioresistance of this functional unit have been proposed. In order to identify therapeutically relevant targets, it is critical that future studies are conducted in model systems that enable these mechanisms to be investigated further.
Mannino Mariella and Chalmers Anthony J., (2011), Radioresistance of glioma stem cells: Intrinsic characteristic or property of the ‘microenvironment‐stem cell unit’?, Molecular Oncology, 5, doi: 10.1016/j.molonc.2011.05.001.