Microenvironmental Regulation of Glioblastoma Radioresponse

Radiotherapy remains a primary treatment modality for glioblastomas (GBM); yet, whereas many GBMs initially respond, the vast majority recurs within 2 years of diagnosis (1). Given that increased total dose fails to improve local control (2) and that the pattern of failure includes a predominance of recurrence within the initial treatment volume (3), GBM cells in situ are considered to be radioresistant. Defining the processes and molecules responsible for this radioresistance should provide a rational basis for designing target-based strategies that enhance GBM radiosensitivity and therapeutic response. Investigations aimed at delineating such mechanisms and identifying targets for radiosensitization have generally focused on in vitro cultures of long-established glioma cell lines. However, the biology of glioma cell lines as reflected by genetic abnormalities, gene expression, and orthotopic growth patterns has little in common with GBM in situ (4). Moreover, the radiosensitivity of such glioma cells is not significantly different from cell lines initiated from tumor types that typically respond to radiotherapy (5).

With respect to a more biologically accurate model system, data now suggest that GBMs are driven and maintained by a subpopulation of clonogenic cells referred to as tumor stem-like cells (TSC). The identification and isolation of GBM TSCs has been primarily based on the stem cell–associated protein CD133 (6), although additional markers have been reported (7). CD133⁺ TSCs have a number of in vitro properties in common with normal neural stem cells including continuous self-renewal; expression of stem cell–related genes and the capacity to at least partial differentiate along neuronal and glial pathways (6, 8, 9). However, in terms of radiosensitivity, we have recently shown that CD133⁺ TSCs in vitro are actually more radiosensitive than established glioma cell lines (10); the applicability of CD133⁺ TSCs as a model of GBM radioresistance is thus unclear. A parameter that may influence radioresponse and obviously not considered in this initial investigation of GBM TSCs (10) is the brain microenvironment. In contrast to the traditional GBM cell lines, when implanted into the brains of immunocompromised mice TSCs grow as invasive neoplasms comprising heterogeneous subpopulations (6, 8, 9, 11–13). Moreover, brain tumor xenografts initiated from TSCs simulate the genotype and gene expression patterns of the GBM from which they originated (9, 12). Given that TSC-initiated intracerebral (ic) xenografts replicate the genotype, phenotype, and in vivo growth pattern of GBMs, we reasoned that they are also likely to reflect the potential influence of the microenvironment on GBM radiosensitivity.

Identifying a role for the microenvironment in GBM radioresponse requires a measure of intrinsic radiosensitivity that allows for the direct comparison between cells grown in vitro and as tumor xenografts. In contrast to tumor growth rate and animal survival, an indicator of radiosensitivity defined at the individual cell level and applicable to in vitro and in vivo models is γH2AX foci expression. It is well established that γH2AX foci correspond to radiation-induced DNA DSBs and that their dispersal correlates with DSB repair (14, 15). Because DSBs are the critical lesion in radiation-induced cell death, γH2AX foci also provide a measure of radiosensitivity (16–18). To determine whether the microenvironment influences the intrinsic radiosensitivity of GBM cells, we have used γH2AX foci to directly compare the radioresponse of GBM TSCs grown in vitro and as intracerebral xenografts. Data presented show that the initial level of radiation-induced γH2AX foci was significantly reduced in tumor cells within intracerebral xenografts and the foci that did form dispersed more rapidly as compared with cells irradiated under the in vitro conditions. These results thus imply that GBM cells grown intracerebral are less susceptible to DSB induction and have an increased capacity to repair DSBs, which then suggests that the brain microenvironment contributes to GBM radioresistance.

Neurosphere forming cultures NSC11 and GBMJ1 were isolated from 2 human GBM surgical specimens as described previously (10, 19). NSC11 was kindly provided by Dr. Frederick Lang (M. D. Anderson Cancer Center) and GBMJ1 was generated at Moffitt Cancer Center from surgical specimen classified as glioblastoma according to WHO criteria (20) and was obtained following informed consent in accordance with the local institutional review board. Neurospheres were maintained in medium consisting of DMEM/F-12 (Invitrogen), B27 supplement (1X; Invitrogen) and human recombinant bFGF and EGF (50 ng/mL each; R&D Systems). To dissociate neurospheres into single cells, spheres were treated with TryplE Express (Invitrogen) for 5 minutes at 37°C, then subjected to mechanical disaggregation and strained through a 40 μm cell strainer (BD). All cultures were maintained at 37°C in an atmosphere of 5% CO₂/7% O₂ (21).

CD133⁺ cells were isolated from NSC11 and GBMJ1 neurosphere cultures by FACS as previously described (10). Both CD133⁺ cell cultures met the criteria for tumor stem-like cells (6) including self-renewal, differentiation along glial and neuronal pathways, expression of stem cell–related genes and formation of brain tumors when implanted in immunodeficient mice (10, 21). For use in an in vitro experiment, CD133⁺ neurosphere cultures were disaggregated into single cells and seeded into poly-l-lysine–coated plates in neural basal media containing EGF and bFGF, that is, stem cell growth media. Under these conditions the TSCs grow as an adherent monolayer maintaining their CD133 expression and stem-like characteristics (10). To induce differentiation and cell cycle arrest, CD133⁺ cell were exposed to medium consisting DMEM/F-12 and 10% FBS for 10 days. Differentiation was defined as the loss of CD133 expression, the gain of expression of GFAP and/or βIII tubulin and cell cycle arrest.

Athymic nude mice (nu/nu, males, Harlan Laboratories) 4–6 weeks of age were used in these studies. To implant tumor cells, mice were anesthetized using isoflurane gas and placed in a small animal stereotactic apparatus (Stoelting). CD133⁺ cells (10⁵) were injected in a total volume of 5 μL at 1.0 mm anterior and 2.0 mm lateral to the bregma to a depth of 3.5 mm at a rate of approximately 1 μL/min. Mice were kept in the Moffitt Cancer Center animal facility and observed every day until the initial signs of morbidity. All experiments were conducted as approved by the Institutional Animal Care and Use Committee and the experiment protocols were approved by the Division of Comparative Medicine, University of South Florida.

For in vitro experiments, the CD133⁺ TSCs in stem cell medium or after differentiation were irradiated as monolayers using XRad 320 X-irradiator (Precision XRay Inc.) at 320 kV and a dose rate of 289.8 cGy/min. For in vivo irradiation, mice were anesthetized using a cocktail of ketamine/xylazine/acepromazine and placed in well-ventilated Plexi glass jigs with shielding for the entire torso of the mouse along with critical normal structures of the head (e.g., ears, eyes, neck). Irradiation was conducted using the same X-ray machine as for the in vitro studies at a dose rate of 289.8 cGy/min. Sham-irradiated mice served as the control group.

For in vitro analyses, CD133⁺ NSC11 and GBMJ1 were seeded onto Lab-Tek CC2-treated tissue culture slides (Thermo Fisher) at least 24 hours before use in an experiment. To visualize foci, cultures were fixed with 4% paraformaldehyde, permeabilized with 0.1% Triton X-100, and blocked with 1% bovine serum albumin (BSA) in PBS containing 5% goat serum. The slides were incubated with primary antibodies (1:1,000) to phospho-H2AX (Upstate Biotechnology) or 53BP1 (BD Transduction Laboratories) overnight at 4°C and followed by secondary antibodies Alexa Fluor 488 goat anti-mouse IgG and Alexa Fluor 594 goat anti-rabbit IgG (1:1,000), respectively (Molecular Probes), and mounted with Prolong Gold antifade reagent containing 4′, 6-diamidino-2-phenylindole (Invitrogen) to visualize nuclei. Cells were analyzed on a Zeiss upright fluorescent microscope. Data presented are the mean ± SD of 3 independent experiments in which 50 cells were evaluated.

For in vivo studies, mice were euthanized by CO₂ inhalation and perfused with 4% paraformaldehyde in PBS (pH 7.4) via cardiac puncture. Brains were then removed and placed in 10% buffered formalin for 48 hours before embedding in paraffin. The paraffin embedded brains were cut into 10 μ thick slices; sections were deparaffinized in xylene and rehydrated in decreasing grades of alcohol. Sections were boiled in citrate buffer and incubated in 1% bovine serum albumin (BSA) in PBS containing 10% goat serum. The slides were then incubated with primary antibodies to phospho-H2AX (Upstate Biotechnology; 1:500) or 53BP1 (BD Transduction Laboratories; 1:500) for overnight at 4°C followed by secondary antibodies Alexa Fluor 488 goat anti-mouse IgG and Alexa Fluor 594 goat anti-rabbit IgG at 1:500, respectively and mounted with Prolong Gold antifade (Invitrogen). Foci analysis was conducted using a Leica confocal microscope (Leica TCS SP5 Confocal) with evaluation of only nonnecrotic tumor tissue. To obtain an accurate foci count, the depth of the nuclei was scanned by focusing manually along the optical axis (z-direction; original magnification, 100×) and images of 10 to 20 slices were recorded per z-stack to map the entire nucleus, essentially as described (18). Data are presented as the mean ± SD of 3–4 mice collected in at least 2 independent experiments in which 50 cells were evaluated.

Differences between control and irradiated groups were evaluated according to Student's t-test with statistical significance defined as P < 0.05.

CD133 expression was evaluated in paraffin embedded tissue as described above using a primary antibody against CD133 (Cell Signaling; 1:500) followed by Alexa Fluor 594 goat anti-rabbit IgG (1:500). After washing in PBS, slides were mounted in antifade. To better visualize the expression of GFAP and βIII tubulin within brain tumor xenografts, analysis was conducted on frozen sections. After cardiac perfusion brains were postfixed overnight, transferred to 10% sucrose (4°C) followed by 20% sucrose (4°C) and embedded in Tissue-Tek (Sakura). Cryostat sections (6 μ) were acetone-fixed and blocked in 10% goat serum in 1% BSA for 1 hour at room temperature and incubated with primary antibodies diluted in 1% BSA [anti-GFAP, 1:200 (Millipore); anti-βIII tubulin, 1:500 (Abcam)] followed by incubation with secondary antibodies (goat anti-mouse conjugated to Alexa Fluor 488 and Alexa Fluor 594 goat anti-rabbit IgG (1:500), respectively, and mounted in antifade. Images were generated using a Leica confocal microscope.

For microarray analysis, 100 ng of total RNA was collected in replicate from intracerebral tumor xenografts and in vitro cultures and converted to cDNA (Ambion). The cDNA samples were amplified and labeled with biotin following the procedure initially described by Van Gelder et al.(22). Hybridization with the biotin-labeled cRNA, staining, and scanning of the microarray chip followed the prescribed procedure outlined in the Affymetrix technical manual as previously described (23). For the oligonucleotide probe arrays, Human Genome U133A 2.0 chips (Affymetrix) were used having >22,000 probesets. The chips were scanned and the output files were visually inspected for hybridization artifacts and then analyzed using Affymetrix Microarray MAS 5.0 software. After MAS 5.0 analysis, effects of in vivo growth were determined by dividing the expression levels from a given intracerebral xenograft (in vivo) by the corresponding CD133⁺ or differentiated in vitro samples. Fold-changes were uploaded into Gene Cluster software (http://bonsai.hgc.jp/∼mdehoon/software/cluster/software.htm), log 2 transformed and arranged in a self-organizing map (SOM; ref. 24), which was visualized using the Treeview software (http://bonsai.hgc.jp/∼mdehoon/software/cluster/software.htm). Genes that were differentially expressed in vivo as compared with in vitro conditions were defined as those with a >2.0 fold-change (increased) and <0.5 fold-change (decreased). The ROS/ARE (antioxidant responsive element) gene list was generated by combining the gene lists for ROS metabolism (25) and ARE containing genes (26) and consisted of 367 individual genes (Supplementary Table 1). The differentially expressed genes were filtered for those target genes and uploaded into Venn diagrams to show the overlapping increased and decreased genes between each experimental calculation.

To determine whether the brain microenvironment influences the intrinsic radiosensitivity of GBM cells, we initially focused on intracerebral xenografts initiated from CD133⁺ NSC11 cells. The intracerebral implantation of CD133⁺ NSC11 cells into nu/nu mice results in the formation of a highly invasive brain tumor containing cells expressing GFAP, βIII tubulin, and CD133 (Fig. 1). Detection of GFAP and βIII tubulin indicates the presence of tumor cells that have differentiated at least partially along astrocytic and neuronal pathways, respectively; CD133 expression suggests the continued presence of TSCs. Whereas 100% of the cells implanted intracerebrally were CD133⁺, at the initial signs of animal morbidity, 11% ± 3.3 (mean ± SD, n = 3) of the cells within a tumor expressed CD133. Thus, these data indicate that the implanted CD133⁺ NSC11 cells proliferate and undergo differentiation along both glial and neuronal pathways to form an invasive, phenotypically heterogeneous tumor, that is, recapitulating GBM in situ.

As an indicator of DSB induction and repair we used γH2AX nuclear foci. Mice bearing NSC11 intracerebral tumors were irradiated (6 Gy) at the initial signs of morbidity and euthanized at times out to 24 hours. Using 10 μ sections, γH2AX foci were quantified in individual nuclei using confocal microscopy and a stacking procedure that allowed for evaluation of whole nuclei in the z-direction. This procedure is necessary to prevent underscoring of foci due to incomplete visualization of the entire nucleus and is essentially as described by Rube et al. in their evaluation of γH2AX foci in mouse tissue (18). Only tumor cells in nonnecrotic regions of the tumor mass were evaluated; representative composite images of γH2AX foci in control and irradiated NSC11 intracerebral xenograft are shown in Figure 2A. The maximum number of γH2AX foci was reached at 0.5 hour after irradiation followed by a rapid decline at 1 hour with a further reduction by 6 hours; at 24 hours there was no significant difference as compared with control (Fig. 2B).

γH2AX foci levels were also evaluated in NSC11 cells grown in vitro (Fig. 2C). For this analysis 2 in vitro growth conditions were evaluated: CD133⁺ NSC11 cells maintained in stem cell medium (TSCs) and CD133⁺ NSC11 cells that had been exposed to 10% FBS for 10 days (differentiated), which results in the loss of CD133 expression and differentiation along astroglial and neuronal pathways (21). Use of both in vitro growth conditions thus approximates the phenotypic heterogeneity of the intracerebral xenografts. Moreover, use of both in vitro models addresses the potential contribution of cell cycle phase distribution in radiation-induced γH2AX foci formation, that is approximately 2-fold more foci are induced in G₂ than in G₁ (15). In the differentiated cultures >93% of cells are in G₁/G₀ versus only approximately 50% in the CD133⁺ cultures (data not shown and ref 10). For differentiated cells in vitro, the number of γH2AX foci induced by radiation (2 Gy) was less than that induced in the CD133⁺ NSC11 cells at each time point evaluated (Fig. 2C), which may be due to the greater percentage of G₁/G₀ cells. Although γH2AX foci returned to control levels by 24 hours after irradiation of the differentiated cultures, in the CD133⁺ TSCs the number remained significantly above control values, similar to our previous results (10). It should be noted that our initial analysis of radiation-induced γH2AX foci in CD133⁺ GBM TSCs was conducted at 20% oxygen (10); the present results were obtained for cells grown at 7% oxygen (21), which had no significant influence on foci induction or dispersal. Comparison of the time courses shown in Figure 2B and C indicates that for cells irradiated as intracerebral xenografts the maximum γH2AX foci dispersal rate occurred between 0.5 and 1 hour postirradiation, whereas for both in vitro models foci dispersal was relatively steady over the 24-hour evaluation period. These data suggest that repair of radiation-induced DSBs was considerably more rapid in tumor cells irradiated under orthotopic conditions.

Comparison of Figure 2B and C also revealed that 6 Gy delivered to intracerebral xenografts resulted in a similar level of γH2AX foci as induced by 2 Gy in vitro. To better illustrate this difference in susceptibility to radiation-induced γH2AX foci the dose response at 0.5 hour for each growth condition is shown in Figure 2D. With respect to in vitro growth conditions, the differentiated cultures contained fewer radiation-induced γH2AX foci at each dose tested than the actively cycling CD133⁺ cultures, again, perhaps due to the differences in cell cycle phase distribution. However, the NSC11 cells grown orthotopically, were substantially less susceptible to γH2AX induction than both in vitro culture models.

53BP1 is a critical protein participating in the radiation-induced DNA damage response that also forms nuclear foci after irradiation (27). Although the retention of 53BP1 at foci is facilitated by γH2AX, its initial recruitment to the site of DSBs does not require γH2AX (28) and thus provides an independent measure of DSB induction (18, 28). After irradiation of CD133⁺ NSC11-initiated xenografts the number of 53BP1 foci reached a maximum at 0.5 hour followed by a rapid decline at 1 hour returning to control levels by 6 hours (Fig. 3A). Irradiation of the NSC11 cells in vitro resulted in 53BP1 foci induction and dispersal similar to that detected for γH2AX. For differentiated cells in vitro, the number of 53BP1 foci induced by radiation (2 Gy) was less than that induced in the CD133⁺ NSC11 cells at each time point evaluated (Fig. 3B). By 24 hours after irradiation 53BP1 foci returned to control levels in differentiated cultures, yet remained significantly above control values in the CD133⁺ cultures. As compared with intracerebral xenografts, the dispersal of 53BP1 foci after irradiation under both in vitro growth conditions was considerably slower, remaining significantly above control levels at 6 hours (Fig. 3B). Consistent with the γH2AX data, these results suggest that the repair of DSBs is more rapid in NSC11 cells grown orthotopically. In addition, the number of 53BP1 foci induced in both in vitro models at 0.5 hour after 2 Gy was similar to that induced by 6 Gy for cells within the intracerebral xenografts. Thus, tumor cells grown orthotopically were less susceptible to radiation-induced 53BP1 foci formation.

To determine whether these results were unique to NSC11 tumors, similar experiments were conducted using CD133⁺ GBMJ1 TSCs. The intracerebral implantation of these cells results in the formation of highly invasive brain tumors with cells expressing GFAP, βIII tubulin as well as CD133 (Fig. 4), similar to NSC11 tumors. Whereas 100% of the GBMJ1 TSCs implanted intracerebrally were CD133⁺, at the initial signs of animal morbidity, the percent CD133⁺ cells was 9.6% ± 1.5 (mean ± SD, n = 3). After irradiation (6 Gy) of intracerebral xenografts, γH2AX foci were readily detectable at 0.5 hour returning to control levels by 6 hours (Fig. 5A). These results were then compared with those obtained from GBMJ1 cells maintained under 2 types of in vitro growth conditions: CD133⁺ cells maintained in stem cell medium (TSCs) and CD133⁺ cells that had been exposed to 10% FBS for 10 days (differentiated; Fig. 5B). In contrast to NSC11 cells, there was no detectable difference in the radiation-induced γH2AX response of CD133⁺ and differentiated GBMJ1 cells in vitro. Whereas >90% of the cells in the differentiated cultures were in G₁/G₀ versus approximately 57% in the CD133⁺ GBMJ1 cultures (data not shown), additional phenotypic changes may be involved affecting γH2AX induction under these 2 in vitro conditions. However, similar to the NSC11 model, the dispersal of γH2AX foci was more rapid after irradiation of GBMJ1 intracerebral xenografts as compared with the in vitro growth conditions. In addition, the susceptibility of cells grown orthotopically to radiation-induced γH2AX foci was substantially less than for GBMJ1 cells irradiated in vitro (2 vs. 6 Gy).

To begin to address the mechanisms responsible for the differences in γH2AX and 53BP1 foci induction as defined above, microarray analysis was used to generate gene expression signatures for NSC11 and GBMJ1 cells grown in vivo as intracerebral xenografts and in vitro under stem cell and differentiated conditions. To define the changes in gene expression resulting from in vivo growth, gene expression signatures of the intracerebral xenografts (in vivo) were directly compared with those generated from their corresponding in vitro cultures (CD133⁺ and differentiated). Using these data, a SOM was generated to compare the in vivo mediated changes in gene expression induced in the NSC11 and GBMJ1 models (Fig. 6A). Within the each of the GBM models, there were a significant number of genes in CD133⁺ and differentiated cultures that were commonly affected by in vivo growth. In addition, although there were differences between the GBM models, there were also a significant number of genes similarly affected by in vivo growth in both NSC11 and GBMJ1. These results suggest that the effects of brain microenvironment on gene expression are not tumor specific and may be of a more general nature.

The gene expression profiles comparing in vivo to in vitro growth were then evaluated in terms of genes that can play a role in radioresponse. Of particular interest were genes involved in ROS scavenging and antioxidant response, which have long been implicated as determinants of radiosensitivity. Accordingly, gene expression signatures as defined in Figure 6A for intracerebral xenografts versus their respective in vitro cultures were interrogated using a list of genes associated with ROS metabolism (25) and antioxidant response as reflected by an ARE containing promoter (Supplementary Table 1; ref. 26). The ROS/ARE genes whose expression was modified at least 2-fold by in vivo growth were identified for each in vitro growth condition (CD133⁺ and differentiated) and GBM model (NSC11 and GBMJ1) with similarities depicted in Venn diagrams (Fig. 6B). Each xenograft contained substantially more upregulated than downregulated ROS/ARE genes than their respective in vitro CD133⁺ cultures; 114 genes were commonly upregulated (top panel). The xenografts also contained substantially more upregulated than downregulated ROS/ARE genes than their respective differentiated in vitro cultures; 93 genes were commonly upregulated (middle panel). To determine whether there were similarities between the 2 in vitro phenotypes in terms of ROS/ARE genes affected by intracerebral growth, the 125 commonly affected genes (up- and downregulated) identified for intracerebral xenografts versus CD133⁺ cells (top panel) were compared with the 110 commonly affected genes identified for intracerebral xenografts versus differentiated cells (middle panel). As shown in the bottom panel of Figure 6B, 64 ROS/ARE genes expressed in CD133⁺ and differentiated cells in vitro were commonly upregulated under intracerebral growth conditions in both GBM models, whereas only 3 genes were downregulated (Supplementary Table 2). These results indicate that a consequence of orthotopic growth of GBM cells is an increased expression of ROS/ARE-related genes, which would be consistent with greater antioxidant capacity and reduced susceptibility to radiation-induced DSBs.

GBMs comprise a diverse set of tumors that are highly variable with respect to histology, genetic abnormalities, and gene expression profiles. However, despite their extensive biological heterogeneity, although some GBMs respond better than others, they all essentially fail radiotherapy (1). This relatively “homogeneous” clinical response in a background of intertumor heterogeneity suggests that the microenvironment may play a significant role in determining their radioresponse. The goal of this study was to test the hypothesis that the brain microenvironment contributes to GBM radioresistance. Toward this end it was necessary to directly compare the intrinsic radiosensitivity of cells grown in vitro and in vivo as orthotopic xenografts.

The gold standard for defining radiosensitivity in vitro is the clonogenic survival assay, which measures the consequences of radiation on the proliferative potential of individual cells. The radioresponse of human brain tumor xenografts, however, is typically defined according to animal survival and more recently using imaging techniques to define tumor growth rate, both of which are highly dependent on the number of tumor cells (clonogens) at the time of irradiation and in essence evaluate the radiosensitivity of the tumor cell population as a whole. Thus, to quantitatively compare the intrinsic radioresponse of tumor cells within intracerebral xenografts to that of cells maintained under in vitro growth conditions we used γH2AX foci, which provides a sensitive measure of radiation-induced DNA DSBs (15). The dispersal of γH2AX foci reflects DSB repair with the residual foci remaining at 24 hours after irradiation having been shown to correlate with radiation-induced cell death, that is, radiosensitivity. Whereas the relationship between residual γH2AX expression and radiosensitivity has been best established using in vitro models (5, 16, 17, 29), the correlation has also been reported after in vivo irradiation of human tumor xenografts (17) as well as mouse normal tissue (18, 30). The data reported here indicate that for 2 GBM TSC lines the dispersal of radiation-induced γH2AX foci was faster in cells within intracerebral xenografts as compared with cells grown in vitro. These results then suggest that the GBM cells have a greater capacity to repair DSB when grown under orthotopic conditions, indicative of radioresistance as compared with their in vitro counterparts. In contrast to the ROS/ARE genes, there was not a consistent increase in the expression of DNA repair-related genes in the in vivo versus in vitro models (data not shown). Thus, at present, the mechanism responsible for the enhanced repair capacity in GBM cells under orthotopic conditions remains to be defined.

GBM cells irradiated under intracerebral conditions were also substantially less susceptible to γH2AX induction than cells grown in vitro. Differences in cell cycle phase distribution are unlikely to account for this effect: similar results were obtained when comparing cells grown orthotopically to cells in vitro that were cycling (those grown in stem cell media) and those arrested in G₀/G₁ (those grown in differentiation media). Alternatively, γH2AX induction in vitro can be overestimated when cells are irradiated on glass slides due the photo-electric effect and the generation of secondary radiation particles (31), which was suggested as a potential explanation for the fewer γH2AX foci induced in mouse normal tissue as compared with mouse cells in vitro (18). However, in our studies radiation was delivered in vitro and in vivo using a 320 kV X-irradiator, eliminating the contribution of the photo-electric effect and the possible overestimation of γH2AX induction in vitro. Finally, whereas a reduction in the maximum number of γH2AX foci induced could also be the result of relative decreases in H2AX expression and/or the expression of H2AX kinases (e.g., ATM, DNA-PK; ref. 5), similar reductions were detected in the initial number 53BP1 foci induced ic, which provides an independent measure of DSB induction (28).

Thus, at this point data suggest that the reduced γH2AX and 53BP1 foci induction in intracerebral xenografts may be the result of fewer DNA DSBs. It has been reported that radiation-induced γH2AX expression was 3–4 times greater in well-oxygenated cells as compared with anoxic cells, with 0.55% oxygen producing a half-maximal response (32). With respect to the study presented here, anoxia as well as 0.55% oxygen would be expected to result in necrosis. Although areas of necrosis are present in the intracerebral xenografts, γH2AX and 53BP1 foci were evaluated only in healthy appearing tumor tissue (nonnecrotic) with a relatively uniform response across tumors. Thus, it seems unlikely that the reduced level of foci induction detected after intracerebral irradiation can be solely attributed to radiobiological hypoxia (∼0.5%; ref. 33). An additional determinant of DNA DSB induction and consequently radiosensitivity is cellular/tissue antioxidant capacity (34). Accordingly, enzymatic and nonenzymatic scavenging of ROS is well established to protect cells from radiation-induced DNA damage. Genes involved in ROS metabolism and antioxidant response were expressed at higher levels in intracerebral xenografts than in the TSCs or their differentiated progeny in vitro. If this gene expression pattern translates to increased antioxidant capacity as described for breast tumor cells in vitro (35), the consequence would be reduced levels of DSBs after irradiation of orthotopic xenografts, consistent with reduced levels of radiation induced γH2AX and 53BP1. Of note, Pollard et al. also recently showed that ROS scavengers reduced the levels of radiation-induced γH2AX foci in vitro, which was accompanied by a reduction in cytotoxicity (36).

The data presented here suggest that the brain microenvironment protects GBM cells from radiation-induced DSBs and facilitates their repair, a situation indicative of a relative decrease in intrinsic radiosensitivity. These results then imply that the microenvironment plays a major a role in mediating the radioresistance of GBMs. Although the relative decrease in susceptibility of intracerebral xenografts to DSB induction can be linked to modifications in gene expression, the specific mechanisms through which the microenvironment regulates radiation-induced DNA damage and repair, and thus radiosensitivity, remain the subject of future investigations. However, based on these, results it seems that for preclinical studies aimed at the identification of targets for GBM radiosensitization and the evaluation of corresponding radiosensitizers it will be necessary to account for the brain microenvironment.

No potential conflicts of interest were disclosed.

This work was supported by grant CA138519 from the National Institute of Health (P.J.T.) and a Cancer Center Support Grant issued to Moffitt Cancer Center from the National Cancer Institute.