Glioblastoma multiforme (GBM) is the most common form of brain tumor with a poor prognosis and resistance to radiotherapy. Recent evidence suggests that glioma-initiating cells play a central role in radioresistance through DNA damage checkpoint activation and enhanced DNA repair. To investigate this in more detail, we compared the DNA damage response in nontumor forming neural progenitor cells (NPC) and glioma-initiating cells isolated from GBM patient specimens. As observed for GBM tumors, initial characterization showed that glioma-initiating cells have long-term self-renewal capacity. They express markers identical to NPCs and have the ability to form tumors in an animal model. In addition, these cells are radioresistant to varying degrees, which could not be explained by enhanced nonhomologous end joining (NHEJ). Indeed, NHEJ in glioma-initiating cells was equivalent, or in some cases reduced, as compared with NPCs. However, there was evidence for more efficient homologous recombination repair in glioma-initiating cells. We did not observe a prolonged cell cycle nor enhanced basal activation of checkpoint proteins as reported previously. Rather, cell-cycle defects in the G1–S and S-phase checkpoints were observed by determining entry into S-phase and radioresistant DNA synthesis following irradiation. These data suggest that homologous recombination and cell-cycle checkpoint abnormalities may contribute to the radioresistance of glioma-initiating cells and that both processes may be suitable targets for therapy. Mol Cancer Ther; 11(9); 1863–72. ©2012 AACR.
Gliomas are the most common adult brain cancers, comprising 80% of diagnosed cases (1). Anaplastic astrocytoma (grade III) and glioblastoma multiforme (GBM; grade IV) tumors are highly lethal (1). Standard therapy is combined local aggressive cytoreductive surgery followed by radiotherapy. Unfortunately, GBM are particularly resistant to conventional treatment, leading to recurrence of brain tumors. A pool of glioma-initiating cells (2), similar to neural progenitor cells (NPC; ref. 3), can survive exogenous damage, such as the lethal double strand breaks (DSB), to repopulate tumor cells following treatment.
DSBs are highly detrimental to the structural integrity of chromosomes. To counteract this damage, normal and cancer cells initiate the DNA-damage response. DSBs are sensed primarily by the MRN complex; Mre11, Rad50, and Nbs1 proteins that rapidly relocates to the site of damage and recruits ATM (ataxia-telangiectasia mutated) kinase, resulting in downstream phosphorylation cascades that initiate cell-cycle arrest and DNA repair (4). There are 2 major DSB pathways engaged during DNA repair. Nonhomologous end joining (NHEJ) is initiated predominantly during the G1-phase of the cell cycle and because the majority of cells are in G1-phase, NHEJ is considered the major pathway for DSB repair. The second repair mechanism occurs during late S- and G2-phases. Homologous recombination is an error-free method of repair using sister chromatids as templates to replace damaged DNA (5). In contrast, NHEJ relies on ligases and excision repair enzymes to adhere broken DNA ends that can introduce spontaneous mutations (6). Depending on the severity of insult, cells with a high mutational burden may initiate apoptosis to minimize damage to the genome and prevent mutations from being maintained through subsequent cell divisions.
Compared with normal cells, cancer cells have enhanced DNA repair pathways, which confer greater survival. Investigations characterizing subpopulations of aberrant stem cells have shown efficient DSB repair through various DNA damage markers, but no specific pathways have been identified to explain the increased survival (7). Recent reports also provided evidence for differences in radiosensitivity between subpopulations of glioma-initiating cells (8–10), suggesting disparity in DNA repair efficiency. Whether DNA repair is a major mechanism of radioresistance in glioma-initiating cells remains controversial. To further understand these DNA repair pathways, we compared the DNA damage response to ionizing radiation (IR) in nontumor forming NPCs (11, 12) and glioma-initiating cells isolated from GBM tumors. The significance of this comparison is the potential for identifying key proteins or pathways unique to glioma-initiating cells. In our study, increased survival of glioma-initiating cells following IR was observed. Although, glioma-initiating cells had a higher proportion of surviving cells, attenuated checkpoint kinase activation suggested inadequate cell-cycle arrest at G1–S, allowing a portion of G1 cells to enter S-phase. NHEJ activity was reduced in several glioma-initiating cells lines, but these differences did not account for altered survival. Interestingly, glioma-initiating cells seemed to favor the homologous recombination repair pathway. These data indicate that glioma-initiating cells have an ineffective G1–S checkpoint allowing damaged cells to enter S-phase and use homologous recombination to repair the genome.
Materials and Methods
Glioma-initiating cells extraction and maintenance
Glioma-initiating cells enrichment from tumors from 3 GBM patients was carried out as described by Piccirillo and colleagues (13); resulting lines were designated as L1b, L2b, and L3b. Briefly, tumors were dissociated into single-cell suspensions before enriching for stem cells/neurospheres in serum-free media supplemented with 20 ng/mL EGF and 10 ng/mL basic fibroblast growth factor. U251 and U87 neurosphere cultures were obtained from Day and colleagues (14) and also underwent the glioma-initiating cells enrichment process. No authentication was carried out by the authors. Neurosphere cultures were maintained at 37°C in 5% CO2 with supplements (12) for 4 to 8 weeks before characterization, as described in Supplementary Methods. The neurosphere media is a selective culture that enriches for progenitors and precursors, but not differentiated cells. All primary and tumor cell lines grown in neurosphere media will contain a mixed population of progenitor and precursor cells. Throughout this article, we will refer to the entire population of cells within the cultures as glioma-initiating cells.
Nontumorigenic NPCs derived from fetal brain tissue are commercially available (ReNcell; Millipore) and were previously characterized by Reynolds and colleagues (11, 12, 15). For experiments, neurosphere cultures were grown for 5 days, before trypsinizing with 0.25% trypsin-EDTA (Invitrogen) in PBS into single-cell suspensions before all indicated treatments (12).
Dissociated neurospheres were prepared as duplicates before IR. Protocol was previously described (16). Details are in Supplementary Methods.
Nonhomologous end joining assay
The NHEJ reporter plasmid pEGFP-N1 (Clontech) was digested with HindIII (New England Biolabs) and purified by gel extraction kit (Qiagen), and aliquots analyzed by gel electrophoresis to confirm complete digestion. Neurosphere lines were transfected using a Gene PulserII (Biorad) at 220 V, 960 μF, infinite resistance. A total of 1 × 106 dissociated cells were transfected with 1 μg of linearized pEGFP-N1, and in parallel with 1 μg circularized pEGFP-N1 as control for transfection efficiency. GFP expression was measured by FACScan (Becton Dickinson) 48 hours later.
DNA pulse labeling
Iododeoxyuridine (IdU) analog (100 μm) was added into neurosphere media. Cells were placed for 30 minutes at 37°C before fixation and cytospinning onto Superfrost plus slides (see Supplementary Methods). Slides were placed for 1 minute in lysis buffer (50 mmol/L Tris-HCl pH8.0, 1 mmol/L EDTA, 0.1% SDS) at room temperature. Slides were rinsed with PBS before incubating in 1 mol/L HCl, for 5 minutes on ice, followed by 2 mol/L HCl, for 10 minutes at room temperature and 10 minutes at 37°C. Immunolabeling is described in Supplementary Methods.
Homologous recombination assay
Neurosphere lines were stably transfected as described above with 1 μg of pDR-GFP and stable integrants selected with 1 to 8 μg/mL puromycin (Sigma; ref. 17). Analysis was carried out 48 hours after I-Sce I−/+ transfection as previously indicated (18).
Fiber assay for DNA replication
The DNA fiber assay was conducted as previously described (19). Total DNA fibers were scored (≥300) for each cell line. The percentage of new initiations was expressed as fold change [new initiations / (continuing + new initiations)].
Long-term growth potential of glioma-initiating cells and NPCs
The gold standard to determine the presence of normal stem or progenitor cells is to evaluate their self-renewal and lineage differentiation (20, 21). Assays for glioma-initiating cells additionally also require determination of tumor formation in an animal model. We first examined whether glioma-initiating cells had stem cell markers and self-renewal capacity comparable with NPCs. The profile of stem cell markers differed between glioma-initiating cells cultures, but U251 and L1b showed near identical marker expression to NPCs (Supplementary Fig. S1A). All cultures except U87 displayed CD49f expression (22). Two other stem cell markers, Sox2 and Nestin, were detected in all glioma-initiating cells and NPCs with 80% to 95% and 78% to 93% positive cells, respectively, by IF (Supplementary Fig. S1A). When glioma-initiating cells and NPCs were plated with 5% fetal calf serum, both cell types began to differentiate within 5 to 7 days (Supplementary Fig. S2A). Neural (β-III tubulin) and glial fibrillary acidic protein [(GFAP) and mitogen-activated protein (MAP)] makers were analyzed 2 weeks later. Glioma-initiating cells expressed all lineage markers with percentage expression being 17% to 40% (β-III tubulin), 17% to 31% (MAP), and 18% to 41% (GFAP) across all glioma-initiating cells cultures. Lineage distribution in differentiated NPC was 31% (β-III tubulin), 18% (GFAP), and 17% (MAP; Supplementary Fig. S2B and S2C). As surface marker expression is only one method to define stem cell populations, we also evaluated long-term self-renewal using serial dilution. Single cells were seeded from passage 1, and expansion between passages was measured (Supplementary Fig. S1B). All lines expanded constantly as shown by an unchanged slope of a straight line.
Finally, to show that glioma-initiating cells could recapitulate tumors in vivo, we used a Scid/Nod tumorigenesis model (23). Intracranially injected tumor neurospheres (U251) and primary glioma (L2b) produced glioblastoma in mice (Supplementary Fig. S1C). Survival curves are shown in Supplementary Fig. S1D. Despite limited CD133 expression, they were capable of forming tumors. This finding is consistent with recent publications showing that both CD133-positive and CD133-negative glioma-initiating cells can induce tumor development (24). Histologic analysis showed tumors with highly dense region of marked nuclear atypia and pseudopalisades-like formation in surrounding tissue, all characteristic of GBM (Supplementary Fig. S1C).
Glioma-initiating cells cultures are radioresistant
We next determined whether glioma-initiating cells cultures had radioresistant properties (25). Following 2Gy IR exposure, cell survival was determined by trypan blue assay (19). All glioma-initiating cells cultures had more viable cells over time compared with NPCs (Fig. 1A; P < 0.05). Survival for glioma-initiating cells cultures ranged from 40% to 60% 96 hours after IR versus 20% for NPCs. glioma-initiating cells cultures also showed increased survival with doses from 2 to 10 Gy (Fig. 1B). Alternate analysis using neurosphere size (23, 26) showed NPCs after 2 Gy had a 0.51-fold reduction in size (Supplementary Fig. S3A). In comparison, glioma-initiating cells had a 0.89- to 0.66-fold reduction, suggesting a lesser impact of IR-induced damage on glioma-initiating cells growth (P < 0.01). At 5 Gy dose, NPCs reduced to 0.38 the original size, whereas the fold change in glioma-initiating cells size was 0.76 to 0.51 (P < 0.01).
We next determined whether glioma-initiating cells survival was due to accelerated repopulation or apoptosis resistance (27, 28). Radiation-induced DNA damage caused comparable cell death in glioma-initiating cells and NPCs at 24 and 48 hours post-IR (Fig. 1C and Supplementary Fig. S3B). We next used CFSE labeling to measure proliferation. CFSE incorporates into the membrane and daughter cells inherit half of the dye after mitosis. Individual colors illustrate cell divisions (generation numbers; Fig. 1D). All glioma-initiating cells proliferated at higher rates than NPCs following irradiation indicating that continuous proliferation rather than apoptosis resistance in glioma-initiating cells accounted for increased survival. In detail, 96 hours post-2 Gy IR showed reduced growth with NPC undergoing 5 to 6 divisions. U251, L2b, and L3b showed little change in proliferation with U251 and L2b capable of 6 to 7 divisions and L3b 5 divisions. At 5 Gy dose, U251 underwent 7 to 8 divisions, L2b and L3b 4 to 5 divisions, whereas NPC proliferation reduced to 3 to 4 divisions. At high dose (10 Gy), U251, L2b, and L3b cells proliferated for 4 to 6 divisions versus 2 to 3 for NPCs.
For long-term proliferation after 2 Gy irradiation, serial passage showed marked reduction in NPCs during initial passages (1–2) before returning to normal growth from passage 3 (Supplementary Fig. S4A). After 5 Gy reduction in growth was observed in glioma-initiating cells lines, but with faster recovery. A reduction in growth was observed in NPCs with recovery from passage 5 (Supplementary Fig. S4B). Conversely, L2b, U251, and U87 showed recovery from passage 3, followed by L1b and L3b at passage 4.
Delayed DSB repair in glioma-initiating cells during early recovery
The survival, serial passage, and neurosphere formation data indicated that glioma-initiating cells might have enhanced DNA repair, allowing rapid growth post-IR. To examine the DNA damage repair response, the levels of proteins central to DSB recognition and response were measured by immunoblotting. Interestingly, U87 and L3b had reduced protein expression compared with NPCs (Fig. 2A). In U87, total ATM and Mre11 levels were reduced (29). Similarly, protein levels of MRN complex members in L3b were diminished. In contrast, L1b and L2b displayed increased MRN protein levels, whereas no significant difference was identified between NPCs and U251. Radiation did not cause any consistent change in levels of DNA damage response proteins that might account for changes in radioresistance. We next investigated whether differences in the efficiency of DNA DSB repair might explain radioresistance by following γH2AX foci formation and resolution (Fig. 2B). Because the maximum foci number at 1 hour differed between cultures, the rate of repair was expressed as fold change by normalizing to their own neurosphere lines. In NPCs, efficient repair was identified by rapid reduction in γH2AX foci at 6 hours (reduced to 17%; Fig. 2C). Conversely, glioma-initiating cells repair was slow at early time points (<6 hours) ranging from 33% to 73% (P < 0.01). Repair of DNA DSB was not completed until 12 to 24 hours in glioma-initiating cells.
NHEJ function does not explain glioma-initiating cells radioresistance
To investigate repair mechanisms responsible for radioresistance, we measured NHEJ activity in vivo by recapitulating the joining of DNA ends with a linearized GFP expression vector (Fig. 3A). NHEJ activity was detected in all neurosphere cultures (Fig. 3B) but was variable between glioma-initiating cells cultures. In U251 and U87, NHEJ efficiency was reduced to 23% (P < 0.018) and 19% (P < 0.024) compared with NPCs (46%). L2b had modestly reduced NHEJ activity to approximately 70% of NPCs, whereas L3b was comparable with NPCs. L1b had higher NHEJ activity (P < 0.034) than NPCs. As phosphorylation of DNA-PKcs correlates with NHEJ activation, DNA-PKcs protein levels and its phosphorylation were measured by immunoblotting. Total DNA-PKcs levels did not change appreciably between glioma-initiating cells and NPC cultures (Fig. 3C). However, DNA-PKcs phosphorylation in U251 and U87 cultures was attenuated and consistent with the reduced levels of NHEJ in these cells.
When treated with DNA-PKcs inhibitor (DNA-PKi) to inhibit NHEJ activity before IR (Fig. 3D), U251 and L2b continued to show better survival (Fig. 3D). U251 at 24 hours showed no significant difference in viable cells between DNA-PKi treated (76%) and IR only (79%). Even at 96 hours, viable cells were similar (47% vs. 53%, respectively). Partial radiosensitivity occurred in L2b, with DNA-PKi treated (35%) and IR alone (46%) at 96 hours. In contrast, NPCs showed increased radiosensitivity. At 24 hours, DNA-PKi–treated NPCs showed 35% viability versus IR alone (57%). Results at 96 hours were similar: DNA-PKi treated (11%) versus IR alone (21%) showing that NPCs are more dependent on NHEJ than glioma-initiating cells.
Increased homologous recombination repair in glioma-initiating cells
Homologous recombination is the second major pathway involved in resolution of DSBs (30). Given the reduced levels of NHEJ for many glioma-initiating cells, we quantified homologous recombination using stably transfected pDR-GFP neurosphere lines containing a functional but interrupted GFP sequence with a single DSB site that is inducible only by I-Sce I expression and a downstream complementary nonfunctional GFP sequence that is required for recombination (Fig. 4A; ref. 31). After I-Sce I expression to create a single DSB, GFP-positive cells were 0.74% for NPCs but homologous recombination repair was approximately 4-fold higher in U251 (3.76%, P < 0.0002) and L2b (3.94%, P < 0.0005). L3b had a less efficient homologous recombination repair (1.18%, P < 0.009). Homologous recombination efficiency was confirmed by measuring Rad51 foci accumulation and disassembly (Fig. 4B and Supplementary Fig. S5; ref. 32). After IR, maximal numbers of Rad51 positive cells (53%–66%) were achieved by 6 hours for L2b and U251, and followed by rapid loss between 12 to 24 hours. In contrast, NPCs showed gradual increase with Rad51-positive cells (26%) at 12 hours and slow loss of foci (12%) at 36 hours, supporting slower homologous recombination activity. L3b was intermediate between NPCs, and U251 and L2b (Fig. 4B). Brca1, a second marker for homologous recombination was also analyzed (Supplementary Fig. S6A; refs. 33, 34). Results showed that appearance and loss of Brac1 foci was similar to Rad51, again supporting rapid homologous recombination in glioma-initiating cells (Supplementary Fig. S6B).
Following treatment with Rad51 siRNA (Fig. 4C), both U251 and L2b showed a drastic decrease in cell viability post-IR (Fig. 4D). At 24 hours, Rad51 siRNA-treated U251 showed approximately 21% viable cells versus control siRNA treated (76%). By 96 hours, Rad51 siRNA-treated U251 had very low cell viability (∼10%) versus control siRNA (55%). L2b behaved similarly; 24 hours, Rad51 siRNA L2b (26%) versus control siRNA (71%) and 96 hours, Rad51 siRNA L2b (9%) versus control siRNA (39%). These results contrasted with NPCs, in which Rad51 and control siRNA treated NPCs showed a similar percentage of viable cells (56% vs. 55%, respectively, at 24 hours and 15% vs. 23% at 96 hours). This clearly showed the greater importance of homologous recombination for the survival of glioma-initiating cells compared with NPCs.
DNA synthesis inhibition after IR is deficient in glioma-initiating cells
As glioma-initiating cells had a higher percentage of cells undergoing homologous recombination, we hypothesized glioma-initiating cells had unperturbed entry into S-phase postirradiation. To examine this, we used IdU pulse labeling to measure DNA synthesis (Fig. 5A). Three glioma-initiating cells were selected (U251, L2B, and L3b) because they showed different efficiencies of homologous recombination. NPCs, U251, L2b, and L3b had 25%, 30%, 29%, and 31% IdU-positive cells, respectively (Fig. 5B). Following IR, IdU-positive cells were reduced 2-fold for NPCs, but in glioma-initiating cells no reduction occurred (30%–32% IdU positive), indicating that cells continued to enter S-phase after IR (P < 0.01). This is supported by asynchronous cell-cycle analysis showing unperturbed entry into S-phase in glioma-initiating cells (P < 0.05) after 5Gy, indicating an attenuated S-phase checkpoint (Fig. 5C).
DNA initiations remain relatively unchanged in glioma-initiating cells after irradiation
We next examined capacity of glioma-initiating cells to initiate DNA synthesis following IR. Cells were pulse-labeled with CldU before irradiation to visualize ongoing DNA synthesis, followed by a pulse with IdU post-IR to detect new initiations (Fig. 6A). Because initiations differed between neurosphere lines, we normalized to unirradiated cells and compared fold change after DNA damage. In NPCs, new initiations reduced by 66% (Fig. 6B). Interestingly, U87, L1b, and L2b presented with relatively unaltered new initiation post-IR (P < 0.0003). Partial inhibition of new initiations occurred in U251 (42% reduction, P < 0.023) and L3b (32% reduction, P < 0.008). But, all glioma-initiating cells maintained a higher level of new DNA replication after IR compared with NPCs. Thus, glioma-initiating cells exhibit radioresistant DNA synthesis, indicating attenuation of the intra-S-phase checkpoint (35).
To investigate further, we examined ATM activation and phosphorylation of downstream substrates (SMC1, Chk1, and Chk2; ref. 36). ATM activity was attenuated in U87 with remaining glioma-initiating cells similar to NPCs (Fig. 6C). U87 also had reduced levels of SMC1, Chk1, and Chk2 proteins, and radiation-induced phosphorylation of these proteins was defective (Fig. 6C). U251 and L1b, which had normal levels of ATM and SMC1, showed reduced radiation-induced phosphorylation of these proteins, including Chk2. Activation of p53 was also measured. Basal p53 level was comparable with NPCs in glioma-initiating cells, except U87 and L1b in which levels were low. p53 phosphorylation was detected in NPCs, U251, L2b, and L3b with a stronger response in U251 (Fig. 6C). No phosphorylation was detected for U87 or L1b, which might be explained by low p53 protein.
Overall, our data suggested that defective checkpoint activation in glioma-initiating cells allows unperturbed S-phase entry resulting in preferential use of the homologous recombination pathway.
The mechanism for radioresistance in glioma has remained unclear because previous studies compared between subpopulations of aberrant glioma stem cells without using normal neuronal stem cells (8, 9). Human neural stem cells are an ideal control, given their normal repair profile and nontumor–forming nature (37). We used the recently developed immortalized human NPCs as a control in this study to allow for comparison between populations with “normal” (NPCs) and aberrant (glioma-initiating cells) radiosensitivity. The NPCs were similar to neural stem cells, as shown by their extensive self-renewal capacity and multipotency, allowing formation of different lineages. NPCs sustain self-renewal and proliferation, without rapidly developing chromosomal abnormalities (11). Given the capacity to compare with NPCs, it seemed relevant to focus on the overall glioma-initiating cells population, while incorporating NPC as a benchmark to determine the efficiency of DNA damage responses.
Cancer stem cells (CSC), expressing CD133 have been implicated as the radioresistant population responsible for tumor development (2, 7). CD133+ glioma cells showed increased survived post-IR versus CD133− tumor cells (7). Chk2 kinase inhibition reversed this radioresistance. However, recent data cast doubt on these early findings. Under hypoxic conditions, CD133− CSCs expressed CD133+ (38, 39). Also, in vivo studies showed that CD133− CSCs drive tumorigenesis in nude rats (24, 40). In our study, we showed Sox2 and Nestin expression in all cultures, but noted differences in CD49f expression between various glioma-initiating cells lines (41, 22). Despite low CD133 expression, all glioma-initiating cells induced tumor formation when injected intracranially into mice, suggesting CD133 expression does not correlate with tumorigenic potential. Further studies also showed similar radioresistant properties between CD133+/− populations following DNA damage (8, 9); measurement of γH2AX foci failed to distinguish significant differences between CD133+ and CD133− CSCs (24, 40). However, time points were limited to 1 hour (maximum foci) and 24 hours (background levels). These time points would not detect differences in the rate of DNA DSB repair, as seen here, where the rate was slower in glioma-initiating cells, though they were more radioresistant than NPCs (Fig. 1A and 1B). Supporting data from self-renewal and neurosphere size comparison also indicates glioma-initiating cells have greater recovery than NPCs after IR. Surprisingly, NHEJ was reduced 2-fold in U251 and U87, the more radioresistant neurosphere cultures. This was borne out in total repair of DNA DSBs, as determined by loss of γH2AX foci. In these experiments, the rate of DSB repair was reduced at initial time points and more breaks remained at later times. Combined data indicates that DNA DSB repair, at least by NHEJ, does not explain radioresistance in glioma-initiating cells, as shown by DNA-PKcs inhibition having a lesser effect on glioma-initiating cells than NPCs (Fig. 3D). In contrast to homologous recombination repair, as determined by pDR-GFP expression, formation and resolution of Rad51 and Brca1 foci was more efficient in the radioresistant cultures (U251 and L3b). Glioma-initiating cells treated with Rad51 siRNA showed increased radiosensitivity versus control siRNA-treated cells. Conversely, Rad51 knockdown had little effect on NPC radiosensitivity confirming the importance of homologous recombination particularly in glioma-initiating cells (Fig. 4D). Therefore increased dependence on homologous recombination coupled with defective cell-cycle checkpoints could explain the increased glioma-initiating cells radioresistance.
Cancer cells are characterized by abnormalities in cell-cycle control, including prolonged activation of the mitotic checkpoint, constitutive activation of the G1–S checkpoint by p53, and activation by the ATM-53BP1-Chk2 pathway. Constitutive activation can cause selective pressure in tumors and subsequent inactivation of DNA damage responses (42, 43). Human embryonal carcinoma cells were shown to be defective for the G1–S checkpoint after DNA damage but exhibited both S-phase and G2–M delay (44). However, when these cells were differentiated with retinoic acid, they failed to arrest in G1 and quickly exited S-phase arresting in G2–M. In our study, glioma-initiating cells (U251, L2b and L3b) continued to enter S-phase after IR at rates comparable with untreated cells, showing failure of the G1–S checkpoint. Glioma-initiating cells also exhibited radioresistant DNA synthesis. In a normal cell, DNA damage accumulating from cells passing through S-phase carrying DNA damage would increase cell death. In contrast, tumor cells circumvent cell-cycle checkpoints and continue undergoing cell divisions with a “DNA damage burden” without inducing cell death, thus resulting in increased survival. Interestingly, this phenomenon has previously been associated with human genetic disorders characterized by radiosensitivity and chromosomal instability and is a determinant of a defective intra-S-phase checkpoint (36). It seems likely that the G2 checkpoint is still intact in glioma-initiating cells. This contrasts with data for embryonic carcinomas, which displayed prominent S- and G2-phase checkpoints (44). Although enhanced basal activation of Chk1 and Chk2 was shown in CD133+ glioma cells (9), we saw no evidence of this in glioma-initiating cells. These results are in keeping with the cell-cycle data because we did not observe a delay at the G1–S checkpoint, as reported for CD133+ glioma cells (9). Bao and colleagues (7) postulated that cell-cycle delay might represent a general mechanism for genome protection in glioma-initiating cells. This pattern of radiation-induced signaling in glioma-initiating cells and its relationship to radioresistance is difficult to explain, as IR-induced activation of ATM kinase appears normal in all glioma-initiating cells except U87. Failure to observe normal radiation signaling through SMC1, Chk1, and Chk2 in U87 can be explained by reduced protein levels. However, reduced signaling through SMC1 in U251 and L1b is intriguing, given normal ATM activation and SMC1 protein levels. Phosphorylation of SMC1 is important for maintaining genome integrity and enhancing cell survival (45). The reduced reliance on SMC1 phosphorylation in glioma-initiating cells may be compensated for enhanced homologous recombination and cell-cycle checkpoint changes.
In summary, we showed that glioma-initiating cells are more radioresistant than NPC, but the extent of resistance is variable. On the basis of these results, we suggest that treatment of patients with drugs (i.e., ATM or ATR inhibitor) specifically targeting the homologous recombination pathway or blocking S-phase transition may be novel therapeutic approaches. This approach may be more effective on GBM than surrounding tissues, as homologous recombination is less used in NPCs.
Disclosure of Potential Conflicts of Interest
No potential conflicts of interest were disclosed.
Conception and design: Y.C. Lim, T.L. Roberts, B.W. Day, S. Kozlov, D.G. Walker, M.F. Lavin
Development of methodology: Y.C. Lim, T.L. Roberts, B.W. Day, A.W. Kijas, M.F. Lavin
Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): Y.C. Lim, B.W. Day, A. Harding, K.S. Ensbey, M.F. Lavin
Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): Y.C. Lim, T.L. Roberts, M.F. Lavin
Writing, review, and/or revision of the manuscript: Y.C. Lim, T.L. Roberts, S. Kozlov, M.F. Lavin
Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): K.S. Ensbey
Study supervision: T.L. Roberts, D.G. Walker, M.F. Lavin
The study was supported by grants from the Briz Brain and Spinal Foundation, and the National Health and Medical Research Council of Australia to M.F. Lavin.
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
The authors thank Professor Reynolds for the glioma-initiating cells lines, Dr. Stacey for CFSE dye, Dr. Richard for pDR-GFP construct, Paula Hall and Grace Chojnowski for assistance with flow cytometry, and Tracey Laing for assistance with typing the manuscript.
Note: Supplementary data for this article are available at Molecular Cancer Therapeutics Online (http://mct.aacrjournals.org/).
- Received December 22, 2011.
- Revision received May 29, 2012.
- Accepted June 29, 2012.
- ©2012 American Association for Cancer Research.