CANCER STEM CELLS. IGF1 Receptor Signaling Regulates Adaptive Radioprotection in Glioma Stem Cells INTRODUCTION

Transcription

1 CANCER STEM CELLS IGF1 Receptor Signaling Regulates Adaptive Radioprotection in Glioma Stem Cells SATORU OSUKA, a,b OLTEA SAMPETREAN, b,c TAKATSUNE SHIMIZU, b,d ISAKO SAGA, b,e NOBUYUKI ONISHI, b EIJI SUGIHARA, b,c JUN OKUBO, b,f SATOSHI FUJITA, b,g SHINGO TAKANO, h AKIRA MATSUMURA, h HIDEYUKI SAYA b,c a Department of Neurosurgery, Graduate School of Comprehensive Human Sciences and; h Department of Neurosurgery, Faculty of Medicine, University of Tsukuba, Tsukuba, Japan; b Division of Gene Regulation, Institute for Advanced Medical Research and; e Department of Neurosurgery, Keio University School of Medicine, Tokyo, Japan; c Core Research for Evolutional Science and Technology, Japan Science and Technology Agency, Tokyo, Japan; d Department of Pathophysiology, Faculty of Pharmaceutical Sciences, Hoshi University, Tokyo, Japan; f Department of Pediatrics, Graduate School of Medicine, University of Tokyo, Tokyo, Japan; g Department of Neurosurgery, Toho University, Ohashi Hospital, Tokyo, Japan Key Words. Insulin-like growth factor 1 Glioblastoma Forkhead box O Cancer stem cell Radioresistance ABSTRACT Cancer stem cells (CSCs) play an important role in disease recurrence after radiation treatment as a result of intrinsic properties such as high DNA repair capability and antioxidative capacity. It is unclear, however, how CSCs further adapt to escape the toxicity of the repeated irradiation regimens used in clinical practice. Here, we have exposed a population of murine glioma stem cells (GSCs) to fractionated radiation in order to investigate the associated adaptive changes, with the ultimate goal of identifying a targetable factor that regulates acquired radioresistance. We have shown that fractionated radiation induces an increase in IGF1 secretion and a gradual upregulation of the IGF type 1 receptor (IGF1R) in GSCs. Interestingly, IGF1R upregulation exerts a dual radioprotective effect. In the resting state, continuous IGF1 stimulation ultimately induces downregulation of Akt/ Disclosure of potential conflicts of interest is found at the end of this article. extracellular-signal-regulated kinases (ERK) and FoxO3a activation, which results in slower proliferation and enhanced self-renewal. In contrast, after acute radiation, the abundance of IGF1R and increased secretion of IGF1 promote a rapid shift from a latent state toward activation of Akt survival signaling, protecting GSCs from radiation toxicity. Treatment of tumors formed by the radioresistant GSCs with an IGF1R inhibitor resulted in a marked increase in radiosensitivity, suggesting that blockade of IGF1R signaling is an effective strategy to reverse radioresistance. Together, our results show that GSCs evade the damage of repeated radiation not only through innate properties but also through gradual inducement of resistance pathways and identify the dynamic regulation of GSCs by IGF1R signaling as a novel mechanism of adaptive radioprotection. STEM CELLS 2013;31: INTRODUCTION Despite recent advances in the planning and delivery of radiotherapy, resistance to ionizing radiation remains a major problem for the treatment of many types of cancer including melanoma, renal cell carcinoma, and malignant brain tumors [1]. Glioblastoma is the most lethal type of primary brain tumor, with affected individuals having a median survival time of <2 years. Its poor prognosis is due to cells that survive intensive treatment including surgery, radiotherapy, and chemotherapy [2]. Glioblastomas recur even in the cases that transiently exhibit complete radiological remission, suggesting that even when treatment-resistant cells are so scarce as to be undetectable by conventional imaging, they possess the ability to regenerate the original tumor and thereby give rise to disease recurrence [3]. These cells have thus been suggested to act as cancer stem cells (CSCs) [4, 5], with their stem-like characteristics such as relative quiescence, high self-renewal capacity, and protection by the niche underlying their radioresistance [6 10]. Indeed, specimens of malignant glioma from patients who had undergone radiotherapy were found to be enriched in CSCs [11]. Furthermore, a regimen of fractionated radiation, similar to that used in clinical practice, increased the size of the CSC population in preclinical mouse models of human cancer [6, 12, 13]. Comparisons of CSCs with nonstem cancer Author contributions: S.O.: concept and design, collection and assembly of data, data analysis and interpretation, and manuscript writing; O.S.: concept and design, collection and assembly of data, data analysis and interpretation, manuscript writing, and final approval of manuscript; T.S., N.O., E.S., S.T., and A.M.: provision of study material and data analysis and interpretation; I.S., J.O., and S.F.: collection and assembly of data; H.S.: concept and design, data analysis and interpretation, manuscript writing, and final approval of manuscript. Correspondence: Oltea Sampetrean, M.D., Ph.D., Division of Gene Regulation, Institute for Advanced Medical Research, Keio University School of Medicine, 35 Shinanomachi, Shinjuku-ku, Tokyo , Japan. Telephone: þ ; Fax: þ ; oltea@a6.keio.jp Received August 31, 2012; accepted for publication December 20, 2012; first published online in STEM CELLS EXPRESS January 17, VC AlphaMed Press /2013/$30.00/0 doi: /stem.1328 STEM CELLS 2013;31:

2 628 IGF1R Signaling Regulates Radioresistance in GSCs cells have implicated several signaling pathways in radioresistance, including the Notch [14], Wnt-b-catenin [15], and Hedgehog [16] pathways. However, in retrospective analyses, it is often not clear whether upregulation of such pathways is inherent to CSCs or whether it occurs by a gradual shift in phenotype that reflects adaptation to repeated radiation. Growth factor signaling is responsive to noxious stimuli such as oxidative stress and ionizing radiation [17]. Signaling by the type 1 receptor for IGF (IGF1R) is activated by chemotherapy, radiotherapy, or stress stimulation in several types of cancer [18 21]. Upregulation of IGF1R signaling has also been found to promote radioresistance in several types of solid tumor [22 24]. In bulk tumor cells, IGF1R-induced activation of PI3K-Akt signaling gives rise to radioresistance by preventing apoptosis and promoting cell survival [25, 26]. Furthermore, IGF1R signaling contributes to the maintenance of stem-like cells in T-cell lymphoma [27] and colorectal cancer [19], and it protects intestinal stem cells from radiation-induced apoptosis [28]. The relations among IGF1R upregulation, radioresistance, and CSCs have remained unclear, however. The purpose of this study was to prospectively analyze the changes induced in CSCs by fractionated radiation with the use of a mouse model of glioblastoma, with the ultimate goal of identifying a targetable factor required for the development of radioresistance in glioma stem cells (GSCs). We found that radiation increased the secretion of IGF1 and resulted in a gradual upregulation of its type 1 receptor (IGF1R) in GSCs. This adaptive change in IGF1R signaling had a dual effect: it mediated an enhancement in stem cell characteristics such as self-renewal and slow proliferation in the resting, nonstimulated GSCs and enhanced survival signaling after acute radiation, inducing a pronounced radioresistance in GSCs. Moreover, we found that inhibition of IGF1R is an effective way to reverse radioresistance in vitro and in vivo. Our findings highlight a novel mechanism by which GSCs adapt in order to escape the consequences of radiation exposure, and they suggest a new target for overcoming radioresistance in glioblastoma. MATERIALS AND METHODS Cell Culture Primary tumors were established by orthotopic implantation of Ink4a/Arf / neural stem and progenitor cells expressing green fluorescent protein (GFP)-tagged human HRasV12 into the brain of wild-type mice [29]. Cells from the primary tumors were isolated at 21 days after implantation, and GFP-positive cells were sorted with a flow cytometer and maintained as tumor spheres (TS) in neural stem cell medium (NSM), consisting of Dulbecco s modified Eagle s medium-f-12 (Sigma, St. Louis, MO, supplemented with recombinant human epidermal growth factor (20 ng/ml) (PeproTech, Rocky Hill, NJ, recombinant human basic fibroblastic growth factor (bfgf) (20 ng/ml) (PeproTech), B27 supplement without vitamin A (Invitrogen, Carlsbad, CA, heparan sulfate (200 ng/ml), penicillin (100 U/ml), and streptomycin (100 ng/ml) (Nacalai Tesque, Kyoto, Japan, In Vitro Fractionated Radiation TS were exposed to 5 Gy of ionizing radiation every 3 or 4 days with the use of an x-irradiator (Hitachi MBR-1520R-3; Hitachi Medical Corporation, Tokyo, Japan, co.jp) at settings of 150 kv and 20 ma. The culture medium was replaced every 3 4 days, and the cells were passaged as appropriate. The TS formed after 12 doses of radiation (total of 60 Gy) were designated TS-RR. Sphere Formation Assay Sphere formation was assayed by suspension of cells in 0.5% agarose in order to prevent aggregation. Low-melting point agarose (Sigma) was diluted with an equal volume of 2 NSM to a final concentration of 1% and allowed to solidify as the bottom layer in six-well plates. Tumor cells were harvested in NSM and plated at equal single-cell densities in a second layer of NSM containing 0.5% agarose. The cell layer was then covered with NSM and the medium was changed every 3 4 days. Two weeks after plating, cells were fixed with 4% paraformaldehyde and stained with toluidine blue O (Sigma). The wells were photographed, and spheres containing >100 cells were counted manually. Colony Formation Assay For evaluation of clonogenic survival, cells were embedded in agarose at densities adjusted to result in a number of spheres approximately equal to that formed by the control group and were allowed to settle for 24 hours before irradiation. For treated cells, mouse IGF1 (50 ng/ml, R&D Systems, Minneapolis, MN, was added 1 hour before irradiation, neutralizing antibody to IGF1 (4 lg/ml, Abcam, Cambridge, MA, was added 2 hours before irradiation, and IGF1R inhibitors picropodophyllin (0.4 lm, picropodophyllin [PPP], Santa Cruz, CA, and AEW541 (2.0 lm, Active Biochemicals, Wan Chai, HongKong, were added 1 hour before irradiation. Cells subjected to irradiation alone were returned to the incubator for 13 days with medium changes every 3 4 days. Cells treated with the soluble agents were incubated for 48 hours with the respective factor, after which the medium was replaced with fresh NSM and the cells were cultured for 11 days with medium changes. Colonies were stained and counted as described above. Establishment of TS and TS-RR Clones Cells derived either directly from TS-RR or from TS after exposure to a single dose (5 Gy) were plated at single-cell density in soft agar and cultured for 2 weeks in order to obtain spheres that could be recognized macroscopically. The spheres were then isolated and dissociated by exposure to trypsin, and the resulting cells were cultured in NSM. Cytokine Array Conditioned medium from TS or TS-RR that had been exposed (or not) to ionizing radiation (5 Gy 2) and then cultured for 72 hours with a medium change at 24 hours was collected for determination of the concentrations of various cytokines with the use of a Mouse Cytokine Antibody Array (AAM-CYT-4, RayBio, Norcross, GA, ELISA for IGF1 IGF1 levels in tumor lysates were measured with the use of a Mouse IGF-1 ELISA Kit (R&D Systems). Immunoblot Analysis Total protein was extracted from cells by their repeated passage through a 25-gauge needle in RIPA buffer (Sigma). Nuclear and cytoplasmic fractions were separated with the use of NE-PER Nuclear and Cytoplasmic Extraction Reagent (Thermo Scientific, Rockford, IL, The separated proteins were transferred to a membrane (Trans-Blot Turbo PVDF Transfer Pack, Bio-Rad, Richmond, CA, with the use of a Trans-Blot Turbo Transfer Starter System (Bio- Rad). Target proteins were detected by consecutive incubation of the membrane with the primary antibodies listed in Supporting Information section, HRP-conjugated secondary antibodies, and enhanced chemiluminescence reagents (Nacalai). The intensity of immunoreactive bands was measured with the use of Multi Gauge Software version 3.1 (Fujifilm, Japan,

3 Osuka, Sampetrean, Shimizu et al. 629 RNAi The retroviral expression vector preps was used to introduce shrna into cells, as previously described [30]. The sequences of the sense oligonucleotides were 5 0 -GTCCTTTCCCTAGCACAC TTA-3 0 for FoxO3a shrna #1, 5 0 -GGCAAGAGCTCTTGGT GGAT-3 0 for FoxO3a shrna #2, and 5 0 -CGTACGCGGAATA CTTCGA-3 0 for luciferase (nonspecific control) shrna. Cells were infected with the retroviral vector as previously described [31] and were then subjected to selection in the presence of puromycin (3 lg/ml). Long-Term IGF1 Treatment Cells were incubated in NSM supplemented with recombinant mouse IGF1 (50 ng/ml, R&D Systems) for up to 27 days. The culture medium was replaced every 3 days, and cells were passaged as appropriate. Animal Experiments Orthotopic implantation of cells was performed as previously described [29]. For radiation experiments, 1,000 viable TS or TS- RR cells were injected into the right forebrain of C57BL/6 mice. The animals were then subjected (or not) to daily whole-brain irradiation (2 Gy) from day 7 to 11 after cell implantation. Radiation was confined to the brain by protection of the rest of the body with a lead shield. The mice were observed daily and killed when moribund. For evaluation of the effect of the IGF1R inhibitor PPP (Santa Cruz Biotechnology, Santa Cruz, CA, TS or TS-RR cells ( ) were injected s.c. into the femur region of BALB/c nu/nu mice on day 0. PPP (20 mg/kg) was injected i.p. twice a day from day 3 to 7 in an aqueous solution containing 0.9% dimethyl sulfoxide (Sigma), 7% N-dimethylacetamide (Sigma), and 10% Cremophor EL (Sigma). Control mice were treated with vehicle only. Tumors were irradiated with a daily dose of 2 Gy from day 3 to 7. Tumor volume was determined by direct measurement with calipers and calculation according to the formula: p/6 (largest diameter) (smallest diameter) [2]. Statistical Analysis Data are presented as means 6 SD unless indicated otherwise. Survival curves were constructed by Kaplan-Meier analysis and were compared with the log-rank test. Other data were analyzed with the paired or unpaired Student s t test as performed with IBM SPSS Statistics 18 software (IBM, Chicago, IL, com). A p value of <.05 was considered statistically significant. Ethical Considerations All animal experiments were performed in accordance with the animal care guidelines of Keio University. Antibodies, Immunofluorescence Analysis, Flow Cytometry, Pyronin-Y Staining, Coculture Experiments, Cell Proliferation Assay, Cell Survival Assay, Phos-tag Immunoblot Analysis, and PKH26 Labeling Assays These methods are described in Supporting Information. RESULTS GSCs with a High Self-Renewal Capability and Low Proliferation Rate Are Enriched After Fractionated Radiation Prospective analysis of the adaptive signaling changes induced in GSCs by fractionated radiation requires an effective method for the identification of these cells as well as elimination of confounding factors such as genetic heterogeneity. To satisfy these conditions, we compared GSCs before and after repeated irradiation in a murine glioblastoma model that we previously established and which allows the purification of GFP-labeled GSCs with a homogeneous genetic background. As described previously [29], primary tumors were formed by orthotopic implantation of Ink4a- and Arf-null neural stem cells that had been transduced with a GFP-tagged form of the oncoprotein HRasV12. The histopathology of these tumors is highly similar to that of human glioblastoma. Tumors were removed, and tumor cells were purified by cell sorting on the basis of GFP fluorescence and then cultured under conditions that promote sphere formation in order to enrich for stem cells [32]. The resulting cell population, designated TS, is serially transplantable, with as few as 10 cells being able to reconstitute tumors in immunocompetent mice [29], suggesting that these cells have the potential to act as GSCs. We then subjected TS to a regimen of fractionated radiation consisting of exposure to a dose of 5 Gy every 3 4 days, with the cells having received 12 doses of radiation (total of 60 Gy) being designated TS-RR. Having established these two cell populations, we first examined whether fractionated radiation had changed the phenotype of TS by comparing TS and TS-RR. Given that fractionated radiation was previously shown to result in enrichment of cells with stem-like characteristics [6, 12], we examined TS and TS-RR for expression of stem cell markers, tumorigenic capacity, proliferation, and self-renewal and differentiation capabilities. Immunofluorescence staining revealed that both TS and TS-RR express markers of immature cells, such as nestin, Nanog, and Sox2, at a high level (Fig. 1A) and that they are both able to differentiate along astrocytic and neuronal, but not oligodendrocytic lineages after serum addition (Supporting Information Fig. S1A). Implantation of 10 viable cells of either type into the right forebrain of wild-type mice resulted in the death of 90% of the animals as a result of tumor formation (Fig. 1B), indicating that both TS and TS-RR are highly tumorigenic. These results confirmed that both TS and TS-RR are GSCs, with a common set of stem-like characteristics. Proliferation assays revealed, however, that the overall rate of cell proliferation was slower for TS-RR than for TS (Fig. 1C). Detailed analysis confirmed that TS-RR exhibit a longer doubling time (Supporting Information Fig. S1B), but no significant changes in cell cycle profile (Supporting Information Fig. S1C) or increase in quiescent cells (Supporting Information Fig. S1D). Furthermore, in sphere formation assays, the number of spheres formed by TS-RR cells was significantly greater than that formed by TS cells (Fig. 1D, 1E), indicating that the self-renewal ability of the former is greater than that of the latter. These results thus showed that fractionated radiation did indeed enrich for cells with enhanced stem-like characteristics. GSCs that Survive Fractionated Radiation Form Radioresistant Tumors We next compared the radiosensitivity of TS and TS-RR. Clonogenic assays showed that the fraction of cells surviving singledose radiation was significantly higher for TS-RR than for TS, regardless of the dose administered (Fig. 2A, 2B). Furthermore, staining of apoptotic cells with annexin V revealed that only 17.6% 6 0.4% of TS-RR cells underwent apoptosis during the 48 hours after single-dose irradiation, compared with a value of 26.0% 6 1.4% for TS cells (Fig. 2C, 2D). Consistent with this finding, the sub-g 1 cell population identified by flow cytometric analysis of DNA content was also markedly larger for TS than for TS-RR after ionizing irradiation (Supporting Information

4 630 IGF1R Signaling Regulates Radioresistance in GSCs Figure 1. Glioma stem cells with a high self-renewal capability and low proliferation rate are enriched after fractionated radiation. (A): Immunofluorescence staining of TS and TS-RR for nestin, Nanog, and Sox2 (red). Nuclei were counterstained with DAPI (blue). Scale bars ¼ 50 lm. (B): Survival curves for wild-type mice (n ¼ 10) after orthotopic implantation of 10 cells each for TS or TS-RR. (C): Cell proliferation analysis for TS and TS-RR as evaluated by WST-8 assays. Data are expressed relative to the value for TS (control) and are means 6 SD from three independent experiments. *, p <.05. (D): Self-renewal ability of TS and TS-RR as evaluated by sphere formation assay at the indicated cell densities. Scale bar ¼ 10 mm. (E): Quantification of sphere number in experiments similar to that in (D). Data represent the number of spheres formed per 1,000 cells seeded and are means 6 SD from three independent experiments. *, p <.05. Abbreviations: DAPI, 4 0,6-diamidino-2- phenylindole. Fig. S2A, S2B). Together, these results indicated that TS-RR are more resistant to ionizing radiation in vitro than are TS. For TS-RR to be considered a truly radioresistant cell population, however, they must exhibit similar properties in vivo. We therefore compared the response of tumors formed by the two types of cells to radiation. Mice orthotopically injected with 1,000 viable cells were subjected to whole-brain irradiation consisting of a daily dose of 2 Gy from day 7 to 11 after cell implantation. Radiotherapy conferred approximately 30% benefit in overall survival for mice injected with TS cells, whereas no such effect was apparent for those injected with TS-RR cells (Fig. 2E), confirming that TS-RR represent a clinically relevant radioresistant cell population. Highly Radioresistant Clones Are Not Present in the Initial GSC Population To investigate whether TS-RR represent the most radioresistant cells from the initial TS population or cells that have undergone an adaptive change and acquired radioresistance as the result of repeated irradiation, we established clones from both TS and TS-RR (Fig. 3A). We obtained 20 TS clones and 17 TS-RR clones by isolating spheres formed from single cells in soft agar. To eliminate cells with little or no radioresistance, we exposed TS to 5 Gy of ionizing radiation before clone isolation. We next analyzed the response of the individual clones to radiation. The fraction of cells that survived single-dose radiation was significantly higher for TS-RR and TS-RR clones than for TS and TS clones. Importantly, no TS clones were more radioresistant than TS-RR (Fig. 3B, 3C; Supporting Information Fig. S3). These data suggested that TS-RR may have acquired radioresistance de novo during repeated irradiation. Fractionated Irradiation Induces Secretion of IGF1 and Upregulation of IGF1R in GSCs One of the main features that distinguish an adaptive response to therapy from an intrinsic one relates to the influence of

5 Osuka, Sampetrean, Shimizu et al. 631 Figure 2. Glioma stem cells that survive fractionated radiation form radioresistant tumors. (A): Clonogenic survival assay for TS and TS-RR subjected (or not, control) to a single dose (6 Gy) of IR. Representative images of colonies formed by surviving cells 13 days after irradiation are shown. Scale bar ¼ 10 mm. (B): Quantification of colony number in experiments similar to that in (A). The fraction of surviving cells after radiation doses of 2, 4, or 6 Gy was determined as means 6 SD for three independent experiments. *, p <.05 versus the corresponding value for TS. (C): Flow cytometric analysis for TS and TS-RR cells stained with allophycocyanin-labeled annexin V and propidium iodide before (control, or Ct) or 48 hours after exposure to 5 Gy of ionizing radiation. (D): Quantification of apoptotic cells in experiments similar to that in (C). Data are means 6 SD from three independent experiments. *, p <.05. (E): Survival curves for mice implanted with 1,000 viable cells (TS or TS-RR) and subjected (or not) to daily exposure to 2 Gy of ionizing radiation from day 7 to 11 after cell implantation (n ¼ 5 for TS-RR without radiation, n ¼ 6 for TS without radiation and TS-RR with radiation, and n ¼ 7 for TS with radiation). Abbreviations: IR, ionizing radiation. external factors. In addition to changes in extracellular matrix and reactive cells, secretion of soluble factors by tumor cells has been shown to modulate response to therapy [33 36]. To address this issue, we performed coculture experiments without direct contact between cell types. These experiments revealed that the apoptotic cell fraction for TS after singledose irradiation was significantly reduced in the presence of TS-RR (Fig. 4A, 4B), suggesting that soluble factors secreted by TS-RR promote radioprotection. Screening for cytokines secreted by TS and TS-RR with the use of an antibody array resulted in the identification of six factors whose secretion by TS was increased after radiation: intercellular adhesion molecule 1, osteopontin, IGF1, insulin-like growth factor-binding protein 2 (IGFBP2), bfgf, and lungkine (Fig. 4C). Among these factors, secretion of IGF1 and IGFBP2 was increased not only in TS after radiation but also in TS-RR compared with TS, with or without radiation (Fig. 4D; Supporting Information Fig. S4A). Given that IGF1 has been implicated in CSC maintenance [19, 27] and tumor radioresistance [22, 24] and has been associated with poor prognosis in glioblastoma [37 39], we selected this growth factor for further analysis. ELISA analysis of tumor lysates showed that IGF1 levels were also higher in tumors formed by TS-RR than in those formed by TS and were increased after a course of radiation therapy in tumors formed by either cell type (Fig. 4E). To examine the relation between IGF1 increase and radiation in GSCs, the abundance of IGF1 downstream factors was evaluated in TS and TS-RR as well as in cells at intermediate stages of the transition between TS and TS-RR. We found that the amount of IGF1R increased gradually during fractionated

6 632 IGF1R Signaling Regulates Radioresistance in GSCs Figure 3. Highly radioresistant clones are not present in the initial glioma stem cell population. (A): Experimental protocol for establishment of TS and TS-RR clones. (B): Clonogenic survival assay for TS and TS-RR clones subjected (or not, control) to irradiation with a single dose of 5 Gy. Representative images of colonies formed 13 days after irradiation are shown. Scale bars ¼ 10 mm. (C): Quantification of colony number in experiments similar to that in (B). Data are expressed as the fraction of surviving cells relative to corresponding nonirradiated controls and are means 6 SD from three independent experiments. *, p <.05 for the mean value for all TS clones versus that for TS-RR. Abbreviations: IR, ionizing radiation. radiation (Fig. 4F), without any compensatory change in the expression level of the insulin receptor (Supporting Information Fig. S4B). Interestingly, the amounts of phosphorylated forms of Akt, extracellular-signal-regulated kinases (ERK), and S6 decreased after repeated radiation exposure, in the absence of an obvious change in the abundance of one of their main negative regulators, phosphatase and tensin homolog. In addition, transcription factor FoxO3a exhibited a gradual increase in abundance that paralleled that of IGF1R, whereas the amounts of FoxO1 and FoxO4 were unchanged or did not manifest a consistent trend (Supporting Information Fig. S4B). To confirm whether radiation can result in Akt signaling downregulation on a background of increased IGF1R levels, we examined the dynamics of the IGF1R and downstream factors after single-dose irradiation. Immunoblot analysis revealed that while IGF1R abundance gradually increased during the first 48 hours after radiation, levels of phosphorylated Akt, ERK, and S6 transiently increased during the first 24 hours, but subsequently decreased to levels lower than the initial ones (Fig. 4G). A similar response was observed after treatment of TS with IGF1 (Fig. 4H), confirming that Akt signaling can indeed be downregulated in the presence of increased amounts of IGF1R, probably by negative feedback,

7 Osuka, Sampetrean, Shimizu et al. 633 Figure 4. Fractionated radiation induces secretion of IGF1 and upregulation of IGF1R in glioma stem cells. (A): Flow cytometric analysis of annexin V staining in TS exposed to ionizing radiation (5 Gy) and cocultured with TS or TS-RR. (B): Quantification of the apoptotic cell population in experiments similar to that in (A). Data are means 6 SD from three independent experiments. *, p <.05. (C): Antibody array analysis of various cytokines in conditioned medium of TS or TS-RR before (control) and after irradiation (5 Gy 2). Sidebars mark cytokines with levels increased after irradiation. (D): Quantification of relative expression levels of IGF1 in (C). (E): IGF1 levels as determined with an ELISA in lysates of tumors formed by TS or TS-RR and subjected (or not, control) to irradiation as in Figure 2E. (F): Immunoblot analysis of IGF1R and downstream signaling molecules in TS exposed to IR (5 Gy) for the indicated number of times. b-actin was probed as a loading control. (G): Immunoblot analysis of IGF1R and downstream signaling molecules in TS at 12, 24, and 48 hours after single-dose irradiation (5 Gy) and (H): 6 and 12 hours after addition of 50 ng/ml of IGF1. Abbreviations: APC, allophycocyanin; IR, ionizing radiation; PI, propidium iodide. and that both radiation and exposure to IGF1 elicit such a response. Fractionated Radiation Induces FoxO3a-Mediated Self-Renewal and Slow Proliferation in GSCs FoxO transcription factors are regulated by signaling from insulin, IGF1, and various cytokines [40 43]. In the absence of phosphorylated Akt, members of the FoxO family are localized to the nucleus in an active form that can induce cell cycle arrest, resistance to oxidative stress, and stem cell maintenance through upregulation of key target genes such as those for p21 Cip1/Waf1,p27 Kip1,andcyclinD [43 45]. Given that TS-RR exhibited a slow proliferation rate and low level of Akt phosphorylation, we examined whether activation of FoxO3a might also be a characteristic of TS-RR. To determine the status of FoxO3a activation, we subjected cytoplasmic and nuclear fractions of TS

8 634 IGF1R Signaling Regulates Radioresistance in GSCs Figure 5. Fractionated radiation induces FoxO3a-mediated self-renewal and slow proliferation in glioma stem cells. (A): Immunoblot analysis of total and phosphorylated forms of FoxO3a in cytoplasmic and nuclear fractions of TS and TS-RR. MEK1/2 and lamin C were also examined as cytoplasmic and nuclear markers. (B): Immunoblot analysis of proteins encoded by FoxO3a target genes in TS and TS-RR. The intensity of the immunoreactive bands was also quantified as indicated. (C): Immunoblot analysis of FoxO3a in TS and TS-RR after infection with retroviral vectors for control or FoxO3a shrnas. (D): Quantification of apoptotic cells by annexin V staining after FoxO3a depletion by RNAi in TS and TS-RR. Data are means 6 SD from three independent experiments. *, p <.05. (E): Self-renewal ability of TS and TS-RR after FoxO3a depletion as evaluated with sphere formation assays. Data are means 6 SD from three independent experiments. *, p <.05. (F): Cell proliferation analysis for TS and TS-RR after FoxO3a depletion as evaluated by live cell counting. Data are expressed relative to the value for control shrna and are means 6 SD from three independent experiments. *, p <.05. (G): Clonogenic survival assay for FoxO3a-depleted TS and TS-RR after exposure to 4 Gy of IR. Representative images of colonies formed 13 days after irradiation are shown. Scale bars ¼ 10 mm. (H): Quantification of colony number in experiments similar to that in (G). Data are expressed as the fraction of surviving cells relative to corresponding nonirradiated controls and are means 6 SD from three independent experiments. *, p <.05. Abbreviations: IR, ionizing radiation. and TS-RR to immunoblot analysis with antibodies to total or Ser 253 -phosphorylated (inactive) forms of FoxO3a. Both the amount of FoxO3a in the nuclear fraction and the nuclear:cytoplasmic ratio of this protein were greater for TS-RR than for TS (Fig. 5A; Supporting Information Fig. S5A). The abundance of phosphorylated FoxO3a in the cytoplasm was also slightly increased for TS-RR compared with TS, possibly reflecting the increase in the overall

9 Osuka, Sampetrean, Shimizu et al. 635 Figure 6. IGF1 receptor signaling regulates adaptive radioprotection in glioma stem cells. (A): Immunoblot analysis of cytoplasmic [C] and nuclear [N] fractions of TS and TS-RR 1, 3, and 6 hours after 5 Gy of ionizing radiation. Internal controls: MEK1/2 cytoplasmic fraction, lamin c nuclear fraction. (B): Immunoblot analysis of IGF1 downstream factors in TS, TS-RR 12 and 24 hours after 5 Gy of ionizing radiation. (C, D): Immunoblot analysis of IGF1 downstream factors in TS and TS-RR treated with or without 5 Gy of ionizing radiation and neutralizing antibody for IGF1. (E): Colony formation for TS and TS-RR after exposure to a single dose (4 Gy) of ionizing radiation and addition of IGF1 (50 ng/ml) to growth factor-depleted medium. Data are expressed as the fraction of surviving cells for TS or TS-RR relative to the corresponding nonirradiated control values and are means 6 SD from three independent experiments. *, p <.05. Representative images of colonies are shown in Supporting Information Figure S6C. (F): Self-renewal ability of TS-RR and TS after long-term treatment with IGF1 (50 ng/ml), as evaluated with sphere formation assays. Data are expressed as the ratio of sphere formation for nontreated TS and are means 6 SD from three independent experiments. *, p <.05. Representative images of colonies are shown in Supporting Information Figure S6D. (G): Clonogenic survival assay for IGF1-treated TS and TS-RR after exposure to 5 Gy of ionizing radiation. Data are expressed as the fraction of surviving colonies relative to corresponding nonirradiated controls and are means 6 SD from three independent experiments. *, p <.05. Representative images of colonies are shown in Supporting Information Figure S6E. (H): Immunoblot analysis of IGF1R and downstream signaling molecules in TS-RR and TS before after 27 days of IGF1 treatment. amount of FoxO3a in TS-RR (Fig. 4F). Consistent with the nuclear localization of FoxO3a in TS-RR, the expression of its target genes Gadd45a, p27, and p21 was increased at the protein level (Fig. 5B). To confirm the biological role for FoxO3a activation in TS-RR, we performed RNA interference experiments in TS and TS-RR with two shrna constructs (shfoxo3a#1 and #2), resulting in depletion of FoxO3a by 65% with shfoxo3a#1

10 636 IGF1R Signaling Regulates Radioresistance in GSCs and by 75% with shfoxo3a#2 (Fig. 5C). Evaluation of cell death by annexin V staining showed that FoxO3a depletion increased the level of apoptosis in both TS and TS-RR without irradiation (Fig. 5D). Moreover, the percentage of apoptotic cells was inversely proportional to the level of FoxO3a. Knockdown of FoxO3a also resulted in a marked attenuation of sphere formation by TS and especially TS-RR (Fig. 5E), indicative of reduced self-renewal capacity. Furthermore, in both types of cells, FoxO3a depletion also resulted in increased proliferation (Fig. 5F; Supporting Information Fig. S5B) without a significant change in cell cycle profile (Supporting Information Fig. S5C, S5D). These results show that FoxO3a plays an important role in stem cell maintenance in the resting state in both types of cells and confirms that FoxO3a activation as a result of fractionated radiation is as a key factor in the enhanced stem cell characteristics of TS-RR. Since FoxO3a regulates transcription of genes involved cell cycle control and DNA damage response, we next asked whether FoxO3a activation was also responsible for the radioresistance of TS-RR. Clonogenic survival assays revealed that the fraction of cells surviving single-dose irradiation was significantly reduced by FoxO3a depletion, but the effect was not TS-RR specific (Fig. 5G, 5H). IGF1 Receptor Signaling Regulates Radioprotection in GSCs Having established that FoxO3a is a key factor of the TS-RR phenotype in the nonstimulated state, but that it does not directly regulate the response to stimulation with single-dose radiation, we assessed the molecular changes occurring immediately after radiation. TS-RR exhibited a decrease in nuclear, active FoxO3a, but a rapid increase in phosphorylated Akt (Fig. 6A), suggesting that Akt activation takes prevalence as modulating factor in the acute response of TS-RR to radiation. To confirm this hypothesis, Akt and downstream factors were analyzed in more detail. Immunoblot analysis revealed that levels of phosphorylated Akt/S6 and ERK increased in TS-RR more than in TS during the first 24 hours after single-dose radiation (Fig. 6B). Furthermore, the increase in phosphorylated Akt was also paralleled by an increase in Bcl-2, suggesting a stronger induction of antiapoptotic effectors in TS-RR. This increase in phosphorylated Akt/ERK/S6 and Bcl-2 in TS- RR after acute radiation could be prevented by treatment with an IGF1 neutralizing antibody (Supporting Information Fig. S6A, S6B; Fig. 6C, 6D), confirming that the changes seen after acute radiation are also IGF1-IGF1R dependent. To further ascertain the role of IGF1 in the TS-RR phenotype, we analyzed TS cells after addition of exogenous IGF1. A treatment of one dose of 50 ng/ml significantly increased cell survival of TS after single-dose radiation (Fig. 6E; Supporting information Fig. S6C). Furthermore, prolonged exposure to IGF1 increased the self-renewal ability of TS cells in a time-dependent manner (Fig. 6F; Supporting Information Fig. S6D). Radioresistance of IGF1-treated TS cells also increased after 21 days of treatment (Fig. 6G; Supporting Information Fig. S6E). At molecular level, IGF1 exposure resulted in an increase in the amount of IGF1R and a decrease of phosphorylated forms of Akt, ERK, and S6 (Fig. 6H). These results indicated that prolonged exposure to IGF1 induces functional and molecular changes similar to the ones induced by repeated radiation, confirming that IGF1/IGF1R signaling is an important regulator of radioresistance in GSCs. Inhibition of IGF1R Signaling Attenuates Radioresistance in GSCs We next asked whether IGF1R signaling inhibition could attenuate the radioresistance of GSCs. Treatment with the IGF1 neutralizing antibody markedly reduced radioresistance (Fig. 7A), with the surviving fraction of TS-RR after irradiation being reduced to the level of that for TS (Fig. 7B). The effect of IGF1R inhibition on radioresistance of TS-RR was verified by the use of two receptor inhibitors, PPP and AEW541. At a concentration of 0.4 lm, PPP reduces cell proliferation by half (Supporting Information Fig. S7A), drastically inhibits phosphorylation of IGF1R (Supporting Information Fig. S7B) and activation of Akt, ERK, and S6 (Supporting Information Fig. S7C), and has no immediate effect on FoxO3a expression or localization (Supporting Information Fig. S7D). PPP significantly reduced radioresistance of TS-RR, as seen in colony formation assays (Fig. 7C; Supporting Information Fig. S7E). The same effect was confirmed for a second IGF1R inhibitor, AEW541 (Supporting Information Fig. S7F, S7G). Finally, we investigated how IGF1R inhibition affects the response to radiation in tumor-bearing mice. PPP, administered intraperitoneally, effectively reduced levels of phosphorylated IGF1R in subcutaneous tumors (Supporting Information Fig. S7H). For tumors formed by TS, both radiation therapy and PPP inhibited tumor growth, with the combination of both treatments showing the greatest effect (Fig. 7D, 7E). For tumors formed by TS-RR, radiation alone, as expected, had a less substantial effect on tumor growth, whereas PPP markedly enhanced the inhibitory action of radiation (Fig. 7D, 7E). Together, these results confirmed that inhibition of IGF1R signaling reduces the radioresistance of GSCs in vivo. DISCUSSION In this study, we analyzed the changes induced in GSCs by fractionated radiation. We found that repeated radiation first caused an increase in IGF1 secretion and a gradual upregulation of IGF1R expression in GSCs. This adaptive change in IGF1R signaling ultimately increased the self-renewal capability and slowed the proliferation rate of GSCs, which subsequently formed highly radioresistant tumors. Furthermore, IGF1R upregulation also potentiated the induction of the canonical Akt signaling pathway after acute radiation, conferring further radioprotection to GSCs. Long-term exposure to IGF1 mimicked the effects of repeated radiation, while targeting of IGF1R resulted in the reversal of radioresistance. GSCs have previously been shown to be more radioresistant than nonstem glioma cells as a result of intrinsic properties such as an increased DNA repair capability and antioxidative capacity. Targeting of such properties has shown promising results with regard to potential therapeutic application [7, 46]. However, these approaches have been only palliative, suggesting the existence of further aspects of GSC radioresistance that have remained elusive. Preclinical studies are often limited by the paucity of effective GSC markers, and clinical studies have been mostly retrospective. Furthermore, recent data indicating that glioblastoma stem cells are not of a single type [47, 48] raise the possibility that not all GSCs are equally radioresistant and further complicate the targeting of individual intrinsic properties. Our approach was to start with an established and characterized population of GSCs and to focus on the changes they incur during fractionated radiation and emergence of the most radioresistant population. Targeting of the pathway responsible for adaptive changes during radiation exposure might be expected to prevent such changes and thereby to block the escape of the cells from the therapeutic consequences of IR. The use of GSCs with a uniform genetic background averts complications such as those arising from differences in

11 Osuka, Sampetrean, Shimizu et al. 637 Figure 7. Inhibition of IGF1R signaling attenuates radioresistance in GSCs. (A): Colony formation for TS and TS-RR after exposure to a single dose (4 Gy) of ionizing radiation together with either neutralizing antibodies to IGF1 or control IgG. Scale bar ¼ 10 mm. (B): Quantification of colony number in experiments similar to that in (A). Data are expressed as the fraction of surviving cells for TS or TS-RR relative to the corresponding control value and are means 6 SD from three independent experiments. *, p <.05. (C): Colony formation for TS and TS-RR after exposure to a single dose (4 Gy) of ionizing radiation together with either PPP or DMSO (control). Data are expressed as the fraction of surviving cells for TS or TS-RR relative to the corresponding control value and are means 6 SD from three independent experiments. *, p <.05. Representative images of colonies are shown in Supporting Information Figure S7E. (D): Subcutaneous tumors formed by TS or TS-RR in mice treated with vehicle (control), ionizing radiation alone, the IGF1R inhibitor PPP alone, or the combination of radiation and PPP. The animals were photographed on day 19 after cell implantation. The tumor margin is indicated by a red dotted line. (E): Quantification of tumor volumes in mice treated as in (D). Data represent means 6 SD (n ¼ 5). Statistical significance and symbols are as follows: *Significant (p <.05), PPP versus PPP þ IR, day 17 to day 21, **Significant (p <.05), IR versus PPP þ IR, day 10 to day 21, ***Significant (p <.001), Control versus IR, day 10 to day 21, n.s.: not significant, Control versus IR, day 7 to day 19, # Significant (p <.05), PPP versus PPP þ IR, day 17 to day 21, ## Significant (p <.001), IR versus IR þ PPP, day 7 to day 21. (F): Proposed model of the mechanisms by which fractionated radiation induces radioresistance in GSCs. Repeated radiation increases IGF1 secretion, induces a gradual increase in IGF1R expression in GSCs (TS), and ultimately a decrease in phosphorylated Akt in the GSCs surviving the regimen of fractionated radiation (TS-RR). The downregulation of Akt signaling enables activation of FoxO3a and a gradual switch toward a slower proliferation rate and enhanced self-renewal through FoxO3 target genes. (G): After acute radiation, the increased secretion of IGF and the relative abundance of IGF1R in TS-RR compared to TS allow a stronger induction of the canonical Akt survival signaling, leading to radioresistance. Abbreviations: GSCs, glioma stem cells; IR, ionizing radiation; PPP, picropodophyllin.

12 638 IGF1R Signaling Regulates Radioresistance in GSCs oncogenes or tumor suppressor genes among the initial cell population and thus allows a focused analysis of signal plasticity. The possibility that repeated radiation only selects the most radioresistant GSCs existing in the initial population does have to be considered, however. Until technical advances allow analysis of radioresistance at the single-cell level for all the cells in a tumor, it will be difficult to exclude this possibility definitively. To address this issue, we analyzed 20 clones each for primary and irradiated GSCs, finding that no single clone derived from the initial population matched the level of resistance of those isolated after radiation. Furthermore, our observations that factors secreted by the irradiated population enhanced the radioresistance of the initial population within 48 hours and that long-term exposure to IGF1 induces a radioresistant phenotype in GSCs are also indicative of an adaptive change in signaling rather than of clonal selection. Secretion of factors by tumor cells has been shown to enhance acquired resistance to therapy in pancreatic cancer in an autocrine manner [49]. We have now shown that radiation induces a change in the amounts of cytokines secreted by GSCs, with six out of 34 cytokines tested showing increased levels. On the basis of the consistency of its radiation-secretion correlation as well as its clinical relevance in stem cell biology and glioblastoma, we chose IGF1 for further investigation and found it to be a crucial factor in the radiation-induced adaptation of GSCs. Importantly, the increase in the secretion of IGF1 was apparent not only in vitro but also in tumors formed by the irradiated GSCs in vivo, indicating that this characteristic is maintained even within the microenvironment of the brain and showing its robustness and clinical relevance. Interestingly, it has been reported that radiation-induced secretion of IGF1 in human fibroblasts upregulates IGF1R/Akt signaling in bystander cells [50]. It is plausible that, in vivo, continuous exposure to IGF1 induces upregulation of IGF1R expression not only in GSCs but also in nonstem cancer cells or even tumor-associated fibroblasts and endothelial cells, which would further increase the therapeutic benefits of targeting IGF1R signaling. Overexpression of soluble IGF1 and its receptor IGF1R has been detected in several types of cancer, including prostate cancer, melanoma, and glioblastoma [37 39, 51 53]. IGF1R signaling has also been found to mediate resistance to chemotherapy and to radiation [18 20, 22, 54]. The mechanisms described to date for such resistance to acute radiation include a joint contribution of the two major pathways that originate from IGF1R, the PI3K-Akt pathway and the Ras- Raf-MEK-ERK pathway [54]. Our results now show that, in the case of fractionated radiation, adaptive changes in IGF1R signaling induce radioresistance in GSCs by two mechanisms, as summarized in Figure 7F, 7G. First, repeated radiation induces a continuous exposure of the GSCs to IGF1 and a gradual increase in IGF1R expression, which is ultimately accompanied by a decrease in phosphorylated Akt and ERK in the emerging GSCs. The downregulation of Akt signaling might reflect the operation of a negative feedback loop induced during repeated radiation or a change in interaction of IGF1 with its binding proteins. In turn, Akt downregulation enables activation of FoxO3a and a gradual switch toward a slower proliferation rate and enhanced self-renewal through FoxO3 target genes. Overall, these adaptive changes result in an enhancement of stem-like characteristics and confer protection to GSCs in their resting state (Fig. 7F). The second effect of the changes in IGF1R signaling is protection after acute radiation (Fig. 7G). Our results show that although single-dose radiation activates Akt in both primary and radioresistant GSCs, the extent of the activation is more pronounced in the latter. The relative abundance of IGF1R, the increased secretion of IGF, and the sudden activation from a suppressed state probably all contribute to induce a stronger Akt activation, along with downstream survival signaling. This dual radioprotective effect of IGF1R signaling in GSCs has critical implications in the outcome of fractionated radiation. Inhibition of IGF1R has been found to sensitize human brain, breast, prostate, colon cancer, osteosarcoma, and lung cancer cell lines to radiation [55 59]. We have now further shown that it also attenuates acquired radioresistance in GSCs, which suggests that IGF1R blockade is an important strategy in the prevention of recurrence after radiotherapy. Moreover, we found that treatment with an IGF1R inhibitor, a class of drug already shown to be clinically safe [60, 61], resulted in marked enhancement of radiosensitivity in tumors formed by the most radioresistant GSCs, suggesting a new target for overcoming therapeutic resistance induced during fractionated radiation in glioblastoma. CONCLUSION This study shows that GSCs escape the toxicity of repeated radiation not only through innate properties but also through gradual upregulation of resistance pathways. IGF1R signaling plays a pivotal role in enhancing GSCs radioresistance by maintaining them in a state of suppressed proliferation, which helps avoid the damage of subsequent radiation insults, and by rapidly inducing survival signaling immediately after radiation. Elucidation of stem-cell-specific adaptive radioprotection mechanisms and identification of targetable key factors are crucial to refinement of radiosensitizing strategies and prevention of tumor relapse. ACKNOWLEDGMENTS We thank K. Arai for secretarial assistance. Satoru Osuka is a research fellow of the Japan Society for the Promotion of Science. This work was supported in part by a Research Fellowship from the Japan Society for the Promotion of Science for Young Scientists (to S.O.) and by a grant from the Ministry of Education, Culture, Sports, Science, and Technology of Japan (to H.S.). DISCLOSURE OF POTENTIAL CONFLICTS OF INTEREST The authors declare no potential conflicts of interest. REFERENCES 1 Malik B, Nie D. Cancer stem cells and resistance to chemo and radio therapy. Front Biosci (Elite Ed) 2012;4: Stupp R, Mason WP, van den Bent MJ et al. Radiotherapy plus concomitant and adjuvant temozolomide for glioblastoma. N Engl J Med 2005;352: Chen J, Li Y, Yu TS et al. A restricted cell population propagates glioblastoma growth after chemotherapy. Nature 2012;488: