Chronically Radiation-Exposed Survivor Glioblastoma Cells Display Poor Response to Chk1 Inhibition under Hypoxia

Glioblastoma is the most malignant primary brain tumor, and a cornerstone in its treatment is radiotherapy. However, tumor cells surviving after irradiation indicates treatment failure; therefore, better understanding of the mechanisms regulating radiotherapy response is of utmost importance. In this study, we generated clinically relevant irradiation-exposed models by applying fractionated radiotherapy over a long time and selecting irradiation-survivor (IR-Surv) glioblastoma cells. We examined the transcriptomic alterations, cell cycle and growth rate changes and responses to secondary radiotherapy and DNA damage response (DDR) modulators. Accordingly, IR-Surv cells exhibited slower growth and partly retained their ability to resist secondary irradiation. Concomitantly, IR-Surv cells upregulated the expression of DDR-related genes, such as CHK1, ATM, ATR, and MGMT, and had better DNA repair capacity. IR-Surv cells displayed downregulation of hypoxic signature and lower induction of hypoxia target genes, compared to naïve glioblastoma cells. Moreover, Chk1 inhibition alone or in combination with irradiation significantly reduced cell viability in both naïve and IR-Surv cells. However, IR-Surv cells’ response to Chk1 inhibition markedly decreased under hypoxic conditions. Taken together, we demonstrate the utility of combining DDR inhibitors and irradiation as a successful approach for both naïve and IR-Surv glioblastoma cells as long as cells are refrained from hypoxic conditions.


Introduction
Glioblastoma remains a significant health problem with being an incurable malignant brain tumor in adults [1]. The standard of treatment for patients diagnosed with glioblastoma has long entailed tumor resection, followed by chemotherapy and radiotherapy as described in the landmark European Organization for Research and Treatment of Cancer (EORTC) Brain Tumor and Radiotherapy Group and the National Cancer Institute of Canada [2,3]. Recent genomic and molecular studies have shown that glioblastoma is the most heterogeneous disease among all cancer types and is composed of several cell populations with multiple genotypic origins [4]. Despite the advances in our understanding of glioblastoma genetics, cell-of-origin, or tumor heterogeneity, the survival rates have remained unchanged during the last decade.
Ionizing radiation (IR)-based radiotherapy is a gold therapeutic cornerstone for glioblastoma patients. It is applied as a fractionated clinical regimen, by administering patients 2 Gy of IR over 5 days/week reaching a total of 60 Gy at the end of 6 weeks. However, despite IR and concomitantly applied chemotherapy with Temozolomide, a DNA alkylating agent, tumor recurrence is observed in majority of the patients. One of the mechanisms behind therapeutic failure is considered to be inherent or acquired therapy resistance of glioblastoma cells. The tumor cells that are radioresistant cannot be efficiently eradicated after a full dose of IR treatment suggesting that tumor cells develop adaptations to the applied therapies by undergoing genetic or epigenetic changes [5,6]. Repopulation by IR-exposed surviving glioblastoma cells during fractionated IR is among the main reasons for radiotherapy-resistant tumor recurrence [7]. To overcome this problem, several approaches for radiosensitization have been investigated, yet none of them has translated into the clinic to improve the radiosensitivity in glioblastoma patients so far [8]. Although different glioblastoma cell lines have been examined in this context, where they were exposed to short-term and low doses of IR, the behavior of glioblastoma cells after long-term and high-dose radiation (total 60 Gy) remains largely unknown [9][10][11][12][13]. Most pre-clinical studies that interrogated the low-dose IR response of glioblastoma cell lines have shown that the radiosensitization effect is achieved by mainly targeting DNA damage repair pathways, tumor microenvironment, and cancer stem cells, but there are conflicting results with respect to obtaining radiation-persistent models in those studies [14,15]. Currently, there are no effective therapies to target long-term-IR-exposed survivor (IR-Surv) cells.
Radiotherapy induces damage to the genetic material of the cell and affects numerous vital cellular mechanisms that may trigger radioresistance with persistent or irreparable DNA damage, activated DNA damage response (DDR), irreversible cell cycle arrest, and oncogene activation, besides several unknown reasons [16][17][18]. Furthermore, metabolic changes occur in response to IR treatment, by stimulating oxidative stress and hypoxic mechanisms. Hypoxia Inducible Factor 1 (HIF-1) stabilization or activation by IR triggers protective processes by regulating downstream target genes that can induce immunosuppressive and antiapoptotic responses [19]. Several studies reported the improvement of radiosensitivity by blocking DDR and hypoxia pathways [20,21]. Although targeting such pathways for glioblastoma therapy has shown promise in animal models, none has so far worked in clinical practice and improved patient survival [22,23]. Therefore, a better understanding of IR response in clinically relevant experimental cell models are needed to mimic radiobiological characteristics of tumors after standard clinically applied therapeutic regimens. Moreover, growing evidence suggests the host immunity and inflammation as two conditions impacting glioblastoma progression, which clinically stratifies patients into two significant outcome groups following the same radiochemotherapy protocols, pointing out the importance of tumor stroma and microenvironment in addition to tumor characteristics [24][25][26][27][28]. As radiotherapy targets not only the tumor but also the adjacent healthy brain tissue, the inflammation and hypoxic changes in stroma and its relationship with glioblastoma require further experimental modeling to resolve related clinical discrepancy.
In this study, we established human IR-Surv glioblastoma cell models in vitro by exposing cells to 40-60 Gy of fractionated radiotherapy. Using established and patientderived cell lines, we selected radiation survivor cells and characterized the phenotypic and transcriptomic alterations in these cells. We demonstrated that DDR and hypoxia pathways have undergone major adaptations in IR-Surv cells in favor of improved DNA repair capacity. We showed that targeting these pathways using chemical inhibitors or switching oxygen conditions along with IR may serve as key therapeutic approaches for radiosensitization of IR-Surv cells and may be applied in the clinic to target recurrent tumors.

Generation of Radiation Survivor (IR-Surv) Glioblastoma Cell Populations
To mimic the standardized radiotherapy protocol used in the clinic and generate clinically relevant irradiation-exposed cell populations, we used three established (U373, T98G, LN229) and one primary (KUGBM8) cell line and irradiated them to a total dose of 40-60 Gy, fractionated by 2 Gy five times a week. Parental cell lines were also passaged with irradiated samples to establish age-matched controls ( Figure 1A). LN229 and T98G cells could survive until a total dose of 30 Gy and KUGBM8 cells until 40 Gy. Cell populations that survived long-term IR exposure were named as IR Survivor (IR-Surv) cells. Since U373 60 Gy and KUGBM8 40 Gy IR-Surv cells persisted longer under high exposure to IR compared to other cell lines, we mainly focused on the characterization of these and their parental pairs. To investigate whether selected IR-Surv cells maintain their persistent phenotype with secondary irradiation, we tested the viability of U373 60 Gy and KUGBM8 40 Gy cells in response to varying amounts of single-dose irradiation with clonogenic assays, which can be considered as a gold standard to assess the long-term effects of chemoradiation studies [30]. To this end, cells were exposed to 2, 4, 6, and 8 Gy of a single dose of irradiation, and colony-forming ability was measured after 14 days (Figure 2A). Accordingly, U373 60 Gy cells exhibited less response to radiation treatment and better colony-forming abilities than their parental pairs after 4, 6, and 8 Gy treatments ( Figure 2B). In contrast, we did not The morphological analysis revealed that irradiation caused a significant increase in cell size in U373 60 Gy cells without affecting nucleus size. Specifically, average cell size was 1693 ± 89 µm 2 in U373 cells and 2734 ± 473 µm 2 in U373 60 Gy cells. Nucleus sizes were comparable in U373 and U373 60 Gy cells, as 149 ± 8 and 165 ± 12 µm 2 , respectively ( Figure 1B,C). Compared to their parental cells, both nucleus and cell size decreased in KUGBM8 40 Gy cells. While cell sizes were 6171 ± 1009 and 4239 ± 912 µm 2 , nucleus sizes were 502 ± 63 and 389 ± 73 µm 2 in KUGBM8 and KUGBM8 40 Gy cells, respectively ( Figure 1D). The percentage of multinucleated cells in the population increased in U373 60 Gy and KUGBM8 40 Gy compared to their parental pairs ( Figure 1E), consistent with the previous reports on HepG2 cells [29]. We have not observed any significant alteration in the morphology of IR-exposed T98G and LN229 cells (Supplementary Figure S1C). Long-term IR exposure also affected proliferation rates; the proliferation rate of U373 60 Gy cells was slower than U373 ( Figure 1F). However, we observed a slightly increased proliferation rate in KUGBM8 40 Gy cells (Supplementary Figure S1A). In addition, there were significant differences in the cell cycle distribution of IR-Surv cells and their parental pairs. The percent number of U373 60 Gy cells in the G1 phase of the cell cycle was higher than its parental pair ( Figure 1G), but we did not observe a significant change in the cell cycle distribution of KUGBM8 cells (Supplementary Figure S1B).
To investigate whether selected IR-Surv cells maintain their persistent phenotype with secondary irradiation, we tested the viability of U373 60 Gy and KUGBM8 40 Gy cells in response to varying amounts of single-dose irradiation with clonogenic assays, which can be considered as a gold standard to assess the long-term effects of chemoradiation studies [30]. To this end, cells were exposed to 2, 4, 6, and 8 Gy of a single dose of irradiation, and colony-forming ability was measured after 14 days ( Figure 2A). Accordingly, U373 60 Gy cells exhibited less response to radiation treatment and better colony-forming abilities than their parental pairs after 4, 6, and 8 Gy treatments ( Figure 2B). In contrast, we did not observe any significant difference in colony-forming abilities of KUGBM8 40 Gy cells ( Figure 2B, Supplementary Figure S1D). We also examined whether there was any crossresistance of IR-Surv cells to Temozolomide (TMZ), the clinically applied chemotherapeutic for glioblastoma [2]. We treated naïve and IR-Surv cells with increasing doses of TMZ and examined different responses of IR-Surv cells. Accordingly, U373 60 Gy cells had a higher IC50 value of TMZ than its parental pair and displayed a TMZ-resistant behavior (IC 50 U373 = 18.81 µM, IC 50 U373 60 Gy = 80.75 µM) ( Figure 2D). However, KUGBM8 40 Gy cells displayed a better response to TMZ than their parental pair. For further elucidation of secondary therapy response, we combined single-dose 4 Gy IR with high-dose (125 µM) TMZ. Similar to previous findings, U373 60 Gy cells had a higher tolerance to TMZ + IR combination, whereas KUGBM8 40 Gy cells were more sensitive ( Figure 2E). There was no observed difference in the TMZ response of IR-Surv T98G or LN1229 cells (Supplementary Figure S1E). Together, we generated clinically relevant cell line models of IR-surviving cells, one of which was derived from a well-known established cell line and the other one from a primary cell line. Despite their few differences, both IR-Surv cell lines displayed refractory behavior to secondary irradiation, mimicking the radioresistance observed in clinical settings.
(IC50U373 = 18.81 μM, IC50U373 60Gy = 80.75 μM) ( Figure 2D). However, KUGBM8 40 Gy cells displayed a better response to TMZ than their parental pair. For further elucidation of secondary therapy response, we combined single-dose 4 Gy IR with high-dose (125 μM) TMZ. Similar to previous findings, U373 60 Gy cells had a higher tolerance to TMZ + IR combination, whereas KUGBM8 40 Gy cells were more sensitive ( Figure 2E). There was no observed difference in the TMZ response of IR-Surv T98G or LN1229 cells (Supplementary Figure S1E). Together, we generated clinically relevant cell line models of IR-surviving cells, one of which was derived from a well-known established cell line and the other one from a primary cell line. Despite their few differences, both IR-Surv cell lines displayed refractory behavior to secondary irradiation, mimicking the radioresistance observed in clinical settings.   To examine the differences in gene networks and pathways between parental and IR-Surv populations, we performed Gene Set Enrichment Analysis (GSEA) over 22,000 identified pathways from different datasets. Pathways such as DNA Repair and doublestranded break repair were upregulated in U373 and KUGBM8 IR-Surv populations. This is not surprising, as surviving long-term exposure to ionizing radiation partly depends on adaptive mechanisms of DNA damage response and repair ( Figure 3B) [31]. Focusing on two of the activated pathways, GOBP Regulation of DNA Repair and Reactome DNA DSB Repair, we observed that the majority of the genes were upregulated in IR-Surv cells To examine the differences in gene networks and pathways between parental and IR-Surv populations, we performed Gene Set Enrichment Analysis (GSEA) over 22,000 identified pathways from different datasets. Pathways such as DNA Repair and double-stranded break repair were upregulated in U373 and KUGBM8 IR-Surv populations. This is not surprising, as surviving long-term exposure to ionizing radiation partly depends on adaptive mechanisms of DNA damage response and repair ( Figure 3B) [31]. Focusing on two of the activated pathways, GOBP Regulation of DNA Repair and Reactome DNA DSB Repair, we observed that the majority of the genes were upregulated in IR-Surv cells ( Figure 3C,D, Supplementary Table S1), suggesting IR-Surv cells rewire DNA damage recognition, response, and repair pathways to adapt to extreme IR exposure, in accordance with previous reports [32]. We then focused on the common DEGs in the IR-Surv populations of the two independent cell lines. Notably, there were 101 upregulated and 318 downregulated common genes ( Figure 3E). In the GSEA analysis, the common upregulated gene sets were not highly significant (with the FDR < 0.05 cut-off); however, the common downregulated gene sets identified Hypoxia-related pathways very significantly ( Figure 3F,G, Supplementary Figure S2D-G). When the significantly common downregulated genes were analyzed individually in Enrichr platform using MSigDB 2021 [33][34][35], "Hypoxia" was again the most significantly altered gene signature (Supplementary Figure S3). In both U373 and KUGBM8 IR-Surv populations, hypoxia-related pathways were the only commonly downregulated pathways, suggesting that cells adopt a less hypoxic signature upon IR exposure as a survival mechanism.
Besides commonalities, cell line-specific DEGs were in abundance in IR-Surv cells. Accordingly, there were 442 upregulated and 485 downregulated unique genes in U373 60 Gy cells, and there were 2040 upregulated and 2835 downregulated unique genes in KUGBM8 40 Gy cells ( Figure 3E), hinting to cell line-specific adaptive mechanisms for IR survival. When GSEA analyses were conducted to identify cell line-specific changes, by excluding commonly altered pathways, we observed that the majorly altered gene sets in U373 60 Gy cells belonged to "interferon response genes", "STAT3 targets", and infection-related pathways (Supplementary Figure S4A). In KUGBM8 cells, the number of significantly altered pathways were higher (with FDR < 0.05 cut-off). Upregulated pathways included cell cycle-related gene sets, such as "Hallmark_G2M_Chekpoint", "Hallmark_E2F_Targets", and "YO_Myc_Targets_Up". In contrast, most downregulated gene sets were extracellular matrix (ECM)-related, such as "Reactome_Collagen_Degradation", "GOCC_Collagen_ containing_ECM", and "GOBP_ECM_Disassembly" among others (Supplementary Figure S4B). Together, these results from transcriptomic analysis suggest that surviving through IR might lead to several different adaptive gene expression changes specific to cell types involved and provide insights into the mechanisms for IR survival and tumor recurrence.

IR-Surv Cells Have Higher DNA Repair Capacity
The overarching goal of radiotherapy is to generate DNA damage, causing genomic instability and leading to the death of tumor cells. Indeed, one major mechanism of survival from radiotherapy is through alteration of DNA damage response and repair [13]. To this end, we examined the generation and repair of DNA double-stranded breaks induced by ionizing radiation by staining for (γH2AX) and Tumor suppressor p53-binding protein 1 (53BP1). After a single dose of 4 Gy irradiation of U373 and U373 60 Gy , presence and clearance γH2AX or 53BP1 levels were examined at 1 h and 6 h ( Figure 4A). We observed that 4 Gy IR exposure increased 53BP1-positive foci in both U373 and U373 60 Gy cells. The foci number decreased to approximately 40% in U373 cells and to around 25% in U373 60 Gy at 6 h ( Figure 4B), indicating a different level of regulation of DBS repair by 53BP1 in IR-Surv cells. As an indicator of DNA DSB burden of cells, 1 h after IR treatment, both U373 and U373 60 Gy cells had elevated γH2AX-positive foci at comparable levels ( Figure 4C). Basal levels of γH2AX foci were higher in U373 60 Gy , plausibly due to the prolonged IR exposure from which the cells survived despite DNA damage (Supplementary Figure S4C). After 6 h, foci number did not change in U373 cells, but the γH2AX number significantly decreased in U373 60 Gy cells ( Figure 4C), suggesting that IR-Surv populations had altered DNA DSB recognition and repair machinery and faster DSB break repair [36]. Furthermore, gene expression levels of several DNA damage response elements, such as ATM, ATR, CHK1, Rad51, and genes associated with Mismatch repair (MMR) were upregulated in U373 60 Gy IR-Surv cells ( Figure 4D). This gene expression signature was not observed in KUGBM8 or other glioblastoma cell lines that were utilized to generate clinically relevant IR-Surv models (Supplementary Figure S4D). Expression of O6-Methylguanosine methyltransferase-MGMT, an important prognostic marker for glioblastoma, was upregulated both at gene and protein levels in U373 60 Gy cells, but not in KUGBM8 40 Gy cells ( Figure 4D,E). The protein levels of activated (phosphorylated) forms of Chk1 and Chk2, γH2AX, and Rad51 were all upregulated U373 60 Gy cells. However, activated or basal Atm and Atr kinase expression levels were lower in U373 60 Gy cells. Among MMR proteins, upregulation of Msh3 and Msh6 were observed in U373 60 Gy cells ( Figure 4E). While some changes in protein levels were consistent with RNA-seq and qRT-PCR results, such as the expression of MGMT, H2AX, Rad51, MSH3, and MSH6, some changes were not directly correlated. However, together with the increased phosphorylated (active) protein levels of Atm, Atr, Chk1, and Chk2, our results highlighted an overall activated DDR state in IR-Surv cells. Methylguanosine methyltransferase-MGMT, an important prognostic marker for glioblastoma, was upregulated both at gene and protein levels in U373 60 Gy cells, but not in KUGBM8 40 Gy cells ( Figure 4D,E). The protein levels of activated (phosphorylated) forms of Chk1 and Chk2, γH2AX, and Rad51 were all upregulated U373 60 Gy cells. However, activated or basal Atm and Atr kinase expression levels were lower in U373 60 Gy cells. Among MMR proteins, upregulation of Msh3 and Msh6 were observed in U373 60 Gy cells ( Figure  4E). While some changes in protein levels were consistent with RNA-seq and qRT-PCR results, such as the expression of MGMT, H2AX, Rad51, MSH3, and MSH6, some changes were not directly correlated. However, together with the increased phosphorylated (active) protein levels of Atm, Atr, Chk1, and Chk2, our results highlighted an overall activated DDR state in IR-Surv cells. DNA damage response and repair pathways are among the most targeted pathways for therapeutic purposes in cancer. Inhibition of central regulators of DDR, such as Atm, Atr, Chk1, and Chk2 is considered a prime therapeutic approach in chemo-or radiosensitization studies [37][38][39]. Based on our observations with IR-Surv cells, which activated DDR to adapt to long-term IR, we interrogated whether their inhibition would sensitize DNA damage response and repair pathways are among the most targeted pathways for therapeutic purposes in cancer. Inhibition of central regulators of DDR, such as Atm, Atr, Chk1, and Chk2 is considered a prime therapeutic approach in chemo-or radiosensitization studies [37][38][39]. Based on our observations with IR-Surv cells, which activated DDR to adapt to long-term IR, we interrogated whether their inhibition would sensitize IR-Surv cells to irradiation. We selected five DDR-related kinase inhibitors (DDRi) (AZD7762, AZD6738, KU55933, BML-277, and LY2603618) targeting Atm, Atr, Chk1, or Chk2 ( Figure 5A). U373 and U373 60 Gy both responded to DDRi in a dose-dependent manner ( Figure 5A,B). Further, 1 µM DDRi radiosensitized both U373 and U373 60 Gy cells, but the degrees of sensitization were different between them, when examined with short-term (7 day) viability assays ( Figure 5C). In addition, we tested the effect of 1 µM DDRi and IR combination treatment on a long-term (14-day long) clonogenic assay. Accordingly, U373 60 Gy cells were slightly less responsive to KU55933 (Atm inhibitor) individual treatment, consistent with our initial findings. Combination treatment of DDRi and single-dose 4 Gy IR were very effective on both U373 and U373 60 Gy cells ( Figure 3D,E). These results suggest IR-Surv cells with increased DNA damage response activity can be sensitized to IR treatment using DDR inhibitors.  Figure 5A). U373 and U373 60Gy both responded to DDRi in a dose-dependent manner ( Figure 5A,B). Further, 1 μM DDRi radiosensitized both U373 and U373 60 Gy cells, but the degrees of sensitization were different between them, when examined with short-term (7 day) viability assays ( Figure 5C). In addition, we tested the effect of 1 μM DDRi and IR combination treatment on a long-term (14-day long) clonogenic assay. Accordingly, U373 60 Gy cells were slightly less responsive to KU55933 (Atm inhibitor) individual treatment, consistent with our initial findings. Combination treatment of DDRi and single-dose 4 Gy IR were very effective on both U373 and U373 60 Gy cells ( Figure 3D,E). These results suggest IR-Surv cells with increased DNA damage response activity can be sensitized to IR treatment using DDR inhibitors.

IR-Surv Cells Have Lower Hypoxic State and Exhibit Poor Response to DDR Inhibition under Further Hypoxia
As hypoxia was identified as the top downregulated pathway from our transcriptomic analysis, we investigated the behavior of IR-Surv cells by exposing them to hypoxic conditions. For this, we examined three conditions, control (normoxia), acute hypoxia (applied for 1 day) and chronic hypoxia (applied for 4 days) ( Figure 6A). As a mimic for irradiation, we used DSB-causing drug Bleomycin [40] and investigated the activation of H2AX under normoxic and hypoxic conditions upon 2 h of Bleomycin treatment. Accordingly, under normoxic conditions, Bleomycin increased γH2AX activation in both U373 and U373 60 Gy cells significantly. The γH2AX activation levels were similar under hypoxia in U373 cells. However, U373 60 Gy cells exhibited slightly less γH2AX activation under hypoxic conditions, suggesting a different mode of adaptation to DNA damage and repair ( Figure 6B). These adaptations have possibly affected cell cycle progression. Chronic hypoxia did not alter cell cycle distribution of U373 60 Gy , only affecting U373 cells through G2/M arrest ( Figure 6C). Downregulated response to hypoxia was also observed at gene expression level. Upon hypoxia, the extent of vascular endothelial growth factor (VEGF) upregulation, which can be considered a biomarker for hypoxia [41], was not the same in IR-Surv cells compared to parental cells. Specifically, upon hypoxia treatment, the VEGF induction was more than 20-fold in U373 parental cells, but it was less than 10-fold in U373 60 Gy IR-Surv cells. In addition, CHK1 upregulation in U373 60 Gy cells was not observed to the same extent as in parental U373 cells ( Figure 6D). To test whether the cells' response to DDRi would be altered under hypoxic conditions, we treated U373 and U373 60 Gy with low-dose (200 nM) Chk1 inhibitors, AZD7762 and LY2603618. To highlight their effects on cell viability, we applied 200 nM of both AZD7762 and LY603618 under normoxic and hypoxic conditions to both U373 and U373 60 Gy cells. While 4 days of Chk1i treatment was very effective on U373 cells, it became more effective under hypoxic conditions. However, similar cell viability was observed for U373 60 Gy cells under both normoxic and hypoxic conditions ( Figure 6E). To test whether this phenotype is temporary or exclusive to short-term hypoxia, we performed a 14-day clonogenic assay in hypoxic conditions ( Figure 6F). After Chk1i treatment, cells were incubated in a hypoxic incubator for 14 days. Clonogenic assay results revealed that U373 60 Gy cells were far less sensitive to Chk1i-hypoxia combination treatment than U373 cells ( Figure 6G). Thus, IR-Surv cells exhibit resistance to Chk1 inhibition, and the sensitization can be achieved through IR combination, but not to sufficient degrees in hypoxic conditions. (G) Representative clonogenic assay images and quantification of U373 and U373 60 Gy cells upon treatment of CHK1 inhibitor AZD7762 and LY2603618 in normoxia and hypoxia, for 14 days. (ns denote not significant, for p-values, *, ** and **** denote p < 0.05, p < 0.01, p < 0.0001, respectively, two-way ANOVA).

Discussion
As high proliferation and infiltration capacity and the ability to adapt and develop resistance to therapies are significant hallmarks of the glioblastomas [42], the main treatment options of surgery, chemotherapy, and radiotherapy are not enough for cure. Despite the refined RT regimens and TMZ administration, recurrence occurs mostly in central high dose radiotherapy field within 90% of patients due to intrinsic or acquired therapy resistance of tumor cells [5,[43][44][45]. Therefore, understanding the molecular (G) Representative clonogenic assay images and quantification of U373 and U373 60 Gy cells upon treatment of CHK1 inhibitor AZD7762 and LY2603618 in normoxia and hypoxia, for 14 days. (ns denote not significant, for p-values, *, ** and **** denote p < 0.05, p < 0.01, p < 0.0001, respectively, two-way ANOVA).

Discussion
As high proliferation and infiltration capacity and the ability to adapt and develop resistance to therapies are significant hallmarks of the glioblastomas [42], the main treatment options of surgery, chemotherapy, and radiotherapy are not enough for cure. Despite the refined RT regimens and TMZ administration, recurrence occurs mostly in central high dose radiotherapy field within 90% of patients due to intrinsic or acquired therapy resistance of tumor cells [5,[43][44][45]. Therefore, understanding the molecular mechanisms behind this adaptive persistence is of utmost priority to design effective therapeutic strategies. In this study, we employed a clinically relevant radiotherapy regimen to investigate the phenotypic alterations of surviving glioblastoma cells to recapitulate the early stages of recurrence and demonstrated that IR-Surv cells had increased DNA damage repair capacity and reduced response to hypoxia and documented for the first time in literature the downregulation of hypoxic signature as well as the lower induction of hypoxia target genes, through functional assays and transcriptomic analysis.
By utilizing DNA damage response-related kinases, we showed that IR-Surv cells have a slight resistance to DNA damage response-related kinase inhibition, but these cells can be sufficiently eradicated by using inhibitors combined with single-dose IR exposure. Furthermore, we showed that both parental and IR-Surv cells reclaim resistance to Chk1 inhibition in hypoxic conditions. Together, our results suggest that IR-Surv cells may become more resistant to DDR inhibition in long-term hypoxic conditions (Figure 7), which provides insight into the future design of effective combinatorial radiotherapy strategies. mechanisms behind this adaptive persistence is of utmost priority to design effective therapeutic strategies. In this study, we employed a clinically relevant radiotherapy regimen to investigate the phenotypic alterations of surviving glioblastoma cells to recapitulate the early stages of recurrence and demonstrated that IR-Surv cells had increased DNA damage repair capacity and reduced response to hypoxia and documented for the first time in literature the downregulation of hypoxic signature as well as the lower induction of hypoxia target genes, through functional assays and transcriptomic analysis. By utilizing DNA damage response-related kinases, we showed that IR-Surv cells have a slight resistance to DNA damage response-related kinase inhibition, but these cells can be sufficiently eradicated by using inhibitors combined with single-dose IR exposure. Furthermore, we showed that both parental and IR-Surv cells reclaim resistance to Chk1 inhibition in hypoxic conditions. Together, our results suggest that IR-Surv cells may become more resistant to DDR inhibition in long-term hypoxic conditions (Figure 7), which provides insight into the future design of effective combinatorial radiotherapy strategies. The radiation response and mechanisms of radioresistance are extensively studied in different cancer types, both primary and recurrent models. The model we exploit mimics the radiotherapy schedules applied to patients, which was previously used in few studies [46,47]. In the generation of radiation-survivor cell lines, we assured to remain faithful to the "Clinically Relevant Radioresistant (CRR) cell line" notion [29]. CRR concept in the generation of radioresistant cancer model is based on conventional fractional RT, exposure to 2 Gy IR once a day for more than a month. The cells were exposed to 2 Gy IR/day for five days; the fraction between two treatments was two days. In those fractioned days, cells were passaged along with parental cells to ensure cells were not under stress besides radiation therapy. This way, we succeeded in generating age-matched parental/irradiation-survivor (IR-Surv) cell line pairs. To examine the survival capacities of glioblastoma cell lines with different genetic landscapes, we utilized one primary cell line generated in our laboratory and three established cell lines [48]. However, the endurance levels of cell lines to total IR treatment were variable. The highest dose exposure was achieved by U373 cells (60 Gy), followed by KUGBM8 (40 Gy) and T98G and LN229 cells (30 Gy). The diversity of endurance of the cells may depend on various factors yet to be discovered. From the analysis of common mutations in glioblastoma, PTEN and p53 status could be among The radiation response and mechanisms of radioresistance are extensively studied in different cancer types, both primary and recurrent models. The model we exploit mimics the radiotherapy schedules applied to patients, which was previously used in few studies [46,47]. In the generation of radiation-survivor cell lines, we assured to remain faithful to the "Clinically Relevant Radioresistant (CRR) cell line" notion [29]. CRR concept in the generation of radioresistant cancer model is based on conventional fractional RT, exposure to 2 Gy IR once a day for more than a month. The cells were exposed to 2 Gy IR/day for five days; the fraction between two treatments was two days. In those fractioned days, cells were passaged along with parental cells to ensure cells were not under stress besides radiation therapy. This way, we succeeded in generating age-matched parental/irradiation-survivor (IR-Surv) cell line pairs. To examine the survival capacities of glioblastoma cell lines with different genetic landscapes, we utilized one primary cell line generated in our laboratory and three established cell lines [48]. However, the endurance levels of cell lines to total IR treatment were variable. The highest dose exposure was achieved by U373 cells (60 Gy), followed by KUGBM8 (40 Gy) and T98G and LN229 cells (30 Gy). The diversity of endurance of the cells may depend on various factors yet to be discovered. From the analysis of common mutations in glioblastoma, PTEN and p53 status could be among indicators of radiation persistence [49]. In the case of PTEN, where KUGBM8 and LN229 are wild-type, T98G cells have missense mutation. In contrast, U373 have null mutation of PTEN, correlating our observed radiation-survivor phenotype with different studies suggesting PTEN loss or mutation leading to radioresistance [50]. Except for KUGBM8 cells, all the cells used in the study were p53 mutant. Like most glioblastoma tumors and established cell lines, radiation response depends not solely on one gene or a pathway but on various components of tumor progression [48,50].
As a representative model, we chose one established cell line and one primary cell line to investigate irradiation-induced alterations. Although U373 60 Gy and KUGBM8 40 Gy share common characteristics such as increased ratio of multinucleated cells and retained colony formation abilities after secondary IR exposure, they exhibit different phenotypes and therapy responses. We have not observed cell cycle arrest at the G1 phase in KUGBM8 40 Gy cells, which is reported as an indicator for radioresistance [38]. Cells may have undergone senescence after irradiation as defense mechanism, in consistency with different reports [51]. In addition, response to TMZ and adjuvant TMZ + IR differed between U373 60 Gy and KUGBM8 40 Gy cells, very likely due to differential MGMT expression levels of these IR-Surv cells. Indeed, according to our transcriptomic analysis, MGMT expression was upregulated in U373 60 Gy cells (LFC = 5.26) but downregulated in KUGBM8 40 Gy (LFC = −1.78). In consistent with high MGMT expression being related to poor response to TMZ [52], IR-Surv cells that upregulated MGMT displayed collateral resistance to TMZ. This suggests that some tumors may respond to TMZ when they are in a naïve state prior to IR treatment; however, prolonged IR treatment may cause the survivor tumor cells to exhibit crossresistance to TMZ, making the chance of recurrence higher.
Our RNA sequencing analyses highlight several differentially expressed pathways in IR-Surv cells. While some of these pathways were shared between two cell lines, some were cell line-specific. For example, in U373 IR-Surv cells, interferon responsive gene networks were altered. However, in KUGBM8 IR-Surv cells, major effects were observed in ECM-related pathways, suggesting possible explanations for the phenotypic differences in U373-derived and KUGBM8-derived IR-Surv cells. It will be of future interest to decipher the relation between these pathways and IR response in our paired cell line models. Despite U373 60 Gy and KUGBM8 40 Gy exhibiting differences in functional assays and gene networks, they also shared few commonly altered pathways. As the overarching goal of radiotherapy is to generate DNA damage directly or indirectly [31,53], our results stand in parallel that IR-Surv cells show notable alterations in DNA damage response and repair pathways. With increased genomic instability, IR-Surv cells alter DNA repair machinery to survive and adapt to constant exposure to IR via increased expression of DNA damage response and repair-related proteins. This adaptation partly relies on increased DNA repair capacity. Radiation-exposed cells also survive with a higher burden of genomic instability, which is not elucidated in our study but prompts future work [54,55]. Our observations suggest that U373 60 Gy cells have a higher burden of DNA damage, yet upon further IR-exposure, they have higher DNA damage clearance and repair capacity than their naïve pair. We utilized a panel of inhibitors targeting DNA damage response-related kinases for radiosensitization. These DDRi inhibitors were studied in different cancer types in various conditions [39,56,57]. Our study showed that individual DDRi treatments were not sufficient for eradication of IR-Surv cells, suggesting that increased DNA repair capacity compensates for inhibition of these kinases.
Ataxia-Telangiectasia Mutated (Atm) kinase is activated explicitly upon DSB generation, which activates downstream effector kinases Checkpoint Kinase 1 (Chk1) and Checkpoint Kinase 2 (Chk2) [58]. The inhibitors of Atm, Chk1, and Chk2 kinases are extensively studied, combined with chemotherapeutic drugs and radiotherapy [20,59,60], even though multiple clinical trials are currently proceeding. As involved in regulating DDR and cell cycle checkpoints, Chk1 is among the ideal targets for radiosensitization [61,62]. Our study is consistent with radiosensitization studies in the context of Chk1 function, as radiation-mediated Chk1 activation can be exploited for radiosensitization of IR-Surv cells.
Indeed, Chk1 activation, as gauged by its gene expression and phosphorylated protein levels in IR-Surv cells, was among the most prominent changes observed in our IR-Surv cells. Therefore, it may be plausible to design therapies with Chk1 inhibitors for recurrent tumors that have high levels of activated Chk1. All in all, there is a great need for the development of new therapeutic strategies with new-generation DDR inhibitors and radiotherapy with lower cytotoxicity and improved efficacy. While several DDR inhibitors offer great promise for future clinical applications, some have failed in early stages of clinical trials. For example, a phase I trial of Chk1 inhibitor AZD7762 with irinotecan in glioblastoma was stopped because of toxicity reports (NCT00473616). To overcome the hurdles of toxicity due to high-dose drug administration, recently, new suggestions have been made en route to design effective DDRi-based combination therapies. For example, the sequence of drug administration has been suggested as an important factor to consider, as naïve tumor cells do not contain any DNA damage and do not become dependent on checkpoint kinases for their survival [37,42]. However, in our study, we show that IR-Surv tumor cells that are derived from naïve cells indeed become very adaptive to increased DNA damage and show dependency on several DDR molecules. Therefore, using inhibitors of DDR, specifically in our case, Chk1i, along with radiotherapy, might offer success in the clinic as the tumors have already completed an adaptation. This would allow for using lower doses of inhibitors and reducing the toxicity associated with drug administration. Our study might serve as a treatment model combining a low dose of DDRi and radiotherapy in both naïve and IR-Surv cell populations.
Therapy resistance of the tumors does not solely depend on the genetic and epigenetic landscape of tumor cells but is highly influenced by the tumor microenvironment [63]. Hypoxia is one of the highly studied concepts in the field of radiobiology, associated with all six R's (Repair, Redistribution, Repopulation, Reoxygenation, Reactivation of immune response, and Radiosensitivity) [64]. Indeed, the tumor core, which is highly hypoxic, is known to be more radioresistant [65,66]. Decreased oxygenation of tumor cells makes radiotherapy ineffective as ionizing radiation fails to generate reactive oxygen species that would lead to DNA damage [67]. The opposite scenario is still not elucidated. In this study, we showed that both IR-Surv cell lines exhibit downregulation of hypoxic gene signature through transcriptomic analysis and by the lower induction of hypoxia target genes, such as VEGF. We showed that cells that escape from chronic radiation did so by several adaptive changes, one of which resulted in a lower hypoxia response. This might seem in contrast with the knowledge in radiobiology at first sight. However, this finding, which is based on two different IR-Surv models, may suggest that tumor cells can find a way to counteract the high pressure exerted by chronic IR and become more vulnerable at the end. IR-Surv cells' response to DDR inhibition, particularly to Chk1 inhibition, was also markedly less under hypoxia in IR-Surv cells [68], suggesting that further applications of DDRi need to take into account the hypoxic nature of tumors for best clinical translation.
Taken together, we generated useful, clinically relevant radiation survivor models that exhibit several major adaptive mechanisms. We showed that efficacy of radiotherapy not only depends on hypoxic conditions but also irradiation-escaped cells that may display resistance to hypoxia. In addition, targeting DDR kinases such as Chk1 is effective on irradiation-escaped cells but the efficacy would decrease under hypoxia. These results could provide insight into designing effective treatment strategies for recurred tumors from radiotherapy.

Generation of Radiation Exposed Cell Lines
Cells were irradiated with 6MV X-Ray at a dose rate of 600 MU/min in Varan iX model linear accelerator, located at the Radiation Oncology Department of Koç University Hospital. To mimic clinically relevant standardized radiotherapy, cells were exposed to 2 Gy IR every day for 4-6 weeks. U373 cell lines were exposed to 60 Gy ionizing radiation (IR) for 6 weeks. Irradiation of T98G, LN229 cell lines were concluded at 30 Gy and KUGBM8 cell irradiation was completed after 4 weeks (40 Gy).

Cell Size and Nuclei Size Analysis
All nuclei and cell size measurements were analyzed in ImageJ. Surface area of the cell size and nuclei were measured one cell at a time and plotted according to the scale. Specifically, to measure area, freehand tool was used to outline the area-of-interest.

Clonogenic Assay
All parental and irradiated cells were seeded as 750 cells/well to 6-well plates as triplicates and exposed to single doses of ionizing radiation of 2, 4, 6, or 8 Gy for each plate and incubated for 14 days. Wells were washed with 1 ×x DPBS twice, and colonies were fixed with the ice-cold methanol treatment for 5 min. After fixation, wells were washed with 1 × DPBS twice and incubated with crystal violet for 15 min. Crystal violet was removed, and plates were washed. After the plates were dried, plates were scanned, and colony densities for each well were quantified with Adobe Photoshop CC 2019 (USA).

Cell Viability Assay
Cells were seeded as 1000 cells/well, treated with the corresponding drug on day 1 and/or exposed to 4 Gy IR treatment on day 3. On day 5, MTT solution (3 mg/mL) was added as 25 µL per well and incubated for 4 h at 37 • C. After incubation, culture medium was aspirated, 100 µL DMSO was added and dissolved. Plate reading was performed in Synergy H1 Reader (BioTek, Winooski, VT, USA) at 570 nm wavelength. Survival was described as a percentage of viable cells of each sample compared with DMSO control groups.

Cell Cycle Assay
Cells were harvested from 6-well plates, and pellets were washed with ice-cold PBS. For fixation, pellets were resuspended with cold ethanol (70%) and incubated for 30 min at 4 • C. After fixation, samples were centrifuged and washed with PBS twice, resuspended in 50 µL PBS containing RNase A (100 µL /mL) and incubated at room temperature for 15 min. Propidium Iodide (PI) (50 µL/mL) was added, and samples were incubated at room temperature for 30 min. Stained samples were analyzed by BD Accuri C6 (BD Biosciences, USA) flow cytometer and 10,000 events were recorded for each sample. For cell cycle analysis in hypoxic cells, The Muse ® Cell Cycle Kit (MCH100106) was used according to the manufacturer's protocol.

H2AX Activation Assay
Cells were harvested from 6-well plates following the corresponding treatment, and H2AX activation was quantified with The Muse ® H2AX Activation Dual Detection Kit (MCH200101) according to the manufacturer's protocol.

Quantitative Real-Time PCR (qRT-PCR)
To determine respective mRNA expression, parental and irradiated cell pellets were collected, and RNA isolation was performed with NucleoSpin. RNA Isolation Kit according to manufacturer's instructions (Macherey-Nagel, Düren, Germany). RNA concentrations were measured with Nanodrop. With reverse transcriptase reaction, 900 ng of cDNA was obtained using M-MLV Reverse Transcriptase (Invitrogen, MA, USA). mRNA expression levels of specific genes were detected by LightCycler. 480 SYBR Green I Master (Roche). Sequences of used primers are listed in Table S2.

Western Blotting
Cell pellets were lysed in an appropriate volume of lysis buffer (1% NP40, 150 mM NaCl, 1 mM EDTA, 50 mM Tris-HCl (pH 7.8), 1 mM NaF) containing 0.1 mM PMSF and 1X protease inhibitor cocktail (complete Protease Inhibitor Cocktail Tablets, Roche). Following 30 min of incubation in the lysis buffer, the lysates were sonicated and centrifuged (12,000 rpm, 4 • C, 15 min). Samples were denatured in 4 × SDS sample buffer at 95 • C for 5 min. For equal protein loading, Pierce™ BCA (Bicinchoninic Acid) Protein Assay (Thermo Fisher, Waltham, MA, USA) was performed, and calculations were performed accordingly. For immunoblotting, equal amounts of protein were separated by SDS-polyacrylamide gel electrophoresis and transferred onto a PVDF membrane by Trans-Blot ® TurboTM RTA Mini PVDF Transfer Kit (Biorad, Philadelphia, PA, USA). Later, the membranes were blocked with 5% non-fat dry milk in TBS-T (20 mM Tris-HCl, pH 7.8, 150 mM NaCl, 0.1%, v/v Tween-20) at RT for 1 h. After blocking, the membrane was incubated with primary antibodies overnight (4 • C). The list of primary antibodies is listed in Table S3. The membrane was washed three times with TBS-T for 15 min. The corresponding appropriate horseradish peroxidase coupled secondary antibodies (Cell Signaling, 1:10,000) were incubated for 1 h, and the membrane was washed three times with TBS-T. Blots were incubated with Clarity TM Western ECL Substrate (Biorad, Philadelphia, PA, USA) and visualized using an Odyssey Scanner (LiCor Biosciences, Lincoln, NE, USA).

Immunofluorescence Staining
Cells were seeded on 24-well plates on coverslips as 20.000 cells/well. After irradiation, the media was removed, and cells were washed with PBS twice. Cells were fixed with 4% PFA for 5 min at room temperature. After PBS wash, fixed cells were treated with 0.1% Triton X-100 for 5 min and washed with PBS. Each well was treated with 250 µL SuperBlock IHC Blocking Solution (ScyTek Laboratories, Logan, UT, USA) at room temperature for 15 min. After rinsing wells with PBS, coverslips were incubated with primary antibodies overnight at 4 • C. Coverslips were washed with PBS again and incubated with respective secondary antibodies at room temperature for 1 h in the dark. The list of primary and secondary antibodies is given in Table S4. Images were taken at Zeiss Axio Imager M1 (Germany) at 40× magnification. Foci numbers were counted for each condition and normalized to untreated control groups.

RNA-Sequencing and Analysis
Total RNAs of irradiated and parental cells were isolated using Macherey-Nagel NucleoSpin ® RNA Isolation Kit. Based on protocols of BGISEQ-500 platform, RNA-seq libraries were prepared. Libraries were sequenced on a BGI seq 500 platform using 20 million single-end reads per sample. For U373 and U373 60 Gy and KUGBM8 and KUGBM8 40 Gy , independent 3 replicate samples were sent for sequencing. Sequenced data were converted to FASTQ files using BGISEQ-500 platform at BGI Genome Sequencing Company (Beijing, China). FASTQ files were uploaded into Genialis Expressions Platform (Genialis, Inc., Boston, MA, USA) and analyzed. According to differential expression data, volcano plots were generated, with log2(FC) = 1 and 0.05 FDR cut-offs. Heat map representations were also used for selected gene sets to visualize differential expressions. Pathway analysis was performed with Enrichr gene list enrichment analysis tool directly linked to Genialis website. Gene Set Enrichment Analysis was performed. The RNA-seq data were deposited in NCBI's Gene Expression Omnibus (GEO), with accession number GSE199862.