DUSP6 regulates radiosensitivity in glioblastoma by modulating the recruitment of phosphorylated DNAPKcs at DNA double-strand breaks

Glioblastoma (GBM) is the most malignant astrocytoma and has a 15-month median survival, despite the aggressive treatment regimen of maximum surgical resection followed by radiotherapy and chemotherapy using temozolomide (Tamimi and Juweid, 2017). Poor survival in GBM is mostly due to the inherent resistance to conventional therapies (Ramirez et al., 2013; Kaur et al., 2015), resulting in tumor recurrence in 90% of cases. With the aim to understand therapy resistance in GBM, we previously developed in vitro radiation survival models from GBM cell lines and patient samples derived from naïve primary GBM tumors. Using these models, we have shown that when GBM cell cultures are exposed to a lethal dose of radiation, a subset of residual cells (RS cells) survives. These cells are transiently non-proliferative and senescent, but eventually resume proliferation to give rise to relapse GBM, similar to recurrent tumors observed in clinical scenarios (Kaur et al., 2015, 2016, 2019, 2020; Rajendra et al., 2018). Recurrent GBMs are more difficult to treat as they are resistant to conventional therapies. Except for a few targeted therapies, such as FDA-approved angiogenic inhibitors (Hundsberger et al., 2017) or Raf and mTOR inhibitors (Lau et al., 2014), there are no effective treatments for recurrent GBM. Thus, it is imperative to identify the active signaling mechanisms conferring a survival advantage to recurrent cells.

Dual-specificity phosphatase 6 (DUSP6) is a member of the mitogen-activated protein kinase (MAPK) phosphatase family, which specifically binds to ERK1 and ERK2 (also known as MAPK3 and MAPK1, respectively; referred to here collectively as ERK) via its N-terminal kinase interaction motif (KIM) and dephosphorylates the T202 and Y204 residues of ERK1, and T185 and Y187 of ERK2 (Ahmad et al., 2018). This specific interaction is essential for inducing conformational changes and catalytic activation of DUSP6 (Ekerot et al., 2008; Bermudez et al., 2011). DUSP6 is known to function as a tumor suppressor in various cancers, including pancreatic cancer, non-small cell lung cancer, esophageal squamous cell carcinoma, nasopharyngeal carcinoma and ovarian cancer. However, in human GBM, thyroid carcinoma, breast cancer and acute myeloid carcinoma, it has been reported to play a pro-oncogenic role (Ahmad et al., 2018). DUSP6 is transcriptionally regulated by its own substrate ERK via its nuclear target Ets1 (Ekerot et al., 2008; Zhang et al., 2010). Although DUSP6 is known to be a cytoplasmic protein on account of its N-terminal conserved nuclear export signal (NES) (Ekerot et al., 2008; Caunt and Keyse, 2013; Kidger and Keyse, 2016), it has been shown to shuttle between the nucleus and cytoplasm, and the presence of an NES mediates the cytoplasmic localization of DUSP6 via CRM-1 (XPO1)-dependent nuclear export pathway (Karlsson et al., 2004; Caunt and Keyse, 2013). Beyond negative regulation of the ERK MAPK signaling pathway and sequestration of dephosphorylated ERK in the cytoplasm (Caunt and Keyse, 2013), there have been limited studies on the non-canonical functions of DUSP6. Although some studies have identified a role for DUSP6 in mediating the DNA damage response (DDR) (Bagnyukova et al., 2013; Wu et al., 2018; Ramkissoon et al., 2019) and negative regulation of reactive oxygen species-induced mitochondrial damage and apoptosis (Ma et al., 2020), the underlying mechanisms of the DDR modulation remain largely unexplored. In our study, we found that DUSP6 has a non-canonical role in modulating DNA repair via the non-homologous end joining (NHEJ) pathway, and that DUSP6 inhibition causes radiosensitization of parent and relapse GBM cells.

DUSP6 is reported to be upregulated at both transcript and protein levels in GBM compared to normal astrocytes, and its overexpression confers resistance to cisplatin therapy, highlighting its tumor-promoting potential in GBM (Messina et al., 2011). Since recurrent GBM is resistant to conventional therapies, we sought to explore the potential role of DUSP6 in recurrent GBM. For this, we first examined DUSP6 protein expression in primary (n=11) and recurrent (n=11) GBM patient biopsies using immunohistochemistry (IHC). A distinct and highly significant upregulation of DUSP6 was observed in recurrent samples compared to levels in the primary samples (P<0.001) (Fig. 1A,B). Of note, a striking observation was that, despite previously being reported as a predominantly cytoplasmic protein, we observed both nuclear and cytoplasmic localization of DUSP6 in primary and recurrent GBM patient samples (Fig. 1A).

In order to understand the survival and resistance mechanisms of parent and relapse GBM, we established an in vitro radiation survival model using naïve GBM patient-derived primary cultures and GBM cell lines (Fig. 1C). Using this model, we demonstrated that a subset of innately resistant ‘residual’ tumor cells survives at 8–16 days post exposure to a single, lethal dose of radiation (the dose at which more than 90% of cells show loss of clonogenic survival potential). Since different cell lines vary in their radiation sensitivity, the lethal doses determined were different for the different cell lines (10 Gy for U87MG, 8 Gy for SF268 and 14 Gy for LN229 and LN18). The residual cells thus obtained were transiently non-proliferative, and the majority of them (80–90%) were senescent (Fig. S1A,B). However, at ∼30–60 days post irradiation, a population of relapse tumor cells emerged from this senescent, residual subset of cells by resuming proliferation. We termed these re-emergent cells the ‘relapse’ population, since they are similar to the clinical scenario of recurrence in GBM (Kaur et al., 2015, 2016, 2019, 2020; Rajendra et al., 2018). We repeated these experiments by subjecting U87MG and SF268 cell lines to 10 Gy and 8 Gy radiation, respectively. The growth of irradiated cells followed the same trend, with relapse cells being generated from the transiently non-proliferative residual cells (Fig. 1D). We also observed that administration of 14 Gy radiation resulted in ∼90% cell death in LN229 and LN18 cell lines (Fig. S1C,D), but they too eventually relapsed after 30–60 days (data not shown).

Importantly, as seen in clinical settings, orthotopic tumorigenicity assays in mice demonstrated that the relapse population generated from the cellular model was indeed more aggressive and formed significantly larger tumors than the parent population (Fig. 1E,F) and exhibited higher tumor burden at the endpoint (day 24; Fig. 1G).

In order to decipher the role of DUSP6 in the relapse GBM cells, we initially checked the protein levels of DUSP6 in the parent, residual and relapse populations generated from four GBM cell lines – U87MG, LN18, LN229 and SF268 – by subjecting them to their respective lethal doses. As shown in Fig. 1H, there were no major alterations in the levels of DUSP6 across all the populations in the cell lines tested. Notably, we observed two distinct bands of DUSP6, which correspond to the two different translation products initiating from two different ATG start sites (Met1 and Met14; Dowd et al., 1998), also reported previously in lung cancer cells (Zhang et al., 2010). Furthermore, we examined the localization of DUSP6 in GBM cell lines. Consistent with our IHC data, both cytoplasmic and nuclear DUSP6 were observed in parent and relapse populations of all the GBM cell lines. Nuclear DUSP6 exhibited a punctate appearance in parent and relapse GBM cells, as well as in the A549 lung cancer cell line, but not in MCF7 breast cancer cell lines (Fig. 1I; Fig. S2A), indicating that nuclear localization of DUSP6 is a significant event in certain tumor types. In order to rule out non-specific binding of the anti-DUSP6 antibody, we performed co-immunofluorescence in LN229 parent cells overexpressing Myc-tagged wild-type DUSP6 (Bagnyukova et al., 2013) using anti-Myc and anti-DUSP6 antibodies. As seen in Fig. S2B, the Myc and DUSP6 signals colocalized in the nucleus and to a greater extent, in the cytoplasm, indicating the specificity of the anti-DUSP6 antibody. Furthermore, we performed western blots in parent and relapse GBM cell lines U87MG and LN229 after fractionating the nucleus and the cytoplasm. Consistent with the immunofluorescence studies, we observed DUSP6 to be present in both the nuclear and cytoplasmic fractions (Fig. 1J), again corroborating our findings that DUSP6 is present in the nucleus of GBM cells. Surprisingly, we found that out of the two translation products of DUSP6, the larger protein was present exclusively in the nuclear fraction, whereas the smaller protein was present exclusively in the cytoplasm, implying that the larger protein might be involved in certain non-canonical nuclear specific functions in GBM cells.

To explore the functional relevance of DUSP6 in GBM, we first used BCI – an allosteric chemical inhibitor of DUSP6 (Wu et al., 2018; Ramkissoon et al., 2019) – to study the differential effect of DUSP6 inhibition on the survival of parent and relapse GBM cells. Initially, parent and relapse cells of the U87MG and LN229 cell lines were treated with increasing concentrations of BCI (1–5 µM) for 72 h. As shown in Fig. 2A and B, there was a dose-dependent decrease in the viability of parent and relapse cells of U87MG, and to a greater extent of LN229, upon BCI treatment. The optimum concentration used for further experiments was 1 µM for U87MG, as there was no loss of viability at this concentration. Since LN229 cells were more sensitive to BCI, 0.5 µM BCI was used in subsequent experiments, as cell viability was greater than 80% at this concentration. Inhibition of DUSP6 catalytic function by BCI at these concentrations was evaluated by analyzing the phosphorylation status of ERK. As expected, in comparison to levels in untreated cells, BCI treatment elevated the levels of phosphorylated ERK (p-ERK) in the relapse population of both cell lines. However, in the parent population, BCI treatment failed to cause any increase in p-ERK levels at the time point examined, in both the cell lines (Fig. 2C), possibly due to intrinsically high levels of p-ERK in the parent population.

Since radiation therapy remains the mainstay of the treatment regime for primary and recurrent GBM, we further asked whether pretreatment with BCI can radiosensitize GBM cells in vitro. For this, we performed a clonogenic radiation survival assay by treating parent and relapse cells of U87MG and LN229 with 0.25 µM and 0.125 µM BCI, respectively, for 30 min before exposure to variable doses of radiation (2–10 Gy). BCI doses chosen were a quarter of the previously optimized doses because higher doses were found to be lethal when combined with radiation in the clonogenic assay. As shown in Fig. 2D,E, pretreatment with BCI reduced the survival of both parent and, more significantly, relapse cells of U87MG and LN229, even at lower doses of radiation (2, 4 and 6 Gy), thereby highlighting the radiosensitizing effect of BCI on GBM cells. Importantly, prolonged treatment with BCI prevented recurrence of parent and relapse U87MG cells after exposure to 10 Gy radiation, the lethal dose for U87MG (Fig. S3A). Although radiation treatment induced necrosis of GBM cells, there was no significant increase in the percentage of necrotic or apoptotic cells upon concomitant administration of BCI in both parent and relapse populations (Fig. S3B,C). This implies that BCI treatment induces sustained growth arrest and prevents resumption of proliferation in parent and relapse cells post exposure to a lethal dose of radiation.

Radiation therapy induces DNA double-strand breaks (DSBs) in cells, which can be repaired via homologous recombination (HR) and/or NHEJ pathways, and if the induced DSBs are left unrepaired they can lead to apoptosis (Seluanov et al., 2010; Nowsheen and Yang, 2012). Since we observed a distinct nuclear localization of DUSP6, we wanted to understand whether BCI treatment-induced radiosensitization in GBM cells was due to an altered DDR mediated by inhibition of DUSP6 nuclear function. For this, we first checked for phosphorylation of the DNA damage sensor protein H2AX (at S139; hereafter referred to as γH2AX) at 45 min (initial time point) and 24 h post irradiation (later time point) in U87MG and LN229 parent and relapse cells treated with 1 µM or 0.5 µM BCI (for U87MG and LN229, respectively) for 30 min, followed by exposure to 10 Gy or 14 Gy radiation (for U87MG and LN229, respectively), the lethal doses standardized for these cell lines, as mentioned above. The time points were chosen on the basis of previous reports that radiation-induced γH2AX levels increase within 30 min and reach maximum levels by around 1 h, followed by decline and disappearance by 24 h (Mariotti et al., 2013; Lee et al., 2019). As expected, γH2AX levels were significantly increased at 45 min post irradiation in both parent and relapse populations; however, surprisingly, the combined treatment of radiation and BCI significantly reduced the activation of γH2AX at 45 min in the parent and relapse populations of both the cell lines. γH2AX levels returned to the basal values at 24 h post irradiation in both the populations, and no differences were observed with BCI pretreatment (Fig. 2F–I). Of note, treatment with BCI alone did not induce any DSBs at both the time points in U87MG and LN229 parent and relapse cells, and hence all the further experiments with BCI were performed in combination with irradiation.

Next, we checked whether the reduced γH2AX levels in the combination therapy group were due to reduced DNA damage. Neutral comet assays performed using cells treated with irradiation and either BCI (IR+BCI) or the vehicle control (IR+vehicle) showed equal DSBs in parent and relapse populations of both the cell lines (Fig. 3A–D), implying that the reduced levels of γH2AX observed after BCI treatment were not due to decreased DNA damage induction in these cells. To explore the mechanism of reduced γH2AX in IR+BCI-treated cells, we first checked the activation of the major kinases that phosphorylate H2AX in response to radiation – DNAPKcs (also known as PRKDC), ATM and ATR – by assaying their activating phosphorylation sites. As shown in Fig. 3E,F, among the three kinases, phosphorylation of DNAPKcs at S2056 (p-DNAPKcs) was most upregulated at 45 min post irradiation in both the cell lines; phosphorylation of ATM at S1981 (p-ATM) was also upregulated in both the cell lines at this time point, but the levels were not as high as those of p-DNAPKcs. Phosphorylation of ATR at T1989 (p-ATR) was not significantly altered at the 45 min time point. However, we did not observe any significant changes in the levels of p-DNAPKcs upon BCI pretreatment in the parent and relapse populations of both the cell lines at 45 min and 24 h post irradiation (compare IR+vehicle and IR+BCI groups; U87MG parent, P=0.690 and P=0.581 for 45 min and 24 h, respectively; U87MG relapse, P=0.739 and P=0.887; LN229 parent, P=0.789 and P=0.233; LN229 relapse, P=0.956 and P=0.917; n=3; two-tailed, unpaired t-test). Similarly, we did not observe any significant changes in the levels of p-ATM upon BCI pretreatment in the parent and relapse populations of both the cell lines at the same time points post irradiation (U87MG parent, P=0.159 and P=0.581 for 45 min and 24 h, respectively; U87MG relapse, P=0.747 and P=0.575; LN229 parent, P=0.344 and P=0.948; LN229 relapse, P=0.279 and P=0.794; n=3; two-tailed, unpaired t-test). Taken together, the data indicate that DUSP6 inhibition does not prevent phosphorylation and activation of these major DNA damage sensor kinases.

In order to identify the major active DSB repair pathway in our parent and relapse GBM cells, we checked for the activation of NHEJ and HR, the most prominent pathways active after DSB formation (Seluanov et al., 2010), using in vivo HR and NHEJ activity assays as described previously (Seluanov et al., 2010; Tilgner et al., 2013; Newman et al., 2015; Salunkhe et al., 2018; Kaur et al., 2020). Indeed, we found significantly higher activity of NHEJ, as compared to HR, in the parent and relapse populations of both U87MG and LN229 (Fig. S4A,B). These data corroborate our earlier finding (Fig. 3E,F) that p-DNAPKcs (a major regulator of NHEJ repair pathway) was the most significantly upregulated kinase post radiation treatment of GBM cells. Since we did not find any significant effect of BCI pretreatment on the levels of p-DNAPKcs post irradiation (Fig. 3E,F), we checked for the recruitment of p-DNAPKcs to DSB sites. Indeed, administration of BCI along with radiation significantly reduced the recruitment of p-DNAPKcs specifically in the relapse population of U87MG cells (Fig. 4A,B) and in both parent and relapse populations of LN229 cells (Fig. 4C,D) at 45 min post irradiation. p-DNAPKcs foci were resolved to near baseline values in the parent and relapse populations of both the cell lines, and no significant differences were observed between co-administration of radiation with DMSO vehicle or BCI by 24 h. These results show that inhibition of DUSP6 significantly reduces recruitment of the major NHEJ protein p-DNAPKcs at radiation-induced DSBs, with the effect being more pronounced in the relapse population. Recruitment of 53BP1 (also known as TP53BP1), another DNA damage mediator protein that promotes NHEJ and inhibits HR (Gupta et al., 2014), was also checked upon combination therapy of radiation with BCI. 53BP1 recruitment was significantly upregulated at 45 min post irradiation in parent and relapse populations of U87MG and LN229 cells. As seen with p-DNAPKcs, recruitment of 53BP1 was also significantly reduced in the relapse population of both the cell lines upon IR+BCI treatment as compared to IR+vehicle (DMSO) treatment, whereas no significant difference was observed in parent cells at 45 min upon IR+BCI treatment. At 24 h, the recruitment levels returned to baseline, and no significant differences were observed between IR+vehicle and IR+BCI treatment groups in U87MG, although the levels significantly decreased in the combination group of IR+BCI in LN229 parent cells (Fig. 4E–H). Taken together, these results suggest that inhibition of DUSP6 significantly reduces the early recruitment of major NHEJ proteins (p-DNAPKcs and 53BP1) to radiation-induced DSB sites, particularly in relapse GBM.

To further validate our findings, we overexpressed wild-type and dominant-negative C293S mutant forms of DUSP6 (Bagnyukova et al., 2013) in the parent and relapse populations of U87MG and LN229 cells (Fig. 5A,B). Overexpression of mutant DUSP6 significantly reduced the clonogenic survival of relapse GBM cells as compared to that of cells harboring wild-type DUSP6, even at lower doses of radiation (Fig. 5C,D). Similar to the effects of DUSP6 inhibitor treatment, overexpression of dominant-negative mutant DUSP6 also significantly reduced the recruitment of p-DNAPKcs to DSB sites at 45 min post irradiation in U87MG relapse cells (Fig. 5E,F) and in LN229 parent and relapse cells (Fig. 5G,H) that returned to baseline values at 24 h post irradiation. These data again corroborate our findings that inhibition of DUSP6 activity modulates the DDR by reducing the recruitment of p-DNAPKcs, particularly in relapse GBM cells, resulting in deregulated DDR signaling and decreased survival upon exposure to radiation. It is noteworthy that upon overexpression of DUSP6 in these GBM cells, the smaller protein was more significantly overexpressed (Fig. 5A,B) on account of its higher translation efficiency (Dowd et al., 1998). As per our earlier findings (Fig. 1J), it is likely that this overexpressed protein is mostly cytoplasmic. Hence, the radiosensitization effects seen with overexpression of the mutant form are somewhat lesser than those seen with BCI pretreatment.

To assess the radiosensitization effects of BCI in vivo, we established orthotopic mouse models of GBM by intracranially injecting U87MG parent and relapse cells expressing luciferase. At ∼6 days post tumor cell injection, these mice were grouped to receive radiation therapy in fractions of 2 Gy daily for a total dose of 10 Gy along with either DMSO vehicle (IR+vehicle) or 25 mg/kg BCI (IR+BCI), administered intraperitoneally every 48 h for the duration of radiotherapy (Fig. 6A). Thus, the combination therapy group received three doses of DMSO vehicle or BCI. As shown in Fig. 6B,C, the combination therapy with radiation and BCI delayed tumor progression of U87MG parent cells, and a significant difference in tumor growth was observed between the IR+vehicle and IR+BCI groups at day 28 post injection. Although the combination therapy of radiation and BCI failed to cause significant reduction in tumor burden of U87MG relapse cells (Fig. 6D,E), the time taken for tumor progression increased by 1 week in the IR+BCI group. Thus, the combination therapy with radiation and BCI significantly increased the progression-free survival (PFS) by 1 week in both the parent and relapse tumors, as compared to combination therapy with radiation and DMSO vehicle (PFS of 3 weeks in the IR+vehicle group versus 4 weeks in the IR+BCI group for parent tumors, and PFS of 2 weeks in the IR+vehicle group versus 3 weeks in the IR+BCI group for relapse tumors; Fig. 6F,G), signifying that BCI treatment can potentially radiosensitize GBM tumors, including recurrent GBM, thus significantly delaying the tumor progression and improving the therapeutic outcome. This finding is particularly important in the context of recurrent GBM, since radiation therapy (IR+vehicle) failed to improve the PFS of relapse tumors, as seen in Fig. 6G (PFS of 2 weeks in the untreated and IR+vehicle groups).

GBM is a highly infiltrative primary brain tumor associated with poor prognosis. Despite an increase in the understanding of this disease, a cure still eludes GBM patients. DNA repair components such as MGMT and PARP-1 are reported to be altered in GBM (Lawrence et al., 2015; Erasimus et al., 2016), affecting the efficacy of chemoradiotherapy, the major therapeutic modality of GBM. Modulation of DNA repair pathways has also been reported to impart resistance to radiotherapy in GBM (Erasimus et al., 2016; Ghorai et al., 2020). Despite many reports, our understanding of DDR regulation in response to radiation treatment remains limited. Here, we show that DUSP6 is upregulated in recurrent GBM and regulates the recruitment of p-DNAPKcs to DSB sites, thus regulating DNA repair and cell survival.

Knockdown of DUSP6 or its inhibition using BCI have been reported to overcome cisplatin resistance in gastric cancer (Wu et al., 2018) and to sensitize cervical cancer cells to anti-EGFR therapy and camptothecin (Bagnyukova et al., 2013). These effects of altering DUSP6 activity have been shown to be mediated via changes in the DDR pathways of the cells (Bagnyukova et al., 2013; Wu et al., 2018; Ramkissoon et al., 2019); however, none of these studies have addressed the mechanisms underlying DUSP6-mediated alterations of the DDR.

Mislocalization of oncoproteins, tumor suppressor proteins and other cancer-related proteins due to deregulation of protein translocation machinery has been reported in various cancers, and this has been found to result in tumor development and metastasis (Wang and Li, 2014) as well as multi-drug resistance (Liang et al., 2003). Recently, we have also shown that altered localization of p65 (also known as RELA) regulates therapy-induced senescence in GBM (Salunkhe et al., 2021). Here, we observed a distinct nuclear localization of DUSP6, which is reported to be a predominantly cytoplasmic protein (Ekerot et al., 2008; Caunt and Keyse, 2013; Kidger and Keyse, 2016). Similar observations have been reported in triple-negative breast cancer, where nuclear DUSP6 is observed in HER2⁺ circulating tumor cells and brain metastases, indicating that nuclear DUSP6 may be associated with brain metastasis (Wu et al., 2019). Moreover, we found that the larger of the two DUSP6 protein forms was specifically localized in the nucleus of GBM cells. We hypothesized that this nuclear DUSP6 plays a role in modulating the DDR by regulating the recruitment of repair proteins to chromatin. Indeed, we have demonstrated that DUSP6 regulates the recruitment of p-DNAPKcs, thus controlling phosphorylation of H2AX and recruitment of 53BP1, which subsequently facilitates DNA repair and cell survival following exposure to radiation. Accordingly, DUSP6 inhibition resulted in decreased DNA damage sensing due to reduced recruitment of p-DNAPKcs, γH2AX and 53BP1, thus compromising the efficiency of DNA damage repair. However, at present, we are yet to decipher the actual mechanism of DUSP6-mediated recruitment of DDR proteins to the chromatin, and no studies have shown the direct regulation of DNA repair proteins such as p-DNAPKcs by DUSP6. We have also demonstrated that despite differences in early recruitment of DDR proteins upon BCI treatment, eventually at 24 h, BCI treated and untreated cells show resolution of DSBs. However, the fact that BCI treatment abrogates long-term survival and proliferation emphasizes the importance of efficient recruitment of DNA repair proteins during the initial stages of DSB repair for the survival of GBM cells.

From a clinical perspective, we have demonstrated radiosensitization of relapse GBM cells upon inhibition of DUSP6 either by using BCI or by overexpressing a dominant-negative mutant form of DUSP6. This is an important preclinical finding, given the fact that at present there are no major treatment modalities for recurrent GBM. Our in vivo studies showed that tumor progression was significantly delayed upon BCI treatment in combination with radiation therapy, again corroborating our in vitro findings, wherein BCI pretreatment induced cytostasis in GBM cells. Future studies aimed at improving the uptake and bioavailability of BCI in the brain may be useful in improving the radiosensitizing effect of BCI on GBM tumors in vivo.

Taken together, as summarized in Fig. 6H, we have demonstrated for the first time that a nuclear form of DUSP6 exists in GBM and that inhibition of DUSP6 catalytic activity causes radiosensitization of GBM cells, particularly in the relapse population, by decreasing the recruitment of DNA repair factors to the radiation-induced DSB sites, thereby resulting in growth inhibition. From a clinical perspective, administration of BCI along with the conventional therapies may prolong the time for recurrence in GBM. Thus, our study provides a rationale for the development of DUSP6 inhibitors in combination with radiation as a potential treatment regimen for primary and recurrent GBM.

The GBM cell line U87MG was procured from ATCC. The SF268 GBM cell line and HEK 293FT cells were kind gifts from Dr Amit Dutt, Integrated Cancer Genomics Lab, ACTREC, Navi Mumbai, India. LN229 and LN18 GBM cell lines were procured from NCCS, Pune, India. All the cells were maintained in DMEM (Gibco) supplemented with 10% fetal bovine serum (Gibco), 200 U/ml penicillin and 100 μg/ml streptomycin at 37°C and 5% CO2 in a humidified incubator. All the cell lines were authenticated by short tandem repeat (STR) profiling.

Cells were seeded at 70% confluence and exposed to their respective lethal doses of γ-radiation using a ⁶⁰Co source. After the initial death phase, ∼8–10 days post irradiation, the cells remained transiently non-proliferative for about 5–15 days, after which they resumed proliferation. The non-proliferative cells were designated as the residual (RS) population of cells, and the cells which outgrew from the RS population were designated as relapse (R) population.

DUSP6 inhibitor BCI [(E)-2-benzylidene-3-(cyclohexylamino)-2,3-dihydro-1H-inden-1-one] was purchased from Merck Millipore (CAS 15982-84-0; catalog number 317496) and reconstituted in DMSO. For immunofluorescence studies and western blot analysis, U87MG cells were treated using 1 µM BCI at least 30 min before exposure to 10 Gy radiation. LN229 cells were treated with 0.5 µM BCI and exposed to 14 Gy radiation. For immunofluorescence studies and western blot analysis, cells were collected at appropriate time points for fixation or protein isolation.

ΔDUSP6 (C293S mutant) and DUSP6 plasmids developed by Dr Igor Astsaturov were procured from Addgene (Addgene plasmids 27977 and 27975, respectively; Bagnyukova et al., 2013). For generation of lentiviral particles, overexpression plasmids were transfected into HEK 293FT cells (kind gift from Dr Amit Dutt, Integrated Cancer Genomics Lab, ACTREC, Navi Mumbai, India) along with psPAX packaging vector (kind gift from Dr Amit Dutt) and pMD2.G envelope vector (kind gift from Dr Amit Dutt) using calcium phosphate. Viral particles were then transduced into U87MG and LN229 cells, and transduced cells were selected using 2–2.5 µg/ml puromycin.

For enumeration of cell viability upon radiation, 50,000 SF268, U87MG, LN229 and LN18 cells were seeded in 60-mm dishes in triplicate and were exposed to 8 Gy, 10 Gy, 14 Gy and 14 Gy doses of radiation, respectively. After every 3 days, the cells were trypsinized and the viable count was determined on a hemocytometer using Trypan Blue dye.

For the proliferation assay upon BCI treatment, 25,000 parent and relapse cells of the U87MG and LN229 cell lines were seeded in 6-well plates and were treated with different concentrations of BCI, ranging from 1 µM to 5 µM. After 72 h of BCI treatment, cells were trypsinized, and the viable count was determined using Trypan Blue.

For the proliferation assay with combination treatment of radiation and BCI, 25,000 parent and relapse cells of the U87MG cell line were seeded in 6-well plates and were treated with DMSO vehicle or 1 µM BCI at least 30 min before exposure to 10 Gy radiation. Cells were trypsinized and the viable count was determined using Trypan Blue after every 72 h. BCI was replenished every 72 h.

For senescence-associated β-galactosidase activity assays, 25,000 cells of parent and residual populations of U87MG, SF268, LN229 and LN18 cell lines were seeded in 24-well plates. The assay was performed using a Senescence β-galactosidase Staining Kit (#9860, Cell Signaling Technologies) as per the kit protocol. Positive staining was analyzed on an Olympus IX71 microscope. Images of more than 100 cells were acquired at 20× objective magnification.

For clonogenic assays, 100 cells of U87MG and 300 cells of LN229 were seeded into 12-well plates and were exposed to different doses of radiation after 30 min pretreatment with 0.25 µM BCI for U87MG and 0.125 µM BCI for LN229. Plates were incubated at 37°C for 7–10 days until colonies consisting of ≥50 cells appeared. Colonies were then fixed using chilled methanol and stained using 0.25% Crystal Violet and counted microscopically.

Cells were seeded on coverslips and exposed to their respective lethal doses of radiation after 30 min pretreatment with DMSO or BCI. Cells were fixed with chilled methanol at −20°C for 10 min followed by phosphate-buffered saline (PBS) wash and then permeabilized with 0.5% Triton X-100 for 15 min at 4°C. After a single PBS wash, cells were incubated with 1% BSA solution in PBST (PBS with 0.1% Tween 20) for 1 h at room temperature, followed by overnight incubation with primary antibody at the dilutions specified in Table S1, at 4°C. After washing thrice for 10 min with PBS, coverslips were incubated with Alexa Fluor-, FITC-, or PE-conjugated anti-rabbit or anti-mouse secondary antibody for 1 h at room temperature. For co-immunofluorescence of Myc tag and DUSP6, sequential staining was performed first for the Myc tag and then for DUSP6. Nuclei were counterstained with DAPI (1.5 μg/ml) for 1 min, washed thrice with PBS and mounted using VECTASHIELD mounting medium (Vector Labs). The cells were visualized under a Zeiss LSM 510 Meta Confocal Microscope, and the staining intensities were analyzed using ImageJ software (NIH, Bethesda, MD).

Cells were lysed using RIPA lysis buffer (50 mM Tris-HCl, pH 8.0, 150 mM NaCl, 5 mM EDTA, 1% NP-40, 0.5% sodium deoxycholate and 0.1% SDS) containing protease and phosphatase inhibitor cocktail (Sigma Aldrich) for 15 min on ice. The supernatant was collected, and 30 µg of protein was loaded onto an SDS–PAGE gel, transferred to nitrocellulose membrane and probed with the specified antibodies at the dilutions mentioned in Table S2. Vinculin was used as a loading control. Immunoreactive proteins were visualized using an enhanced chemiluminescence (ECL) reagent (BioRad). Nuclear cytoplasm fractionation was performed as described previously (Dimauro et al., 2012), and western blotting was performed using 30 µg protein from each fraction.

IHC staining of DUSP6 (1:50; Invitrogen, 720405) was performed on paraffin sections of GBM patient biopsies (obtained from the tissue biorepository, ACTREC, Navi Mumbai, India, after approval from the TMC-ACTREC Institutional Ethics Committee), using a Vectastain ABC kit (Vector laboratories) and developed using DAB peroxidase substrate kit (Vector laboratories) as per the manufacturer's instructions. Staining was visualized under a Zeiss Axio Imager Z1 upright microscope, and intensities were scored using IHC Profiler (Varghese et al., 2014). Since these patient samples were collected retrospectively from the tissue repository, waiver of consent was granted for procuring the samples by the institutional ethics committee.

Cells were harvested at the indicated time points. Further steps were followed as previously described (Salunkhe et al., 2018). Comet tails were stained with 2.5 µg/ml propidium iodide solution and visualized under a Zeiss Axio Imager Z1 upright microscope. Percentage tail DNA was calculated using Open Comet software in ImageJ.

HR and NHEJ reporter plasmids were a kind a gift from Dr Vera Gorbunova, Department of Biology, University of Rochester, NY (Seluanov et al., 2010). These plasmids are fluorescent reporter constructs used to quantitatively measure HR and NHEJ activity in a cell. Briefly, these constructs contain an engineered GFP gene, which has a rare-cutting ISce-I endonuclease site for DSB induction. Initially, the constructs are GFP negative due to the presence of an additional exon or mutations. HR- or NHEJ-mediated repair of ISce-I-induced DSBs results in restoration of functional GFP. Thus, the number of GFP-positive cells corresponds to HR or NHEJ activity in the cells. In this assay, transfection with TdRed plasmid (a kind gift from Dr Amit Dutt, Integrated Cancer Genomics Laboratory, ACTREC, India) aids in determining the transfection efficiency of cells. Thus, the ratio of GFP-positive to TdRed-positive cells determines the overall repair efficiency of cells. Parent and relapse cells of U87MG and LN229 were seeded in 96 well plates at ∼80% confluency and were transfected with 100 ng of ISce-I digested linearized HR and NHEJ plasmids using Lipofectamine 3000 (Invitrogen). TdRed was used as transfection control. After 72 h of transfection, the number of GFP- and TdRed-positive cells were determined using an Attune Nxt flow cytometer, and their ratios were calculated to obtain repair efficiency.

Parent and relapse cells of U87MG and LN229 cell lines were seeded in 60 mm dishes and were treated with 1 µM or 0.5 µM of BCI 30 min before exposure to 10 Gy or 14 Gy radiation, respectively. BCI was replenished after every 72 h and, post day 13 of radiation, the cells were harvested and used for an apoptosis assay by staining for Annexin V and propidium iodide as per the kit manufacturer's protocol (Cell Signaling Technologies). The percentage of positive cells was identified using an Attune Nxt flow cytometer.

Luciferase-expressing U87MG cells were generated by transduction with lentiviral particles from pLenti-CMV-Puro-Luc plasmid (w168-1), developed by Dr Eric Campeau and Dr Paul Kaufman (Addgene plasmid 17477; Campeau et al., 2009). For tumorigenicity studies, 0.5 million luciferase-expressing U87MG cells were injected intracranially into the brains of 6–8-week-old male nude mice (Mus musculus) according to previously described protocol (Ozawa and James, 2010). The coordinates for injection were 2 mm right and 1 mm posterior to the bregma, at a depth of 2.5 mm. Mice were monitored at specific intervals for tumor progression by bioluminescence imaging, for which 150 mg/kg D-luciferin potassium salt (Biosynth AG) was administered intraperitoneally. About 10–12 min post administration of luciferin, bioluminescence imaging was performed using IVIS Spectrum. Intensity values were calculated using Living Image software (Perkin Elmer), and the total flux values expressed in photons/sec were plotted.

For inhibitor studies, 0.5 million luciferase-expressing U87MG parent and relapse cells were injected intracranially into 6–8-week-old female nude mice and were grouped to receive either radiation in combination with DMSO (vehicle) or radiation with BCI (25 mg/kg). At ∼5–6 days post injection, mice were subjected to whole-brain fractionated radiotherapy of 10 Gy (2 Gy for 5 days). DMSO or BCI were administered intraperitoneally at the intervals of 48 h, starting on day 1 of radiotherapy. Treatment response was monitored every week using bioluminescence imaging, and mice were euthanized upon symptoms of distress such as weight loss by <20% of the original, hunchback posture, loss of mobility, head tilting to one side, etc. All the animal experiments were performed as per protocols approved by the Institutional Animal Ethics committee of ACTREC, Navi Mumbai, India.

Significance of intensity scores from immunofluorescence studies was determined using one-way ANOVA with Tukey's multiple comparison test. IHC studies were analyzed using the Mann–Whitney U-test. Significance of Kaplan–Meier plots was calculated using log-rank test. Significance values of all the other studies were determined using two-tailed, unpaired Student's t-test.

We acknowledge Ketaki Patkar for assistance in generating reagents.