The HIF1α/JMY pathway promotes glioblastoma stem-like cell invasiveness after irradiation

Human glioblastoma (GBM) is the most common primary malignant brain tumor. A minor subpopulation of cancer cells, known as glioma stem-like cells (GSCs), are thought to play a major role in tumor relapse due to their stem cell-like properties, their high resistance to conventional treatments and their high invasion capacity. We show that ionizing radiation specifically enhances the motility and invasiveness of human GSCs through the stabilization and nuclear accumulation of the hypoxia-inducible factor 1α (HIF1α), which in turn transcriptionally activates the Junction-mediating and regulatory protein (JMY). Finally, JMY accumulates in the cytoplasm where it stimulates GSC migration via its actin nucleation-promoting activity. Targeting JMY could thus open the way to the development of new therapeutic strategies to improve the efficacy of radiotherapy and prevent glioma recurrence.


Results
γ-radiation increases the migration velocity and invasive capacity of human GSCs. We used time-lapse videomicroscopy to characterize the motility patterns of two human GSC lines: TG1N and TG16, which were obtained from patients with high-grade gliomas 28,29 . Since then they were systematically cultured as tumorospheres in defined stem cell culture conditions, allowing them to keep their GSC properties including their capacity to generate intracerebral tumors in immunodeficient mice ( Supplementary Fig. S1A).
Twenty-four hours after plating on laminin substrate, TG1N and TG16 cells adopted a bipolar and elongated shape (Supplementary Fig. 1B) and displayed high motility (mean velocities of 26.3 ± 0.6 µm/h and 25.7 ± 1.1 µm/h, respectively) without a predefined direction ( Supplementary Fig. S1C, Supplementary Movies S1 and S2), consistently with random motility pattern with high velocity previously reported for other GSC lines 30 .
We then determined the effects of different ionizing radiation doses ranging from 0 to 3 Gy on the motility pattern of TG1N and TG16 cells. In agreement with the well-known radiation-resistance of GSCs 23,29 , quantification of activated caspase-3 and -7 in irradiated cultures by ELISA revealed minimal increases in apoptosis at 24 h post-irradiation, even at the highest dose (Supplementary Table S1). This was further confirmed by using IncuCyte Cytox Reagent to assess cell death by videomicroscopy at different times after irradiation (Supplementary Table S2). Flow cytometric analysis with propidium iodide DNA staining at 24 h post-irradiation revealed no effect of 0.5 Gy irradiation on the cell cycle of TG1N and TG16 and only a low G2/M accumulation after 3 Gy in cultures of both cell lines (Supplementary Table S3). Similarly, the colony formation assay revealed that only the dose of 3 Gy significantly impairs clonogenicity of both TG1N and TG16 cells ( Supplementary Fig. S2).
GSC migration velocity was measured over periods of 4 h ranging from 8-28 h post-irradiation. We showed dose-dependent increases of migration velocity of irradiated cells as compared to that of unirradiated controls, which remained stable during this period of time (Fig. 1A). No increase was detected after 0.1 Gy, whereas the highest increase was observed at 8-12 h after 3 Gy irradiation (1.34-and 1.23-fold increases for TG1N and TG16, respectively, ***p < 0.001; Fig. 1A). Migration velocity decreased thereafter at the highest dose probably due to the cell cycle alterations reported above (Supplementary Table S2 and S3). By contrast, we showed that 0.5 Gy induced a persistent increase in the migration velocity of the two cell lines, which remained detectable up to 52 h post-irradiation (Fig. 1B).
Irradiated (0.5 Gy) GSCs significantly explored a wider territory than unirradiated controls as shown by cumulative traces (Fig. 1C) and mean square displacement (MSD) measurements ( Supplementary Fig. S3A,B) of cells tracked from 24 to 28 h or 24 to 26 h post-irradiation respectively. This occurred without any change in directional persistence estimated either by the end-point method (defined as the ratio of the distance between two points by the actual trajectory; Supplementary Fig. S3C,D) or over time (every 10 min, Supplementary  Fig. S3E,F).
We then tested the effect of radiation on GSCs invasiveness using the Matrigel invasion chamber assay. As shown in Fig. 2A, 0.5 Gy significantly increased the invasiveness of both GSC lines (TG1N: 221 ± 41% and TG16: 125 ± 11%).
To further explore in vivo the effects of radiation on GSC invasiveness, irradiated (0.5 Gy) TG1N and TG16 cells were stereotaxically injected into the striatum of adult Nude mice (Fig. 2B). Serial coronal brain slices obtained two days after engraftment revealed that human nestin-positive cells exhibited a greater dispersion in the coronal plane, when cells were irradiated prior to injection compared to unirradiated controls (Fig. 2C).
Altogether, our data showed that sublethal doses of irradiation stimulate both the motility and invasive capacity of human GSCs.
Radiation-induced migration of GSCs depends on a rapid and transient nuclear accumulation of HIF1α. HIF1α has been shown to play a key role as a transcription factor in hypoxia-induced migration/ invasion of several glioblastoma cell lines 9,31-35 . Since HIF1α nuclear accumulation has been previously reported to be induced by ionizing radiation in tumor cells 36 , we investigated whether HIF1α could be involved in the radiation-induced migration/invasion of GSCs.
In normoxia, HIF1α is hydroxylated by prolyl hydroxylase (PHD) leading to its recognition by the von Hippel-Lindau protein and subsequent ubiquitination and targeting to the proteasome for rapid degradation 37 . PHD destabilization under hypoxic conditions allows the accumulation of HIF1α and its translocation to the nucleus 38 , where it forms a heterodimeric transcription factor complex with HIF1ß and binds the promoter regions of target genes 39 .
To investigate the role of HIF1α in radiation-induced migration, we treated our cells with Deferoxamin (DFO), an iron chelator known to stabilize HIF1α 40 . As shown in Fig. 3A-C, 100 µM DFO induced the nuclear accumulation of HIF1α in 86 ± 6% of TG1N cells compared to 12 ± 7% of controls (***p < 0.001). Similar data were obtained with TG16 cells (Supplementary Fig. S4D).
Strikingly, we found a similar nuclear accumulation of HIF1α in 70% of irradiated TG1N cells at 1 h postirradiation, compared to control cells (Fig. 3A,B, ***p < 0.001). This increase remained transient since the percentage of HIF1α-positive cells, as well as nuclear HIF1α intensity returned to control levels at 4 h post-irradiation (Fig. 3A,B and Supplementary Fig. S4A,B). Confirming the radiation-induced activation of the HIF1α pathway, we showed an increased transcriptomic expression of well-known HIF1α target genes in irradiated-GSCs 41 ( Supplementary Fig. S5A).
We next showed that the radiation-induced accumulation of HIF1α did not involve transcriptional regulation, since HIF1α mRNA levels remained unchanged in irradiated cells compared to unirradiated controls, consistently with the determinant role of post-transcriptional modifications in HIF1α accumulation 42 ( Supplementary  Fig. S4C). Interestingly, DFO treatment increased cell velocity ( Fig. 3D; ***p < 0.001), mimicking the effects of irradiation ( Fig. 3D; **p < 0.01). Similar data were obtained with TG16 cells (Supplementary Fig. S4E). www.nature.com/scientificreports/ To further investigate the importance of HIF1α on radiation-induced migration, we then used YC1, a nitric oxide-independent activator of soluble guanylyl cyclase described to indirectly block HIF1α expression at the post-transcriptional level 43 . We first checked the inhibition efficiency of 50 µM YC1 on HIF1α expression induced by DFO in our cells and under the culture conditions used for these cells (Supplementary Fig. S5B). To this end, we used HIF1α knockdown TG1N GSCs generated by lentiviral vector transduction of small-hairpin RNAs against HIF1α (shHIF1α), which dramatically decreased the HIF1α mRNA basal levels; Supplementary  Supplementary Fig. S4E).
Finally, we treated TG1N and TG16 cells with specific siRNAs that decreased HIF1α mRNA expression by 88-93% (Fig. S4F). As shown in Fig. 3E and S4G, HIF1α knockdown in TG1N and TG16 cells inhibited radiation-induced migration compared to control siRNA (siCt)-transfected cells. Altogether, these data strongly demonstrate the key role for HIF1α in the radiation-induced GSC migration.

Stimulation of the Hif1α/JMY pathway increases radiation-induced GSC migration. Nuclear
HIF1α is known to bind to hypoxia response elements (HRE) present in the promoters of a large number of genes 44 ; these genes encode proteins critical for many important cellular processes, including migration 45 . Junction-mediating and regulatory protein (JMY) is one of the genes whose transcription is driven by HIF1α under hypoxic conditions 46 . JMY has also been reported to enhance cell motility and invasion via its ability to induce actin nucleation 47,48 .
Immunofluorescence revealed that JMY was significantly up-regulated both in TG1N and TG16 cells 24 h after 0.5 Gy irradiation (Fig. 4A,B; ***p < 0.001). RT-qPCR showed an increase in JMY mRNA levels detectable from 8 h post-irradiation and persisting thereafter ( Fig. 4C and Supplementary Fig. S6A). As shown in Supplementary Fig. S7, the activation kinetic of JMY after irradiation is comparable to that of other well-known HIF1α target genes 41 . Moreover, measurement of JMY promoter activity in TG1N cells using luciferase assays confirmed that irradiation induced the activation of the JMY promoter (Fig. 4D).
Strikingly, stabilizing HIF1α levels with DFO increased the expression of JMY, whereas blocking HIF1α with YC1 prevented the irradiation-induced increase in JMY in both TG1N and TG16 cells ( Fig. 4E and Supplementary Fig. S6B).
Finally, to further investigate the importance of HIF1α in the radiation-induced increase of JMY, we assessed JMY mRNA expression in HIF1α knockdown TG1N and TG16 GSCs (Supplementary Fig. S5C and S6C) and their respective controls 18 h after irradiation (0.5 Gy). As shown in Fig. 4F and Supplementary Fig. S6D, JMY mRNA expression increased after irradiation in shCt GSCs cells, but remained unchanged in both HIF1α knockdown TG1N and TG16 GSCs. Therefore these data clearly demonstrate that the transcriptional activation of JMY is dependent on HIF1α in irradiated GSC.
We next investigated the effects of JMY knockdown using a specific siRNAs that decreased by 64 and 70% JMY mRNA expression in TG1N and TG16 cells respectively compared to siCt-transfected cells ( Supplementary  Fig. S6E). JMY knockdown did not alter the basal migration rates of TG1N ( www.nature.com/scientificreports/ expressing a negative control shRNA (shCt-TG1N cells) (Fig. Supplementary S6G). As reported above using siJMY, the JMY knockdown did not alter the in vitro basal migration rates of TG1N, whereas it completely abolished the radiation-induced migration (Fig. 4H). The stable JMY knockdown prevented also the radiationinduced invasion capacity of TG1N cells, as estimated in the invasion chamber test (Fig. 4I). ShJMY-TG1N cells and shCt-TG1N cells were then stereotaxically injected into the striatum of nude mice just after irradiation as described above. Analysis of serial coronal brain slices obtained two days after engraftments revealed that contrary to shCt-TG1N cells, 0.5 Gy radiation prior to injection did not increase the dispersion of shJMY-TG1N cells (Fig. 4J).
Altogether, these results demonstrate that radiation-induced migration of GSCs is linked to HIF1α-dependent cytoplasmic accumulation of JMY.
The role of JMY in cell motility has been attributed to its actin nucleation-promoting activity 46,48 . We thus quantified F-actin in irradiated GSCs by measuring Alexa-596 phalloidin staining. Interestingly, 24 h after irradiation (i.e. at the peak of radiation-induced migration (Fig. 1), we showed a significant increase in cellular content of F-actin in irradiated, as well as DFO-treated GSCs (Fig. 5A-D). By contrast, HIF1α inhibition by YC1 (Fig. 5A-D) or by siRNAs (Fig. 5E,F), as well as the knockdown of JMY (Fig. 5E,F), prevented both the increase of F-actin and the radiation-induced migration (Figs. 3E and 4G, Supplementary Fig. S4G and S6F). (C) YC1 (50 µM) treatment 2 h prior to irradiation (0.5 Gy) prevented the radiation-induced nuclear accumulation of HIF1α in TG1N GSCs. One hour after irradiation or DFO treatment (100 µM), nuclear HIF1α fluorescence intensity was determined in at least 50 cells per condition (**p < 0.01 and ***p < 0.001). (D) Effects of YC1 or DFO treatments on TG1N GSCs migration velocity 24 h PI. Migration velocity was expressed as percentages of the unirradiated control and calculated from at least 80 cells for each condition (**p < 0.01 and ***p < 0.001). (E) HIF1α knockdown prevented the radiation-induced migration of TG1N cells. TG1N GSCs were transfected with a siRNA targeting HIF1α (siHIF1α) or a scramble control (siCt) and irradiated (0.5 Gy) 24 h later. Twentyfour hours after irradiation, migration velocity was determined and expressed as percentage of the unirradiated control. Data were obtained from at least 70 cells per condition (***p < 0.001).

Radiation-induced migration is related to GSC stemness.
We finally investigated the dynamic behavior of our cell lines cultured under differentiating (diff) conditions (medium supplemented with 10% FBS without FGF2 and EGF) that let them loss their stem cell properties including their capacity to generate brain tumors in immunodeficient mice 29 . No obvious morphological changes were observed in diffTG1N which maintained a stable (diffTG1N) migration velocity compared to their parental cells (Fig. 6A). In contrast, diffTG16 cells presented with a markedly flattened cytoplasm and a significant decrease in migration velocity compared to the parental GSC lines (Fig. 6A).
Interestingly, ionizing radiation did not increase migration velocity (Fig. 6B) nor the expression of HIF1α (Fig. 6C) and JMY (Fig. 6D) in the differentiated cell lines, suggesting that the radiation-induced stimulation of cell motility is specific to GSCs due to the lack of activation of the HIF1α/JMY pathway in differentiated cells.

Discussion
GSCs are thought to play crucial roles in GBM relapse 23,25 . Here, we report that sublethal doses of irradiation enhance the motility and invasiveness of human GSCs through HIF1α nuclear accumulation that in turn increases the cytoplasmic actin nucleator JMY. Our study is thus in line with previous reports showing that ionizing radiation promotes migration and invasion in various cancer cell lines 15 , including human glioma cell lines [5][6][7][8][9][10][11][12][13][14]16,17,49 . However, to our knowledge, our study is the first to show not only that human GSCs are prone to this radiation effect, but also that this is specifically linked to their stemness properties, as radiation did not induce this effect on "differentiated" glioma cells. This link may explain why we were able to use a lower dose (0.5 Gy) to enhance GSC motility compared to the dose of 1 Gy 10-12 or greater [5][6][7][8][9]13,14,16,17 required to stimulate non-stem-like human glioma cell lines.
Previous studies have reported that hypoxia may increase the migration/invasion capacity of various glioblastoma cell lines in a HIF1α-dependent manner [31][32][33][34][35]50 , suggesting that HIF1α is involved in therapeutic failure and GBM relapse 49 . Our study shows that the key factor linking the radiation-induced enhancement of GSCs to stem-like properties is the fact that GSCs are much more prone than differentiated cells to the activation of HIF1α in response to radiation. This may also explain why we were able to show that 0.5 Gy enhanced the migration of GSCs, whereas Kim et al. have reported that 6 Gy-irradiation in either single or fractioned doses, but not 2 Gy irradiation, was able to induce HIF1α stabilization and stimulate the migration of U87 and U373 glioma cells 9 .
We have shown that radiation induces the migration of GSCs through a HIF1α-dependent cytoplasmic accumulation of JMY. Indeed, whereas we cannot exclude the involvement of other HIF1α-dependent pathway due to the well-known pleiotropic effects of HIF1α, our data demonstrate that JMY was absolutely required for this radiation-induced effect. JMY was initially described as a transcriptional cofactor cooperating with p300/CBP to augment p53 signaling during the DNA damage response 27,51 . JMY has been reported to accumulate in the nucleus after exposure to ultraviolet light, etoposide and actinomycin, promoting p53-mediated apoptosis. Since TG1N, but not TG16 52 , are p53 proficient, the role of JMY in radiation-induced migration is not related to p53. Moreover, the ionizing radiation doses used in this study did not trigger the nuclear accumulation of JMY and induced very low levels of GSC apoptosis, suggesting that JMY does not act as a transcriptional cofactor in radiation-induced cell migration. Rather, our data show that JMY is functioning through its previously described role in cell motility under hypoxic conditions by controlling actin dynamics via its nucleation-promoting activity 46,47 . HIF1α has been widely considered a prominent cancer drug target due to its role in the regulation of multiple survival pathways in solid hypoxic tumors. However, targeting HIF1α is highly challenging and may induce severe side effects due its multiple functions [53][54][55][56] . In this context, specific targeting of JMY could provide new therapeutic perspectives to limit radiation-induced migration of GSCs and hence prevent tumor recurrence following radiotherapy.

Materials and methods
Human glioma stem-like cell (GSC) lines and treatments. The TG1N and TG16 GSC lines were obtained from surgical resections carried out at Sainte Anne Hospital (Paris, France) on patients with highgrade gliomas according to the WHO classification 28,29,52 . Since then they were systematically cultured as tumorospheres in defined stem cell culture condition (serum-free Dulbecco's Modified Eagle Medium DMEM/F12 supplemented with B27 without vitamin A (1X, Invitrogen), heparin (5 µg/mL, Stem Cell Technologies), human recombinant epidermal growth factor (EGF, 20 ng/ml, Sigma) and human basic fibroblast growth factor (FGF-2, 20 ng/ml, Sigma)) at 37 °C in an atmosphere containing 5% CO 2 . Every week, cells were mechanically dissociated after a 10 min incubation at room temperature with the Accutase cell dissociation reagent (Sigma) and reseeded at 0.5 × 10 6 cells per T75 flask.
Generation of stable JMY-deficient GSCs was performed using lentivirus vectors constructed as follow: different hybrids of primers encoding the expected shRNA sequences were annealed by mixing 500pmoles each in NEB2.1 1 × buffer (New England Biolabs, Ipswich, MA). After 5 min at 100 °C, the annealing occurred slowly by cooling down the heat block for overnight in a styrene box. Thus, the different hybrids of primers were inserted in pTRIP-MND-GFP-H1-SanDI 59 under control of H1 promoter by complementary single-strand annealing 60 . Briefly, the plasmid was digested by SanDI (ThermoFisher Scientific, Waltham, MA) and was treated with T4 DNA polymerase (0.75U) for generating single-strand sequence complementary to single strand extended sequences present in primer hybrids. The annealing reaction was transformed in DH5α-T1R homemade competent cells. Positive clones were validated by DNA sequencing.
One day after platting on laminin substrate (5 µg/mL; Sigma), GSCs were transduced with a pool of the three lentiviral vectors at a MOI 5 (MOI-defined as the number of lentiviral particles able to transduce used per HEK-293). Transduced GSCs expressing GFP were then Fac-sorted based on GFP expression and thereafter maintained in culture.

Reverse transcription-quantitative PCR (RT-qPCR). RNA was extracted using the RNeasy Plus Mini
Kit or the RNeasy plus Micro Kit (Qiagen) according to the manufacturer's instructions. Isolated RNAs were transcribed into cDNA using the High capacity RNA to cDNA Master Mix (Applied Biosystems). Quantitative PCR reactions were performed in 96-well plates in triplicate using SYBR Green Master Mix (Applied Biosystems). The primers used are listed in Supplemental Table S4.
Luciferase JMY reporter assay. GSCc (15 × 10 3 cells) were electroporated using the Neon® transfection system (Thermo Fisher Scientific) with either 50 ng of the control (empty) pLightSwitch_empty_Prom vector (ref #S790005) or the pLightSwitch Prom reporter plasmid for the JMY gene promoter (#S719700; SwitchGear Genomics), then immediately transferred in 96-well plates previously coated with laminin (5 µg/mL; Sigma). GSCc were irradiated (0.5 Gy) 24 h after electroporation and Luciferase reporter activity was determined at different time points using the LightSwitch Dual Assay System (SwitchGear Genomics) according to the manufacturer's instructions.
Intracerebral grafts. Swiss nu/nu mice were maintained with access to food and water ad libitum in a colony room kept at a constant temperature (19-22 °C) and humidity (40-50%) on a 12:12 h light/dark cycle. All animal-related procedures were performed in compliance with the European Communities Council Directive of 22th September 2010 (EC/2010/63) and were approved by Comité d'Ethique en Expérimentation Animale, Direction de la Recherche Fondamentale, CEA (authorization #A12-029 and #A16-002; CEtEA-CEA DRF IdF).
Cell grafting was performed as previously described 63 using a stereotaxic apparatus (David Kopf model 900 Small Animal Stereotaxic Instrument). 100,000 dissociated cells (2 µL) were inoculated into the two hemispheres using a 33G Hamilton needle (Hamilton Bonaduz) at the following coordinates: anteroposterior: + 0.5 mm, dorsoventral: − 3 mm and lateral: + 1.5 (control cells) and − 1.5 mm (irradiated cells). Forty-eight hours after grafting, animals were deeply anesthetized and intracardially perfused with 4% paraformaldehyde in PBS. Brains were removed, postfixed overnight, cryoprotected with 10% sucrose/PBS, and frozen in dry ice-cooled isopentane. A cryostat (Leica CM3050S) was used to prepare serial coronal brain sections (14 µm) with an inter-slice spacing of 60 µm. These sections were mounted in order to analyze the dispersion of grafted cells by immunofluorescence staining with an anti-human nestin antibody (MAB1259, 1/400; R&D Systems, Fig. 2b) or immunodetection of GFP expression as previously described 29,63 . Images were acquired at 10 × magnification using NIS Elements software with a Pathfinder-Nikon motorized microscope (Nikon Instruments Inc.).
GSC dispersion in the coronal plane was calculated as the sum of the surfaces in µm 2 occupied by human nestin-positive or GFP-positive cells in the different coronal slices analyzed.

Statistical analyses.
All values are reported as the mean ± SEM. Statistical significance for two groups was assessed by the unpaired Mann-Whitney test or t-test. For comparison between more groups, a non-parametric ANOVA was performed followed by post hoc tests. As previously reported 64 , a two-way ANOVA with time and condition was used to compare MSD data. Statview (Abacus Concepts) and Prism Graphpad 7.1 software programs were used. Statistical significance levels are denoted as follow: *p < 0.05, **p < 0.01 and ***p < 0.001.