The m6A RNA Demethylase ALKBH5 Promotes Radioresistance and Invasion Capability of Glioma Stem Cells

Simple Summary Glioblastoma stem cells (GBMSCs), which are particularly radio-resistant and invasive, are responsible for the high recurrence of glioblastoma (GBM). Therefore, there is a real need for a better understanding of the mechanisms involved in these processes and to identify new factors that might be targeted to radiosensitize GBMSC and decrease their invasive capability. Here, we report that the m6A RNA demethylase ALKBH5, which is overexpressed in GBMSCs, promotes their radioresistance by controlling the homologous repair. ALKBH5 was also involved in GBMSC invasion. These data suggest that ALKBH5 inhibition might be a novel approach to radiosensitize GBMSCs and to overcome their invasiveness. Abstract Recurrence of GBM is thought to be due to GBMSCs, which are particularly chemo-radioresistant and characterized by a high capacity to invade normal brain. Evidence is emerging that modulation of m6A RNA methylation plays an important role in tumor progression. However, the impact of this mRNA modification in GBM is poorly studied. We used patient-derived GBMSCs to demonstrate that high expression of the RNA demethylase, ALKBH5, increases radioresistance by regulating homologous recombination (HR). In cells downregulated for ALKBH5, we observed a decrease in GBMSC survival after irradiation likely due to a defect in DNA-damage repair. Indeed, we observed a decrease in the expression of several genes involved in the HR, including CHK1 and RAD51, as well as a persistence of γ-H2AX staining after IR. We also demonstrated in this study that ALKBH5 contributes to the aggressiveness of GBM by favoring the invasion of GBMSCs. Indeed, GBMSCs deficient for ALKBH5 exhibited a significant reduced invasion capability relative to control cells. Our data suggest that ALKBH5 is an attractive therapeutic target to overcome radioresistance and invasiveness of GBMSCs.


Introduction
Glioblastoma (GBM) is the most aggressive primary brain tumor in adults. These tumors exhibit diffuse infiltration into brain tissue, rendering surgical resection difficult and incomplete. GBM are also particularly resistant to adjuvant treatments after surgery. Despite the combination of radiotherapy and chemotherapy, most GBM recurs and median survival remains around 15 months [1]. It is speculated that this high recurrence is due to the presence of glioblastoma stem cells (GBMSCs), which are characterized by their ability of self-renewal, the overexpression of stem cell markers, their pluripotent aptitude to differentiate into neural lineages, and their high tumorigenic potential in vivo [2,3]. GBMSCs are particularly chemo-radioresistant and are also characterized by a high capacity toma biopsy specimens.
To assess a potential off-target effect of the ALKBH5 siRNA, we used a second siRNA and showed in GBMSCs a high inhibition of ALKBH5 expression at the mRNA and protein level as well as similar results on clonogenic survival assays under IR (Figure S1A-C). Whereas demethylases, such as ALKBH5, remove N6-methylation of adenosine, the m6A modification is catalyzed by methyltransferases, particularly METTL3 and METTL14. We therefore compared, by real-time PCR, the relative expression of these methyltransferases with that of ALKBH5 in GBMSC. Results in Figure S1D show a 3-to 5fold higher expression of ALKBH5 compared to MTTL3 and MTTL14 in the GBMSCs used in this study.

Targeting ALKBH5 in GBMSCs Inhibits the Expression of Key DNA Damage Response Genes
Radioresistance is determined by the ability of cancer cells to repair radiation-induced DNA damages an upregulated DNA damage repair (DDR) has been reported in GBMSCs [31][32][33]. We therefore investigated the role of ALKBH5 in the DDR by evaluating if this m6A RNA demethylase regulates the expression of target genes involved in different DNA repair pathways. In the first steps of the DDR, the checkpoint kinases CHK1 and To determine whether ALKBH5 affects radiation sensitivity, we performed 3-D clonogenic survival assays with increasing doses of irradiation (IR) ( Figure 1C-E). The survival fractions after IR were significantly decreased in all GBMSCs transfected with the specific ALKBH5 siRNA compared to the control siRNA, indicating that downregulation of ALKBH5 mRNA expression radiosensitizes GBMSCs derived from human glioblastoma biopsy specimens.
To assess a potential off-target effect of the ALKBH5 siRNA, we used a second siRNA and showed in GBMSCs a high inhibition of ALKBH5 expression at the mRNA and protein level as well as similar results on clonogenic survival assays under IR (Figure S1A-C). Whereas demethylases, such as ALKBH5, remove N6-methylation of adenosine, the m6A modification is catalyzed by methyltransferases, particularly METTL3 and METTL14. We therefore compared, by real-time PCR, the relative expression of these methyltransferases with that of ALKBH5 in GBMSC. Results in Figure S1D show a 3-to 5-fold higher expression of ALKBH5 compared to MTTL3 and MTTL14 in the GBMSCs used in this study.

Targeting ALKBH5 in GBMSCs Inhibits the Expression of Key DNA Damage Response Genes
Radioresistance is determined by the ability of cancer cells to repair radiation-induced DNA damages an upregulated DNA damage repair (DDR) has been reported in GBM-SCs [31][32][33]. We therefore investigated the role of ALKBH5 in the DDR by evaluating if this m6A RNA demethylase regulates the expression of target genes involved in different DNA repair pathways. In the first steps of the DDR, the checkpoint kinases CHK1 and CHK2 are activated by the effector kinases ATR and ATM, resulting in the initiation of cell cycle checkpoints, cell cycle arrest, and DNA repair [34]. Following ALKBH5 knockout with specific siRNA, we observed a significant inhibition of CHK1 mRNA expression compared to a scramble control ( Figure 2A).
Similar results were obtained with a second ALKBH5 siRNA ( Figure S1E). A high inhibition of CHK1, when ALKBH5 is downregulated, was also confirmed at the protein level by Western blot analysis ( Figure 2B). In addition, we observed, in response to IR, an increase in CHK1 phosphorylation that was inhibited by ALKBH5 siRNA (Figure 2B), indicating that the activation of CHK1 following IR-induced DNA damage can be overcome by blocking the m6A RNA demethylase. In contrast, blocking ALKBH5 did not affect the expression of CHK2, ATR, or ATM. Since the results in GBMSCs showed that ALKBH5 could regulate CHK1 expression, we then analyzed the correlation between ALKBH5 and CHK1 expression using the glioblastoma database of the Cancer Genome Atlas (TCGA). As shown in Figure 2C, a significant positive correlation was observed between the expression of the two genes.
Cells use two main pathways to repair double-strand breaks (DSBs), homologous recombination repair (HRR) and nonhomologous end joining (NHEJ). We next examined the role of ALKBH5 in the expression of key DNA damage repair players involved in these two pathways. First, we observed in the primary cell lines transfected with specific ALKBH5 siRNA compared to a scramble control a significantly reduced expression of several genes involved in HRR. In particular, Rad51, XRCC2, BRCA2, EXO1, and BRIP1 were downregulated ( Figure 3A,B). Similar results were obtained on Rad51 expression with a second ALKBH5 siRNA ( Figure S1E). We next analyzed the effect of IR on the expression of these HRR target genes in GBMSCs. As shown in Figure 3C,D, IR significantly increased Rad51, XRCC2, BRCA2, EXO1, and BRIP1 mRNA levels, whereas the induction by radiations was abrogated when ALKBH5 was inhibited by specific siRNA, indicating that blocking this m6A RNA demethylase was able to overcome the upregulation of DDR genes induced by IR. In contrast, the expression of the key genes involved in NHEJ, including Ku70, Ku80, and DNA-PKs, was not affected by ALKBH5 knockdown (data not shown).
that blocking this m6A RNA demethylase was able to overcome the upregulation of DDR genes induced by IR. In contrast, the expression of the key genes involved in NHEJ, including Ku70, Ku80, and DNA-PKs, was not affected by ALKBH5 knockdown (data not shown).   The recombinase Rad51 is the central protein involved in the homologous recombination of DNA during double-strand break repair [35]. Previous reports have confirmed a high expression of Rad51 in GBMSCs and its important role in resistance to radiotherapy [4,33,36,37]. As mentioned above, Rad51 mRNA expression can be regulated by ALKBH5. We then evaluated Rad51 levels by Western blot in GBMSCs following ALKBH5 knockdown. We observed that IR induced an upregulation of Rad51 protein levels, which was attenuated by ALKBH5 inhibition ( Figure 4A). Phosphorylation of the Ser-139 residue of The recombinase Rad51 is the central protein involved in the homologous recombination of DNA during double-strand break repair [35]. Previous reports have confirmed a high expression of Rad51 in GBMSCs and its important role in resistance to radiotherapy [4,33,36,37]. As mentioned above, Rad51 mRNA expression can be regulated by ALKBH5. We then evaluated Rad51 levels by Western blot in GBMSCs following ALKBH5 knockdown. We observed that IR induced an upregulation of Rad51 protein levels, which was attenuated by ALKBH5 inhibition ( Figure 4A). Phosphorylation of the Ser-139 residue of the histone H2AX (γ-H2AX) is an early cellular response to DSBs and its quantification is a very sensitive method of detecting DSBs [38]. To confirm that ALKBH5 could affect the repair of IR-induced DSBs, we examined the expression levels of the DSBs marker γ-H2AX by Western blot analysis. As expected, γ-H2AX increased rapidly following IR (3 h) and returned to basal levels at 24 h in control cells, indicating an efficient DDR process in GBMSC neurospheres ( Figure 4B). However, in cells transfected with an ALKBH5 siRNA, the DNA damages persisted significantly 24 h post-IR ( Figure 4B), indicating that blocking ALKBH5 can delay DDR, improving the radiotherapy efficacy. The transcriptional factor FOXM1 is known to play an essential role in cell cycle, proliferation, DNA replication, and stem cell maintenance [39]. We have recently shown that FOXM1 is a key regulator of GBMSC radioresistance [27]. In addition, Zhang et al. reported the regulation of FOXM1 expression by ALKBH5 [14]. We therefore examined whether the activation of FOXM1 following IR could be regulated by ALKBH5. As shown in Figure 4C, we observed that IR induced an upregulation of FOXM1 protein levels, which was attenuated by ALKBH5 inhibition.
Cancers 2021, 13, x 7 of 17 the histone H2AX (γ-H2AX) is an early cellular response to DSBs and its quantification is a very sensitive method of detecting DSBs [38]. To confirm that ALKBH5 could affect the repair of IR-induced DSBs, we examined the expression levels of the DSBs marker γ-H2AX by Western blot analysis. As expected, γ-H2AX increased rapidly following IR (3 h) and returned to basal levels at 24 h in control cells, indicating an efficient DDR process in GBMSC neurospheres ( Figure 4B). However, in cells transfected with an ALKBH5 siRNA, the DNA damages persisted significantly 24 h post-IR ( Figure 4B), indicating that blocking ALKBH5 can delay DDR, improving the radiotherapy efficacy. The transcriptional factor FOXM1 is known to play an essential role in cell cycle, proliferation, DNA replication, and stem cell maintenance [39]. We have recently shown that FOXM1 is a key regulator of GBMSC radioresistance [27]. In addition, Zhang et al. reported the regulation of FOXM1 expression by ALKBH5 [14]. We therefore examined whether the activation of FOXM1 following IR could be regulated by ALKBH5. As shown in Figure 4C, we observed that IR induced an upregulation of FOXM1 protein levels, which was attenuated by ALKBH5 inhibition.

Blocking ALKBH5 Expression Represses Invasion of GBMSCs
The role of ALKBH5 in tumor cell invasion remains unclear. To investigate if ALKBH5 could be involved in this process in GBMSCs, we performed 3-D invasion assays in Matrigel, as described in the methods, using invasive primary neurospheres from three different patients in which ALKBH5 was knocked down with a specific siRNA. As shown in Figure 5, all GBMSC spheroids deficient for ALKBH5 exhibited a significant reduced invasion capability relative to control spheroids.

Blocking ALKBH5 Expression Represses Invasion of GBMSCs
The role of ALKBH5 in tumor cell invasion remains unclear. To investigate if ALKBH5 could be involved in this process in GBMSCs, we performed 3-D invasion assays in Matrigel, as described in the methods, using invasive primary neurospheres from three different patients in which ALKBH5 was knocked down with a specific siRNA. As shown in Figure 5, all GBMSC spheroids deficient for ALKBH5 exhibited a significant reduced invasion capability relative to control spheroids.  The transcriptional co-activator Yes-associated protein 1 (YAP1) is known to regulate the expression of target genes involved in the interactions between the extracellular matrix and cancer cells, and plays an important role in cellular invasion [40]. YAP1 is often highly activated in cancer stem cells and recently, Yu et al. demonstrated that YAP1 signaling is required for migration and invasion of GBMSCs [41].
We first analyzed the correlation between ALKBH5 and YAP1 expression using the glioblastoma database of the Cancer Genome Atlas, (TCGA). As shown in Figure 6A, a significant positive correlation was observed between the expression of the two genes. We therefore determined if ALKBH5 could regulate YAP1 expression in GBMSCs. Following ALKBH5 knockout in GBMSC with specific siRNA, we observed a significant inhibition of YAP1 mRNA expression compared to a scramble control ( Figure 6B). A high inhibition of YAP1, when ALKBH5 is downregulated, was also confirmed at the protein level by Western blot analysis ( Figure 6C). Surprisingly, the expression of YAP1 paralog TAZ (transcriptional co-activator with PDZ-binding motif) was not altered by ALKBH5 knockout (data not shown).
Verteporfin was originally designed as a photosensitizer; however, without photoactivation, it can disrupt the interaction and the transcriptional activity of YAP/TEAD and thus inhibit YAP-mediated effects, leading to reduced YAP1 signaling [42,43]. In addition, verteporfin was previously used in GBM to show the role of YAP1 on cell growth, survival, and resistance to chemotherapy [44][45][46]. First we performed a dose response curve to evaluate the effect of verteporfin on cell viability ( Figure 6D). The doses 1 and 2 µM which do not affect cell viability were chosen to assess the impact of verteporfin on cell invasion. Verteporfin was added to the medium of neurospheres before Matrigel solidification. 24 h after Verteporfin treatment at 1 and 2 µM, we observed a significant decrease of Matrigel invasion by GBM cells ( Figure 6E). To confirm the specific role of YAP1 in GBMSC, we also used YAP1 siRNA, previously validated in GC2 [28] in the 3D invasion assay and observed a similar decrease in Matrigel invasion by GBMSC ( Figure S1F).

Molecular Signatures Associated with High ALKBH5 Expression
In a previous study, Cui et al. reported low levels of m6A RNA modification in GBMSCs [12]. In addition, Zhang et al. showed that blocking ALKBH5 in GBMSCs, using an shRNA strategy, affects m6A global levels [14], suggesting that high expression of ALKBH5 might be more widely associated with general pathways involved in hallmarks of cancers. Analysis of the TCGA database based on ALKBH5 expression (the 25% highest versus 25% lowest expression as described in the methods) revealed the activation of several molecular pathways, including growth factors activity, inflammatory and immune response, and more particularly numerous molecular signatures related to cell motility and adhesion to the extracellular matrix (Table S1).

Discussion
The m6A mRNAs methylation in cancers is a very recent field of investigation. Methyltransferases METTL3/METTL14 and their co-factors, also known as m6A "writers", catalyze the formation of m6A whereas the demethylases FTO and ALKBH5, named m6A "erasers", selectively remove m6A modifications. Evidence is emerging that modulation of m6A RNA methylation plays an important role in different aspects of tumor biology, including growth, tumorigenesis, and self-renewal of cancer stem cells [7][8][9][10]. In brain tumors, the two m6A mRNA demethylases FTO and ALKBH5 have been clearly associated to pro-oncogenic functions, promoting cell proliferation, self-renewal, and the progression of GBMSC-initiated tumors [11][12][13][14][15]. In contrast, the role of the methyltransferases in GBM is more controversial. In accordance with the pro-oncogenic role of demethylases described above, Cui et al. reported a tumor-suppressor role of the methyltransferases METTL3/METTL4 by demonstrating that knocking down these enzymes enhances GBMSC growth and self-renewal [12]. In contrast Visvanathan described a pro-oncogenic role of METTL3 in glioma stem cells [47].
The role of ALKBH5 in radioresistance of GBM or other cancer types has never been reported. To our knowledge, this study is the first one demonstrating that this m6A mRNA demethylase is responsible for the resistance to radiotherapy. It is well demonstrated that GBMSCs contribute to glioma radioresistance through an increase in their DNA repair capacity [31][32][33]36]. Here, we show for the first time that key DNA damage repair players involved in homologous recombination, which is the repair pathway preferentially used by GBMSCs, can be downregulated by blocking ALKBH5, leading to GBMSC radiosensitization. This major finding highlights the potential of ALKBH5 as a therapeutic target in the treatment of GBM radioresistance since inhibiting this RNA demethylase can block multiple partners of the homologous recombination, including the key players of this pathway, RAD51 and CHK1.
Although several papers have reported that targeting RAD51 in GBMSCs using inhibitors or siRNA increased the radiosensitivity of GBMSCs [33,36,37,48,49], our study is the first demonstrating that blocking ALKBH5 expression in GBMSCs decreased both the basal and IR-induced expression of RAD51, and inhibited the ability of GBMSCs to repair DNA damages in response to IR, leading to radiosensitization of the cells. In addition to RAD51, we observed several other DNA repair genes induced by IR and involved in HR that were also downregulated by targeting ALKBH5 in GBMSCs. We observed a decrease in the BRCA2, BRIP1, EXO1, and XRCC2, 4 genes, whose high expression in GBM has been correlated with poor patient survival prognosis [50,51]. We also validated the checkpoint kinase 1 (CHK1) involved upstream of the HR pathway, as a target of ALKBH5. CHK1 has also been shown to be upregulated in GBMSCs and preferentially activated following IR. Inhibition of CHK1 increased IR-induced DNA damages and radiosensitivity of GBMSCs [4,52].
The results of our study are in accordance with the pro-tumoral role previously reported for m6A mRNA demethylases and demonstrate that ALKBH5 overexpression in GBMSCs may contribute to the efficiency of DNA damage repair, leading to radioresistance.
The role of RNA demethylases in radioresistance of cancer cells suggests that coadministration of RNA demethylases inhibitors with radiotherapy may offer a potential therapeutic approach to overcome GBM radioresistance. Several RNA demethylases inhibitors have already been developed with a potent anticancer effect. More particularly, FTO inhibitors, meclofenamic acid (MA), or its ethyl ester form (MA2) have been recently identified as specific FTO inhibitors that have shown antitumor activity on myeloid leukemia (AML) and GBM, prostate, and uterine cervical cancers [12,[53][54][55][56]. R-2-hydroxyglutarate (R-2HG), an oncometabolite resulting from the conversion of αketoglutarate by mutant IDH1/2 enzymes, has also been shown to specifically inhibit FTO and proliferation of leukemic or GBM cells [13]. More recently, a study also reported the efficacy of a specific ALKBH5 inhibitor, ALK04, in reducing tumor growth of melanoma cells [57]. Several other compounds, including (rhein, MO-I-500, Entacapone, NCDPCB, CHTB), have also been described as FTO or ALKBH5 potential inhibitors, although their anti-tumoral action has not been demonstrated [58]. Interestingly, Malacrida et al. recently reported that a new sodium channel blocker inhibited the migration of the GBM cell line U87. Although they do not demonstrate the role of ALKBH5 in the migration of the U87 cells, they showed an off-target effect of this sodium channel blocker on ALKBH5, suggesting that this compound may represent a new potential ALKBH5 inhibitor [59].
In addition to its role in radioresistance, we also demonstrated in this study that ALKBH5 contributes to the aggressiveness of GBM by favoring the invasion of GBMSCs. Indeed, GBMSC spheroids deficient for ALKBH5 exhibited a significant reduced invasion capability relative to control spheroids. This is the first report showing that ALKBH5 promotes GBMSC invasion. Our results emphasize the importance of ALKBH5 as a target to efficiently reduce GBMSC invasion. Indeed, blocking ALKBH5, which regulates the expression of both YAP1 and FOXM1, two transcription factors overexpressed in GBMSCs and previously involved in invasion, might have a higher impact on GBMSC invasion than blocking these two factors independently. In addition, the involvement of ALKBH5 in the invasive capability of GBM cells highlighted by our results is reinforced by the analysis of the TCGA database, which showed, associated with ALKBH5 high expression, several molecular signatures related to cell motility, adhesion, or extracellular matrix organization. These data suggest that in addition to YAP1 or FOXM1 that might contribute to ALKBH5-mediated invasion, numerous other factors related to cell invasion may be involved.
These results are in accordance with previous reports, which showed that the two m6A mRNA demethylases, ALKBH5 and FTO, promote cell proliferation and invasion in breast, gastric, and ovarian cancers [20,21,60,61]. However, it seems that ALKBH5 and FTO can have an opposite role in other cancer types, suggesting a complex regulation system. In particular, these demethylases suppress malignancy and invasion of pancreatic and colorectal cancers [24,25]. In addition, controversial results have been published in lung cancers. Whereas Jin et al. showed that ALKBH5 expression decreases tumor growth and metastasis by reducing YAP expression [62], other authors reported an increase in proliferation and invasion of lung cancer cells mediated by ALKBH5 or FTO [18,19,22].
In summary, our study highlights the crucial role of m6A mRNA demethylases in regulating radioresistance of GBM. Our results demonstrate that ALKBH5 inhibition, by decreasing HR, which is preferentially used by GBMSCs, might be a novel approach to radiosensitize GBMSC. In addition, we also showed that ALKBH5 might be an attractive therapeutic target to overcome invasiveness of GBMSCs.

GBM Patient-Derived Cells
GBM biopsy specimens were obtained from the Neurosurgery Department at Toulouse University Hospital as part of a clinical protocol (PI Pr. E. Cohen-Jonathan-Moyal) approved by the Human Research Ethics Committee (ethical code 12TETE01, ID-RCB number 2012-A00585-38, date of approval: 07-05-2012). Written informed consents were obtained from all the patients. Tumors were classified as GBM according to the WHO. All GBM used in this study were IDH1wt. The primary neurospheres were obtained from GBM specimens as described by Avril et al [63] and maintained in DMEM-F12 (GIBCO) supplemented with B27 and N2 (LifeTechnologies), FGF-2, and EGF (Peprotech) at 37 • C in a 5% CO2 incubator. Neurospheres were used during less than 12 passages to keep cell characteristics. The GBMSCs used in the study have been previously characterized [27][28][29][30].

Cells Irradiation
Cells were exposed to different doses of IR, as indicated, using an irradiator XRad Smart plus (Actemium NDT-Products and Systems, Le Plessy, France).

SiRNA Transfection, RNA Extraction, Reverse Transcription, and Real-Time PCR
The siRNA against ALKBH5(1) and YAP1 or the scramble control were purchased from Qiagen (Courtaboeuf, France) and the siRNA ALKBH5(2) was purchased from Ambion (Thermo Fisher, Illkirch, France). Transfection was performed with Lipofectamine RNAi Max (Invitrogen, Thermo Fisher, Illkirch, France) following the manufacturer's protocol. The RNeasy RNA isolation Kit (Qiagen, Courtaboeuf, France) was used to extract total RNA. Reverse transcription was performed using the Prime Script RT Reagent kit (TAKARA, St Germain en Laye, France). Real-time PCR was performed using the ABI-Stepone+ (Applied Biosystems, Thermo Fisher, Illkirch, France). Normalization was done with GAPDH.

Three-Dimensional Clonogenic Assay
Primary neurospheres from different patients expressing a specific siRNA (si-ALKBH5) or a scramble control (si-Scr) were seeded in 96-well plates (100 cells/wells, 12 wells per condition). After 24 h, cells were treated or not with different doses of gamma rays (2 to 6 Gy). Then, 8-10 days post-IR, the number of neurospheres/well with more than 50 cells was measured. The surviving fraction (SF) was calculated taking into account the plating efficiency in the non-irradiated condition (PE = spheres number/seeded cells number × 100).

Three-Dimensional Tumor Spheroid Invasion Assay
Three-dimensional tumor spheroid invasion assays were performed as described by Vinci M. et al. [64]. Briefly, 1000 to 2000 cells were seeded into ultra-low attachment 96-well round-bottom plates to allow formation of 3-D spheroid. Then, 72 h later, the spheroid were transfected with specific siRNA or a scramble control using Lipofectamine RNAi Max (Invitrogen, Thermo Fisher, Illkirch, France) following the manufacturer's protocol. Then, 24 h after, the transfection half of the growth medium was removed and replaced by the Matrigel. In the experiments with Verteporfin, 72 h after spheroid formation, the inhibitor was added to the medium just before addition of the Matrigel. After 1 h at 37 • C, when the Matrigel was solidified, growth medium was added to the wells. Images were taken at T0 and T24 h using a Nikon microscope and the Nikon software NIS Elements. The spheres area and the area covered by the invading cells was measured using the Image J software.

Genes Correlations
The correlations between ALKBH5 expression and CHK1 or YAP1 were obtained by the co-expression analysis in the TCGA database. Links between biomarkers were assessed using Spearman's rank correlation coefficients and their associated p-values. ALKBH5 expression was plotted on the x-axis, and the expression levels of other genes of interest were represented on the y-axis.

Molecular Signatures Associated with ALKBH5 Expression
The quantile normalized RNAseq gene expression of the glioblastomas cohort (GBM) was downloaded from TCGA. Only samples (n = 442) from G4 stage [65] were kept. Based on their ALKBH5 expression, the two groups of samples with the 25% lowest and the 25% highest expression were selected and defined (n = 220). The differential gene expression was tested using Wilcoxon or T-test after a normality test [66] and p-values were corrected using the Benjamini-Hochberg method [67]. Genes upregulated in the ALKBH5-high group with a log2 Fold Change > 0.5 were selected and compared to the database C5 [68] using Autocompare ZE with Zelen exact (ZE) p-value calculation [66,69].

Conclusions
The notion that m6A RNA methylation is an important process in the regulation of key genes involved in carcinogenesis is very recent. This epigenetic modification modulates mRNA expression, leading to pro-or anti-oncogenic effects. In GBM, demethylases are overexpressed, promoting GBMSC growth. In the present study, we showed that demethylases affect other important aspects of GBM development and recurrence: radioresistance and invasion of GBMSCs, suggesting that these enzymes might represent interesting targets in the treatment of GBM.