Epigenetic Deregulation of the Histone Methyltransferase KMT5B Contributes to Malignant Transformation in Glioblastoma

Glioblastoma multiforme (GBM) is the most common and aggressive type of brain tumor in adulthood. Epigenetic mechanisms are known to play a key role in GBM although the involvement of histone methyltransferase KMT5B and its mark H4K20me2 has remained largely unexplored. The present study shows that DNA hypermethylation and loss of DNA hydroxymethylation is associated with KMT5B downregulation and genome-wide reduction of H4K20me2 levels in a set of human GBM samples and cell lines as compared with non-tumoral specimens. Ectopic overexpression of KMT5B induced tumor suppressor-like features in vitro and in a mouse tumor xenograft model, as well as changes in the expression of several glioblastoma-related genes. H4K20me2 enrichment was found immediately upstream of the promoter regions of a subset of deregulated genes, thus suggesting a possible role for KMT5B in GBM through the epigenetic modulation of key target cancer genes.

Glioblastoma multiforme (GBM) is the most common and aggressive type of brain tumor in adulthood. Epigenetic mechanisms are known to play a key role in GBM although the involvement of histone methyltransferase KMT5B and its mark H4K20me2 has remained largely unexplored. The present study shows that DNA hypermethylation and loss of DNA hydroxymethylation is associated with KMT5B downregulation and genome-wide reduction of H4K20me2 levels in a set of human GBM samples and cell lines as compared with non-tumoral specimens. Ectopic overexpression of KMT5B induced tumor suppressor-like features in vitro and in a mouse tumor xenograft model, as well as changes in the expression of several glioblastoma-related genes. H4K20me2 enrichment was found immediately upstream of the promoter regions of a subset of deregulated genes, thus suggesting a possible role for KMT5B in GBM through the epigenetic modulation of key target cancer genes.

INTRODUCTION
Glioblastoma multiforme (GBM) is a malignant grade IV glioma which represents the most common and aggressive primary brain tumor among adults. The first line of treatment consists of surgical resection followed by radiotherapy and/or chemotherapy with the alkylating agent temozolomide (TMZ). Nevertheless, the median overall survival after treatment is only 14 months (Weller et al., 2015). Hence there is a critical need to gain a deeper understanding of this frequently fatal and heterogeneous brain cancer, and search for new therapeutic targets.
During tumorigenesis, epigenetic alterations affect different types of cancer-related genes involved in cell cycle control, DNA repair, apoptosis, or cell signaling (Feinberg et al., 2016), among others. Moreover, they can also occur at epigenetic regulator genes, thereby triggering a chain of genome-wide massive epigenetic alterations. As examples of the latter are, on the one hand, epigenetic disruption through the downregulation of different micro RNAs that modulate histone deacetylases (HDACs) expression has been reported in several cancers, such as hepatocellular carcinoma (Yuan et al., 2011), lymphoplasmacytic lymphoma (Roccaro et al., 2010), tongue squamous cell carcinoma (Kai et al., 2014), and breast cancer (Wu et al., 2014). On the other hand, disruption of the methylation signature at the promoter of genes encoding histone deacetylases HDAC4, HDAC5, and HDAC9 has been described in papillary thyroid carcinoma (White et al., 2016). In the context of neuroblastoma and glioma, epigenetic inactivation via hypermethylation at promoter CpG island (CpGi) of the histone methyltransferase gene NSD1 has been defined as the mechanism responsible for the altered histone methylation landscape observed in both types of tumor (Berdasco et al., 2009). In the same vein, it has been shown that DNA methyltransferases (DNMTs) encoded by DNMT1 and DNMT3B are overexpressed in glioma due to, respectively, an aberrant histone code or an aberrant methylation pattern at their promoters (Rajendran et al., 2011). Recently, our laboratory has demonstrated that the expression of the epigenetic enzyme TET3 is aberrantly downregulated in GBM through epigenetic mechanisms, leading to a genome-wide reduction of 5-hydroxymethylcytosine (5hmC) levels (Carella et al., 2020).
Thus, besides genetic features (Brennan et al., 2013), DNA methylation/hydroxymethylation and histone modifications are major epigenetic mechanisms regulating gene expression and genomic stability whose alterations have been widely described in GBM (Hegi et al., 2005;Fernandez et al., 2018;García et al., 2018;Carella et al., 2020). In addition, the brain has a unique DNA methylation/hydroxymethylation landscape, characterized by the highest levels of 5-methylcytosine (5mC) and 5hmC of all human tissues (Nestor et al., 2012). In general terms, 5mC in promoters contributes to create a repressive chromatin environment and decreases gene expression, while 5hmC located within the body of genes may activate transcription (Chen et al., 2016). The balance between the two marks modulates a plethora of biological processes and has been extensively reported to be dysregulated in GBM (Johnson et al., 2016;López et al., 2017;Fernandez et al., 2018;García et al., 2018;Carella et al., 2020). In addition to the epigenetic marks of DNA, many studies have addressed how the modulation of chromatin structure via histone modifications is affected in the context of GBM (Liu et al., 2010). They have mostly focused on the function of histone modifying enzymes, specifically histone lysine demethylases (KDMs) and histone deacetylases (HDACs), and the use of different inhibitors (Singh et al., 2015).
However, to date little attention has been paid to the role of histone lysine methyltransferases (KMTs) in GBM (Gursoy-Yuzugullu et al., 2017). Methylation at lysine 20 of histone H4 (H4K20me) is frequently altered in cancer (Behbahani et al., 2012). However, the contribution of the epigenetic enzymes regulating these histone posttranslational modifications is still not fully understood. A preliminary genome-wide screening of candidate genes altered in GBM revealed that the gene encoding the histone methyltransferase KMT5B (alias SUV420H1) is frequently hypermethylated and hypo-hydroxymethylated in this tumor type. KMT5B employs H4K20me1 as a substrate, giving rise to H4K20me2 . In the present study we investigated the molecular basis of the deregulation of this epigenetic enzyme as well as its functional consequences and its possible tumoral role in GBM.

Human Brain and Glioblastoma Samples
Non-tumoral human brain (n = 28) and GBM samples (n = 37) were obtained from the Tumor Bank of the Institute of Oncology of Asturias (Asturias, Spain), the Neurological Tissue Bank of the Clinic Hospital (IDIBAPS, Barcelona, Spain), and the Biobank of the Virgen de la Salud Hospital (BioB-HVS, Toledo, Spain). Informed written consent was obtained from all patients involved in this study. Tissue collection and all analyses were conducted in accordance with the Declaration of Helsinki and approved by the Clinical Research Ethics Committee of the Principality of Asturias (Spain) (date of approval 14/10/2013, project identification code: 116/13).

DNA Methylation Analysis With High-Density Array and Whole Genome Bisulfite Sequencing
Data corresponding to the oxBS and BS conversion of brain or GBM samples were obtained from our previously reported HumanMethylation 450K array (E-MTAB-6003) (Fernandez et al., 2018). Processed WGBS data were obtained from Arrayexpress (E-MTAB-5171) (Raiber et al., 2017). Both datasets were filtered to highlight the CpG coverage along the KMT5B gene. Estimations of 5mC and 5hmC levels were calculated with the R/bioconductor package ENmix (version 1.12.3) using the oxBS.MLE method (Xu et al., 2016).

DNA Extraction
DNA was extracted from freshly frozen GBM, control brains and human GBM cell lines with standard phenol-chloroformisoamyl alcohol protocol. Samples were stored at −20 • C until further analysis.

Bisulfite Pyrosequencing
Bisulfite modification of DNA was performed with the EZ DNA methylation-gold kit (Zymo Research) following the manufacturer's instructions. The set of primers for PCR amplification and sequencing was designed using the specific software PyroMark assay design (version 2.0.01.15) (Supplementary Table 1). After PCR amplification, pyrosequencing was performed using Pyro-Mark Q24 reagents, vacuum prep workstation, equipment, and software (Qiagen).

5-Hydroxymethylcytosine Immunoprecipitation and Quantitative Real-Time Polymerase Chain Reaction Assay
EpiQuik Hydroxymethylated DNA Immunoprecipitation Kit (hMeDIP, Epigentek) was used for immunoprecipitation of 5hmC in 11 samples (corresponding to 5 brain samples, 5 GBM samples and LN-229 GBM cell line), according to supplier's protocol. Input, non-specific IgG and 5hmC-enriched

KMT5B Transfection
LN-229 cells were stably transfected with either expression vectors encoding human KMT5B (pEGFP-KMT5B) or with empty vector (pEGFP-C1), kindly provided by Professor Miki Hieda (Ehime Prefectural University of Health Sciences, Matsuyama, Japan) and constructed as described previously (Yokoyama et al., 2014), using Lipofectamine 3000 (Invitrogen, L3000015) according to the manufacturer's instructions. Geneticin-resistant cells were selected 72 h after transfection with 1 mg/mL G-418 (Sigma-Aldrich, A-1720), after which subcloning was carried out by limiting dilution, and several clones overexpressing KMT5B were obtained. KMT5B expression was tested in positive clones by qRT-PCR using specific primers for KMT5B transcription variant 1 detection (Supplementary Table 1).
DNA was purified and used for quantitative real-time PCR with SYBR Green 2X PCR Master Mix (Applied Biosystems) and the primers for the gene loci listed in Supplementary Table 1. Input chromatin DNA was used to create a standard curve and determine the efficiency of amplification for each primer set in a StepOnePlus TM Real-Time PCR machine (Applied Biosystems). All samples were measured in triplicate. IgG was used as negative control. ChIP data were analyzed and are shown in the results as the percentage relative to the input DNA amount by the equation:

Immunohistochemistry
The immunohistochemical analyses of KMT5B protein levels in three GBM samples and paired-normal brain tissue were performed using the EnVision FLEX Mini Kit (DAKO, K8024) and Dako Autostainer system. Briefly, paraffin embedded tissue (3-5 µm) was deparaffinized, rehydrated, and then epitopes were retrieved by heat induction (HIER) at 95 • C for 20 min at pH 6.0 (DAKO, GV805) in the Pre-Treatment Module, PT-LINK (DAKO). Sections were incubated with anti-KMT5B antibody (sc-169462, Santa Cruz) at 1:100 dilution in EnVision TM FLEX Antibody Diluent (DAKO, K8006) for 60 min after blocking endogenous peroxidase with EnVision TM FLEX Peroxidase-Blocking Reagent (DAKO, DM821). Signal was detected using diaminobenzidine chromogen as substrate after incubation with Dako EnVision TM FLEX/HRP (DAKO, DM822). Sections were counterstained with hematoxylin. Appropriate negative and positive controls were also tested. After the complete process, sections were dehydrated and mounted with permanent medium (Dako mounting medium, CS703). Images were captured using a Nikon Eclipse Ci microscope equipped with a 20x objective and a camera (Nikon Instruments Europe B. V.).

Cell Proliferation Assay
Cell proliferation rates were measured using xCELLigence Real Time Cell Analyzer (RTCA, ACEA Biosciences). Quadruplicates of mock or stably-transfected LN-229 clone cells (15 × 10 3 ) were seeded onto analyzer specific plates. Cell impedance was measured every 2 h for eight consecutive days through microelectric biosensors located at the base of the plate wells. Cell proliferation was represented by Cell Index, slope, and doubling time parameters.

Colony Formation Assay
Colony formation assays of LN-229 clones stably-transfected with KMT5B or empty vectors were conducted as described by Franken et al. (2006), with minor modifications. Briefly, cells were seeded at two different densities (100 and 200 cells) onto P-6 wells. After incubation for 14 days, stable colonies were fixed and stained with a mixture of 6.0% glutaraldehyde (Sigma-Aldrich, 340855) and 0.5% crystal violet (Sigma-Aldrich, C3886). After washing, the number of colonies was counted manually. The results were expressed as mean Plating Efficiency (PE), a parameter which represents the ratio between the number of grown colonies versus the number of cells seeded at each different density.

Statistical Analysis
All statistical analyses were conducted using the R statistical programming language (version 3.4.0). Specific analyses are described in the corresponding section.

KMT5B Shows Altered 5mC/5hmC Pattern in GBM
On the basis of data from our previously reported 450 K Illumina methylation array of 11 GBM and 5 non-tumoral brain samples (Fernandez et al., 2018), we performed a genome-wide screening of candidate genes which are altered in GBM. We found that KMT5B, a gene encoding a histone methyltransferase, had altered DNA methylation and hydroxymethylation patterns. Specifically, compared to healthy brain tissue, GBM showed higher levels of 5mC and greatly reduced levels of 5hmC in the CpGi at the promoter and several CpG sites within the KMT5B gene body (Figure 1A and Supplementary Table 3).
Comparison with publicly available whole genome bisulfite sequencing (WGBS) data, with a deeper CpG coverage along the KMT5B locus (Raiber et al., 2017), confirmed a 5mC/5hmC imbalance in GBM as compared to non-tumoral brain ( Figure 1B and Supplementary Table 4).
We validated 5mC array data by means of the bisulfite pyrosequencing of five representative CpG sites at KMT5B promoter in the same specimens included in the array, as well as in a larger set of samples consisting of: 27 GBM, 4 non-tumoral brains, and 4 human GBM cell lines. Pyrosequencing results confirmed the 5mC pattern observed in the array (Figure 1C and Supplementary Table 5). With respect to 5hmC array data, we validated those results using a different technique that relied not on oxidative bisulfite conversion, but on DNA immunoprecipitation with an antibody against 5hmC. We used an hMeDIP Kit (see "Materials and Methods" section) with 5 GBM, 5 non-tumoral brain samples and the human GBM cell line LN-229. qRT-PCR was used to analyse the samples and the results confirmed the extensive loss of 5hmC affecting the KMT5B locus in GBM (Wilcoxon rank sum test, * * p < 0.01; Figure 1D).

KMT5B Expression Is Downregulated in a Subset of Human GBM Samples and in GBM Cell Lines
Next, we wanted to investigate whether the aberrant methylation landscape of KMT5B could have an effect at the transcriptional regulation level. In order to determine the expression status of KMT5B in GBM, we measured its mRNA levels by means of qRT-PCR in 13 GBM samples, four human GBM cell lines, and 23 non-tumoral control brains (seven corresponding to gray matter and 16 to white matter samples). The results showed that KMT5B mRNA expression was significantly downregulated in a subset of GBM samples compared with non-tumoral brains ( Figure 1E). Comparison with gene expression array data obtained from different brain datasets in ArrayExpress (Günther et al., 2008;Sim et al., 2009;Grzmil et al., 2011) confirmed that downregulation of KMT5B affects a subset of GBMs (Supplementary Figure 1

, Supplementary Table 8, and Supplementary Methods).
To validate our observations at the protein level, we subjected three GBM samples and their paired non-tumoral brain tissue to immunohistochemistry (IHC). The results were in line with the mRNA expression data and revealed that KMT5B protein expression was high in non-tumoral brain tissues and low to absent in GBM ( Figure 1F).
These findings point to a potential epigenetic silencing of KMT5B through the methylation of the promoter and hypohydroxymethylation of the gene body in GBM.

Correlation Between Epigenetic Regulation and KMT5B Gene Expression
To further explore the possible functional link between the epigenetic alterations described above and KMT5B repression, we FIGURE 1 | 5mC and 5hmC profiling along the KMT5B locus in brain and glioblastoma multiforme (GBM) samples and analysis of KMT5B expression in brain, GBM and GBM cell lines. (A) Line plots represent percentage of average 5mC (upper panel) and 5hmC values (lower panel) corresponding to 25 CpG positions in the KMT5B gene for five control brain and eleven GBM samples, depicted in blue and violet, respectively. Data were obtained from a genome-wide Illumina HumanMethylation 450 K array (E-MTAB-6003). Welch t-tests were applied for each CpG and those that were significant are depicted in red (q-value < 0.05). (B) WGBS dataset with high content profiling of 5mC and 5hmC (accession number E-MTAB-5171) confirmed the altered 5mC/5hmC pattern of KMT5B. CpGs depicted in red represent those with changes =15%. Lower diagrams in panels (A,B) represent the relative location of the analyzed CpG sites along KMT5B locus. (C,D) Biological validation of the methylation and hydroxymethylation values in five KMT5B promoter CpGs using, respectively, bisulfite pyrosequencing and hMeDIP. The gray dotted line in panel (D) indicates 5hmC levels in GBM cell line LN-229. Wilcoxon rank sum tests were applied (*p < 0.05; **p < 0.01; ***p < 0.001). (E) Box plot represents KMT5B mRNA relative expression in 16 white matter and 7 gray matter control brains, 13 GBM samples and 4 GBM cell lines, measured by qRT-PCR in relation to GAPDH and represented as Fold Change relative to white-matter control brains. Wilcoxon rank sum tests were applied (*p < 0.05; **p < 0.01; ***p < 0.001). (F) Representative images of KMT5B protein levels detected in GBM samples and their paired non-tumoral brain tissue by immunohistochemistry (IHC), shown at 20× magnification. Only non-tumoral brain sections showed positive KMT5B nuclear staining. performed a pharmacological restoration of 5mC/5hmC levels in human GBM cell line LN-229. To this end, we incubated LN-229 cells with 5-aza-2 -deoxycytidine (5-AZA-dC) either alone or together with vitamin C. 5-AZA-dC is an epigenetic drug, widely used in clinical practice, that decreases 5mC levels through DNA methyl-transferases (DNMTs) inhibition, and increases 5hmC levels through upregulating the expression of Ten-Eleven-Translocation proteins (TETs) (Sajadian et al., 2015). Vitamin C, a TET cofactor, has been shown to increase intragenic 5hmC levels and to act synergistically with 5-AZA-dC in vitro (Sajadian et al., 2016). The treatments carried out here caused a significant decrease in 5mC and an increase in 5hmC levels, as confirmed at two intragenic CpG sites (CpG in the array CG27086672 and the next downstream CpG position) by means of bisulfite pyrosequencing (Wilcoxon rank sum tests, * p < 0.05; * * p < 0.01; Figure 2A and Supplementary Table 5). At the mRNA level, KMT5B expression was also upregulated, especially when both drugs were present (Welch t-test, * * p < 0.01; Figure 2B).
Taken together, these data indicate that aberrant epigenetic regulation contributes to KMT5B repression in GBM.

Ectopic Expression of KMT5B Reduces the Tumorigenic Potential of Glioblastoma Cells in vitro
To delve into the possible role of KMT5B epigenetic repression in GBM, we stably transfected either a KMT5B expression plasmid or an empty vector (mock) into the human GBM cell line LN-229, which exhibits hypo-hydroxymethylation and downregulation of KMT5B. Two clones, namely KMT5B #3 and KMT5B #7, were selected among those that were drug resistant, and the expression of KMT5B was confirmed by qRT-PCR ( Figure 3A). Impedancebased proliferation assays showed that KMT5B-transfected clones grew less than control cells ( Figure 3B). Overexpression of KMT5B also reduced the clonogenicity of the clones compared with the mock transfected cells (Figure 3C). In addition, an MTT assay demonstrated that KMT5B overexpression caused a significant decrease in cell viability ( Figure 3D). Cell cycle analysis confirmed that KMT5B overexpression induced G2/M arrest, as was demonstrated previously by Evertts et al. (2013), thus explaining the decreased cell proliferation and viability of the overexpressing clones (Supplementary Figure 2).
As expected, the increased expression of KMT5B elevated the levels of its histone mark H4K20me2, concomitantly reducing those of H4K20me1, as seen by immunofluorescence (IF) in stably-transfected clones versus mock (Figures 4A,B and Supplementary Figures 3A,B). Conversely, inhibition of KMT5B enzymatic activity with the selective drug A-196 (Bromberg et al., 2017) increased monomethylated and decreased dimethylated forms of H4K20 in stably-transfected clones, thus recapitulating the histone H4K20 methylation landscape of LN-229 GBM cell line (Figures 4C,D and Supplementary Figures 3C,D).

KMT5B Overexpression Reduces Tumor Growth in Nude Mice
In light of the above data, we sought to further characterize the role of KMT5B expression in GBM in vivo. To this end, we subcutaneously injected stably-transfected KMT5B or mock LN-229 cells into nude mice. Tumor volume was notably smaller in mice flanks receiving KMT5B-overexpressing LN-229 clones compared to those injected with empty vector transfected cells (General linear model, * * * p-value < 0.001; Figure 5A). At the time of sacrifice, 56 days after tumor-xenograft implantation, the average weight of tumors generated by KMT5B-overexpressing clones was significantly lower than those from control cells (Wilcoxon rank sum tests; * p < 0.05; * * p < 0.01; Figure 5B). The tumor xenografts thereby revealed that KMT5B overexpression also slows tumorigenesis in vivo, suggesting that KMT5B could play a possible role as a putative tumor suppressor in GBM.

Overexpression of KMT5B Affects Genome-Wide Gene Expression in GBM Cells
In an attempt to shed light on the molecular pathways and mechanisms by which KMT5B decreases GBM tumorigenesis, we subjected our stably-transfected overexpressing or mock LN-229 clones to RNAseq (Supplementary Table 6). KMT5B has been proposed to impact transcriptional regulation by means of its product H4K20me2 (Miao and Natarajan, 2005;Vakoc et al., 2006). Our results showed that overexpression of KMT5B changed the transcriptome of GBM cells through up-or downregulation of hundreds of genes (Figure 6 and Supplementary Table 7). Specifically, for clone KMT5B#3 a total of 725 genes were significantly affected, corresponding to 333 upregulated and 392 downregulated ones. Regarding KMT5B#7, expression changes implicated a total of 962 genes of which 484 were upregulated and 478 downregulated. The Venn diagram in Figure 6B represents the overlap of differentially expressed genes in KMT5B clones as compared to the LN-229 control cells (mock). The two clones shared a total of 315 differentially expressed genes. Among the upregulated ones, CDH11 (cadherin-11) has been associated with blocking the invasion of GBM cells in vitro (Delic et al., 2012). Interestingly, several of those genes had been previously described to be overexpressed and implicated in GBM tumorigenesis, including: CA9 (carbonic anhydrase 9) (Boyd et al., 2017), PDL1 (Programmed Death Ligand 1, also known as CD274) (Jacobs et al., 2009), and IL13RA2 (Interleukin 13 Receptor Subunit Alpha 2). We focused on IL13RA2 as this gene encodes a monomeric IL4-independent and high grade glioma-associated IL13 receptor (Mintz et al., 2002). IL13RA2 is overexpressed in most patients with GBM but not in normal brain (Brown et al., 2013). As expected, the gene-disease associations platform DisGeNET related the abovementioned genes with malignant glioma. Interestingly, it also FIGURE 4 | Immunofluorescence analysis of mono-and dimethylation of H4K20. Representative images of mock and KMT5B-transfected clones immunostained with (A) anti-H4K20me1 or (B) anti-H4K20me2 antibodies (green). In panels (C,D) representative images of KMT5B-transfected clones immunostained with (C) anti-H4K20me1 or (D) anti-H4K20me2 antibodies (green) after treating the cells with KMT5B inhibitor A-196 for 48 h. Nuclei were counterstained with DAPI (blue). Bars: 10 µm.
highlighted the association of certain other genes with both malignant glioma and kidney diseases, thus pointing to a yet-undescribed hypothetical role of KMT5B in such renal pathologies ( Figure 6C).
To verify the RNAseq results and assess the involvement of KMT5B and H4K20 methylation in the regulation of IL13RA2 and CDH11 expression, we performed a ChIP assay for H4K20me2 in our clones. The analysis of immunoprecipitated DNA by qRT-PCR demonstrated that H4K20me2 is enriched immediately upstream of the IL13RA2 and CDH11 transcription start sites in the KMT5B-overexpressing clones, but not in mock-transfected GBM cells ( Figure 6D). Additionally, we conducted pairwise correlation analyses between KMT5B and candidate genes using the cBioPortal platform (Cerami et al., 2012) with microarray data obtained from the comprehensive characterization of glioblastoma performed by the TCGA consortium (Cancer Genome Atlas Research Network, 2008). Considering all the samples from the GBM dataset (n = 206), KMT5B mRNA expression showed inverse correlation with IL13RA2 and direct correlation with CDH11, which is in line with our RNAseq and qChIP results (Supplementary Figure 4, upper panels). When grouping the samples by their corresponding GBM gene expression profile (Classical, Mesenchymal, Neural, and Proneural), we observed a similar trend in correlations across all subtypes, and they were particularly significant in the neural GBM subtype (Supplementary Figure 4). Altogether, these data suggest that IL13RA2 and CDH11 could represent direct targets of KMT5B-mediated gene expression regulation.

DISCUSSION
The involvement of KMTs in GBM pathogenesis has not as yet been fully elucidated. Referring back to our data from a 450K Infinium Illumina methylation array reported in 2018 (Fernandez et al., 2018), we found that the KMT5B gene shows an altered DNA methylation and hydroxymethylation profile in GBM as compared to non-tumoral brain samples. The present study aimed to investigate the contribution of this chromatin modifying gene to the etiology of GBM. We obtained evidence that it could exert an anti-tumoral effect, likely through its histone mark H4K20me2.
Expression analyses revealed that KMT5B hypermethylation and hypo-hydroxymethylation correlated with its decreased expression in some of our GBM specimens in comparison with normal brain tissue strongly indicating that the epigenetic downregulation of KMT5B might be relevant for a subset of GBM tumors. In fact, gene inactivation by DNA hypermethylation of CpGis at gene promoters is a well-established epigenetic hallmark in cancer (Koch et al., 2018). As for DNA hydroxymethylation, its presence at gene bodies is positively correlated with gene expression (Jin et al., 2011b), while its loss has been described in several cancers, including GBM (Jin et al., 2011a). This was reinforced in this work through the pharmacological reestablishment of 5mC and 5hmC levels, using a DNMTs inhibitor and a TETs cofactor, which successfully rescued KMT5B expression in our human GBM cell line LN-229. Nevertheless, unlike Sajadian et al. (2016), we did not find differences in the magnitude of changes in 5mC and 5hmC levels using 5-AZA-dC alone or when in combination with vitamin C. This apparent discrepancy might be explained by the fact that we specifically evaluated the methylation/hydroxymethylation state of two intragenic CpGs rather than considering global levels of those marks as in the study cited.
At the functional level, overexpression KMT5B reduced cell proliferation, cell viability and clonogenic potential in vitro, and tumor growth in vivo in mice xenografts. Interestingly, the loss of proliferative capacity induced by KMT5B overexpression in our clones was not related to apoptotic mechanisms, but rather to a G2/M cell cycle blockade. This is in agreement with a previous study showing that KMT5B overexpression causes G2 arrest (Evertts et al., 2013), and is consistent with the widely described role of methylated forms of H4K20 in cell cycle progression .
Regarding the histone mark deposited by KMT5B, H4K20me2, its levels were confirmed to be diminished in mock-transfected GBM cells, and to recover upon KMT5B expression in stably-transfected clones, as demonstrated through immunohistochemical analyses. Concurrently, levels of the histone mark H4K20me1, which serves as KMT5B substrate, varied according to the expression of this enzyme, i.e., they were very low in KMT5B-expressing clones and higher in mocktransfected ones. We further confirmed the H4K20 patterns by treating the cells with A-196, a drug which causes KMT5B enzymatic inhibition. Indeed, the treatment reverted the levels of both marks in KMT5B stably-transfected clones, increasing H4K20me1 and decreasing H4K20me2 to a comparable extent as seen in LN-229 GBM cells.
Furthermore, ectopic expression of KMT5B was associated with transcriptional changes of several GBM-related genes, as seen by the results of RNAseq analyses. The role of H4K20 monomethylation in gene expression has been controversial for some time due to studies associating this mark with both transcriptional repression (Nishioka et al., 2002) and activation (Talasz et al., 2005;Vakoc et al., 2006). However, more recent works and all the genome-wide ChIP analyses to date have rendered compelling results linking this mark with transcription activation (Barski et al., 2007;Li et al., 2011) and, indeed, it is one of the histone modifications that is most strongly correlated with active transcription (Barski et al., 2007;Karliae et al., 2010;Veloso et al., 2014). By contrast, to date, the function of H4K20 dimethylation in gene expression has scarcely been addressed. That said, studies carried out in human cells suggest that H4K20me2 could be related to gene silencing (Miao and Natarajan, 2005;Vakoc et al., 2006;Gwak et al., 2016). The results presented here highlight the importance of KMT5B epigenetic downregulation in GBM, which leads to a distorted H4K20 methylation pattern. This is in line with the roles for  Supplementary Table 2). DNA eluted from the ChIP assay was amplified by qRT-PCR with primer sets mapping to IL13RA2 or CDH11 promoters (listed in Supplementary Table 1). The enrichment of H4K20me2 at the indicated regions in KMT5B clones was compared to the same regions in mock control cells (Welch t-tests; *p < 0.05; **p < 0.01; ***p < 0.001).
H4K20me1 and H4K20me2 in gene expression mentioned above, and tie in with the concept of epigenetic de-repression of oncogenes. While it is well-known that the epigenetic silencing of tumor suppressor genes occurs during tumorogenesis, the role of epigenetic de-repression of oncogenes due to the loss of repressive chromatin marks has remained partially unexplored . Our experiments showed that the epigenetic downregulation of KMT5B impairs the balance between H4K20me species in human GBM cell line LN-229. In turn, this may cause global transcriptomic changes, given our finding that overexpression of KMT5B restored H4K20me2 patterns and decreased the expression of several oncogenes, such as IL13RA2, a tumor-restricted receptor (Mintz et al., 2002). IL13 is a pleiotropic Th2-derived cytokine implicated in inflammation and immunomodulation, and is known to induce apoptosis in different cell types, including GBM (Hsi et al., 2011). IL13RA2 binds IL13 with a higher affinity than the physiological IL13RA1/IL4RA receptor but prevents apoptosis as it does not transduce the signal for STAT6 pathway activation (Kawakami et al., 2001). Although the functional significance of IL13RA2 expression is not fully understood, it is considered an inhibitory or decoy receptor in GBM (Rahaman et al., 2002) that contributes to tumor growth (Hsi et al., 2011) and several studies have found that its overexpression is associated with higher glioma grades and poor patient prognosis (Brown et al., 2013;Han and Puri, 2018). Investigations in pancreatic, colorectal, and ovarian cancer have shown that IL13RA2 overexpression promotes tumor migration and invasion (Fujisawa et al., 2009(Fujisawa et al., , 2012Barderas et al., 2012). More recent studies in GBM have also pointed to a role for IL13RA2 in stimulating cell growth and metastasis (Tu et al., 2016). As such, IL13RA2 has become an attractive therapeutic target in GBM, even though the regulatory mechanisms of its expression are currently unknown (Sharma and Debinski, 2018). By means of RNAseq and ChIP experiments, our study suggests that KMT5B and its mark H4K20me2 might be involved in IL13RA2 regulation since overexpression of KMT5B in our clones caused H4K20me2 deposition upstream of the IL13RA2 promoter and a concomitant downregulation of IL13RA2 mRNA.
Notwithstanding, we observed that some genes were upregulated upon KMT5B overexpression. Among them CDH11, which encodes Cadherin 11, has been previously reported to block GBM cell invasion in vitro (Delic et al., 2012). One plausible explanation is that H4K20me2 might function as both transcriptional activator or repressor depending on the cellular context, in concert with other associated histone modifications and readers. That dual role on gene expression modulation would fit with the abundance and ubiquitous distribution of this mark on chromatin Schotta et al., 2008). Future research is guaranteed in order to elucidate the context dependant function of H4K20me2. Consequently, it is possible that the anti-tumoral effect of KMT5B observed in our study would not be due to the overexpression of this epigenetic modifier itself, but rather might be mediated by the re-establishment of levels of its product H4K20me2 in the promoter region of key GBM-related genes, such as IL13RA2 and CDH11.
The dynamic and reversible nature of H4K20me2 opens the possibility of targeting its demethylase (KDM), via smallmolecule inhibitors, to restore this histone mark in GBM. Indeed, in recent years, histone modifiers, such as KDMs, are envisaged as attractive druggable targets for GBM therapy . Although H4K20me2 is the most abundant form of H4K20 in eukaryotic cells , only a H4K20me2 demethylase has been described so far. It is LSD1n, a neuronspecific splicing variant of the flavin-dependent monoamine oxidase LSD1 with acquired new specificity against H4K20me2 and H4K20me1 (Wang et al., 2015). However, caution is needed when considering LSD1n as a potential therapeutic target using a hypothetical specific KDM inhibitor. On the one hand, there is increasing evidence that histone modifications do not act alone, but influence one another. Hence, modulation of H4K20me2 levels via a LSD1n inhibitor could potentially affect the modification status of other histone residues on the vicinity with unknown effects. On the other hand, non-histone targets of LSD1n should also be taken into account. Moreover, it has been reported that LSD1n promotes neuronal activity-regulated gene expression and plays a critical role in neurite morphogenesis, spatial learning, long-term memory formation, and modulation of emotional behavior (Zibetti et al., 2010;Wang et al., 2015;Rusconi et al., 2016). All of the above reveals the challenging pleiotropic consequences and the complexity of the epigenetic regulation via KDM inhibitors within CNS. In the future, a deeper understanding of the roles and LSD1n targets other than histones, as well as the transcriptional networks regulated by H4K20me2, or even the discovery of other yet unknown specific demethylases, could boost the development of novel specific KDM inhibitors and inform novel strategies in cancer therapy.
Taken together, our results suggest that when KMT5B expression and its histone mark H4K20me2 are perturbed in GBM through deregulated epigenetic mechanisms, this process has a genome-wide effect on gene transcription and may contribute to malignant transformation.

DATA AVAILABILITY STATEMENT
The datasets presented in this study can be found in online repositories. The names of the repository/repositories and accession number(s) can be found below: ENA, PRJEB39613.

ETHICS STATEMENT
The studies involving human participants were reviewed and approved by Clinical Research Ethics Committee of the Principality of Asturias (Spain). The patients/participants provided their written informed consent to participate in this study. The animal study was reviewed and approved by Animal Ethics Committee of University of Oviedo (Asturias, Spain).