Acta Neuropathologica (2019) 138:1053–1074 https://doi.org/10.1007/s00401-019-02062-4 ORIGINAL PAPER Epigenetic loss of RNA‑methyltransferase NSUN5 in glioma targets ribosomes to drive a stress adaptive translational program Maxime Janin1 · Vanessa Ortiz‑Barahona1 · Manuel Castro de Moura1 · Anna Martínez‑Cardús1 · Pere Llinàs‑Arias1 · Marta Soler1 · Daphna Nachmani2 · Joffrey Pelletier3 · Ulrike Schumann4 · Maria E. Calleja‑Cervantes1 · Sebastian Moran1 · Sonia Guil1 · Alberto Bueno‑Costa1 · David Piñeyro1 · Montserrat Perez‑Salvia1 · Margalida Rosselló‑Tortella1 · Laia Piqué1 · Joan J. Bech‑Serra5 · Carolina De La Torre5 · August Vidal6,7 · María Martínez‑Iniesta8 · Juan F. Martín‑Tejera8 · Alberto Villanueva8 · Alexandra Arias9 · Isabel Cuartas9 · Ana M. Aransay10 · Andres Morales La Madrid11,12 · Angel M. Carcaboso13 · Vicente Santa‑Maria11,12 · Jaume Mora11 · Agustin F. Fernandez14 · Mario F. Fraga15 · Iban Aldecoa16 · Leire Pedrosa17 · Francesc Graus18 · Noemi Vidal6 · Fina Martínez‑Soler19 · Avelina Tortosa19 · Cristina Carrato20 · Carme Balañá21 · Matthew W. Boudreau22 · Paul J. Hergenrother22 · Peter Kötter23 · Karl‑Dieter Entian23 · Jürgen Hench24 · Stephan Frank24 · Sheila Mansouri25 · Gelareh Zadeh25 · Pablo D. Dans26 · Modesto Orozco26,27 · George Thomas3,28,29 · Sandra Blanco30,31 · Joan Seoane7,9,32 · Thomas Preiss4,33 · Pier Paolo Pandolfi2 · Manel Esteller1,7,27,32,34 Received: 20 May 2019 / Revised: 8 August 2019 / Accepted: 13 August 2019 / Published online: 19 August 2019 © The Author(s) 2019 Abstract Tumors have aberrant proteomes that often do not match their corresponding transcriptome profiles. One possible cause of this discrepancy is the existence of aberrant RNA modification landscapes in the so-called epitranscriptome. Here, we report that human glioma cells undergo DNA methylation-associated epigenetic silencing of NSUN5, a candidate RNA methyl- transferase for 5-methylcytosine. In this setting, NSUN5 exhibits tumor-suppressor characteristics in vivo glioma models. We also found that NSUN5 loss generates an unmethylated status at the C3782 position of 28S rRNA that drives an overall depletion of protein synthesis, and leads to the emergence of an adaptive translational program for survival under condi- tions of cellular stress. Interestingly, NSUN5 epigenetic inactivation also renders these gliomas sensitive to bioactivatable substrates of the stress-related enzyme NQO1. Most importantly, NSUN5 epigenetic inactivation is a hallmark of glioma patients with long-term survival for this otherwise devastating disease. Keywords Glioma · Epitranscriptomics · RNA methylation · Clinical outcome Introduction Much effort has been devoted to the study of how events leading to aberrant RNA transcription regulation generate Cellular homeostasis requires accurate regulation of protein cancer transcriptomes [35]. However, there is an important levels [46]. In pathological states, such as cellular trans- disjunction between overall transcriptional activity and pro- formation, the proteome undergoes massive changes [16]. tein abundance [49]. Defects in protein degradation, such as those affecting the von Hippel–Lindau ubiquitin ligase complex [52] and the endoplasmic reticulum-associated Maxime Janin and Vanessa Ortiz-Barahona have contributed protein degradation pathway [25], might partially explain equally to this article. this discrepancy, but a picture has emerged of how protein Electronic supplementary material The online version of this levels of key cancer genes are dysregulated by alterations in article (https​://doi.org/10.1007/s0040​1-019-02062​-4) contains de novo protein synthesis [43, 45, 48]. Aberrant translational supplementary material, which is available to authorized users. control occurs during the origin, maintenance, and progres- sion of tumoral cells [43, 45, 48]. The enhanced growth of * Manel Esteller mesteller@carrerasresearch.org cancer cells often requires an increase in global protein syn- thesis that it is strongly linked to increased ribosome activity Extended author information available on the last page of the article 13 Vol.:(0123456789) 1054 Acta Neuropathologica (2019) 138:1053–1074 [43, 45, 48]. However, it has recently become clear that (Gibco). Human grade III glioma cell lines SW1088 and this picture is even more complex than previously thought, BT142 mut/- were purchased from the American Type because oncogenic signaling and stress-related conditions Culture Collection and MOG-G-CCM was purchased from can enhance translation of specific transcripts [2, 47], even Sigma-Aldrich. These cell lines were cultivated regarding to in the context of restricted overall protein synthesis [5, 6]. suppliers recommandations. All cell lines were authenticated Although several cellular factors may be involved in the by short tandem repeat profiling (LGS Standards SLU) and aforementioned phenotypes with respect to RNA transla- tested for the absence of mycoplasma. tion and protein production, it is likely that alterations in components of the machinery of ribosome biogenesis and DNA methylation analyses protein synthesis play a central role in these processes. In this regard, mutations in ribosomal proteins such as RPS19, DNA methylation status of the 5′-end promoter-associated RPL5, RPS26, and RPL11 occur in Diamond–Blackfan ane- CpG Island of the NSUN5 gene was determined by bisulfite mia, a syndrome that it is characterized by increased suscep- genomic sequencing and DNA methylation microarrays. For tibility to cancer [7]. Interestingly, in X-linked dyskeratosis all methods, genomic DNA was bisulfite-converted using an congenita (X-DC), another syndrome related to increased EZ DNA Methylation Gold kit (Zymo Research, Orange, cancer risk, the main mutated gene is DKC1 [18], a pseu- CA, USA). For bisulfite genomic sequencing, a minimum douridine synthase that modifies ribosomal RNA (rRNA) at of eight single clones were interrogated for each sample the post-transcriptional level [37, 53]. The defect in DKC1 and the methylation frequency was calculated in each case. gives rise to an altered internal ribosome entry site-depend- Results were analyzed with BioEdit software and methyl- ent translational program [3, 4, 21, 53]. Most importantly, in ated cytosines were mapped using BSMap software. NSUN5 addition to pseudouridine, ribosomal RNA exhibits another CpG methylation status in pediatric gliomas was determined post-transcriptional modification: the presence of 5-methyl- by pyrosequencing using the primer sequences shown in the cytosine [11, 39]. Neither the existence of distorted rRNA supplementary Key Resources Table. The DNA methylation 5-methylcytosine profiles in human cancer nor the role of array used was the Infinium HumanMethylation450 Bead- RNA methyltransferases for 5-methylcytosine in rRNA for Chip (Illumina). Raw fluorescence intensity values were nor- cellular transformation has been properly characterized [11, malized with Illumina Genome Studio software (V2011.1) 39]. Here, we show that NSUN5, acting as the RNA meth- using ‘control normalization’ with background correction. yltransferase for the C3782 position of human 28S rRNA, is Normalized intensities were then used to calculate DNA an enzyme with tumor-suppressor properties that undergoes methylation levels (beta values). epigenetic loss in gliomas, leading to an overall deficiency in protein synthesis and the initiation of a specific translational Expression analyses program for adaptation to cellular stress conditions. Most important, and as it also occurs with the isocitrate dehydro- For real-time quantitative reverse transcription-PCR experi- genase (IDH1) mutations [32, 50], the presence of NSUN5 ments, total RNA was extracted using the SimplyRNA kit DNA methylation-associated silencing in human gliomagen- (Promega) on a Maxwell RSC device (Promega) and ret- esis identifies patients with good clinical outcome, which is rotranscribed using the ThermoScript RT-PCR system an exceptional feature of people affected by this tumor type. (ThermoFisher) with oligo(dT) primers. The reaction was carried out following the methods for the use of SYBR Green (Applied Biosystems), and GAPDH was used as Materials and methods housekeeping genes to enable normalization. Primers for qPCR are listed in the supplementary Key Resources Table. Cell lines Reactivation treatments with the demethylating agent 5-aza- 2′-deoxycytidine (AZA; Sigma) were performed at 0.5 µM Human glioma cell lines DBTRG-05MG, M059J, A172, for 72 h. For immunoblotting assays, we extracted cell pel- and LN229 were purchased from the American Type Cul- lets and brain white matter samples with 300 µL of RIPA ture Collection; CAS-1 was obtained from The Biological buffer containing protease and phosphatase inhibitors cock- Bank of the IRCCS Azienda Ospedaliera Universitaria San tail cOmplete™ (Roche). Antibodies used in this study are Martino—IST Istituto Nazionale per la Ricerca sul Cancro; described in supplementary Key Resources Table. and KS-1 was purchased from the Japanese Collection of Research Bioresources Cell Bank. Cell lines were all cul- NSUN5 transfection tivated in Dulbecco’s Modified Eagle’s Medium (DMEM) supplemented with 10% v/v fetal bovine serum (FBS) The cDNA sequence of NSUN5 containing a FLAG- (Innovative, Novi, MI) and 1% v/v penicillin/streptomycin Tag in the amino-terminal end was cloned into the 13 Acta Neuropathologica (2019) 138:1053–1074 1055 pLVX-IRES-ZsGreen1 expression plasmid (Clontech camera from IVIS. All mouse experiments were approved Laboratories) between the EcoRI and XbaI restriction by and performed according to the guidelines of the Insti- sites. Lentiviruses containing this construct were produced tutional Animal Care Committee of the Vall d’Hebron by cotransfecting HEK-293T cells with the recombinant Research Institute in agreement with the European Union pLVX-IRES-ZsGreen1, psPAX2 (Addgene), and pMD2.G and national directives. 3 × 105 A172 and LN229 (EV or (Addgene), using j­etPRIME® Transfection Reagent (Poly- NSUN5) or DBTRG-05MG (scramble or shNSUN5) cells plus Transfection) and following the supplier instructions. were stereotactically inoculated into the corpus striatum of The transfection cocktail was removed after 6 h and replaced the right brain hemisphere (1 mm anterior and 1.8 mm lat- by fresh medium. After 72 h, viral containing supernatant eral to the lambda; 2.5 mm intraparenchymal) of 8-week-old was collected, 0.45 µM-filtered, and stored at 4 °C before Balb/c nude mice (Janvier Labs). We inoculated 7 animals infection. The recombinant product was randomly inserted per group. Tumor progression was monitored by biolumines- by lentiviral transduction in the genome of the A172 and cence measurements using the Xenogen ­IVIS® Spectrum. LN229 glioma cell lines. After 5 passages, the green fluo- Mice were euthanized at day 17 after inoculation. rescent cells were sorted by FACS and cultured in DMEM medium supplemented with 10% FBS and 1% v/v penicillin/ Xenografts growth streptomycin (Gibco). For primers sequences, please refer to supplementary Key Resources Table. All mouse experiments were approved and performed in accordance with the guidelines of the Institutional Animal NSUN5 depletion Care Committee of the Bellvitge Biomedical Research Insti- tute. For tumor growth into the brain, 3 × 105 or 1.5 × 106 Four different sequence gene-specific short hairpin RNA cells, respectively, for LN229 (EV or NSUN5) and DBTRG- molecules (shRNAs) for NSUN5 mRNA were designed and 05MG (scramble or shNSUN5) were inoculated into the cor- transduced into DBTRG-05MG and CAS-1 glioma cell line pus striatum of 9 (scramble and shNSUN5) or 10 (EV and (for sequences please refer to supplementary Key Resources NSUN5) mice for each cell line individually. Table). shRNA against the MSS2 yeast mRNA (not present in mammals) was used as scrambled (control). All shRNA NSUN5 immunoprecipitation and RT‑qPCR molecules were ligated into pLVX-shRNA2-ZsGreen plas- mid (Clontech). 10 ug of each encoding plasmid was mixed Immunoprecipitations were carried out from total LN229 with 7.5 μg of ps-PAX2 and 2.5 μg of PMD2.G plasmid extracts (transfected with empty vector or NSUN5) over- (Addgene), using j­etPRIME® Transfection Reagent (Poly- night with 10 μl of Anti-Flag⨁ M2 Magnetic Beads plus Transfection). Upon 10 min of RT incubation, the trans- (Sigma) in RIP buffer (150 mM KCl, 25 mM Tris pH7.4, fection mix was added dropwise on a 10 cm culture plate 5 mM EDTA, 0.5 mM DTT, 0.5% NP40, protease inhibi- containing HEK293-TLV lentiviral packaging cells at 80% tors). Beads were then washed three times with RIP buffer, confluence. After 72 h, medium with high-titer lentiviral and 10% of beads were boiled in Laemmli buffer 1× and particles was 0.45 μm filtered. DBTRG-05MG and CAS-1 loaded on a protein gel to check immunoprecipitation effi- glioma cell lines were cultured in virus containing medium ciency. The pulled-down RNA from the remaining 90% of for 24 h. After 5 passages, the green fluorescent cells were beads was extracted by adding 1 mL of TRIzol⨁ Reagent sorted by FACS and cultured in DMEM medium supple- (ThermoFisher). After phenol extraction and isopropanol mented with 10% FBS and 1% v/v penicillin/streptomycin precipitation, a DNase treatment step was performed, and (Gibco). the final pellet was resuspended in water. Equal amounts of each sample were then retrotranscribed and analyzed by Brain tumor xenografts quantitative PCR (Applied Biosystems 7900HT Fast Real- Time PCR System). The equivalent amount of input RNA IVIS images was processed in parallel to estimate pull-down efficiency. All mouse experiments were approved and performed in RNA methylation analyses accordance with the guidelines of the Institutional Animal Care Committee of the Vall d’Hebron Research Institute. RNA bisulfite conversion and sequencing was performed 1 × 106 cells were stereotactically inoculated into the cor- following Epigentek Methylamp RNA bisulfite conversion pus striatum of the right brain hemisphere of 9-week-old Kit instructions. The resulting bisulfite-modified RNA was athymic mice Nude-Foxn1nu mice (Charles River Labora- converted in cDNA using the ThermoScript RT-PCR system tories). To estimate the size of tumors, the luciferase activity (ThermoFisher) with random hexamer primers. PCR primers of inoculated tumor cells was quantified in a Xenogen–CCD for the sequence of interest are available in supplementary 13 1056 Acta Neuropathologica (2019) 138:1053–1074 Key Resources Table. The amplicon PCR band was purified corresponding to Human (version of April 2016) a list of and ligated into ­pGEM®-T Easy vector (Promega), trans- common contaminants and all the corresponding decoy formed, and sequenced as described previously in DNA entries. Resulting data files were filtered for FDR < 1%. methylation analyses. Bisulfite RNA deep-sequencing (Bis- seq) was performed as described in Suppl. Methods. Deoxynyboquinone and isobutyl‑deoxynyboquinone drug assays Global determination of protein synthesis IC50 studies in the glioma cell lines upon drug exposure Overall protein synthesis in the different experimental con- were performed using the sulforhodamineB (SRB) assay. ditions was assessed by the incorporation of O-propargyl- Briefly, 48 h after exposure to nine increasing concentra- puromycin (OP-Puro) into nascent proteins using the Click- tions of deoxynyboquinone (2 nM–15 µM), culture medium iT® OPP Reagent (Thermo Fisher) [5] and [3H] leucine was removed and 100 µL of 10% trichloroacetic acid was incorporation. added to the wells to fix cells for 1 h at 4 °C. Cells were then washed twice with distilled water and stained with 100 µL of Assessment of translational efficiency 0.4% SRB in 1% acetic acid during 30 min, light-protected, and washed twice with 1% acetic acid. SRB was then solu- Translational efficiency for each RNA was determined by bilized in Tris base (10 mM; pH 10.0) and 540 nm-optical normalizing the ribosome profiling deep-sequencing data density was determined using a microplate reader (Perkin (Ribo-seq) to transcript length and total transcript abundance Elmer Viktor 3). according to deep-sequencing RNA transcriptome (RNA- seq) data, as described in Suppl. Methods. Patients Gene functional‑enrichment analysis and sequence DNA methylation data in the discovery set of glioma cases motifs recognition was collected from The Cancer Genome Atlas (TCGA) Data Portal (https​://tcga-data.nci.nih.gov/tcga/). For the Gene sets were used to perform a gene set over-representa- initial glioma validation cohort, 115 formalin-fixed paraffin- tion analysis over the GO-biological processes and KEGG embedded (FFPE) tumor tissues from glioma patients were pathways included in the public GSEA signature database retrospectively collected in four different centers (Hospital collections and the DAVID web service. The top gene clus- Universitari de Bellvitge, Hospital Germans Trias i Pujol, ters resulting from the hypergeometric test with an FDR Hospital Clínic de Barcelona and University Hospital Basel) adjusted P value < 0.05 was finally considered. 5′UTR from 1989 to 2018 and were histologically reviewed. Most DNA sequences from selected genes were used to perform of patients had received treatment based on temozolamide a motif recognition analysis. RegRNA 2.0 and FIMO soft- combined or not with radiotherapy, and molecular analyses ware from MEME suite 4.12.0 were used to perform CERT, of IDH1 mutational status and MGMT methylation were TOP, PRTE, and uORF motif scanning. IRESPred web available. Co-deletion 1p19q status was only available in 23 service and Infernal software were used to identify IRES of the 115 cases. Patients gave their informed consent to par- motifs. G4-quadruplexes were identifies using QGRS Map- ticipate in the research, which had received ethical approval per software. from the review board of each institution. The expanded glioma validation cohort included 303 additional glioma Stable isotope labeling by amino acids in cell culture patients from which NSUN5 methylation, IDH1 mutation, (SILAC) co-deletion of 1p19q, MGMT methylation, and progression- free survival are available [1, 22]. These cases had signed SILAC was performed as previously described [25]. SILAC written informed consent and were histologically reviewed labeling was performed using SILAC-Lys8- Arg10-Kit as described in the respective publications [1, 22]. media (Silantes). Peptide mixes were analyzed using an OrbitrapFusion Lumos mass spectrometer (Thermo Scien- Statistical analysis tific, San Jose, CA, USA) coupled to an EasyLC (Thermo Scientific (Proxeon), Odense, Denmark). All data were The associations between variables were assessed by χ2 acquired with Xcalibur software v3.0.63. Proteome Discov- tests, Fisher’s exact test, Mann–Whitney test, Welch’s t test, erer software suite (v2.0, Thermo Fisher Scientific) and the Wilcoxon paired test, or Spearman correlation whenever Mascot search engine (v2.5, Matrix Science (1)) were used indicated. Kaplan–Meier plots and log-rank test were used for peptide identification and quantification. Samples were to estimate Overall Survival (OS). Statistical analysis was searched against a SwissProt database containing entries performed using SPSS for Windows (Armonk, NY) and 13 Acta Neuropathologica (2019) 138:1053–1074 1057 GraphPad Prism 5 (La Jolla, CA) for Windows. P values (Fig. 1b). Beyond glioma-derived malignancies, the NSUN5 less than 0.05 were considered statistically significant. All promoter CpG island was mostly found to be unmethylated statistical tests were two-sided. Methylation and expression in the other tumor types available in our data set, with the values for GBM and LGG TCGA primary tumor samples exception of thyroid cancer (4 of 16, 25%), biliary tract (1 of were obtained from the NCI’s Genomic Data Commons 5, 20%), autonomic ganglia (6 of 33, 18%), and soft-tissue (GDC). The values corresponding to glioma cell lines were sarcoma (3 of 23, 13%) (Fig. 1a and Suppl. Dataset 1a). The obtained from COSMIC cell line database. Correlations NSUN5 promoter CpG island was unmethylated in all the were obtained by calculating Spearman’s rank correlation normal human tissues studied (Suppl. Dataset 1d). test and the associated rho coefficient. Having identified the aforementioned NSUN5 CpG island methylation profiles, we studied in greater detail their asso- Data availability ciation with the possible transcriptional inactivation of the NSUN5 gene at the RNA and protein levels in glioma cell All the obtained deep-sequencing data have been deposited lines. We performed bisulfite genomic sequencing of mul- at the SRA (https​://www.ncbi.nlm.nih.gov/sra) with the fol- tiple clones in the glioblastoma cell lines DBTRG-05MG, lowing links: MO59J, CAS-1, A172, LN229, and KS-1 using primers that RNA-Seq: BioProject: PRJNA395552 encompassed the transcription start site-associated CpG https ​ : //datav ​ i ew.ncbi.nlm.nih.gov/objec ​ t /PRJNA ​ 3 9555​ island. We found the 5′-end region of NSUN5 of the LN229, 2?revie​wer=60ndb​08i1g​6na5n​koaai​ad9a6​o A172, and KS-1 glioma cell lines to be hypermethylated in comparison with normal brain white matter (Fig. 1c), bsRNA-Seq: BioProject: PRJNA395575 whereas the DBTRG-05MG, CAS-1, and MO59 J cells were https​://www.ncbi.nlm.nih.gov/biopr​oject​/39557​5 unmethylated (Fig. 1c). These results validated the DNA methylation patterns obtained by the microarray approach Ribo-Seq: BioProject: PRJNA395570 (Fig. 1d). Normal astrocytes, fetal brain, and different brain https​://www.ncbi.nlm.nih.gov/biopr​oject​/39557​0 regions (such as frontal cortex, posterior cingulate cortex, and hippocampus) were unmethylated for NSUN5 in all cases (Dataset 1e). Positive and negative technical controls Results for the DNA methylation assays are shown in Suppl. Data- set 1f. Most notably, the NSUN5-hypermethylated glioma Identification of NSUN5 CpG island promoter cell lines LN229, A172 and KS-1 minimally expressed hypermethylation‑associated transcriptional the NSUN5 RNA transcript and protein, as determined by silencing in glioma cells quantitative real-time PCR and western blot, respectively (Fig. 1e). Conversely, the glioma cell lines unmethylated at To identify candidate genetic and epigenetic defects in the NSUN5 promoter (DBTRG-05MG, CAS-1, and MO59J) the proposed ribosomal RNA cytosine methyltransferases expressed NSUN5 RNA and protein (Fig. 1e). Treatment NSUN5, NSUN1 (28S rRNA C4447), and NSUN4 (12S of the methylated glioblastoma cell lines with the DNA- mitochondrial rRNA C841) in human tumorigenesis, we first demethylating agent 5-aza-2-deoxycytidine restored NSUN5 data-mined a collection of 1001 human cancer cell lines in expression at the RNA and protein levels (Fig. 1f) in asso- which we had recently determined the exome mutational, ciation with hypomethylation events at the promoter CpG transcriptomic, gene copy number, and DNA methylation island (Suppl. Fig. S1a). We extend the study to three grade profiles [20]. The available genomic data did not indicate the III glioma cell lines: BT142 mut/−, harboring a homozygous presence of NSUN5, NSUN1, or NSUN4 mutations, ampli- R132H mutation [26], and SW1088 and MOG-G-CCM, fications, or deletions in the cell lines considered (Suppl. wild type for IDH1. We found DNA methylation-associated Dataset 1). Although genetic defects in the described genes silencing of NSUN5 in BT142 mut/- cells, whereas the two were not found, transcriptional silencing by promoter CpG remaining cell lines (SW1088 and MOG-G-CCM) were island hypermethylation is another mechanism for inactivat- unmethylated and expressed NSUN5 (Suppl. Fig. S1b and ing genes in human cancer [12]. The NSUN1 and NSUN4 c). Treatment of the methylated cell line BT142 mut/- with promoter-associated CpG islands were unmethylated in the the DNA methylation inhibitor restored NSUN5 expression cell lines studied (Suppl. Datasets 1b and c). In contrast, the at the RNA and protein levels (Suppl. Fig. S1d). The study NSUN5 promoter CpG island was methylated in 38% (21 of 16 patient-derived xenograft (PDX) from primary glio- of 55) of the glioma-derived cell lines studied (Fig. 1a and mas engrafted in nude mice showed that NSUN5 was hyper- Suppl. Dataset 1a). Data-mining of microarray expression methylated in three cases (19%). RNA was available for four results [20] showed that NSUN5 hypermethylation was asso- cases (two unmethylated and two methylated) and NSUN5 ciated with transcript downregulation in the glioma cell lines hypermethylation was associated with diminished transcript 13 1058 Acta Neuropathologica (2019) 138:1053–1074 Fig. 1  Transcriptional silencing of NSUN5 by promoter CpG island square, respectively, and percentage of methylation is indicated on hypermethylation in human glioma cells. a Percentage of NSUN5 the right. d DNA methylation profile of the CpG island promoter for methylation in the Sanger panel of cancer cell lines by tumor type. the NSUN5 gene analyzed by the 450 K DNA methylation microar- b NSUN5 methylation is associated with loss of the transcript in the ray. Single CpG absolute methylation levels (0–1) are shown. Green, glioma cell lines from Sanger (n = 48). Correlation analysis between unmethylated; red, methylated. Data from the studied six glioma methylation beta values and expression Z-score are shown. The P cell lines, brain white matter and nineteen normal brain samples are value corresponding to Spearman’s rank correlation test and the asso- shown. e NSUN5 expression levels in glioma cell lines determined ciated rho coefficient are indicated in the figure. c Bisulfite genomic by real-time PCR (data shown represent mean ± S.D. of biological sequencing of NSUN5 promoter CpG Island in glioma cells lines and triplicates) and western blot. f Expression of the NSUN5 RNA tran- brain white matter. CpG dinucleotides are represented as short verti- script and protein was restored in the A172, LN229 and KS-1 cells by cal lines and the transcription start site (TSS) is represented as a long treatment with the demethylating drug 5-aza-2′-deoxycytidine (AZA). black arrow. Single clones are shown for each sample. Presence of an Data shown represent the mean ± S.D. of biological triplicates and P unmethylated or methylated cytosine is indicated by a white or black values were obtained by the Mann–Whitney test. **P < 0.01 levels (Suppl. Fig. S1e). Overall, these results indicate that the growth of glioma cells in vivo by performing intrac- cancer-specific promoter CpG island hypermethylation-asso- erebral mouse xenotransplantation experiments using ciated silencing of the NSUN5 gene occurs in glioma cells. LN229 and A172 GFP-Luc cell lines, and quantified tumor Once we had demonstrated the existence of NSUN5 growth over time by measuring luciferase activity. Tumors CpG island hypermethylation-associated transcriptional obtained from NSUN5-transfected LN229 and A172 cell loss in glioma cell lines, we studied its contribution to the lines showed a significant reduction in size at 10 and 17 day growth of these cells. To this end, we analyzed the effect post-inoculation in comparison with the respective empty- of the restoration of NSUN5 expression in the NSUN5- vector-transfected cells (Fig. 2b, c). In the reverse experi- hypermethylated glioblastoma cell lines LN229 and A172. ment, we observed that tumors arising from unmethylated Upon efficient transfection of NSUN5 (Fig. 2a), we studied DBTRG-05MG cells undergoing NSUN5 shRNA-mediated 13 Acta Neuropathologica (2019) 138:1053–1074 1059 Fig. 2  Restoration of NSUN5 impairs glioma tumor growth in vivo. a mice inoculated with the indicated LN229 and A172 tumors after Western blot to show efficient restoration of NSUN5 protein expres- 17 DPI. d LN229-EV and LN229-NSUN5 cells were injected in the sion upon stable transfection in A172 and LN299 glioma cells and left or right flank of 10 mice, respectively. Tumor volume measured efficient depletion of NSUN5 protein expression in NSUN5-shRNA over time (left panel) and tumor weight upon sacrifice (right panel) DBTRG-05MG glioma cells. EV empty vector. An equal number of are shown. P values obtained by Student’s t test. Error bars show the indicated A172 and LN299 cells populations were stereotactically means ± s.d. e Scramble and NSUN5-shRNA-depleted DBTRG- inoculated into the brain of athymic mice. The size of the tumors was 05MG cells were injected in the left or right flank of 10 mice, respec- estimated at 10 and 17 days post-inoculation (DPI) by the quantifi- tively. Tumor volume measured over time (left panel) and tumor cation of luciferase activity in the tumor cells. b Scatter plots show- weight upon sacrifice (right panel) are shown. P values obtained by ing the individual size of the indicated LN229 and A172 tumors after Student’s t test. Error bars show means ± s.d 10 and 17 DPI. c Representative images of the luciferase signal from depletion showed a significant increase in size in comparison (Suppl. Fig. S2e), whereas transfection-mediated recovery with those originated from scramble shRNA cells (Suppl. of NSUN5 expression in LN229 hypermethylated cells was Fig. S2a, b and c). The study of subcutaneously injected associated with longer overall survival (Suppl. Fig. S2d). cancer cells into nude mice provided similar results: tumors NSUN5 restoration by transfection in hypermethylated glio- originated from NSUN5-transfected LN229 cells had a blastoma cells also diminished proliferation in vitro (Suppl. significantly lower volume and weight than empty-vector- Fig. S3a), whereas its shRNA-mediated downregulation in transfected-derived tumors (Fig. 2d), whereas NSUN5 NSUN5 unmethylated cells increased cell viability (Suppl. shRNA-mediated depletion in the unmethylated DBTRG- Fig. S3b). The glioma-related genes IDH1, TET1, or TET3 05MG cell line increased the volume and weight of the are wild type in the studied cells lines (COSMIC) and we derived tumors (Fig. 2e). Comparing NSUN5 shRNA- observed a similar expression of the three proteins in the depleted vs scramble DBTRG-05MG cells injected ortho- three NSUN5 unmethylated and methylated glioma cell lines topically in the brain of nude mice, we found that NSUN5 (Suppl. Fig. S3c). The expression of these proteins did not downregulation was associated with shorter overall survival change upon NSUN5 restoration in LN229 cells or depletion 13 1060 Acta Neuropathologica (2019) 138:1053–1074 in DBTRG-05MG cells (Suppl. Fig. S3c). These results sug- in LN229 and A172 cells did not change the levels of the gest that NSUN5 has a specific tumor-suppressor role in other two candidate human ribosome RNA methyltrans- gliomas. ferases, NSUN1 and NSUN4 (Suppl. Fig. S5b). We also studied the stable knockdowns of NSUN5 by the short hair- NSUN5 is the RNA methyltransferase for the C3782 pin RNA (shRNA) method in the NSUN5-expressing and position of 28S rRNA unmethylated glioma cell lines DBTRG-05MG and CAS- 1. Upon effective NSUN5 shRNA-mediated downregula- The activity and targets of NSUN5 in human cells have not tion, we observed hypomethylation at the C3782 28S rRNA been formally identified, but the yeast S. cerevisiae (Rcm1) site in DBTRG-05MG and CAS-1 cells (Fig. 3e). The link [13, 38, 40] and worm C. elegans (nsun-5) [38] homologues between NSUN5 epigenetic silencing and C3782 28S rRNA are responsible for methylating C2278 in the 25S rRNA and hypomethylation was also observed in the grade III glioma C2381 in the 26S rRNA, respectively. Phylogenetic analyses cell lines: BT142 mut/- cells, harboring DNA methylation- show that the C2278 and C2381 positions in the yeast and associated silencing of NSUN5, showed an unmethylated worm 25S and 26S rRNAs, respectively, correspond to the C3782 28S site, whereas SW1088 and MOG-G-CCM (both C3782 position of the human 28S rRNA (Fig. 3a). This site unmethylated and expressing NSUN5) showed a methylated is placed in the vicinity of the peptidyltransferase center C3782 28S rRNA position (Suppl. Fig. S1f). of the large ribosomal subunit [36, 38], suggesting that it We also moved from the candidate approach to a non- is relevant for core ribosomal function. Thus, we set out biased genomic setting to identify the spectrum of NSUN5 to demonstrate that NSUN5 is responsible for methylating targets. To map NSUN5-dependent modified cytosine sites C3782 28S rRNA in our glioma models. across the human transcriptome, we coupled bisulfite con- For this purpose, we first showed that 28S rRNA repre- version of cellular RNA with next-generation sequencing sents a binding target of NSUN5, since immunoprecipita- (bsRNA-seq) [44] in NSUN5-transfected LN229 glioma tion of NSUN5 specifically retrieved this RNA sequence cells and empty-vector-transfected cells. Upon efficient res- in the RNA pull-down assay (Fig. 3b). We next performed toration of NSUN5 expression, we observed that the only comparative bisulfite sequencing of multiple clones of the cytosine site in RNA that reached methylation levels of over 28S rRNA extracted from the NSUN5 unmethylated glio- 90% following NSUN5 transfection was the C3782 posi- blastoma cell lines DBTRG-05MG, MO59J, and CAS-1, tion of the 28S rRNA (empirical Bayes moderated t test, and the NSUN5 hypermethylated and transcriptionally P = 0.033). Interestingly, the specificity of the approach was silenced LN229, A172, and KS-1 glioblastoma cell lines. further confirmed by identifying that the only other methyl- We observed that while the NSUN5-expressing DBTRG- ated cytosine in 28S rRNA, the C4447 position correspond- 05MG, MO59J, and CAS-1 cells showed high levels of ing to the C4099 position in mouse 28S rRNA [23] and sug- C3782 28S rRNA methylation, the LN229, A172, and KS-1 gested to be mediated by the other RNA-methyltransferase cells undergoing NSUN5 epigenetic inactivation displayed a NSUN1 [40], was fully methylated in both empty-vector and lack of C3782 28S rRNA methylation (Fig. 3c). Treatment NSUN5-transfected cells (Suppl. Fig. S5c). Thus, all these of the NSUN5 methylated glioblastoma cell lines with the results indicate that NSUN5 is the human C3782 28S rRNA DNA-demethylating agent 5-aza-2-deoxycytidine not only methyltransferase and that epigenetic silencing of NSUN5 in restored NSUN5 expression (Fig. 1f), but also recovered glioma cells prevents the formation of this chemical modifi- C3782 28S rRNA methylation (Suppl. Fig. S4a). The LN229 cation in the rRNA transcript considered here. and A172 cells, showing NSUN5 epigenetic silencing, also To evaluate the structural impact of removing the recovered C3782 28S rRNA methylation upon restoration of methyl group of the C3782 position of the 28S rRNA, we NSUN5 activity by transfection (Fig. 3d). Importantly, we simulated by means of Molecular Dynamics (MD) a sub- constructed an NSUN5 catalytic inert form due to the pres- region of the human 28S subunit around 5mC3782 (see ence of multiple engineered mutations (Suppl. Fig. S4b and “Materials and methods” and Fig. 4a–c). The simulation c) that was unable to restore C3782 28S rRNA methylation of the system with methylated C3782 (labelled r30A5mC) (Suppl. Fig. S4d). The mung bean nuclease protection assay showed a stable methyl-π interaction between methylated (Suppl. Methods) confirmed the RNA bisulfite sequencing C3782 and C3781 (Fig. 4d), while this interaction was lost data for C3782 28S rRNA methylation (Suppl. Fig. S4e, in the unmethylated system (r30AC). The removal of the f and g). LN229 cells with NSUN5 epigenetic loss were methyl group altered significantly the C3781-G3810 base depleted for C3782 28S rRNA methylation in both nuclear pair and the conformation of the bulge at position C3809 and cytoplasmic fraction, whereas NSUN5 transfection (Fig. 4d). The bulge interacts with a small hairpin (resi- restored the methylation status at both compartments (Suppl. dues 3742–3778) which is located at the edge of the 28S Fig. S5a). To demonstrate the specificity of our experimen- subunit, exposed to the P-site of the ribosome (Fig. 4a–d). tal approach, we showed that the transfection of NSUN5 In the unmethylated system, the hairpin undergoes a 13 Acta Neuropathologica (2019) 138:1053–1074 1061 Fig. 3  NSUN5 epigenetic inactivation abrogates the methylation antibody (left panel), followed by analysis of the retrieved RNA by of the C3782 position of human 28S rRNA. a Top, RNA sequence quantitative RT-PCR (right panel). c RNA bisulfite sequencing of the alignment of the conserved human 28S rRNA C3782 position (black 28S rRNA in glioma cells lines according to NSUN5 epigenetic sta- square) in the corresponding 26S, 25S and 28S rRNAs orthologues tus. Cytosines are represented as short vertical lines and the C3782 from C. elegans, S. cerevisiae and M. musculus. Below, Protein site is represented as a long black arrow. Single clones are shown for sequence alignment of human NSUN5 with its recognized rRNA each sample. Presence of an unmethylated or methylated cytosine is 5-methylcytosine RNA-methyltransferase orthologues in C. Elegans, indicated by a white or black square, respectively. d RNA bisulfite S. Cerevisiae and M. musculus. Highlighted in black and grey the sequencing of the 28S rRNA in empty-vector (EV) and NSUN5- identical and physicochemically similar (scoring > 0.5 in the Gonnet transfected LN229 and A172 glioma cells. e RNA bisulfite sequenc- PAM 250 matrix) residues, respectively. The aligned region includes ing of the 28S rRNA in scramble and NSUN5-shRNA-depleted the RNA-methyltransferase domain. b NSUN5 interaction with 28S DBTRG-05MG and CAS-1 glioma cells. For CAS-1, western-blot to rRNA. Total extracts from LN229 cells, either transfected with empty show efficient NSUN5 depletion is shown above vector (EV) or NSUN5 were immunoprecipitated with an anti-Flag conformational change moving away from the P-site cav- tertiary complex rRNA–tRNA–mRNA (Fig. 4f). The dis- ity, showing an average displacement of 2.8 Å with respect tortion on the hairpin observed in the simulation upon the to the methylated form (Fig. 4e). The role of this small removal of the methyl group (r30AC system) might be suf- hairpin is to interact directly through several residues with ficient to impair normal protein synthesis by altering the the tRNA molecule at the P-site, and even with the mRNA P-site conformation. [13], being crucial to ensure the structural stability of the 13 1062 Acta Neuropathologica (2019) 138:1053–1074 Fig. 4  Methylation of the C3782 position of human 28S rRNA medi- structures taken from the simulations were aligned and overlapped ates the structural stability of the tertiary complex rRNA–tRNA– to highlight structural differences. Methyl–π interaction is lost on mRNA and its loss leads to a structural distortion at the edge of the the unmethylated system (r30AC, red), leading to a significant con- 28S subunit in the P-site cavity. a Structure of the Human 80S ribo- formational change in the hairpin defined by residues 3754–3778. e some taken from PDB code 6EK0. Proteins and RNA molecules Root-Mean-Squared-Deviation (RMSD) of the heavy atoms in the are depicted in light-blue and white cartoons respectively, while P-site hairpin (residues 3754–3778) along time. The hairpin in the Mg2 + are represented by pink spheres. b System considered in the r30A5mC system adopted a conformation in average 5.4 ± 0.8 Å apart MD simulations: 30 Å around residue C3782, which is depicted by from the cryoEM structure, while the unmethylated systems (r30AC) its van der Waals radii (see “Materials and methods”). c Residues deviates 8.2 ± 1.2 Å. f Eukaryotic 80S ribosome taken from the cry- constrained at the experimental positions (nucleobases in green and oEM structure (5.5 Å-resolution) deposited with the PDB code 4V6I, backbone in orange), and residues completely free to move during the highlighting the interaction between the P-site hairpin (nucleobases simulation (violet). d Local (left) and “long-range” (right) structural in green and backbone in orange), the tRNA (violet), and the mRNA distortions observed in MD simulations after removal of the methyl (red) group at position C3782 of the 28S subunit. Two representative 13 Acta Neuropathologica (2019) 138:1053–1074 1063 NSUN5 epigenetic loss restrains overall protein stress: the activation of alternative translational programs synthesis that are involved in the adaptive response to stress [24, 45]. To identify the existence of a candidate translational pro- We did not observe major effects in ribosome distribution, gram that directly depends on the loss of C3782 28S rRNA rRNA maturation and total rRNA levels comparing empty- methylation upon NSUN5 epigenetic inactivation, we per- vector and NSUN5-transfected LN229 cells (Suppl. Fig. formed both RNA-Seq and Ribo-seq analyses comparing S5d and e); thus, we wondered about more specific ways empty-vector and NSUN5-transfected LN229 cells in the that NSUN5 and C3782 loss could affect protein levels in absence of H ­ 2O2 stress. Glioma cells with restored NSUN5 cancer cells. Although global translation rates are generally expression presented RNA abundance differences in less increased in cancer cells [45], tumors may encounter hostile than 2% transcripts (255 upregulated and 244 downregu- microenvironments during their growth due to various con- lated) remaining the 98% transcript abundance unchanged ditions such as hypoxia, nutrient deprivation, and genotoxic (absolute log2 fold change > 2) (Fig. 5d). stress [24]. One way that cancer cells survive in such stress On the other hand, we analyzed the differences in the settings is to reduce overall translation, thereby prevent- ribosome occupancy for the same conditions. We observed ing this highly energy-consuming process and avoiding the 7.2% transcripts (1028 over-occupied and 2009 down-occu- accumulation of damaged and misfolded proteins [45]. Thus, pied) with ribosome profiling differences (P value < 0.01 we wondered whether the loss of the methylated site in the and absolute log2 fold change > 0.5) (Fig. 5d). RNAs with 28S rRNA mediated by NSUN5 epigenetic inactivation in enhanced translational efficiency were defined as those glioma cells could act in this manner. To measure global unaltered in the RNA-seq, but upregulated in Ribo-seq [5, protein synthesis, we quantified the incorporation of O-pro- 51]. We identified 1987 transcripts with greater transla- pargyl–puromycin (OP–Puro) into nascent proteins [5] in tion efficiency in the LN229 empty-vector-transfected cells our experimental models. Using H ­ 2O2 in the medium to pro- (harboring NSUN5 epigenetic silencing) in comparison duce oxidative stress, we found that the three glioblastoma with NSUN5-transfected cells (Fig. 5d, Suppl. Table S1). cell lines harboring NSUN5 epigenetic loss (LN229, A172, We confirmed these observations by calculating transla- and KS1) show overall lower global protein synthesis than tion efficiency values from over-dispersion analysis of both the three NSUN5 unmethylated cell lines (DBTRG-05MG, RNA-seq and Ribo-seq measurements by applying a general MO59 J and CAS-1) (Fig. 5a). Importantly, the transfection- linear model (GLM) (Suppl. Methods). mediated recovery of NSUN5 expression in the NSUN5 We observed no over-representation of known regula- epigenetically deficient LN229 cells significantly increased tory motifs in the 5′-untranslated regions of these NSUN5- global protein synthesis (Fig. 5b). Upon another stress con- controlled transcripts, which included CAP-independent dition, nutrient deprivation, the NSUN5 epigenetically defi- elements, such as internal ribosome entry sites (IRES), or cient LN229 cells transfected with the empty vector also CAP-dependent elements such as G-quadruplexes, cyto- showed low overall level of protein synthesis, whereas the sine-enriched regulator of translation (CERT), terminal transfection-mediated recovery of NSUN5 expression sig- oligopyrimidine tract (TOP), pyrimidine-rich translational nificantly increased global protein synthesis (Fig. 5c). Simi- sequences (PRTE) and upstream Open Reading Frames lar results were obtained using [3H] leucine incorporation to (uORFs) (P value = 0.8722 Welch Two sample t test) (Suppl. measure global protein synthesis: the restoration of NSUN5 Fig. S6a). The use of a bicistronic reporter plasmid encoding activity in the epigenetically inactivated LN229 cell signifi- both Firefly (IRES) and Renilla (CAP) luciferases further cantly enhanced [3H] leucine incorporation and, thereby, demonstrated the effect of NSUN5 on CAP-dependent and global protein synthesis (Fig. 5b, c). NSUN5 shRNA-medi- CAP-independent translation (Fig. 5e). To better character- ated downregulation in the NSUN5 unmethylated CAS-1 ize the described set of 1987 RNAs with significantly higher cell line had the opposite effect (Suppl. Fig. S5f). Overall, translational efficiency in NSUN5 epigenetically deficient the results indicate that the transcriptional inactivation of cells, we developed a gene functional annotation by perform- NSUN5 by promoter CpG island and the mediated loss of ing gene set enrichment analysis (GSEA). Using KEGG and methylated C3782 in 28S rRNA were associated with a GO signature collections, we observed an over-representa- restriction of overall protein synthesis. tion of cellular pathways and biological terms related to can- cer, cell cycle, and the ribosome. Following the GSEA and NSUN5 epigenetic loss promotes a specific mRNA sequence motifs recognition analysis described in “Materials translational program for stress survival and methods”, we applied a hypergeometric test with a FDR adjusted P value < 0.05 over the TE candidate genes using It is appropriate to mention that, in addition to the inhibi- DAVID and GSEA signature database collections. Most tion of global protein synthesis described above, human strikingly, we noted an enhancement of ontologies related to tumors manifest another method for coping with cellular cellular response to stress and metabolic adaptation (Fig. 5f). 13 1064 Acta Neuropathologica (2019) 138:1053–1074 Fig. 5  NSUN5 epigenetic loss is associated with depletion of global seq) and ribosome-protected RNA (Ribo-seq) deep-sequencing pro- protein synthesis and the emergence of a stress-response transla- files to identify those RNAs with enhanced translational efficiency in tional program. a NSUN5 unmethylated glioma cell lines DBTRG- NSUN5 deficient cells. 1987 RNAs that did not change in the RNA- 05MG, MO59J, and CAS-1 show higher overall protein synthesis seq of LN229 cells upon NSUN5-transfection were upregulated in the assessed by OP-Puro under oxidative stress (100 mM H ­ 2O2) than Ribo-seq of empty-vector-transfected cells indicating enhanced trans- the NSUN5 methylated cells (A172, LN229 and KS-1). b Restora- lational efficiency. e NSUN5 affects both CAP-dependent and CAP- tion of NSUN5 function by transfection in epigenetically inactive independent translation according to the use of a reporter plasmid LN299 cells increases overall protein synthesis under oxidative stress encoding for Firefly (IRES) and Renilla (CAP) luciferases. f Gene set (100 mM ­H2O2) assessed by OP-Puro. Enhancement of global protein enrichment analysis (GSEA) of the RNAs with increased translational synthesis upon NSUN5 recovery in LN299 cells is also observed by efficiency in NSUN5 deficient cells (hypergeometric test with a FDR the [3H] leucine incorporation assay. c Similar results were obtained adjusted P value < 0.05) upon nutrient deprivation. d Comparison of the total RNA (RNA- To confirm this scenario, in which NSUN5 epigenetic transfection in LN229 cells (Suppl. Table S2). With the aim silencing promotes a global reduction in protein synthe- to improve the characterization of the proteins with signifi- sis, but a specific protein production program emerges cantly different expression upon NSUN5 transfection, we that is associated with enhanced translational efficiency developed a gene functional annotation by performing gene to deal with cellular stress, we carried out stable isotopic set enrichment analysis (GSEA). Following the GSEA and labeling of amino acids in cell culture (SILAC) in empty- sequence motifs recognition analysis described in “Materi- vector-transfected LN229 cells in comparison with NSUN5 als and methods”, we applied a hypergeometric test with stably transfected LN229 cells. At the level of resolution an FDR adjusted P value < 0.05 over the candidate genes employed in the SILAC analyses, 128 proteins were identi- using DAVID and GSEA signature database collections. fied that had significantly different expression upon NSUN5 Using KEGG and GO signature collections, we observed 13 Acta Neuropathologica (2019) 138:1053–1074 1065 an over-representation of cellular pathways and biological shRNA-depleted DBTRG-05MG cells in comparison terms related to cell signaling, response to stress by mis- with scramble shRNA cells (Suppl. Fig. S6b and Suppl. match repair and RNA processing (Suppl. Fig. S5g). Most Table S4). We confirmed these observations by calculating proteins (73%; 94 of 128) were found to be upregulated upon translation efficiency values from over-dispersion analysis restoration of NSUN5, whereas the remaining 27% (34 of of both RNA-seq and Ribo-seq measurements by applying 128) were downregulated (Suppl. Table S2). Western-blot a general linear model (GLM) (Suppl. Methods). validation of three downregulated proteins upon NSUN5 With the aim to improve the characterization of the transfection is shown in Suppl. Fig. S5h. Thus, these data are described set of 787 RNAs with significantly higher transla- consistent with the idea that the loss of NSUN5 in glioma tional efficiency in NSUN5 shRNA-mediated deficient cells, cells restricts general protein synthesis. They also suggest we developed a gene functional annotation by performing the existence of a selective synthesis of specific proteins gene set enrichment analysis (GSEA). Using KEGG and GO in the context of NSUN5 deficiency. In this regard, 26% signature collections, we observed an over-representation (9 of 34) of these proteins were derived from RNAs that of cellular pathways and biological terms related to cancer showed increased translational efficiency upon NSUN5 loss and the ribosome. Following the GSEA and sequence motifs (Fig. 5d). We selected one of these targets, the stress-acti- recognition analysis described in “Materials and methods”, vated protein NQO1 [9], for western-blot and quantitative we applied a hypergeometric test with an FDR adjusted P reverse transcription–PCR analyses, which confirmed that value < 0.05 over the TE candidate genes using DAVID and its NSUN5 regulation occurred at the mRNA translational GSEA signature database collections. As we had previ- level and was not associated with a difference in RNA levels ously described for the LN229 model (Fig. 5f), we noted an (Fig. 6a). In addition, NQO1 was enriched in the polysome enhancement of ontologies related to cellular response to fraction of the empty-vector LN229 cells (where NSUN5 stress and metabolic adaptation (Suppl. Fig. S6c). NQO1, is epigenetically silenced) in comparison with NSUN5- that was characterized above as having an enhanced trans- transfected LN229 cells (Fig. 6b). Further development of lational efficiency (Suppl. Table S1 and Fig. 6a, b) and additional RNA-seq experiments comparing empty-vector overexpression (Suppl. Table S2 and Fig. 6a) in LN229 and NSUN5-transfected LN229 cells, but in the presence of cells with NSUN5 epigenetic loss, was also identified as a ­H2O2 stress, confirmed that NSUN5 transfection had almost target of enhanced translational efficiency in DBTRG-05MG no effect in the transcriptome, remaining 99.5% of the tran- cells with shRNA-mediated depletion of NSUN5 (Suppl. scripts unchanged (Suppl. Table S3). Ribo-seq analysis in Table S4). Western-blot and quantitative reverse transcrip- these cells, undergoing H ­ 2O2 stress, further confirmed that tion–PCR analyses confirmed that NQO1 upregulation at the NQO1 showed greater translation efficiency in the LN229 protein level in NSUN5 depleted cells was not associated empty-vector-transfected cells (having NSUN5 epigenetic with a difference in RNA levels (Suppl. Fig. S6d). Alto- loss) in comparison with NSUN5-transfected cells (Suppl. gether, these data are consistent with the idea that the loss Table S1). We also performed RNA-seq in the absence of of NSUN5 in glioma cells restricts general protein synthesis ­H2O2 in another glioblastoma cell line, A172, that it also while activating the selective synthesis of specific stress- shows NSUN5 epigenetic inactivation obtaining near identi- related proteins such as NQO1. cal results: 99% of transcripts levels were unchanged upon NSUN5 transfection, including NQO1 (Suppl. Table S3). NSUN5 DNA methylation‑associated silencing The above-described scenario was also validated in confers sensitivity to drugs that target NQO1 loss-of-function experiments. To identify the existence of a candidate translational program that directly depends on The finding that NSUN5 epigenetic silencing was associated NSUN5 loss, we performed both RNA-Seq and Ribo-seq with NQO1 overexpression prompted us to study whether analyses in NSUN5 shRNA-depleted vs scramble shRNA these glioma cells might be more vulnerable to compounds DBTRG-05MG cells. Glioma cells with depleted NSUN5 targeting this particular stress-related protein. Drugs that expression presented RNA abundance differences in less generate reactive oxygen species have been suggested to than 1% transcripts, remaining the 99% transcript abundance be useful for pushing cancer cells over an oxidative stress unchanged (absolute log2 fold change > 2) (Suppl. Fig. S6b threshold and into cell death [15]. In this regard, deoxynybo- and Suppl. Table S4), whereas for ribosome occupancy, quinone (DNQ) and isobutyl–deoxynyboquinone (IB-DNQ) we observed a 2.2% transcripts with ribosome profiling dif- are NQO1 bioactivatable substrates that cause the produc- ferences (absolute log2 fold change > 0.5) (Suppl. Fig. S6b tion of large amounts of ROS that cannot be tolerated by the and Suppl. Table S4). RNAs with enhanced translational cancer cells, leading to their death [19, 27, 30]. Thus, we efficiency were defined as those unaltered in the RNA-seq, assessed whether an enhanced growth inhibitory effect of but upregulated in Ribo-seq [5, 51]. We identified 787 these compounds was dependent on NSUN5 methylation. transcripts with greater translation efficiency in NSUN5 First, we demonstrated by western-blot analyses that the 13 1066 Acta Neuropathologica (2019) 138:1053–1074 Fig. 6  NSUN5 epigenetic silencing activates stress-related protein (DBTRG-05MG, MO59 J, and CAS-1). Drug-response curves were and confers growth inhibition sensitivity to NQO1-targeting mol- generated using GraphPad Prism software and analyses were per- ecules. a Validation of NSUN5 translational regulation of the iden- formed with the drc R package. For each cell line and the drug, we tified stress-related target NQO1 expression at the RNA level deter- fit a four-parameter generalized log-logistic model. Comparison of mined by RNA-seq counts (left) and real-time quantitative PCR the IC50 values calculated from the slopes were obtained by means (middle) do not change upon NSUN5 transfection, but NQO1 expres- of a z test (P < 0.0001). e IC50 determination using the SRB assay sion decreased at the protein level (right) upon NSUN5 restoration. in NSUN5-transfected LN299 cells in comparison with empty-vector- b qPCR shows enrichment of the NQO1 transcript in the polysome transfected cells (EV) shows enhanced resistance to DNQ-mediated fraction of the empty-vector LN229 cells in comparison to NSUN5 growth inhibition in the cells with restored NSUN5 expression. Com- transfected-LN229 cells. c NQO1 expression levels in glioma cell parison of the IC50 values calculated from the slopes was obtained lines determined by western blot according to NSUN5 methylation by means of a z test (P < 0.001). f Kaplan–Meier analysis of Survival status. d IC50 determination using the SRB assay in the glioma cell according treatment conditions (IB-DNQ treated vs mock group) in a lines grouped by NSUN5 methylation status. Black dashed curves set of mice models with implanted tumors derived from the NSUN5 represent the 95% confidence band for each group. Glioma cells har- methylated cell line A172 (left) or the NSUN5 unmethylated cell line boring NSUN5 methylation-associated NQO1 overexpression (A172, DBTRG (right). Significance of the log-rank test is shown. Results of LN229 and KS-1) show increased sensitivity to deoxynyboquinone the univariate Cox regression analysis are represented by the hazards (DNQ) and IB-DNQ in comparison to NSUN5 unmethylated cells ratio (HR) and 95% confidence interval (CI) three studied glioma cell lines with NSUN5 hypermethyla- unmethylated glioma cell lines (z test, P < 0.0001) (Fig. 6d). tion (LN229, A172, and KS-1) showed high levels of NQO1, Furthermore, we extended these assays to NSUN5 meth- whereas minimal expression of NQO1 was observed in the ylated LN229 glioma cells, where we have exogenously three NSUN5 unmethylated cell lines (DBTRG-05MG, restored NSUN5 activity and thus depleted the NQO1 pro- MO59 J, and CAS-1) (Fig. 6c). These data confirmed the tein (Fig. 6a). We observed that upon NSUN5 transfection above-described role of NSUN5 expression in the regula- the cells become significantly more resistant to the NQO1 tion of NQO1 protein levels (Fig. 6a). Most importantly, bioactivatable substrate in comparison to empty-vector- the determination of the corresponding IC50 values, accord- transfected cells (z test, P < 0.001) (Fig. 6e). Similar results ing to the SRB assay for cell viability, showed that NSUN5 were also found in vivo for cancer cells injected orthotopi- hypermethylated glioma cells were significantly more sen- cally in the brain of nude mice (Fig. 6f). The use of the sitive to the DNQ and IB-DNQ drugs than the NSUN5 NQO1 bioactivatable substrate increased survival in those 13 Acta Neuropathologica (2019) 138:1053–1074 1067 nude mice with tumors originated from NSUN5 hypermeth- hypermethylation in human primary gliomas. The analy- ylated A127 glioma cells in comparison with the mock- ses of the collection of human gliomas available from The treated group (Fig. 6f), whereas no differences in survival Cancer Genome Atlas (TCGA) (https​://tcga-data.nci.nih. between drug and mock treatment were observed for mice gov/tcga/), which used the same DNA methylation micro- carrying tumors derived from the NSUN5 unmethylated array platform as that used here for our initial Sanger cell cell line DBTRG-05MG (Fig. 6f). Thus, the occurrence of line screening, showed the presence of NSUN5 promoter NSUN5 methylation pinpoints glioma cells that are more CpG island methylation in 28% (190 of 681) of gliomas sensitive to the cytotoxic effect of NQO1 substrates. of different grades (Fig. 7a and Suppl. Table S5). TCGA RNA-sequencing data showed that NSUN5 hypermethyla- NSUN5 epigenetic loss is a hallmark of human tion was associated with transcript downregulation across primary gliomas with good clinical outcome glioma samples (Spearman’s rank correlation, ρ = − 0.7, P < 10−5) (Fig. 7b). The histological grade based on the Once we had determined in glioma cell lines and mouse degree of anaplasia, as set out in the World Health Organi- models the effects of NSUN5 epigenetic silencing, we zation (WHO) Classification of Tumors of the Central Nerv- studied the impact of NSUN5 promoter CpG island ous System, has a major effect on the natural history of the Fig. 7  NSUN5 epigenetic inactivation occurs in human primary univariate Cox regression analysis are represented by the hazards gliomas in association with good clinical outcome. a Percentage of ratio (HR) and 95% confidence interval (CI). d Kaplan–Meier analy- NSUN5 methylation in the TCGA data set of primary tumors accord- sis of OS in IDH1 wild-type gliomas according to NSUN5 methyla- ing to cancer type (top) and according to the cellular grade of the tion status. e Kaplan–Meier analysis of OS in gliomas without 1p/19q glioma (down). b NSUN5 promoter CpG island methylation is asso- deletion according to NSUN5 methylation status. f Kaplan–Meier ciated with the loss of the transcript among all cellular grades in pri- analysis of OS in gliomas with unmethylated MGMT according to mary glioma, in low-grade glioma and in glioblastoma. c Kaplan– NSUN5 methylation status. For all graphs, the P value corresponds to Meier analysis of overall survival (OS) across overall glioma grades the Log-Rank test. Cox regression univariate analysis is represented and in low- and high-grade glioma with respect to NSUN5 methyla- by HR with a 95% CI tion status. Significance of the log-rank test is shown. Results of the 13 1068 Acta Neuropathologica (2019) 138:1053–1074 disease. Low-grade gliomas (LGGs; grades I, II, and III) HR = 1.95; 95% CI = 1.16–3.27) (Fig. 7d). We further have a more favorable prognosis, whereas the high-grade expanded this initial validation cohort with the inclusion gliomas (grade IV), also known as glioblastoma multiforme of two independent glioma cohorts of 303 patients, where (GBM), have an extremely unfavorable outcome [32, 50]. NSUN5 methylation status is available from the same DNA We observed the enrichment of NSUN5 hypermethylation in methylation microarray used in the TCGA glioma cases low-grade gliomas (34%, 180 of 527) relative to GBM (6%, [1, 22]. Suppl. Table S5 shows the clinicopathological and 10 of 154) (Fisher’s exact test, P = 3.8 ­10−9) (Fig. 7a). TCGA molecular characteristics of the overall extended validation RNA-sequencing data showed NSUN5 hypermethylation to cohort. Related to clinical outcome in this large validation be associated with transcriptional downregulation in low- cohort, only PFS was available from the three set of cases and high-grade gliomas (Fig. 7b). Beyond adult gliomas, we (pyrosequencing plus the two data-mined cohorts), and thus, also detected NSUN5 hypermethylation in pediatric brain it was the assessed variable. We confirmed that NSUN5 pro- tumors, such as diffuse intrinsic pontine glioma (DIPG) moter hypermethylation was associated with increased PFS (50%, 5 of 10) and medulloblastoma (25%, 2 of 8). Impor- (log-rank; P value 1­ 0−5; HR = 3.03; 95% CI = 2.35–3.74) tantly, we were able to obtain RNA from three DIPG tumors, (Fig. 7e). Importantly, the impact of NSUN5 methylation where we observed that the two NSUN5 hypermethylated on PFS is maintained when the glioma validation samples cases showed hypomethylation of the C3782 position of are divided into grades. NSUN5 methylation is associated the 28S rRNA, while the NSUN5 unmethylated patient dis- with extended PFS in low-grade gliomas (LGG, n = 183) played high levels of rRNA methylation at this site (Suppl. (log-rank; P value 1­ 0−5; HR = 5.07; 95% CI = 3.46–7.41) Fig. S6e and f). (Fig. 7e) and glioblastoma (GBM, n = 235) (log-rank; P Given the known relation between glioma grade and value ­10−5; HR = 2.14; 95% CI = 1.58–2.91) (Fig. 7e). clinical outcome, we wondered whether NSUN5 hyper- We examined whether other genetic and epigenetic altera- methylation also conferred any prognostic value. We tions of clinical relevance in glioma [32, 50] helped define observed that NSUN5 hypermethylation was associated patients with NSUN5 methylation-associated extended sur- with increased overall survival (OS) in all glioma grades vival. In this regard, the presence of IDH1 mutations in glio- in the TCGA cohort [log-rank; P < 10 −5; hazard ratio mas is known to be a predictor of improved outcome [32, 50]. (HR) = 5.17, 95% CI = 2.78–9.61] (Fig. 7c). Interestingly, We found this to be the case for the gliomas included in the the value of NSUN5 hypermethylation as a predictor of bet- TCGA repository: IDH1 mutations are associated with longer ter outcome was observed for low- and high-grade gliomas OS (Suppl. Fig. S8a). Interestingly, for those cases with a (Fig. 7c). NSUN5 hypermethylation was associated with wild-type IDH1, which are all supposed to have an adverse higher OS in LGG (log-rank; P = 0.044; HR = 2.05; 95% prognosis, the presence of NSUN5 hypermethylation high- CI = 1.01–4.18) and GBM (log-rank; P = 0.012; HR = 8.26; lighted a subset of cases with long OS (Suppl. Fig. S8a). A 95% CI = 1.15–59.50) (Fig. 7c). These data are particularly similar clinical observation than the one observed for IDH1 relevant in the context of GBM because of its dismal out- mutants was found related to the co-deletion of chromosome come in 90% of patients. However, the remaining 10% of arms 1p and 19q, another predictor of good outcome [32, individuals are long-term survivors, for whom the molecular 50], that it was also associated with longer OS in the gliomas factors involved have not been fully identified [32, 50]. from the TCGA repository (Suppl. Fig. S8b). Importantly, Low levels of the NSUN5 RNA transcript were also asso- for those patients without 1p/19q co-deletion, which are all ciated with increased OS in the TCGA cohort (log-rank; expected to have a poor outcome, NSUN5 hypermethylation P < 10−5; HR = 4.57; 95% CI = 2.40–9.04) (Suppl. Fig. S7a). pinpointed a subgroup with improved OS (Suppl. Fig. S8b). Upon division by grade, low levels of NSUN5 RNA tran- Interestingly, an additional complementary molecular sce- script were also associated with increased OS in LGG (log- nario was discovered related to another key marker in human rank; P = 0.014; HR = 2.37; 95% CI = 1.16–4.84) (Suppl. glioma that is also used in clinical practice: the DNA repair Fig. S7b). A trend between low expression of NSUN5 and enzyme MGMT [32, 50]. Epigenetic inactivation of MGMT shorter OS was found in GBMs, but due to the small number is linked to an enhanced response to chemotherapy-alkylating of cases with low expression (n = 2) according to the estab- agents and increased OS in gliomas [10, 17]. We found this to lished cutoff, it lacked statistical power (Suppl. Fig. S7c). be the case for the TCGA gliomas: MGMT hypermethylation We confirmed the TCGA glioma findings in a newly is associated with longer OS (Suppl. Fig. S8c). In addition, obtained validation cohort of 115 primary gliomas (Suppl. here again, for those gliomas unmethylated at MGMT, which Table S5), where NSUN5 methylation status was determined are supposed to show poor outcome, NSUN5 hypermeth- by the pyrosequencing assay: NSUN5 promoter hypermeth- ylation defined a subset with extended OS (Suppl. Fig S8c). ylation was associated with and increased OS (log-rank; The above-described TCGA results were mostly confirmed P = 0.043; HR = 1.90; 95% CI = 1.01–3.58) and extended in our expanded glioma validation cohort. First, in this set of progression-free survival (PFS) (log-rank; P = 0.010; cases, we confirmed that the IDH1 mutation, the co-deletion 13 Acta Neuropathologica (2019) 138:1053–1074 1069 of chromosome arms 1p and 19q and MGMT hypermethyla- RNA molecules such as N6-methyladenosine, N1-methyl- tion were all associated with longer PFS (Suppl. Fig. S8d, adenosine, pseudouridine, or 5-methylcytosine [10, 11, 27]. e, and f). Importantly, for those patients without 1p/19q co- The chemical modifications of RNA, referred to in a global deletion, NSUN5 hypermethylation pinpointed a subgroup context as the epitranscriptome, carry out functions criti- with improved PFS (Suppl. Fig. S8e). Related to IDH1 muta- cal to cellular homeostasis, contributing to RNA stability, tional status, we observed a trend between NSUN5 hyper- export, localization, structure, splicing, processing, recod- methylation and longer PFS in the cases with a wild-type ing, targeting, and translation [8, 11, 14, 34]. We are at the IDH1 that did not reach statistical significance (Suppl. Fig. dawn of our understanding of how disruption of the epitran- S8d). Interestingly, NSUN5 hypermethylation defined a sub- scriptome can initiate and promote tumorigenesis, but the set with extended PFS for both MGMT unmethylated and emerging findings indicate that altered RNA modification hypermethylated patients (Suppl. Fig. S8f). patterns play a fundamental role in cancer. For both TCGA and validation cohorts, we observed that The occurrence of mutations in the pseudouridine syn- NSUN5 methylation was associated with other biomarkers thase DKC1 in X-linked dyskeratosis congenita (X-DC), a of clinical benefit such as IDH1 mutation (Fisher’s exact test disorder characterized by cancer susceptibility [18], is par- P < 10−5), co-deletion of 1p/19q (Fisher’s exact test P < 10−5), ticularly interesting, because it directly affects a key cellu- and MGMT methylation (Fisher’s exact test P < 10−5) (Suppl. lar element, the ribosome. DKC1 is the pseudouridine syn- Table S5). Importantly, all these three last molecular markers thase that mediates the introduction of this modified base are also significantly associated between them (IDH1 muta- in ribosomal RNA [37, 53]. Mice that are hypomorphically tion/co-deletion 1p/19q, Fisher’s exact test P < 10−5; IDH1 mutant for DKC1 exhibit decreased pseudouridine in their mutation/MGMT methylation, Fisher’s exact test P < 10−5; RNAs and a defect in translation that particularly affects co-deletion 1p/19q/MGMT methylation, Fisher’s exact test mRNAs containing Internal Ribosome Entry Sites (IREs) P < 10−5) (Suppl. Table S6). It is also worth to mention that [3, 4, 18, 21, 53]. Translational control is also impaired in NSUN5 methylation was not associated with the glioma CpG human X-DC patients [3, 53]. These results indicate that island methylator phenotype (gCIMP) (Fisher’s exact test, ribosomes are not passive structures in cellular transforma- P = 0.094) (Suppl. Table S5) (Suppl. Methods), a common tion, but actively contribute to the phenotype of cancer cells. phenomenon in IDH1 mutant patients [29]. As expected, This idea is also supported by the existence of specialized the presence of the gCIMP phenotype in the TCGA glioma ribosomes and complex machineries associated with these data set is associated with improved clinical outcome (Suppl. structures and rRNAs [41, 42], in addition to the presence Fig. S9a). Strikingly, when we stratified the TCGA data set of mutations in ribosomal proteins in cancer-associated syn- according to gCIMP status and NUSN5 methylation, we dromes [7]. Significantly for rRNA, in addition to pseudouri- observed that for those cases with a lack of the gCIMP phe- dine, another modification, that of 5-methylcytosine, seems notype, which are all supposed to have a poor clinical out- to have a central regulatory role [31]. This RNA chemical come, the presence of NSUN5 hypermethylation highlighted modification also acts at the level of tRNAs, where loss of a subset of cases with long overall survival (Suppl. Fig. S9b). the 5-methylcytosine RNA-methyltransferase NSUN2 drives Finally, and most importantly, to determine the value of a global reduction in protein synthesis [5, 6, 34] and altered NSUN5 methylation as a candidate independent biomarker, translational programs [5, 6], in a similar manner to that we performed multivariate analysis for NSUN5 methyla- which occurs upon the loss of DKC1 [3, 4, 37, 53]. tion status and the described known biomarkers of clinical In this study, we combined genomic, epigenomic, and outcome (IDH1 mutational, 1p/19q co-deletion and MGMT epitranscriptomic approaches to characterize the 5-methyl- methylation status) for the both studied glioma cohorts. cytosine RNA methyltransferase, NSUN5, and showed how Multivariate Cox regression analysis showed that NSUN5 its epigenetic silencing produces a hypomethylation event at hypermethylation was an independent predictor of OS in a particular position of the 28S rRNA that depletes the over- the TCGA glioma cohort (HR = 0.50; 95% CI = 0.26–0.94; all protein synthesis, but engages glioma cells in a particular P = 0.032) (Fig. 8a) and an independent predictor of PFS translational program for survival under conditions of cellu- in the interrogated validation cohort of glioma patients lar stress. Interestingly, we also show by molecular dynamics (HR = 0.46; 95% CI = 0.35–0.61; P < 0.001) (Fig. 8b). that the methylation of the C3782 position of human 28S rRNA could mediate the structural stability of the tertiary complex rRNA–tRNA–mRNA, leading it loss to a structural Discussion distortion at the edge of the 28S subunit in the P-site cavity that might impair the physiological protein synthesis [28]. More than 100 distinct nucleotide modifications have been The observed “translational paradox”, the emergence of reported in the RNA molecule across the range of all life a translational program in the context of globally depleted forms [10, 11, 27], many of which are also present in human protein production, is a well-known mechanism used by cells 13 1070 Acta Neuropathologica (2019) 138:1053–1074 Fig. 8  Forest plots of the multivariable Cox regression analysis for clinical outcome in the glioma cohorts studied for NSUN5 methylation status taking into account different prognostic factors. P values (P) correspond to hazard ratios (HR), with a 95% of confidence interval (95%CI), associated with OS. Co-variables with associated P value under 0.05 were considered as independent prognostic factor (*P < 0.05, **P < 0.01, ***P < 0.001). a Overall Survival (OS) multivar- iate Cox regression analysis in the TCGA set of glioma patients (n = 497). b Progression-free survival (PFS) multivariate Cox regression analysis in the expanded validation set of glioma patients (n = 418) to exert a strong and timely influence on gene expression to transformed cells. NSUN5 methylation-mediated overex- deal with stress conditions [24]. As tumors progress, they pression of NQO1 can also constitute a therapeutic oppor- encounter many obstacles and unfriendly environments that tunity in these gliomas, because they become more sensitive frequently produce oxidative, replicative, genotoxic, proteo- to ROS-induced death upon the use NQO1-bioactivatable toxic, or metabolic stress [45]. Even oncogenic mutations molecules [19, 27, 30]. Thus, NSUN5 DNA methylation- that promote tumorigenesis in the early stages subsequently associated inactivation, through its linked defect in 28S become a nuisance if they generate high levels of ROS that rRNA methylation, represents a bona fide mechanism that are toxic to cancer cells [24]. One of the main adaptations explains how human gliomas adapt to challenging cellular of tumors to these conditions is the global inhibition of pro- stress conditions by depleting overall synthesis while also tein synthesis that shuts down a highly energy-demanding promoting specific translational programs. Furthermore, process and at the same time prevents the accumulation of although we did not observe NSUN5 genomic defects in misfolded or damaged proteins [45]. However, there is a the studied models, more comprehensive studies are required simultaneous enabling of translation of selected mRNAs to discard additional genetic alterations. that are vital for the adaptive cell response [24]. Another This role of NSUN5 epigenetic silencing in the toleration illustrative case is provided by NQO1, a multifunctional of cellular stress settings resembles another key molecular antioxidant enzyme regulated by the Keap1/Nrf2/ARE path- event in gliomagenesis: the occurrence of IDH1 mutations. way [9], whose translational efficiency increases upon the We have demonstrated that the loss of NSUN5 activity loss of NSUN5 in glioma, as reported here, to overcome extends cell viability in vivo and in vitro, but it is associ- the many types of oxidative stress encountered by these ated with improved clinical outcome. In a similar manner, 13 Acta Neuropathologica (2019) 138:1053–1074 1071 the loss of IDH1 normal catalytic activity contributes to under the FIS PI16/01278 Project (to J. Seoane), the Integrated Pro- gliomagenesis, but it is also associated with good OS. Once ject of Excellence no. PIE13/00022 (ONCOPROFILE) (to M. Estel- ler), CIBER 2016 CB16/12/00312 (CIBERONC) (to M. Esteller), again, we can consider how cancer cells avoid death when co-financed by the European Development Regional Fund, ‘A way to confronted by damaging stress conditions, because, as we achieve Europe’ ERDF, the AGAUR—Catalan Government (Project have shown for NSUN5, the disruption of IDH1 is another No. 2009SGR1315 and 2014SGR633) (to M. Esteller), the Cellex way of dealing with metabolic and hypoxic stress [33]. Thus, Foundation (to M. Esteller), Obra Social “La Caixa” (to M. Esteller), the CERCA Program and the Health and Science Departments of the those patients with gliomas exhibiting NSUN5 epigenetic Catalan Government (Generalitat de Catalunya) (to M. Esteller) and silencing, like those carrying IDH1 mutations, have tumor a grant from the National Health and Medical Research Council of cells that probably are on the verge of death, whose last Australia (APP1061551, to TP). M.W. Boudreau is a member of the resort in their struggle to survive is to restrict overall protein NIH Chemistry-Biology Interface Training Program (T32-GM070421). synthesis and to instigate particular emergency translational programs to deal with multiple stress conditions. Compliance with ethical standards In conclusion, NSUN5 loss in glioma provides a link Conflict of interest M. Esteller reports personal fees from Ferrer Inter- between an epigenetic event, the promoter CpG island national and Quimatryx, outside of the submitted work. hypermethylation of its promoter, and an epitranscriptomic event, the hypomethylated status of the C3782 position of Open Access This article is distributed under the terms of the Crea- the human 28S rRNA. Most significantly, we have shown tive Commons Attribution 4.0 International License (http://creat​iveco​ that NSUN5 inactivation is a likely pathway used by gli- mmons​.org/licen​ses/by/4.0/), which permits unrestricted use, distribu- tion, and reproduction in any medium, provided you give appropriate oma cells to overcome hostile stress conditions by restrain- credit to the original author(s) and the source, provide a link to the ing high energy-consuming global protein synthesis while Creative Commons license, and indicate if changes were made. stimulating the translation of adaptive proteins such as NQO1. These findings, if validated in larger sets of prospec- tive clinical studies, might be useful for the management of References glioma patients, because, in the context of a disease, whose prognosis is usually dismal, the NSUN5 epigenetic lesion 1. Bady P, Kurscheid S, Delorenzi M, Gorlia T, van den Bent MJ, contributes to identify those patients who are likely to have Hoang-Xuan K et al (2018) The DNA methylome of DDR genes and benefit from RT or TMZ in IDH mutant low-grade glioma a good outcome. treated in EORTC 22033. Acta Neuropathol 135:601–615 2. Barna M, Pusic A, Zollo O, Costa M, Kondrashov N, Rego E Acknowledgements We would like to thank the IDIBAPS Biobank et al (2008) Suppression of Myc oncogenic activity by ribosomal and the IGTP-HUGTP Biobank (PT13/0010/0009), integrated in the protein haploinsufficiency. Nature 456:971–975 Spanish National Biobank Network at the Instituto de Salud Carlos 3. Bellodi C, Kopmar N, Ruggero D (2010) Deregulation of onco- III, and to the Xarxa de Bancs de Tumors de Catalunya (XBTC), sup- gene-induced senescence and p53 translational control in X-linked ported by the Pla Director d’Oncologia de Catalunya, for sample and dyskeratosis congenita. EMBO J 29:1865–1876 data procurement. We acknowledge technical support from the Austral- 4. Bellodi C, Krasnykh O, Haynes N, Theodoropoulou M, Peng G, ian Cancer Research Foundation Biomolecular Resource Facility at Montanaro L et al (2010) Loss of function of the tumor suppressor JCSMR, ANU. We thank CERCA Program/Generalitat de Catalunya DKC1 perturbs p27 translation control and contributes to pituitary for their institutional support. tumorigenesis. Cancer Res 70:6026–6035 5. Blanco S, Bandiera R, Popis M, Hussain S, Lombard P, Aleksic J Author contributions Conception and design: MJ, VO-B, and ME. et al (2016) Stem cell function and stress response are controlled Development of methodology: MJ, VO-B, PLA, MS, DN, JP, US, by protein synthesis. Nature 534:335–340 MEC-C, SM, SG, AB-C, MP-S, MR-T, LP, JJB-S, CDLT, JFM-T, AV, 6. Blanco S, Dietmann S, Flores JV, Hussain S, Kutter C, Humphreys AA, IC, AMA, AFF, MFF, MWB, PJH, PK, KDE, PDD, MO, and SB. P et al (2014) Aberrant methylation of tRNAs links cellular stress Acquisition of data (provided animals, acquired and managed patients, to neuro-developmental disorders. EMBO J 33:2020–2039 provided facilities, etc.): AMC, AMLM, AMC, VS-M, JM, LP, FG, 7. Danilova N, Gazda HT (2015) Ribosomopathies: how a com- NV, FM-S, AT, CC, CB, JH, SF, SM, GZ, GT, JS, TP, PPP, and ME. mon root can cause a tree of pathologies. Dis Model Mech Analysis and interpretation of data (e.g., statistical analysis, biosta- 8:1013–1026 tistics, and computational analysis): AM-C, MCM, DP, and MEC-C. 8. Davalos V, Blanco S, Esteller M (2018) SnapShot: messenger Writing, review, and/or revision of the manuscript: all authors. Admin- RNA modifications. Cell 174:498–498.e1 istrative, technical, or material support (i.e., reporting or organizing 9. Dinkova-Kostova AT, Talalay P (2010) NAD(P)H:quinone accep- data and constructing databases): AM-C, MCM, DP, MEC-C, SM, and tor oxidoreductase 1 (NQO1), a multifunctional antioxidant US. Study supervision: ME. enzyme and exceptionally versatile cytoprotector. Arch Biochem Biophys 501:116–123 Funding This work was supported by a European Research Coun- 10. Esteller M (2008) Epigenetics in cancer. N Engl J Med cil (ERC) Advanced Grant under the European Community’s Sev- 358:1148–1159 enth Framework Program (FP7/2007-2013)/ERC Grant Agreement 11. Esteller M, Garcia-Foncillas J, Andion E, Goodman SN, Hidalgo No. 268626—EPINORC project (to M. Esteller), the Ministerio de OF, Vanaclocha V et al (2000) Inactivation of the DNA-repair Economía y Competitividad (MINECO) under Grant No. SAF2014- gene MGMT and the clinical response of gliomas to alkylating 55000-R (to M. Esteller) and the Instituto de Salud Carlos III (ISCIII), agents. N Engl J Med 343:1350–1354 13 1072 Acta Neuropathologica (2019) 138:1053–1074 12. Esteller M, Pandolfi PP (2017) The epitranscriptome of noncoding gliomas—implications for classification and therapy. Nat Rev Clin RNAs in cancer. Cancer Discov 7:359–368 Oncol 14:434–452 13. Gigova A, Duggimpudi S, Pollex T, Schaefer M, Koš M (2014) A 33. Reitman ZJ, Yan H (2010) Isocitrate dehydrogenase 1 and 2 muta- cluster of methylations in the domain IV of 25S rRNA is required tions in cancer: alterations at a crossroads of cellular metabolism. J for ribosome stability. RNA 20:1632–1644 Natl Cancer Inst 102:932–941 14. Gilbert WV, Bell TA, Schaening C (2016) Messenger RNA modi- 34. Roundtree IA, Evans ME, Pan T, He C (2017) Dynamic RNA modi- fications: form, distribution, and function. Science 352:1408–1412 fications in gene expression regulation. Cell 169:1187–1200 15. Gorrini C, Harris IS, Mak TW (2013) Modulation of oxidative 35. Roychowdhury S, Chinnaiyans AM (2016) Translating cancer stress as an anticancer strategy. Nat Rev Drug Discov 12:931–947 genomes and transcriptomes for precision oncology. CA Cancer J 16. Hanashs S, Taguchi A (2010) The grand challenge to decipher the Clin 66:75–88 cancer proteome. Nat Rev Cancer 10:652–660 36. Rozenski J, Crainm PF, McCloskey JA (1999) The RNA modifica- 17. Hegi ME, Diserens AC, Gorlia T, Hamou MF, de Tribolet N, Weller tion database: 1999 update. Nucl Acids Res 27:196–197 M et al (2005) MGMT gene silencing and benefit from temozolo- 37. Ruggero D, Grisendi S, Piazza F, Rego E, Mari F, Rao PH et al mide in glioblastoma. N Engl J Med 352:997–1003 (2003) Dyskeratosis congenita and cancer in mice deficient in ribo- 18. Heiss NS, Knight SW, Vulliamy TJ, Klauck SM, Wiemann S, somal RNA modification. Science 299:259–262 Mason PJ et al (1998) X-linked dyskeratosis congenita is caused 38. Schosserer M, Minois N, Angerer TB, Amring M, Dellago H, Har- by mutations in a highly conserved gene with putative nucleolar reither E et al (2015) Methylation of ribosomal RNA by NSUN5 is functions. Nat Genet 19:32–38 a conserved mechanism modulating organismal lifespan. Nat Com- 19. Huang X, Dong Y, Bey EA, Kilgore JA, Bair JS, Li LS et al (2012) mun 6:6158 An NQO1 substrate with potent antitumor activity that selec- 39. Shafik A, Schumann U, Evers M, Sibbritt T, Preiss T (2016) The tively kills by PARP1-induced programmed necrosis. Cancer Res emerging epitranscriptomics of long noncoding RNAs. Biochim 72:3038–3047 Biophys Acta 1859:59–70 20. Iorio F, Knijnenburg TA, Vis DJ, Bignell GR, Menden MP, Schu- 40. Sharma S, Yang J, Watzinger P, Kötter P, Entian KD (2013) Yeast bert M et al (2016) A landscape of pharmacogenomic interactions Nop2 and Rcm1 methylate C2870 and C2278 of the 25S rRNA, in cancer. Cell 166:740–754 respectively. Nucl Acids Res 41:9062–9076 21. Jack K, Bellodi C, Landry DM, Niederer RO, Meskauskas A, 41. Shi Z, Barna M (2015) Translating the genome in time and space: Musalgaonkar S et al (2011) rRNA pseudouridylation defects affect specialized ribosomes, RNA regulons, and RNA-binding proteins. ribosomal ligand binding and translational fidelity from yeast to Annu Rev Cell Dev Biol 31:31–54 human cells. Mol Cell 44:660–666 42. Shi Z, Fujii K, Kovary KM, Genuth NR, Röst HL, Teruel MN et al 22. Kickingereder P, Neuberger U, Bonekamp D, Piechotta PL, Götz M, (2017) Heterogeneous ribosomes preferentially translate distinct Wick A et al (2018) Radiomic subtyping improves disease stratifi- subpools of mRNAs genome-wide. Mol Cell 67:71–83 cation beyond key molecular, clinical, and standard imaging char- 43. Silvera D, Formenti SC, Schneider RJ (2010) Translational control acteristics in patients with glioblastoma. Neuro Oncol 20:848–857 in cancer. Nat Rev Cancer 10:254–266 23. Legrand C, Tuorto F, Hartmann M, Liebers R, Jacob D, Helm M 44. Squires JE, Patel HR, Nousch M, Sibbritt T, Humphreys DT, Parker et al (2017) Statistically robust methylation calling for whole-tran- BJ et al (2012) Widespread occurrence of 5-methylcytosine in scriptome bisulfite sequencing reveals distinct methylation patterns human coding and non-coding RNA. Nucl Acids Res 40:5023–5033 for mouse RNAs. Genome Res 27:1589–1596 45. Sulima SO, Hofman IJF, De Keersmaecker K, Dinman JD (2017) 24. Leprivier G, Rotblat B, Khan D, Jan E, Sorensen PH (2015) Stress- How ribosomes translate cancer. Cancer Discov 10:1069–1087 mediated translational control in cancer cells. Biochim Biophys 46. Toyama BH, Hetzers MW (2013) Protein homeostasis: live long, Acta 1849:845–860 won’t prosper. Nat Rev Mol Cell Biol 14:55–61 25. Lopez-Serra P, Marcilla M, Villanueva A, Ramos-Fernandez A, 47. Truitt ML, Conn CS, Shi Z, Pang X, Tokuyasu T, Coady AM et al Palau A, Leal L et al (2014) A DERL3-associated defect in the (2015) Differential requirements for eIF4E dose in normal develop- degradation of SLC2A1 mediates the Warburg effect. Nat Commun ment and cancer. Cell 162:59–71 5:3608 48. Truitt ML, Ruggero D (2017) New frontiers in translational control 26. Luchman HA, Chesnelong C, Cairncross JG, Weiss S (2013) Spon- of the cancer genome. Nat Rev Cancer 17:332 taneous loss of heterozygosity leading to homozygous R132H in a 49. Vogel C, Marcotte EM (2012) Insights into the regulation of protein patient-derived IDH1 mutant cell line. Neuro Oncol 15:979–980 abundance from proteomic and transcriptomic analyses. Nat Rev 27. Lundberg AP, Francis JM, Pajak M, Parkinson EI, Wycislo KL, Genet 13:227–232 Rosol TJ et al (2017) Pharmacokinetics and derivation of an anti- 50. Weller M, Wick W (2014) Neuro-oncology in 2013: improving cancer dosing regimen for the novel anti-cancer agent isobutyl- outcome in newly diagnosed malignant glioma. Nat Rev Neurol deoxynyboquinone (IB-DNQ), a NQO1 bioactivatable molecule, 10:68–70 in the domestic felid species. Invest New Drugs 35:134–144 51. Wolfe AL, Singh K, Zhong Y, Drewe P, Rajasekhar VK, Sang- 28. Noller HF (2005) RNA structure: reading the ribosome. Science hvi VR et al (2014) RNA G-quadruplexes cause eIF4A-dependent 309:1508–1514 oncogene translation in cancer. Nature 513:65–70 29. Noushmehr H, Weisenberger DJ, Diefes K, Phillips HS, Pujara K, 52. Wu J, Jiao Y, Dal Molin M, Maitra A, de Wilde RF, Wood LD et al Berman BP et al (2010) Identification of a CpG island methylator (2011) Whole-exome sequencing of neoplastic cysts of the pancreas phenotype that defines a distinct subgroup of glioma. Cancer Cell reveals recurrent mutations in components of ubiquitin-dependent 17:510–522 pathways. Proc Natl Acad Sci USA 108:21188–21193 30. Parkinson EI, Hergenrother PJ (2015) Deoxynyboquinones as 53. Yoon A, Peng G, Brandenburger Y, Zollo O, Xu W, Rego E et al NQO1-activated cancer therapeutics. Acc Chem Res 48:2715–2723 (2006) Impaired control of IRES-mediated translation in X-linked 31. Popis MC, Blanco S, Frye M (2016) Posttranscriptional methylation dyskeratosis congenita. Science 312:902–906 of transfer and ribosomal RNA in stress response pathways, cell differentiation, and cancer. Curr Opin Oncol 28:65–71 Publisher’s Note Springer Nature remains neutral with regard to 32. Reifenberger G, Wirsching HG, Knobbe-Thomsen CB, jurisdictional claims in published maps and institutional affiliations. Weller M (2017) Advances in the molecular genetics of 13 Acta Neuropathologica (2019) 138:1053–1074 1073 Affiliations Maxime Janin1 · Vanessa Ortiz‑Barahona1 · Manuel Castro de Moura1 · Anna Martínez‑Cardús1 · Pere Llinàs‑Arias1 · Marta Soler1 · Daphna Nachmani2 · Joffrey Pelletier3 · Ulrike Schumann4 · Maria E. Calleja‑Cervantes1 · Sebastian Moran1 · Sonia Guil1 · Alberto Bueno‑Costa1 · David Piñeyro1 · Montserrat Perez‑Salvia1 · Margalida Rosselló‑Tortella1 · Laia Piqué1 · Joan J. Bech‑Serra5 · Carolina De La Torre5 · August Vidal6,7 · María Martínez‑Iniesta8 · Juan F. Martín‑Tejera8 · Alberto Villanueva8 · Alexandra Arias9 · Isabel Cuartas9 · Ana M. Aransay10 · Andres Morales La Madrid11,12 · Angel M. Carcaboso13 · Vicente Santa‑Maria11,12 · Jaume Mora11 · Agustin F. Fernandez14 · Mario F. Fraga15 · Iban Aldecoa16 · Leire Pedrosa17 · Francesc Graus18 · Noemi Vidal6 · Fina Martínez‑Soler19 · Avelina Tortosa19 · Cristina Carrato20 · Carme Balañá21 · Matthew W. Boudreau22 · Paul J. Hergenrother22 · Peter Kötter23 · Karl‑Dieter Entian23 · Jürgen Hench24 · Stephan Frank24 · Sheila Mansouri25 · Gelareh Zadeh25 · Pablo D. Dans26 · Modesto Orozco26,27 · George Thomas3,28,29 · Sandra Blanco30,31 · Joan Seoane7,9,32 · Thomas Preiss4,33 · Pier Paolo Pandolfi2 · Manel Esteller1,7,27,32,34 1 14 Cancer Epigenetics and Biology Program (PEBC), Bellvitge Institute of Oncology of Asturias (IUOPA), Instituto Biomedical Research Institute (IDIBELL), L’Hospitalet, de Investigación Sanitaria del Principado de Asturias Barcelona, Catalonia, Spain (ISPA-FINBA), 33011 Oviedo, Spain 2 15 Department of Medicine and Pathology, Cancer Research Nanomaterials and Nanotechnology Research Center Institute, Beth Israel Deaconess Cancer Center, Beth Israel (CINN‑CSIC), Universidad de Oviedo-Principado de Deaconess Medical Center, Harvard Medical School, Boston, Asturias, Oviedo, Spain MA, USA 16 Pathology‑Brain Bank, Hospital Clínic de 3 Molecular Mechanisms and Experimental Therapy Barcelona-CDB-August Pi i Sunyer Biomedical Research in Oncology Program, Metabolism and Cancer Group, Institute (IDIBAPS), Barcelona, Catalonia, Spain Bellvitge Biomedical Research Institute (IDIBELL), 17 Glioma and Neural Stem Cell Group and Translational L’Hospitalet, Barcelona, Catalonia, Spain Genomics and Targeted Therapeutics in Solid Tumors Team, 4 EMBL‑Australia Collaborating Group, Department IDIBAPS, Barcelona, Catalonia, Spain of Genome Sciences, The John Curtin School of Medical 18 Clinical and Experimental Neuroimmunology, IDIBAPS, Research, The Australian National University, Garran Road, Barcelona, Catalonia, Spain Acton, ACT​2601, Australia 19 5 Faculty of Medicine and Health Sciences‑Bellvitge, Proteomics Unit, Bellvitge Biomedical Research Institute Universitat de Barcelona, IDIBELL, L’Hospitalet del (IDIBELL), L’Hospitalet, Barcelona, Catalonia, Spain Llobregat, Barcelona, Catalonia, Spain 6 Department of Pathology, Hospital Universitari de Bellvitge, 20 Department of Pathology, Hospital Universitari Germans IDIBELL, L’Hospitalet del Llobregat, Barcelona, Catalonia, Trias i Pujol, Universitat Autònoma de Barcelona, Badalona, Spain Catalonia, Spain 7 Centro de Investigacion Biomedica en Red Cancer 21 Catalan Institute of Oncology, Hospital Germans Trias i (CIBERONC), Madrid, Spain Pujol, 08916 Badalona, Barcelona, Catalonia, Spain 8 Program Against Cancer Therapeutic Resistance (ProCURE), 22 Department of Chemistry, University of Illinois Catalan Institute of Oncology (ICO), Bellvitge Institute at Urbana-Champaign, Urbana, IL 61801, USA for Biomedical Research (IDIBELL), Oncobell Program, 23 L’Hospitalet del Llobregat, Barcelona, Catalonia, Spain Institute for Molecular Biosciences, Goethe University, 9 Frankfurt, Germany Vall d’Hebron Institute of Oncology (VHIO), Vall d’Hebron 24 University Hospital, Universitat Autònoma de Barcelona, Division of Neuropathology, Institute of Medical Genetics Barcelona, Catalonia, Spain and Pathology, University Hospital Basel, 4031 Basel, 10 Switzerland Centro de Investigación Bimédica en Red de Enfermedades 25 hepáticas y Digestivas (CIBERehd), CIC bioGUNE, 801A Division of Neurosurgery, Toronto Western Hospital, Bizkaia Technology Park, 48160 Derio, Spain University of Toronto, Toronto, ON, Canada 11 26 Department of Pediatric Hematology and Oncology, Hospital Joint IRB‑BSC Program on Computational Biology, Sant Joan de Déu, Barcelona, Catalonia, Spain Institute for Research in Biomedicine (IRB Barcelona), 12 The Barcelona Institute of Science and Technology (BIST), Pediatric Neuro‑Oncology Unit, Department of Pediatric 08028 Barcelona, Catalonia, Spain Hematology and Oncology, Hospital Sant Joan de Déu, 27 Barcelona, Catalonia, Spain Department of Biochemistry and Biomedicine, University 13 of Barcelona, 08028 Barcelona, Spain Preclinical Therapeutics and Drug Delivery Research 28 Program, Fundacio Sant Joan de Deu, Barcelona, Catalonia, Physiological Sciences Department, School of Medicine Spain and Health Sciences, University of Barcelona (UB), Barcelona, Catalonia, Spain 13 1074 Acta Neuropathologica (2019) 138:1053–1074 29 32 Division of Hematology/Oncology, Department of Internal Institucio Catalana de Recerca i Estudis Avançats (ICREA), Medicine, University of Cincinnati Medical School, Barcelona, Catalonia, Spain Cincinnati, OH 45267‑0508, USA 33 Victor Chang Cardiac Research Institute, 30 CIC bioGUNE, 801A Bizkaia Technology Park, Darlinghurst (Sydney), NSW 2010, Australia 48160 Derio, Spain 34 Josep Carreras Leukaemia Research Institute 31 Molecular Mechanisms Program, Centro de Investigación (IJC), Ctra de Can Ruti, Camí de les Escoles s/n, del Cáncer and Instituto de Biología Molecular y Celular 08916 Badalona, Barcelona, Catalonia, Spain del Cáncer, Consejo Superior de Investigaciones Científicas (CSIC), University of Salamanca, 37007 Salamanca, Spain 13