Methylome Analyses of Three Glioblastoma Cohorts Reveal Chemotherapy Sensitivity Markers Within DDR Genes

Zurich Open Repository and Archive University of Zurich Main Library Strickhofstrasse 39 CH-8057 Zurich www.zora.uzh.ch

Year: 2020

Methylome analyses of three glioblastoma cohorts reveal chemotherapy sensitivity markers within DDR genes

Kessler, Tobias ; Berberich, Anne ; Sadik, Ahmed ; Sahm, Felix ; Gorlia, Thierry ; Meisner, Christoph ; Hoffmann, Dirk C ; Wick, Antje ; Kickingereder, Philipp ; Rübmann, Petra ; Bendszus, Martin ;Opitz, Christiane ; Weller, Michael ; van den Bent, Martin ; Stupp, Roger ; Winkler, Frank ; Brandes, Alba ; von Deimling, Andreas ; Platten, Michael ; Wick, Wolfgang

Abstract: BACKGROUND: Gliomas evade current therapies through primary and acquired resistance and the effect of temozolomide is mainly restricted to methylguanin-O6-methyltransferase promoter (MGMT) promoter hypermethylated tumors. Further resistance markers are largely unknown and would help for better stratification. METHODS: Clinical data and methylation profiles from the NOA-08 (104, elderly glioblastoma) and the EORTC 26101 (297, glioblastoma) studies and 398 patients with glioblastoma from the Heidelberg Neuro-Oncology center have been analyzed focused on the predictive effect of DNA damage response (DDR) gene methylation. Candidate genes were validated in vitro. RESULTS: Twenty-eight glioblastoma 5’-cytosine-phosphat-guanine-3’ (CpGs) from 17 DDR genes negatively correlated with expression and were used together with telomerase reverse transcriptase (TERT) promoter mutations in further analysis. CpG methylation of DDR genes shows highest association with the mesenchymal (MES) and receptor tyrosine kinase (RTK) II glioblastoma subgroup. MES tumors have lower tumor purity compared to RTK I and II subgroup tumors. CpG hypomethylation of DDR genes TP73 and PRPF19 correlated with worse patient survival in particular in MGMT promoter unmethylated tumors. TERT promoter mutation is most frequent in RTK I and II subtypes and associated with worse survival. Primary glioma cells show methylation patterns that resemble RTK I and II glioblastoma and long term established glioma cell lines do not match with glioblastoma subtypes. Silencing of selected resistance genes PRPF19 and TERT increase sensitivity to temozolomide in vitro. CONCLUSION: Hypomethyla- tion of DDR genes and TERT promoter mutations is associated with worse tumor prognosis, dependent on the methylation cluster and MGMT promoter methylation status in IDH wild-type glioblastoma.

DOI: https://doi.org/10.1002/cam4.3447

Posted at the Zurich Open Repository and Archive, University of Zurich ZORA URL: https://doi.org/10.5167/uzh-191405 Journal Article Published Version

The following work is licensed under a Creative Commons: Attribution 4.0 International (CC BY 4.0) License.

Originally published at: Kessler, Tobias; Berberich, Anne; Sadik, Ahmed; Sahm, Felix; Gorlia, Thierry; Meisner, Christoph; Hoffmann, Dirk C; Wick, Antje; Kickingereder, Philipp; Rübmann, Petra; Bendszus, Martin; Opitz, Christiane; Weller, Michael; van den Bent, Martin; Stupp, Roger; Winkler, Frank; Brandes, Alba; von Deimling, Andreas; Platten, Michael; Wick, Wolfgang (2020). Methylome analyses of three glioblastoma cohorts reveal chemotherapy sensitivity markers within DDR genes. Cancer Medicine, 9(22):8373-8385. DOI: https://doi.org/10.1002/cam4.3447

2 Received: 4 June 2020 | Revised: 30 July 2020 | Accepted: 14 August 2020 DOI: 10.1002/cam4.3447

ORIGINAL RESEARCH

Methylome analyses of three glioblastoma cohorts reveal chemotherapy sensitivity markers within DDR genes

1Clinical Cooperation Unit Neurooncology, Abstract German Cancer Consortium (DKTK), German Cancer Research Center (DKFZ), Background: Gliomas evade current therapies through primary and acquired resist- Heidelberg, Germany ance and the effect of temozolomide is mainly restricted to methylguanin-O6-methyl- 2 Department of Neurology and transferase promoter (MGMT) promoter hypermethylated tumors. Further resistance Neurooncology Program of the National Center for Tumor Diseases, Heidelberg markers are largely unknown and would help for better stratification. University Hospital, Heidelberg, Germany Methods: Clinical data and methylation profiles from the NOA-08 (104, elderly 3Brain Tumor Metabolism, DKTK, DKFZ, glioblastoma) and the EORTC 26101 (297, glioblastoma) studies and 398 patients Heidelberg, Germany with glioblastoma from the Heidelberg Neuro-Oncology center have been analyzed 4Department of Neuropathology, Heidelberg University Hospital, Heidelberg, focused on the predictive effect of DNA damage response (DDR) gene methylation. Germany Candidate genes were validated in vitro. 5Clinical Cooperation Unit Neuropathology, Results: Twenty-eight glioblastoma 5'-cytosine-phosphat-guanine-3' (CpGs) from DKTK, DKFZ, Heidelberg, Germany 17 DDR genes negatively correlated with expression and were used together with 6European Organization for Research telomerase reverse transcriptase (TERT) promoter mutations in further analysis. and Treatment of Cancer Headquarters, Brussels, Belgium CpG methylation of DDR genes shows highest association with the mesenchymal 7Institute for Clinical Epidemiology and (MES) and receptor tyrosine kinase (RTK) II glioblastoma subgroup. MES tumors Biometry, Tübingen, Germany have lower tumor purity compared to RTK I and II subgroup tumors. CpG hypo- 8 Faculty of Biosciences, Heidelberg methylation of DDR genes TP73 and PRPF19 correlated with worse patient survival University, Heidelberg, Germany 9 in particular in MGMT promoter unmethylated tumors. TERT promoter mutation is Department of Neuroradiology, Heidelberg University Hospital, Heidelberg, Germany most frequent in RTK I and II subtypes and associated with worse survival. Primary 10Department of Neurology, University glioma cells show methylation patterns that resemble RTK I and II glioblastoma and Hospital and University of Zurich, Zurich, long term established glioma cell lines do not match with glioblastoma subtypes. Switzerland Silencing of selected resistance genes PRPF19 and TERT increase sensitivity to te- 11The Brain Tumor Center, Erasmus MC Cancer Institute, Rotterdam, The mozolomide in vitro. Netherlands 12Feinberg School of Medicine, Northwestern University, Chicago, IL, USA

This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and reproduction in any medium, provided the original work is properly cited. © 2020 The Authors. Cancer Medicine published by John Wiley & Sons Ltd.

Cancer Medicine. 2020;9:8373–8385. wileyonlinelibrary.com/journal/cam4 | 8373 8374 | KESSLER ET AL.

13Department of Medical Oncology, Conclusion: Hypomethylation of DDR genes and TERT promoter mutations is asso- Azienda USL-IRCCS Institute of Neurological Sciences, Bologna, Italy ciated with worse tumor prognosis, dependent on the methylation cluster and MGMT 14Clinical Cooperation Unit promoter methylation status in IDH wild-type glioblastoma. Neuroimmunology and Brain Tumor Immunology, DKTK, DKFZ, Heidelberg, KEYWORDS Germany DNA damage response (DDR), methylation profiling, methylguanin-O6-methyltransferase 15 Department of Neurology, Medical promoter (MGMT), telomerase reverse transcriptase (TERT) Faculty Mannheim, MCTN, Heidelberg University, Mannheim, Germany

Correspondence Wolfgang Wick, Neurology Clinic & Neuroonology Program at the National Center for Tumor Diseases, Im Neuenheimer Feld 400, D-69120 Heidelberg, Germany. Email: [email protected] heidelberg.de

Funding information The molecular analyses have been supported by the German Cancer Research Center, the NCT Heidelberg and by a grant from the German Research Foundation (DFG, SFB 1389, TP A03 to TK and WW). The clinical conduct of the study has been supported by F Hoffmann—La Roche Ltd.

1 | INTRODUCTION gene methylation in IDH wild-type glioblastomas dependent on the clinically relevant MGMT promoter methylation is Diffuse gliomas regularly evade current therapies and sev- unknown. Also, various inhibitors of DDR components are eral mechanisms behind resistance and sensitivity have not in preclinical and clinical development allowing to further only recently been discovered with high-throughput efforts exploit the concept of synthetic lethality.9,10 demonstrating the molecular faith of gliomas at recurrence,1 Recent studies suggested a benefit from a methylated but also novel targeted approaches proposing a novel radiation MGMT promoter when receiving temozolomide chemother- sensitivity biomarker.2 Classically, methylation of the methyl- apy only in patients with tumors with additional promoter guanin-O6-methyltransferase promoter (MGMT) promoter is mutation of the DDR gene telomerase reverse transcriptase known to predict response to alkylating chemotherapy, with (TERT).11,12 We identified a subset of glioblastoma patients at best minimal responses for patients with an unmethylated by methylation clustering, who benefitted most from temo- MGMT promoter at diagnosis and progression.3,4 The NOA-08 zolomide treatment when the MGMT promoter is methylated. study demonstrated an impressive survival benefit in MGMT These patients showed enhanced TERT expression.6 In ad- promoter methylated elderly patients with temozolomide treat- dition, TERT promoter mutation was found to be one of the ment compared to radiotherapy alone,5 but this may potentially earlier but not earliest events of gliomagenesis and leads to be restricted to a specific molecular tissue context.6,7 increased TERT expression rather than TERT methylation.13 Nevertheless, disease control for more than a few years Therefore, methylation of DDR genes and mutation status is achieved in barely 15%-20% of younger glioblastoma pa- in particular for TERT may alter the sensitivity to standard tients with tumors with a hypermethylated MGMT promoter radiochemotherapy treatment especially in tumors lacking only. Therefore, further markers to better stratify patients for MGMT promoter methylation. In these tumors prognostic treatment response prediction and to decide for chemo- or ra- factors are rarely known and would be important for patient diotherapy are of great interest. Recent approaches suggested stratification. We here explored the potential of DDR methyl- that the DNA damage response (DDR) gene methylome could ation as further prognostic markers. We based or hypothesis facilitate to predict resistance vs sensitivity to radio- and che- on 450 genes that have been identified as DDR genes14 and motherapy in World Health Organization grade II, IDH mu- restricted the analysis to genes where methylation negatively tant gliomas.8 These markers have not been validated in an impacted expression. Of the DDR genes, TERT is unique in independent data set yet and an effect of differential DDR the sense that TERT promoter mutations mainly influence KESSLER ET AL. | 8375 expression, and therefore, for TERT not methylation but the The original study population consisted of 373 patients. The promoter mutations status was included. Based on these as- investigated biomarker cohort consists of 104 patients (radio- sumptions, we aimed to identify prognostic markers of DDR therapy group: 53, temozolomide group: 51). genes in a combined analysis of three large well-documented glioblastoma cohorts from the NOA-087 and EORC 2610115 studies as well as a patient cohort from our Heidelberg Neuro- 2.1.2 | EORTC 26101 Oncology Center spanning a variety of conditions. Promising candidates are validated in vitro. The EORTC 26101 study randomized patients with progressive glioblastoma between bevacizumab (BEV) with lomustine (CCNU) and CCNU alone.15 The original study cohort con- 2 | METHODS sisted of 596 patients. The investigated biomarker cohort where methylation array data and paraffin tissue for evaluating TERT 2.1 | Patient cohorts mutations status was available consisted of 297 patients.

Two glioma studies and a well-documented cohort of patients treated in Heidelberg with study grade follow up that was 2.1.3 | Heidelberg cohort used as an exploratory, hypothesis-generating data set were included in the present work. Altogether this study involves About 398 patients from the Heidelberg Neuro-Oncology 799 patients with diffuse IDH wild-type glioma. Regression center diagnosed with IDH wild-type glioblastoma between analyses were performed either for each study separately or 07/2014 and 01/2018 based on histopathological and molecu- combined for all three studies together with correction for the lar characteristics. This cohort includes 43 patients from the confounding effect of the study, as indicated. NCT Neuro Master Match (N2M2) pilot study extensively characterized with whole-genome sequencing, whole-exome sequencing (WES), RNAseq, and methylation analysis.16 2.1.1 | NOA-08 Inclusion of patients into this analysis is covered by a local Heidelberg ethics vote (no. S307/2019). The NOA-08 study compared radiotherapy (RT) with temozo- Patients of the two clinical study biomarker cohorts (NOA- lomide (TMZ) chemotherapy in elderly patients (age at diag- 08 and EORTC-26101) are comparable to the original study nosis >65 years) with anaplastic astrocytoma or glioblastoma.5 population regarding survival times and treatment (Table 1).

TABLE 1 Comparison of Biomarker Biomarker cohort Full study cohort cohorts in the present analysis with the original study cohorts. NOA-08 Patient number [n (%)] 104 (28%) 373 (100%) Overall survival [median (95% CI)]a 11.2 (9.6-13.8) 8.7 (8.0-9.8) Event-free survival [median (95% CI)]a 4.1 (3.7-4.5) 4.4 (3.6-5.5) TMZ group [n (%)] 51 (49%) 195 (52%) RT group [n (%)] 53 (51%) 178 (48%) EORTC-26101 Patient number [n (%)] 297 (50%) 596 (100%) Lomustine first group [n (%)] 117 (39%) 231 (39%) BEV ± Lomustine first group [n (%)] 180 (61%) 365 (61%) Overall survival [median (95% CI)]a 8.6 (7.9-9.9) 8.9 (8.2-9.6) Progression-free survival [median (95% CI)]a 3.0 (2.8-3.7) 2.9 (2.8-3.0) Heidelberg cohort Patient number [n] 398 NA Overall survival [median (95% CI)]a 24.9 (19.2-31.0) NA Progression-free survival [median (95% CI)]a 8.2 (7.2-9.2) NA Patients with RT + TMZ [n (%)] 252 (63%) NA

Abbreviations: BEV, bevacizumab; RT, radiotherapy; TMZ, temozolomide. aSurvival times are given in months. 8376 | KESSLER ET AL. 2.2 | Illumina HumanMethylation450 and 2.5 | Cell culture and in vitro assays HumanMethylationEPIC arrays A detailed description of in vitro assays is given in the The Illumina Infinium HumanMethylation450 (450k) Methods S1. Generation and maintenance of established bead chip and MethylationEPIC kits were used to obtain glioma cell lines and primary glioma cell cultures was per- the DNA methylation status at >450 000 and >850 000 formed using standard methods as described previously.21,22 5'-cytosine-phosphat-guanine-3' (CpG) sites, respectively (Illumina), according to the manufacturer's instructions at the Genomics and Proteomics Core Facility of the German 2.6 | Data analysis Cancer Research Center (DKFZ) in Heidelberg, Germany from fresh frozen and paraffin embedded tissue of study Statistical analyses were carried out with R 3.5.1 (www.r- cohorts as well as selected primary patient-derived cul- project.org, RRID:SCR_001905) and Microsoft Excel 2016 tures. Samples were analyzed using the R (www.r-proje (Microsoft Corporation; RRID:SCR_016137). If not oth- ct.org) based methylation pipeline “ChAMP” 2.10.1.17 In erwise stated, a P < .05 was considered as significant and brief, filtering was done for multihit sites, SNPs, and XY marked with a “*.” No outliers have been excluded. Correction chromosome-related CpGs, then, data were normalized for testing of multiple CpGs or genes was performed using with a BMIQ based method and analyzed for batch effects the Benjamini-Hochberg procedure. The performances of with a singular value decomposition algorithm. Batch ef- the multivariate proportional hazards models were calculated fects related to the tissue used (paraffin embedded [FFPE] using a c-index concordance statistic in R. If not otherwise vs fresh frozen [KRYO]) were corrected using ComBat. indicated bar charts show mean values and standard devia- MGMT promoter methylation status was determined by the tion of at least three independent experiments. For survival algorithm of Bady et al.18 The classifier score and associa- analysis, we used custom adaptations of the R packages “sur- tion with specific classifier types (RTK I, II, and MES) was vival” (version 2.42-6) and “survminer” (version 0.4.3). A performed using the Neuropathology 2.0 tool described in Cox proportional hazards model for univariate and multivari- Capper el al.19 Custom scripts based on the R packages ate analysis was used to assess the correlation between CpG “minfi” (version 1.26.2) and “conumee” (version 1.14.0) methylation and survival as implemented in the “coxph” were implemented for copy-number variation profiling and function. Clustering of methylation array samples was done visualization. as described before6 using the ConsensusClusterPlus package in R. Dimensionality reduction and network analysis were also performed with R and are described in the Methods 2.3 | Tumor purity estimation S1. All figures were produced using R-based packages.

Tumor purity estimation was performed on normalized beta values of methylation data with the R package InfiniumPurity 2.7 | Data availability version 1.3.1.20 Methylation raw and processed data are made accessible via the Gene Expression Omnibus (GEO) database (https:// 2.4 | Selection of functional DDR gene www.ncbi.nlm.nih.gov/geo/; RRID:SCR_005012) under the methylation CpGs GEO accession numbers GSE12 2920 (NOA-08 biomarker cohort), GSE14 3755 (EORTC 26101 biomarker cohort), Potential DDR genes were obtained from a 450 putative gene GSE12 2994 and GSE143842 (both Heidelberg cohort). list that was published previously.14 CpGs present in the 450k TCGA data from glioblastoma patients can be accessed via methylation array in the promoter region were tested for CDC portal (portal.gdc.cancer.gov). negative association with expression in The Cancer Genome Atlas (TCGA, RRID:SCR_003193) data set obtained from the National Cancer Institute Genomic Commons Data Portal 3 | RESULTS (GDC Portal, portal.gdc.cancer.gov). CpGs with a correlation r < −.3 and a Benjamini-Hochberg adjusted P < .05 were re- 3.1 | The functional DDR methylome of IDH garded as “functional” in a sense that high methylation cor- wild-type glioblastoma relates with low RNA expression. About 28 CpGs from 17 genes met this criterion and were used for further analysis. The Figure S1 shows the workflow of the project and study co- thereafter used term “functional DDR CpGs” refers to these horts used. DDR CpGs were derived from a list of 450 previ- 28 CpGs. ously published expert-curated human DDR genes.14 We then KESSLER ET AL. | 8377

TP73 cg13943358 TP73 cg03050669 TP73 cg23163013 PARP3 cg12554573 RPA3 cg19739774 chr cohort 25 Heidelberg SMC5 cg14289863 20 EORTC 26101 PARP4 cg20765408 15 NOA−08 10 MGMT XRCC3 cg25999604 5 methylated 0 unmethylated XRCC3 cg23193616 feature classifier TSS1500 XRCC3 cg18597188 APA TSS200 REACT 5'UTR CUL4A cg26825849 GBM MES CCND3 cg17945153 beta value GBM Midline 0.8 GBM MYCN PRPF19 cg00778920 0.6 GBM RTK I 0.4 GBM RTK II EXO1 cg16121177 0.2 H3 K27M GEN1 cg11330941 LGG DNT CNS NB FOXR2 POLR2E cg26528128 INFLAM LGG GG EXO1 cg04706276 no match CHEK1 cg00554702 MVP cg26990835 CSNK1E cg13110239 CSNK1E cg04728789 MGMT cg14194875 MGMT cg12434587 POLE4 cg20919922 POLE4 cg02307033

CHEK1 cg00554702 PRPF19 cg00778920 EXO1 cg04706276 EXO1 cg16121177 CSNK1E cg04728789 chr cohort 25 Heidelberg CSNK1E cg13110239 20 EORTC 26101 15 NOA−08 XRCC3 cg23193616 10 MGMT XRCC3 cg25999604 5 methylated 0 unmethylated XRCC3 cg18597188 feature classifier TSS1500 CUL4A cg26825849 APA TSS200 REACT 5'UTR GEN1 cg11330941 GBM MES z-score POLR2E cg26528128 GBM Midline 4 GBM MYCN RPA3 cg19739774 2 GBM RTK I 0 GBM RTK II SMC5 cg14289863 −2 H3 K27M MGMT cg12434587 −4 LGG DNT CNS NB FOXR2 MGMT cg14194875 INFLAM LGG GG MVP cg26990835 no match POLE4 cg02307033 POLE4 cg20919922 PARP3 cg12554573 PARP4 cg20765408 TP73 cg23163013 TP73 cg03050669 CCND3 cg17945153 TP73 cg13943358

FIGURE 1 Functional DNA damage response 5'-cytosine-phosphat-guanine-3' (CpG) methylation in clinical study cohorts. Heatmaps of functional CpGs of all three studies (NOA-08, n = 104, EORTC 26101 , n = 297 and Heidelberg, n = 398) combined showing normalized methylation beta values (A) and row scaled values (B). 5ʹ-UTR, 5ʹ untranslated region; chr, chromosome; TSS200, 0-200 base pairs upstream of transcription start site; TSS1500, 200-1500 base pairs upstream of transcription start site; a full list of classifier abbreviations can be found in the Supporting Information correlated RNA expression of these genes with CpG meth- methylation values of functional DDR CpGs together with ylation in the TCGA glioblastoma data set and negatively methylation classifier assignments of all 799 IDH wild-type correlated CpGs (r < −.3, P.adj < .05, see Section 2) were re- tumors from the EORTC26101, NOA-08, and Heidelberg garded as functional and used for further analysis. In total, 28 cohorts. Hierarchical clustering divides the CpGs into four CpGs from 17 genes were identified (Table S1). Additionally, different groups, those with generally low methylation, high TERT expression correlates with TERT promoter mutations methylation, and two groups with a broad range of methyla- in samples from Heidelberg, for which expression and mutation values (Figure S3). tion data were available (Figure S2). Mutation frequencies of Principle component analysis of DDR methylation re- DDR genes excluding TERT were rather low in the TCGA veals two main directions. Dimension 1 (24.3% variance) is glioblastoma data set with only three genes exceeding a pre- dominated by XRCC3, CUL4A, and CSK1E and dimension 2 defined 3% mutation frequency in the TCGA glioblastoma (11.1% variance) by POLE4 and TP73 (Figure 2A). Tumor data set (STAG [5%], TP53 [53%], ATRX [9%]) all without purity was estimated from methylation data and is strongly harboring functional CpGs in their promoter. Similar results associated with dimension 1, highlighting the importance of were obtained in the WES Heidelberg glioblastoma IDH tumor purity on methylation profiles (Figure 2B). Tumors wild-type cohort (n = 43). Figure 1A,B show a heatmap with with mesenchymal (MES) classifier assignment differed 8378 | KESSLER ET AL.

samples - PCA ABGpCs - PCA

1.0 6 ● ● ● ●

● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● 0.5 ● ● ● ● 3 ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ●● ● ● ●● ● ● ● ● ● ● ● ● ● ● ●●

) ● ● ● ● ● ● ● ● ● ●● ● ● cont. ) ● ● ● ● ● ● ● ● ● ● ● ●●● ●● ● ● ● ● ● ● ● ● av. meth. beta XRCC3 cg23193616 ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ●● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● XRCC3 cg25999604 ● ● ●● ● ● ● ● ● ● ● ● ● ●● ●●●● ● ● ● ● ● ● ●●● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● XRCC3 cg18597188 ● ●● ● ● ● ●● ●●● ● ●● ●● ● ●● ●● 7.5 ●● ● ●● ● ● ● ● ● ● ● ● ● ● 0.8 POLR2E cg26528128 ● ● ● ● ● ●●● ● ● ●● ●● ● ● ● ●● ● ● ●● ● ● ● ● ● ●● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ●● ●● ● ● ● ● ● ● ● 0.0 CSNK1E cg13110239 ●● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ●●●●● ●● ● ● ● ● ● ●● ● ●● ● ● ● ● ● ●● ●●● ● ● ● ● 0.6 GEN1 cg11330941 ● ● ● ● ● ● ● ● CUL4A cg26825849 5.0 ● ● ● ●● ● ● ● ● ●● ● ● ● 0 ● ● ● ● ● ● ● ● ● ● ● CHEK1 cg00554702 ● ● ● ● ● ● ● ●● ● ●●● ● ● ●● ● ● PRPF19 cg00778920 ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ●●● ● ●●●● ● MGMT cg14194875 ● ● ● ● ● ●●● ● ● ● ●● ● ● ●● ● 0.4 ● ● ● ●● ● ● ● ●● ● ● ● ● ● ● ● ● ● ●● ● ● CSNK1E cg04728789 2.5 ●● ● ● ● ● ● ● ● ● SMC5 cg14289863 ● ● ● ●● ● ● ●● ●● ● ● ● MGMT cg12434587 ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ●● ● ● ●● ●● ● ● ●● ●● ● ● ● ● ●●● ● ● ● ● 0.2 ● ● ● ● ● ● ●● ● ●● ● ● ● ● ● ● ● EXO1 cg16121177 ● ●● ● ● ● ● ● ●● ● ● ● ● EXO1 cg04706276 ● ●● ● ● ● ●● ● ● ● ● ● PCA2 (11.1% PARP4 cg20765408 ● ● ● ● ● ●● ●● ● ● ●● ● ● MVP cg26990835 PCA2 (11.1% ● ● ● ● ● ● ● ● ●●● ● ● ● ● ● ● ● ● ● ● ● ●● ● ●● ● ● ● ● ● RPA3 cg19739774 ● ●● ● ● ● ●● ● ●● ● ● ●●● ● ● ●● ● ● ● ● ●● ● ● ● ● ● ●● ●● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● PARP3 cg12554573 ● ●● ● ● ● ● CCND3 cg17945153 ● ● ● ● ● ● ● ● ● ● ● −0.5 TP73 cg13943358 ● ●●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● POLE4 cg02307033 TP73 cg23163013 ● ● ● ● ● ● ● −3 ● ● ● ● ● ● ● POLE4 cg20919922 ● TP73 cg03050669 ●● ● ● ● ● ● ● ●●●

● ●

−1.0 ●

−1.0 −0.5 0.00.5 1.0 −5 05 PCA1 (24.3%) PCA1 (24.3%)

C samples - PCA D Individuals − PCA 50 6

APA APA CNS NB FOXR2 CNS NB FOXR2

3 GBM MES GBM MES GBM Midline GBM Midline ) ) GBM MYCN 0 GBM MYCN GBM RTK I GBM RTK I GBM RTK II GBM RTK II

0 H3K27M H3K27M INFLAM INFLAM LGG DNT LGG DNT PCA2 (13.0% PCA2 (11.1% LGG GG LGG GG REACT REACT −50 −3 no match no match

−5 05 −80−40 040 PCA1 (24.3%) PCA1 (15.7%)

E EORTC-26101 Heidelberg NOA-08

SMC5 cg14289863 SMC5 cg14289863 RPA3 cg19739774 XRCC3 cg23193616 XRCC3 cg23193616 TP73 cg13943358 XRCC3 cg25999604 XRCC3 cg25999604 PARP3 cg12554573 EXO1 cg04706276 CUL4A cg26825849 XRCC3 cg18597188 chromosome XRCC3 cg18597188 CUL4A cg26825849 EXO1 cg16121177 EXO1 cg16121177 PRPF19 cg00778920 TP73 cg03050669 20 CSNK1E cg04728789 EXO1 cg04706276 TP73 cg23163013 10 SMC5 cg14289863 CSNK1E cg13110239 CHEK1 cg00554702 0 CHEK1 cg00554702 EXO1 cg16121177 XRCC3 cg23193616 EXO1 cg04706276 CSNK1E cg04728789 GEN1 cg11330941 feature PRPF19 cg00778920 CSNK1E cg13110239 CSNK1E cg04728789 CSNK1E cg13110239 TSS1500 TP73 cg23163013 TP73 cg23163013 XRCC3 cg25999604 TSS200 TP73 cg13943358 TP73 cg03050669 XRCC3 cg18597188 5'UTR RPA3 cg19739774 TP73 cg13943358 CUL4A cg26825849 PARP3 cg12554573 GEN1 cg11330941 PRPF19 cg00778920 TP73 cg03050669 RPA3 cg19739774 z-score GEN1 cg11330941 CCND3 cg17945153 PARP4 cg20765408 2 POLR2E cg26528128 CCND3 cg17945153 PARP3 cg12554573 0 POLE4 cg02307033 PARP4 cg20765408 CCND3 cg17945153 POLE4 cg20919922 POLR2E cg26528128 POLE4 cg02307033 -2 MVP cg26990835 POLE4 cg02307033 CHEK1 cg00554702 POLR2E cg26528128 MVP cg26990835 MGMT cg12434587 MGMT cg12434587 POLE4 cg20919922 MVP cg26990835 PARP4 cg20765408 MGMT cg12434587 POLE4 cg20919922 MGMT cg14194875 MGMT cg14194875 MGMT cg14194875 ES K I K K I K K I K K II K II K II RT RT RT RT RT MM RT GB GBM MES GBM MES GBM GBM GBM GBM GBM GBM F G

● ● ● ● ● ●● ● ●● ● ● ●● ● ● ● ●● ● ●●● ●● ● ● ●● ● ● ●● ●● ●●● ● ●●●●● ● ●●●● ● ●● ●●● ●●●●● ● ●● ● ● ●●●● ● ● ● ● ● APA ● ●●● ●● ●● ●● ● ● ●● ● ●●● ●●● ● ● ● ● ● ●● ●● ● ● ● ● ● 75 ●●●●●●●●● ●● ●● ● ● ●●● ●●●● ● ●● ● ● ● ●● ● ●● ●●●●● ● ● CNS NB FOXR2 ● ● ●● ●● ● ●●● ●●●●●●●● ●●●●● ● ●●● ● ● ● ●● ●●● ● ● ●● ●●● ● ● ● ● ● ● ●● ● ● ●●● ● ● ● ●●● ●●●●●● ● ● ●● ● ●●●●● ● ●● ●●● ●●● ●●●●● ● ●● ●● GBM MES ● ● ●●●●●●● ●●●● ●● ●● 75 ● ●● ●● ● ● ● ● ● ● ●● ●●●●●●● ● ●●●●● ● ● ●●●● ●●● ●● ●● ●●●●●●●●●●● ●●●●● ● ● ● ●●●●● ●● ●● ●●● ● ● ● ●● ●●●●●●● ●● ● ●● ● ● GBM Midline ● ●● ●● ● ●●● ●● ●● ● ● ● ● ●●●●● ●● ●● ●●● ● ● tumor purity ● ● ●● ●●●●● ● ● ● ● ●● ● ●●●●●● ●●●● ● ●● ● ●●●● ●●●●●● ●● ●● ●● ● ●●●●●●●● ● ● ● GBM MYCN ● ● ● ● ● ●●● ●● ●●● ●●● ● ● ● ● ● ● ●●●●●●● ●●●● ● ●● ●●●●●● ●●●●● ●● ● ● ● ● ● ●● ●●● ● ● ● ● ● ● ●●● ●●●● ● ● ● ●● ●● ●●● ●● ● ●● ● 0.8 ● ●●● ●● ●● ● ●●● ●● ● GBM RTK I ● ● ●● ● ●● ● ●● ● ●● ●● ●●● ● ● ● ● ● ●●●● ●● ● ● ●● ● ●● ● ● ● ● ● ● ● ● ● ●● ●● ● ● ● 0.6 GBM RTK II ● ● ● ● ● ● ● ●● ● ●● ●● ● ● ● ● ● ● ● ●● ● ● ●●● ● 50 ● ● ● ● ●● ●●●● 50 ● ● ● ● ●● 0.4 H3K27M ● ●● ● ● ● ● ● ● ● ● ● ●●●●● ● ● ●● ● ●● INFLAM ● ● ● ● ● ● 0.2 ● ● ● ● ● ● ● ● ● tumor purity [%] ● LGG DNT tumor purity [%] ● ● ● ● ● LGG GG ● ● ● ● ● ● no match ● 25 ● REACT ●

25 ● ●

0.20.4 0.6 mean XRCC3 methylation [beta] classifier KESSLER ET AL. | 8379 from the RTK I and RTK II tumors (Figure 2C,D). We an- PRPF19 (P = .005-.04, c-index = 0.71-0.73) and TP73 alyzed the association of the methylation of single DDR (P = .02-.04, c-index = 0.71-0.72) methylation were sig- CpGs with classifier assignments in the study cohorts, with nificantly associated with better OS and PFS (Data S2). a cutoff P < .001 (Figure 2E). Distribution of gene methyl- C-indices for multivariate analyses ranged for significant ation was similar in all three investigated glioblastoma co- CpGs between 0.66 ± 0.02 and 0.73 ± 0.02 confirming on horts. Methylation of POLE4 and MVP was highest in RKT average good performance of the models (see respective II tumors, whereas most other genes including XRCC3 and Supporting Information Data). CSNK1E were hypermethylated in mesenchymal classified The recent NOA-08 long term study analysis showed tumors. Mean MGMT methylation was lowest in mesenchy- worse outcome of patients in the RTK I subgroup in our mal tumors; however, highest percentage of MGMT unmeth- Heidelberg and the NOA-08 biomarker cohort7 as well as ylated tumors was found in the RTK I subgroup. Comparison the best prognosis for patients with MGMT promoter meth- between average DDR methylation and global methylation ylated tumors with a RTK II classifier assignment. Though revealed similar patterns with RTK I tumors having lowest with a global P-value of .058 not significant, there was a methylation levels (P < 2.2 × 10−16 for both MES and RTK trend also in the EORTC 26101 recurrent glioblastoma study II against RTK I, Figure S4A,B). toward worse survival of patients with the low methylated Tumor purity might influence the prognostic effect of RTK I tumors (Figure 5A). Furthermore, recent preliminary CpG methylation. Mesenchymal tumors had an average lower data link EGFRvIII to a better prognosis in MGMT meth- tumor purity compared to RTK I or RTK II tumors (median: ylated tumors.23 In concordance, we found EGFRvIII and 0.57 vs 0.79 and 0.75, P < 1 × 10−57 for both comparisons, EGFR amplification both highest in the RTK II subgroup Figure 2F). This may additionally explain the on average (Figure 5C,D) as one potential factor driving chemotherapy lower MGMT promoter methylation level. Conclusively, sensitivity in RTK II tumors. XRCC3, CUL4A, and CSK1E methylation highly correlated Focusing on subtype-specific CpG markers, we identified with tumor purity (Figure 2G; Figure 4SC-F). only MGMT promoter methylation specifically correlating with improved survival in RTK II tumors, MVP was prognostic in RTK I tumors and no CpG was prognostic in MES 3.2 | DDR methylome and interaction subgroup tumors (Data S3). with therapy outcome in glioblastoma The NOA-08 study with the chemotherapy vs radiotherapy regimen additionally offers the chance of finding CpGs A univariate cox proportional hazard model was applied to that are specifically associated with benefit from either treat- identify survival associated functional CpGs. There were ment. MGMT, GEN1, PARP4, and CSNK1E were found to be five genes identified in the exploratory Heidelberg cohort, associated with better survival in the chemotherapy group, which were associated with overall survival (OS). Of these, whereas TP73 and CCND3 methylation were linked to ra- only MGMT promoter methylation was correlated with better diotherapy sensitivity in the OS analysis, however, none of prognosis consistently in all three studies (Data S1). these reached significance after correction for multiple test- In a pooled multivariate cox analysis of all three stud- ing (Data S4). For PFS, only MGMT was prognostic in the ies seven genes prove to be independent markers in both chemotherapy group. OS and progression-free survival (PFS) after correcting for MGMT promoter methylation status and multiple testing (Data S1). Beforehand analysis revealed that the criterion 3.3 | The association of TERT of proportionality was not violated. Further analysis sep- promoter mutation with methylation arated by MGMT promoter methylation revealed that four profiles and survival of the DDR genes (TP73, CSNK1E, EXO1, and PRPF19) showed significant association with OS and PFS only in In the two cohorts with available TERT status, TERT pro- MGMT unmethylated tumors (Data S2). Moreover, after moter mutation was found in 377 of 455 (83%) patients. addition of tumor purity into the multivariate model, only Differences were noted between both cohorts (EORTC 26101

FIGURE 2 DNA damage response (DDR) methylation profiles. A, Principle component analysis (PCA) showing 5'-cytosine-phosphat- guanine-3' (CpG) direction. B, PCA with samples colored by Classifier assignment with the 25 DDR CpGs (C) and the 5000 most variable CpGs (D). E, Methylation of DDR CpGs according to the three most abundant glioblastoma subgroups (MES, RTK I, and RTK II). F, Tumor purity according to tumor subtype. G, Example of correlation between tumor purity and XRCC3 methylation. 5ʹ-UTR, 5ʹ untranslated region; av., average; cont., contribution to a principle component; meth., methylation; TSS200, 0-200 base pairs upstream of transcription start site, TSS1500: 200-1500 base pairs upstream of transcription start site; PCA1, principle component 1; PCA2, principle component 2; a full list of classifier abbreviations can be found in the Supporting Information 8380 | KESSLER ET AL.

0,703

100

0,586 TERT C228T C250T number wildtype 50 0,642

0,2 0,221 0,193 0,296 0,097 00333, 766,0 002, 8,0 022, 06, ,0 0,062 0015004682, 17,0 001 00, 5,0 001 00333, 766,0 0 APA GBM MES GBM MidlineGBM MYCN GBM RTK IGBM RTK II H3 K27M INFLAM LGG DNTLGG GG no match REACT classifier BC

primary tumor - HD cohort recurrent tumor - EORTC 26101 cohort p = 0.013 p = 0.26 1.00 ++++++++++++++ 1.00 ++ ++ ++ ++ + + + C228T C228T + C250T + C250T 0.75 + wildtype 0.75 + wildtype ction] ++ + ction] ++ ra + ra +

al [f + +++ 0.50 0.50 + + ++

+ + + [f all survival + + 0.25 + 0.25 ++++ + + ++ overall surviv over + ++ + + + +++ + + 0 0 + 0200 400600 8001000 0200 400600 8001000 time [days] time [days] Number at risk (number censored) Number at risk (number censored)

TERT=C228T 92 (0) 39 (20) 14 (25)5 (27)2 (28)1 (29) TERT=C228T 189 (0)39 (1) 4 (10) 1 (11) 1 (11) 0 (11) TERT=C250T 22 (0)7 (6)2 (7) 1 (7) 1 (7)1 (7) TERT=C250T 74 (0) 12 (1) 2 (3) 0 (4) 0 (4) 0 (4) TERT=wild−type 44 (0) 30 (6) 13 (10)9 (10)5 (10)4 (10) TERT=wildtype 34 (0) 13 (0) 1 (6) 0 (6) 0 (6) 0 (6) 0 200 400 600800 1000 0200 400 600 800 1000 Time [days] Time [days] D F 21 22 20 1 19 MEgrey60 −0.052 0.06 0.13 −0.026 0.078 −0.064 0.032 −0.2 −0.73 0.39 0.22 −0.16 −0.005 18 (0.4) (0.3) (0.03) (0.7) (0.2) (0.3) (0.6) (6e−04) (2e−53) (7e−12) (2e−04) (0.006) (0.9) 2 MEtan 0.032 −0.11 −0.19 −0.097 −0.0088 −0.029 −0.064 0.12 0.18 −0.14 −0.057 0.049 0.21 (0.6) (0.07) (0.001) (0.1) (0.9) (0.6) (0.3) (0.04) (0.003) (0.02) (0.3) (0.4) (3e−04) 1 17 −0.14 −0.057 −0.11 0.0041 −0.048 0.051 −0.0017 0.57 0.27 0.18 −0.43 0.08 0.012 MEdarkgreen (0.02) (0.3) (0.06) (0.9) (0.4) (0.4) (1) (4e−27) (6e−06) (0.002) (2e−14) (0.2) (0.8) −0.1 −0.04 −0.094 −0.035 −0.042 0.081 0.06 0.52 0.37 0.05 −0.39 0.089 0.075 MEdarkred (0.08) (0.5) (0.1) (0.6) (0.5) (0.2) (0.3) (1e−21) (2e−10) (0.4) (7e−12) (0.1) (0.2) 16 −0.023 −0.2 −0.24 −0.13 −0.09 0.024 −0.12 −0.11 0.2 −0.35 0.18 0.1 0.032 MEpurple (0.7) (9e−04) (3e−05) (0.03) (0.1) (0.7) (0.04) (0.08) (6e−04) (1e−09) (0.002) (0.09) (0.6) 3 −0.025 −0.12 −0.16 −0.11 −0.14 0.12 −0.058 −0.052 0.25 −0.34 0.13 0.067 0.046 MEorange (0.7) (0.04) (0.006) (0.07) (0.02) (0.05) (0.3) (0.4) (2e−05) (3e−09) (0.02) (0.3) (0.4) 15 −0.074 −0.078 −0.12 −0.043 −0.066 0.07 −0.0013 0.15 0.18 −0.095 −0.047 0.046 −0.041 MEblue (0.2) (0.2) (0.04) (0.5) (0.3) (0.2) (1) (0.01) (0.003) (0.1) (0.4) (0.4) (0.5) −0.051 −0.011 −0.038 −0.053 −0.071 0.032 −0.075 −0.17 −0.09 −0.13 0.21 0.024 −0.02 MEred (0.4) (0.9) (0.5) (0.4) (0.2) (0.6) (0.2) (0.005) (0.1) (0.03) (3e−04) (0.7) (0.7) −0.028 −0.098 −0.15 −0.15 −0.11 0.077 −0.06 −0.028 0.12 −0.25 0.13 0.055 0.099 14 MEblack (0.6) (0.1) (0.01) (0.01) (0.08) (0.2) (0.3) (0.6) (0.04) (2e−05) (0.03) (0.4) (0.1) 4 −0.056 −0.14 −0.19 −0.13 −0.11 0.11 2.2e−05 −0.021 0.081 −0.24 0.19 −0.0081 −0.0022 0.5 MElightgreen (0.4) (0.02) (0.002) (0.03) (0.08) (0.06) (1) (0.7) (0.2) (4e−05) (0.001) (0.9) (1) MEmidnightblue −0.031 −0.18 −0.28 −0.2 −0.13 0.15 0.0067 0.057 0.25 −0.34 0.13 0.054 0.1 13 (0.6) (0.003) (1e−06) (8e−04) (0.02) (0.01) (0.9) (0.3) (2e−05) (3e−09) (0.03) (0.4) (0.09) 0.051 −0.041 0.017 −0.038 0.021 −0.079 −0.094 −0.18 −0.023 −0.058 0.079 0.042 −0.023 MElightcyan (0.4) (0.5) (0.8) (0.5) (0.7) (0.2) (0.1) (0.002) (0.7) (0.3) (0.2) (0.5) (0.7) 0.15 −0.066 −0.1 −0.028 −0.091 0.035 −0.1 −0.53 0.37 −0.68 0.41 0.11 −0.052 MEpink (0.01) (0.3) (0.1) (0.6) (0.1) (0.6) (0.08) (6e−23) (7e−11) (7e−43) (8e−13) (0.06) (0.4) 5 0.16 −0.0038 0.013 −0.067 −0.034 0.00087 −0.059 −0.88 −0.098 −0.57 0.71 −0.02 −0.072 12 MEturquoise (0.007) (0.9) (0.8) (0.3) (0.6) (1) (0.3) (8e−120) (0.1) (2e−27) (5e−50) (0.7) (0.2) 0.033 −0.033 −0.034 0.036 −0.0048 −0.029 −0.057 0.1 0.13 −0.0099 −0.11 0.084 0.0072 MEwhite (0.6) (0.6) (0.6) (0.5) (0.9) (0.6) (0.3) (0.1) (0.03) (0.9) (0.06) (0.2) (0.9) MEpaleturquoise −0.089 0.062 0.063 0.023 0.038 −0.028 0.021 0.48 0.13 0.25 −0.39 0.017 0.06 11 (0.1) (0.3) (0.3) (0.7) (0.5) (0.6) (0.7) (3e−18) (0.03) (2e−05) (6e−12) (0.8) (0.3) 6 −0.03 −0.084 −0.12 −0.036 −0.064 0.066 −0.0049 0.53 0.62 −0.076 −0.48 0.1 0.13 MEdarkorange (0.6) (0.2) (0.04) (0.5) (0.3) (0.3) (0.9) (5e−23) (1e−33) (0.2) (2e−18) (0.1) (0.03) 0 MEroyalblue 0.1 −0.087 −0.046 −0.088 −0.061 0.033 −0.055 −0.18 0.47 −0.51 0.12 0.023 0.15 10 (0.1) (0.1) (0.4) (0.1) (0.3) (0.6) (0.4) (0.002) (3e−17) (5e−21) (0.04) (0.7) (0.01) 7 −0.023 0.017 −0.0089 0.0071 −0.01 0.1 0.15 0.45 0.23 0.19 −0.39 −0.025 −0.0062 9 MEviolet (0.7) (0.8) (0.9) (0.9) (0.9) (0.09) (0.01) (7e−16) (7e−05) (0.001) (6e−12) (0.7) (0.9) 8 −0.075 −0.046 0.039 0.042 −0.034 0.13 0.15 0.45 0.53 −0.02 −0.44 0.046 0.065 MEsaddlebrown (0.2) (0.4) (0.5) (0.5) (0.6) (0.03) (0.01) (1e−15) (1e−22) (0.7) (2e−15) (0.4) (0.3) E −0.085 −0.036 −0.014 0.044 −0.016 0.15 0.22 0.57 0.44 0.15 −0.53 −0.03 0.057 MEskyblue (0.2) (0.6) (0.8) (0.5) (0.8) (0.01) (2e−04) (1e−26) (8e−15) (0.01) (1e−22) (0.6) (0.3) −0.15 −0.047 −0.1 0.041 0.012 0.027 0.066 0.94 0.37 0.44 −0.81 0.088 0.074 MEbrown (0.01) (0.4) (0.09) (0.5) (0.8) (0.7) (0.3) (2e−179) (6e−11) (6e−15) (4e−79) (0.1) (0.2) −0.14 −0.013 0.051 0.056 −0.063 0.16 0.16 0.61 0.34 0.27 −0.58 −0.13 0.11 MEdarkgrey (0.02) (0.8) (0.4) (0.4) (0.3) (0.007) (0.008) (2e−31) (3e−09) (4e−06) (7e−28) (0.03) (0.07) −0.077 0.015 0.034 0.089 −0.05 0.071 0.029 0.52 0.3 0.2 −0.49 0.073 0.037 MEsteelblue (0.2) (0.8) (0.6) (0.1) (0.4) (0.2) (0.6) (9e−22) (3e−07) (6e−04) (7e−19) (0.2) (0.5) −0.5 0.00014 −0.013 −0.027 −0.018 −0.025 0.063 0.06 0.38 0.38 −0.02 −0.33 0.077 0.095 MElightyellow (1) (0.8) (0.7) (0.8) (0.7) (0.3) (0.3) (3e−11) (3e−11) (0.7) (1e−08) (0.2) (0.1) 0.019 −0.031 −0.035 −0.11 −0.071 0.076 −0.00062 0.031 0.18 −0.16 −0.0036 0.018 0.087 MEmagenta (0.7) (0.6) (0.6) (0.08) (0.2) (0.2) (1) (0.6) (0.002) (0.008) (1) (0.8) (0.1) 0.049 −0.068 −0.081 −0.024 0.032 −0.03 0.0077 0.028 0.21 −0.11 −0.074 0.051 0.05 MEgreen (0.4) (0.3) (0.2) (0.7) (0.6) (0.6) (0.9) (0.6) (5e−04) (0.06) (0.2) (0.4) (0.4) −0.0062 −0.061 −0.098 −0.01 0.0081 0.027 0.06 0.37 0.4 −0.066 −0.31 0.14 0.11 MEcyan (0.9) (0.3) (0.1) (0.9) (0.9) (0.7) (0.3) (1e−10) (2e−12) (0.3) (8e−08) (0.02) (0.06) −0.011 −0.1 −0.13 0.0051 0.0021 0.031 0.056 0.34 0.47 −0.1 −0.31 0.092 0.074 MEdarkturquoise (0.9) (0.09) (0.03) (0.9) (1) (0.6) (0.4) (2e−09) (4e−17) (0.08) (1e−07) (0.1) (0.2) −0.051 −0.12 −0.18 −0.06 −0.11 0.14 0.038 0.41 0.58 −0.19 −0.32 0.1 0.099 MEsalmon (0.4) (0.05) (0.003) (0.3) (0.08) (0.02) (0.5) (4e−13) (2e−28) (0.001) (4e−08) (0.09) (0.1) −0.055 −0.077 −0.16 −0.046 0.013 0.0097 0.039 0.51 0.38 −0.0092 −0.37 0.13 0.17 MEgreenyellow (0.4) (0.2) (0.007) (0.4) (0.8) (0.9) (0.5) (3e−21) (2e−11) (0.9) (1e−10) (0.02) (0.005) −0.092 −0.016 −0.097 0.016 0.045 −0.0074 0.068 0.66 0.42 0.13 −0.53 0.14 0.072 −1 MEyellow (0.1) (0.8) (0.1) (0.8) (0.5) (0.9) (0.3) (1e−39) (1e−13) (0.02) (5e−23) (0.02) (0.2) −0.042 −0.12 −0.18 −0.011 −0.047 0.083 0.055 0.57 0.68 −0.099 −0.5 0.15 0.067 MEgrey (0.5) (0.05) (0.002) (0.9) (0.4) (0.2) (0.4) (3e−27) (1e−42) (0.1) (7e−20) (0.01) (0.3)

tum TRET tum sex PFS OS MGMT C228TC250T tumor purityGBM RTK I GBM MES GBM RTK II GBM MidlineGBM MYCN chr2 Mb 176.98 177 KESSLER ET AL. | 8381

88%, Heidelberg cohort 72%). TERT promoter mutation was methylation profile in cell culture being most conclusive predominantly found in the three main glioblastoma groups with a pediatric plexus tumor while copy number variation (MES, RTK I, and RTK II). For TERT wild-type tumors 21% (CNV) profiles still allow identification as derived from belong to these groups, as well as 14% of the TERT mutated glioblastoma (Table S2; Figure S6). T-SNE analysis of do (Figure 3A). TERT wild-type status was associated with methylation patterns shows clear separation between cell better PFS in the Heidelberg cohort, but not in the recurrent lines and primary cultures as well as IDH wild-type glioma EORTC 26101 cohort (Figure 3B,C). We restricted the anal- samples (Figure 4B). ysis to RTK I, RTK II, MES, Midline, and MYCN18 tumors PRFP19 methylation was associated with survival in and found 500 differentially methylated CpGs compared to the combined analysis and in particular in MGMT promoter >10 000 differential CpGs in the unrestricted analysis un- unmethylated, therefore, mainly temozolomide resistant, tu- derling the effect of the glioma classification over TERT mors and retained significance after controlling for tumor mutation. Differentially methylated regions showed high purity. Therefore, besides TERT promoter mutation, PRPF19 overlap between C228T and C250T tumors (Figure 3D). On methylation might be an interesting prognostic marker. We chromosome 2, a cluster of DMRs in the promoter regions of independently confirmed a negative correlation between HOXD genes was identified (Figure 3E). A weighted-gene PRPF19 methylation and expression (Figure 4C; r = −.39) correlation network analysis (WGCNA) identified 30 mod- and positive correlation of TERT mutation and TERT expres- ules of CpGs within the EORTC 26101 data set (Figure 3F). sion in the subset of the Heidelberg cohort with available ex- Most modules correlated strongly with tumor purity and dif- pression data (Figure S2). Low methylation/high expression ferentiated between the RTK I, RTK II, and MES subtypes. primary glioma cultures were picked for lentiviral gene ex- Highest correlation to TERT promoter mutation status was pression modulation of PRPF19 and TERT. Knockdown of conclusively found in a module specific for the RTK I pheno- PRPF19 and TERT was validated via quantitative real-time type with CpGs restricted to chromosome 2 in the HOXD12, PCR (Figure S7) and resulted in an enhanced response to te- HOXD4, HOXD3, and MIR10B promoters. However, no spe- mozolomide treatment. Cell cycle analysis revealed a higher cific survival associated module was correlated with TERT proportion of G2 arrested cells after temozolomide treatment promoter mutation. Copy-number analysis revealed associa- in tumor cells deficient of PRFP19 and TERT compared to tion of TERT mutation with amplification of chromosome 7 equally treated tumor cells transfected with respective control and loss of chromosome 10, for example, with a typical glio- vector (Figure 4D,E,H). Similarly, clonogenicity of tumor blastoma phenotype in differential and WGCNA analysis. cells treated with temozolomide was reduced particularly in PRPF19 and TERT knockdown cells (Figure 4F,G,I). These effects were not observed for irradiation with both knock- 3.4 | DDR genes and therapy response in downs (Figure S8). A summary with relevant findings is glioblastoma cell lines and primary glioma given in Figure 5A,B. cell cultures

TERT status, DDR methylome, MGMT promoter meth- 4 | DISCUSSION ylation and methylation patterns in primary glioblastoma cell cultures and cell lines are depicted in Figure 4A. With this cross-study analysis based on large recent cohorts Hierarchical clustering separated adherently growing of glioblastoma patients, we provide evidence for a prognos- cell lines from primary cells based on DDR methyla- tic role of DDR genes including DDR methylome and TERT tion. All samples harbor either TERT promoter mutation promoter mutation status. In our view, this holds several im- C228T or C250T, all but one (9/10) are MGMT methylated portant implications: (Table S2). Methylation is retained in cell culture as all Besides the well described DDR gene MGMT, we identi- but one (5/6) primary cell lines show a matching profile fied DDR genes for which methylation is linked to survival, with one of the main glioblastoma methylation subgroups. in particular in patients with MGMT promoter unmethylated Of note, adherent cell lines grown in serum change their tumors. This is of particular interest as these tumors at best

FIGURE 3 Correlation of telomerase reverse transcriptase (TERT) status with glioblastoma subgroups. A, TERT promoter mutation status and tumor classifier subgroup. Survival analysis of patients with primary glioblastoma (B) Heidelberg cohort and recurrent glioblastoma (C) EORTC 26101 cohort. D, Visualization of genome distribution of differential methylated regions between TERT promoter mutated and wild-type tumors. E, Differential methylated regions on chromosome 2 in C225T (upper row) and C250T (lower row) tumors. F, Heatmap of different 5'-cytosine- phosphat-guanine-3' modules as the result of the WGCNA analysis and relationship to several tumor-specific markers. C225T, C250T, mutation location upstream of the TERT transcription start site; chr2, chromosome 2; PFS, progression-free survival; Mb, megabase; OS, overall survival, a full list of classifier abbreviations can be found in the Supporting Information 8382 | KESSLER ET AL.

A B

TP73 cg13943358 chr TP73 cg03050669 25 RPA3 cg19739774 20 15 PARP4 cg20765408 10 SMC5 cg14289863 5 50 0 XRCC3 cg25999604 CCND3 cg17945153 feature TSS1500 PARP3 cg12554573 TSS200 MGMT cg14194875 5'UTR MGMT cg12434587 z-score cell line MVP cg26990835 5 0 IDHwt glioma

POLE4 cg20919922 PC2 −5 IDHmut glioma POLE4 cg02307033 −10 0 primary culture XRCC3 cg18597188 type XRCC3 cg23193616 cell line CUL4A cg26825849 primary glioma cell CSNK1E cg04728789 CSNK1E cg13110239 TERT C228T EXO1 cg04706276 C250T CHEK1 cg00554702 EXO1 cg16121177 classifier GB RTK I PRPF19 cg00778920 GB RTK II POLR2E cg26528128 plexus tumor ped B -50 no match GEN1 cg11330941 L2 T1 S24 U87 BG5 T325 A172 P3XX

LN229 LN308 -50 -25 025 PC1

C D E

LN308 - cell cycle S24 - cell cycle ● * r = -0.39 30 * control * control * shPRPF19_1 30 shPRPF19_1 70 shPRPF19_2 shPRPF19_2 ●

● ● ● 20 ● 20 60 ● ● ● ● ●

● ●

● 10 50 ● ● ● 10 PRPF19 expression ● ● cells in G2 phase [%] cells in G2 phase [%]

● ● 40 0 0 0 5 0100 0.05 0.10 0.15 0.20 PRPF19 methylation [beta value] temozolomide [µM] temozolomide [µM]

F G H LN308 - clonogenicity S24 - clonogenicity LN308 - cell cycle * * control 80 control 6 shPRPF19_1 shPRPF19_1 control shPRPF19_2 shPRPF19_2 30 shTERT_1 shTERT_2 60 * * 4 * * 20 40

2 clonogenicity [%]

clonogenicity [%] 10 20 cells in G2 phase [%]

0 0 0 0 5 0 5 0100 temozolomide [µM] temozolomide [µM] temozolomide [µM]

I LN308 - cloogenicity control 6 shTERT_1 shTERT_2

* 4 *

2 clonogenicity [%]

0 0 5 temozolomide [µM] KESSLER ET AL. | 8383

FIGURE 4 DNA damage response (DDR) genes and therapy response in glioblastoma cells. A, DDR methylome, telomerase reverse transcriptase (TERT) status, MGMT promoter methylation and methylation patterns in primary glioblastoma cell cultures and cell lines. B, T-SNE analysis of cell lines, primary cell cultures and IDH wild-type samples. C, Correlation between PRPF19 methylation and expression. D, Cell cycle analysis in LN308 cells, silenced for PRPF19 or transfected with respective control vector, after treatment with 5 µmol/L of temozolomide or dimethyl sulfoxide (DMSO) as control. Two different knockdown constructs were used for analysis. E, Cell cycle analysis in S24 cells, silenced for PRPF19 or transfected with respective control vector, after treatment with 100 µmol/L of temozolomide or DMSO as control. Two different knockdown constructs were used for analysis. F, Clonogenicity of LN-308 cells, silenced for PRPF19 or transfected with respective control vector, after treatment with 5 µmol/L of temozolomide or DMSO as control. Two different knockdown constructs were used for analysis. G, Clonogenicity of S24 cells, silenced for PRPF19 or transfected with respective control vector, after treatment with 100 µmol/L of temozolomide or DMSO as control. Two different knockdown constructs were used for analysis. H, Cell cycle analysis in LN308 cells silenced for TERT or transfected with respective control vector, after treatment with 5 µmol/L of temozolomide or DMSO as control. Two different knockdown constructs were used for analysis. I, Clonogenicity in LN-308 silenced for TERT or transfected with respective control vector, after treatment with 5 µmol/L of temozolomide or DMSO as control. Two different knockdown constructs were used for analysis. For panels D-I the mean value and SD of three independent experiments is shown. *P < .05. 5ʹ-UTR: 5ʹ untranslated region; chr, chromosome; TSS200, 0-200 base pairs upstream of transcription start site; TSS1500, 200-1500 base pairs upstream of transcription start site

have very limited response to standard chemotherapy with member of the TP53 gene family and overexpressed in a vari- temozolomide and since further prognostic molecular factors ety of cancers.26,27 Its regulation is complex and not fully un- have remained elusive. In our analysis, we identified methyl- derstood, especially the association between methylation and ation of the two DDR genes PRPF19 and TP73 significantly gene expression is a controversy in different cancers,26,28 but associated with better OS and PFS in particular in patients a regulatory role in chemotherapy response and probably sen- with MGMT promoter unmethylated tumors in our multivar- sitivity through DNA methylation has been described.29 We iate analysis. PRPF19 was previously reported to be involved here describe a negative association between methylation and in DDR24 and has a potential role in oncogenesis,25 but lit- expression for the prognosis relevant CpG cg13943358 and tle is known about its function in glioblastoma. TP73 is a cg2316013 in glioblastoma. The association in chemotherapy response makes both genes very plausible markers for glioma prognosis in the absence of MGMT methylation as a sensitiv- DNA damage response ity factor against chemotherapy, however, formal testing for A methylation analysis (450 genes) TERT promoter a predictive effect of the methylation levels of both genes on mutation chemotherapy response was not in the intention of the study and the analysis comparing different treatment methods in the NOA-08 cohort lacks sufficient sample size for difference functional methylation (17 genes) detection, and therefore, regarded as exploratory. Functional evidence for chemosensitivity, however, was validated for PRPF19 by gene silencing in glioblas-

survial (7 genes) toma cells. Depletion of PRPF19 expression resulted in the + tumor purity survival anticipated sensitization to temozolomide in glioblastoma MGMT - (PRPF19, TP73) (TERT) cell lines and primary cell cultures, and may therefore, B be investigated further as a predictive marker in MGMT TERT k/d TMZ promoter unmethylated glioblastoma. A limitation to use PRPF19 + k/d PRPF19 methylation as prognostic marker is its overall rel- RT atively low methylation, but combination with expression may improve prognostic relevance. The TP73 gene was not FIGURE 5 Summary of relevant findings. A, Methylation functionally analyzed in this study, but represents an at- analysis of 450 DNA damage response (DDR) genes revealed 17 tractive area for further research. The effect of upfront al- functional DDR genes of which in seven genes hypomethylation kylating agents vs targeted treatments in MGMT promoter showed association with reduced survival. Hypomethylation of unmethylated glioblastoma on its prognostic impact could PRPF19 and TP73 was associated with worse survival in patients be answered by subgroup analysis of our currently recruit- with MGMT promoter unmethylated tumors with adjustment for 2 2 30 tumor purity. TERT promoter mutations were correlated with ing N M clinical study. methylation groups and survival times. B, PRPF19 and TERT k/d- RTK I glioblastomas remain a less understood, poorly induced sensitivity in glioblastoma cells toward temozolomide but performing group that have lower DDR and overall meth- not radiotherapy. k/d, knock down; MGMT, unmethylated MGMT ylation levels. Only MVP methylation was prognostic in promoter; TMZ, temozolomide; RT, radiotherapy this group, however, overall promoter methylation of this 8384 | KESSLER ET AL. gene was low challenging its suitability as a robust marker. may finally yield benefit especially for the heavily under- Further studies might dive deeper into the differences es- served patient population with tumors having an unmethyl- pecially between RTK I and II tumors and their differential ated MGMT promoter. methylation profiles. Further limitations of this methylation analysis approach ACKNOWLEDGMENTS include the heterogeneity of the three well-documented patient We are grateful to F. Hoffmann—La Roche Ltd., the cohorts used that might reduce sensitivity for markers poten- German Cancer Research Center (DKFZ) and the NCT tially present only in certain subgroups, but enables to cover a Heidelberg and the German Research Foundation. We variety of conditions for detection of strong universal markers. thank all the patients and relatives for participating and all Even the NOA-08 study compared radiotherapy vs chemother- physicians, nurses, and stuff for care and thorough docu- apy the methylation analysis was not powered nor intended to mentation. We thank the IT core facility of the DKFZ for detect chemosensitivity of certain CpGs. Furthermore, the co- providing cluster computing capacities for the methylation horts based on the two large studies EORTC 26101 and NOA- analysis. We thank Laura Dörner and Antje Habel from 08 were subsets of the original study population based on the the Department of Neuropathology, Heidelberg University availability of tissue for methylation array analysis. Therefore Hospital for support with tissue handling and TERT pro- a sampling bias that often tends toward a better prognosis in moter mutation analysis. Open access funding enabled and the biomarker cohorts cannot be fully excluded. The approach organized by ProjektDEAL. to include only CpGs with negative correlation of methylation with expression in glioblastoma ensures a higher chance of CONFLICT OF INTEREST finding functionally relevant genes from a small defined set, TK, AB, AS, FS, TG, CM, DH, AW, PH, PR, MB, CO, but may not be exhaustive as also a subset of CpGs with pos- RS, FW, AB, AD, and MP reported no conflict of interest. itive correlations of methylation and expression or CpGs not Michael Weller: Research grants from Abbvie, Adastra, captured by the stringent threshold could be robust prognostic Bristol Meyer Squibb (BMS), Dracen, Merck, Sharp & factors though complex regulations. Dohme (MSD), Merck (EMD), Novocure, Piqur and TERT promotor mutation is the main factor facilitating Roche, and honoraria for lectures or advisory board partici- TERT expression and several studies reported differential out- pation or consulting from Abbvie, Basilea, Bristol Meyer comes based on TERT mutation and MGMT promoter methyla- Squibb (BMS), Celgene, Merck, Sharp & Dohme (MSD), tion,6,11,12 this should be taken with caution as these might have Merck (EMD), Novocure, Orbus, Roche and Tocagen. included patients with nonglioblastoma methylation groups as Martin van den Bent: Consulting for Celgene, BMS, Agios, a potential confounder. Here, we demonstrated that TERT mu- Boehringer, Abbvie, Bayer, Carthera, Nerviano, Genenta. tation is associated with worse survival in well characterized Wolfgang Wick has received study drug support from cohorts and silencing of TERT expression in glioma tumor cells Apogenix, Pfizer, and Roche and consulted for MSD and was associated with an enhanced response to temozolomide Roche with all financial reimbursement to the University treatment. Clinic. This study furthermore holds implications for preclinical models. Primary glioma cultures nicely retain the glioblasto- AUTHOR CONTRIBUTIONS ma-like methylation state, whereas cell lines change to a meth- Study concept and design: TK, WW; acquisition of molecu- ylation profile most consistent with a pediatric plexus tumor. lar and in vitro data: TK, AB, FS, DH, PR, WW; data acquisi- Although we have observed similar results for our functional tion of clinical studies: TK, FS, TG, CM, AW, MW, MB, RS, studies between primary glioma cells and adherent cell lines FW, AB, AD, MP, WW; analysis and interpretation of data: and both models cluster outside the patient tumor samples, TK, AB, AS, WW; statistical analysis: TK; study supervi- the methylation profiling strongly encourages the use of pri- sion and coordination: WW; writing the manuscript: TK, AB, mary cell lines as an appropriate model glioma for glioma WW; critical revision of manuscript for intellectual content: biology as they retain a well-preserved glioma methylation AS, FS, TG, C.M, DH, AW, PK, PR, MB, CO, MW, MB, phenotype. Of note, we have not observed a glioblastoma RS, FW, AB, AD, MP. MES primary cell line in our sample. This might be because of the relatively low number of primary cell lines (n = 9), but ORCID the lower tumor purity in MES glioblastomas might prevent Tobias Kessler https://orcid.org/0000-0001-8350-7074 detection of this phenotype in cell culture. Ahmed Sadik https://orcid.org/0000-0002-0328-1990 In summary, low methylation of DDR genes and TERT Dirk C. Hoffmann https://orcid.org/0000-0003-1370-1933 promoter mutation are associated with worse prognosis in Alba Brandes https://orcid.org/0000-0002-2503-9089 glioblastoma patients and current studies on DDR inhibitors Michael Platten https://orcid.org/0000-0002-4746-887X with and without other cytotoxic or immunological therapies Wolfgang Wick https://orcid.org/0000-0002-6171-634X KESSLER ET AL. | 8385

REFERENCES 18. Bady P, Sciuscio D, Diserens A-C, et al. MGMT methylation 1. Körber V, Yang J, Barah P, et al. Evolutionary trajectories of analysis of glioblastoma on the Infinium methylation BeadChip IDH(WT) glioblastomas reveal a common path of early tumori- identifies two distinct CpG regions associated with gene silenc- genesis instigated years ahead of initial diagnosis. Cancer Cell. ing and outcome, yielding a prediction model for comparisons 2019;35:692-704.e12. across datasets, tumor grades, and CIMP-status. Acta Neuropathol. 2. Ma J, Benitez JA, Li J, et al. Inhibition of nuclear PTEN tyrosine 2012;124:547-560. phosphorylation enhances glioma radiation sensitivity through at- 19. Capper D, Jones DTW, Sill M, et al. DNA methylation-based tenuated DNA repair. Cancer Cell. 2019;36:690-691. classification of central nervous system tumours. Nature. 3. Hegi ME, Diserens AC, Gorlia T, et al. MGMT gene silencing 2018;555:469-474. and benefit from temozolomide in glioblastoma. N Engl J Med. 20. Zheng X, Zhang N, Wu HJ, Wu H. Estimating and accounting for 2005;352:997-1003. tumor purity in the analysis of DNA methylation data from cancer 4. Weller M, Tabatabai G, Kastner B, et al. MGMT promoter methyl- studies. Genome Biol. 2017;18:17. ation is a strong prognostic biomarker for benefit from dose-inten- 21. Lemke D, Weiler M, Blaes J, et al. Primary glioblastoma cultures: sified temozolomide rechallenge in progressive glioblastoma: the can profiling of stem cell markers predict radiotherapy sensitivity? DIRECTOR trial. Clin Cancer Res. 2015;21:2057-2064. J Neurochem. 2014;131:251-264. 5. Wick W, Platten M, Meisner C, et al. Temozolomide chemother- 22. Kessler T, Sahm F, Blaes J, et al. Glioma cell VEGFR-2 confers apy alone versus radiotherapy alone for malignant astrocytoma in resistance to chemotherapeutic and antiangiogenic treatments in the elderly: the NOA-08 randomised, phase 3 trial. Lancet Oncol. PTEN-deficient glioblastoma. Oncotarget. 2015;6:31050-31068. 2012;13:707-715. 23. Struve N, Brend T, Ott L, et al. EGFRvIII: a predictive marker for 6. Kessler T, Sahm F, Sadik A, et al. Molecular differences in IDH Temozolomide response in O6-methylguanine-DNA methyltrans- wildtype glioblastoma according to MGMT promoter methylation. ferase negative glioblastoma cells and tumor xenografts. Neuro Neuro Oncol. 2018;20:367-379. Oncol. 2016;18(Suppl 4):iv33. 7. Wick A, Kessler T, Platten M, et al. Superiority of temozolomide 24. Maréchal A, Li J-M, Ji X, et al. PRP19 transforms into a sensor of over radiotherapy for elderly patients with RTK II methylation RPA-ssDNA after DNA damage and drives ATR activation via a class, MGMT promoter methylated malignant astrocytoma. Neuro ubiquitin-mediated circuitry. Mol Cell. 2014;53:235-246. Oncol. 2020;22(8):1162-1172. 25. Yin J, Zhu JM, Shen XZ. New insights into pre-mRNA pro- 8. Bady P, Kurscheid S, Delorenzi M, et al. The DNA methylome of cessing factor 19: a multi-faceted protein in humans. Biol Cell. DDR genes and benefit from RT or TMZ in IDH mutant low-grade gli- 2012;104:695-705. oma treated in EORTC 22033. Acta Neuropathol. 2018;135:601-615. 26. Yao Z, Di Poto C, Mavodza G, Oliver E, Ressom HW, Sherif ZA. 9. Brown JS, O'Carrigan B, Jackson SP, Yap TA. Targeting DNA DNA methylation activates TP73 expression in hepatocellular car- repair in cancer: beyond PARP inhibitors. Cancer Discov. cinoma and gastrointestinal cancer. Sci Rep. 2019;9:19367. 2017;7:20-37. 27. Stiewe T, Zimmermann S, Frilling A, Esche H, Putzer BM. 10. Lord CJ, Ashworth A. The DNA damage response and cancer ther- Transactivation-deficient DeltaTA-p73 acts as an oncogene. apy. Nature. 2012;481:287-294. Cancer Res. 2002;62:3598-3602. 11. Arita H, Yamasaki K, Matsushita Y, et al. A combination of TERT 28. García-Martínez A, Sottile J, Sánchez-Tejada L, et al. DNA meth- promoter mutation and MGMT methylation status predicts clini- ylation of tumor suppressor genes in pituitary neuroendocrine tu- cally relevant subgroups of newly diagnosed glioblastomas. Acta mors. J Clin Endocrinol Metab. 2019;104:1272-1282. Neuropathol Commun. 2016;4:79. 29. Ohmura H, Ito M, Uchino K, et al. Methylation of drug resis- 12. Nguyen HN, Lie A, Li T, et al. Human TERT promoter mutation tance-related genes in chemotherapy-sensitive Epstein-Barr vi- enables survival advantage from MGMT promoter methylation in rus-associated gastric cancer. FEBS Open Bio. 2020;10:147-157. IDH1 wild-type primary glioblastoma treated by standard chemo- 30. Wick W, Dettmer S, Berberich A, et al. N2M2 (NOA-20) phase I/II radiotherapy. Neuro Oncol. 2017;19:394-404. trial of molecularly matched targeted therapies plus radiotherapy in 13. Abou-El-Ardat K, Seifert M, Becker K, et al. Comprehensive patients with newly diagnosed non-MGMT hypermethylated glio- molecular characterization of multifocal glioblastoma proves blastoma. Neuro Oncol. 2019;21:95-105. its monoclonal origin and reveals novel insights into clonal evolution and heterogeneity of glioblastomas. Neuro Oncol. SUPPORTING INFORMATION 2017;19:546-557. Additional supporting information may be found online in 14. Pearl LH, Schierz AC, Ward SE, Al-Lazikani B, Pearl FM. the Supporting Information section. Therapeutic opportunities within the DNA damage response. Nat Rev Cancer. 2015;15:166-180. 15. Wick W, Gorlia T, Bendszus M, et al. Lomustine and bevacizumab How to cite this article: Kessler T, Berberich A, in progressive glioblastoma. N Engl J Med. 2017;377:1954-1963. Sadik A, et al. Methylome analyses of three 16. Pfaff E, Kessler T, Balasubramanian GP, et al. Feasibility of real-time molecular profiling for patients with newly diagnosed glioblastoma glioblastoma cohorts reveal chemotherapy sensitivity without MGMT promoter hypermethylation-the NCT Neuro Master markers within DDR genes. Cancer Med. Match (N2M2) pilot study. Neuro Oncol. 2018;20:826-837. 2020;9:8373–8385. https://doi.org/10.1002/cam4.3447 17. Duarte RC, Morrison A. Dynamic stability of sequential stimu- lus representations in adapting neuronal networks. Front Comput Neurosci. 2014;8:124.