Integrated pathway analysis identifies prognostically relevant subtypes of glioblastoma characterized by abnormalities in multi‐omics

IA18: Tumor suppressor gene silencing by somatic mutations and promoter methylation leads to genomic instability in head and neck squamous cell carcinoma

Recent molecular characterization of gliomas has uncovered somatic gene variation and DNA methylation changes that are associated with etiology, prognosis, and therapeutic response. Here we describe genomic profiling of gliomas assessed for associations between genetic mutations and patient outcomes, including overall survival (OS) and recurrence-free survival (RFS). Mutations in a 50-gene cancer panel, 1p19q co-deletion, and MGMT promoter methylation (MGMT methylation) status were obtained from tumor tissue of 293 glioma patients. Multivariable regression models for overall survival (OS) and recurrence-free survival (RFS) were constructed for MGMT methylation, 1p19q co-deletion, and gene mutations controlling for age, treatment status, and WHO grade. Mutational profiles of gliomas significantly differed based on WHO Grade, such as high prevalence of BRAF V600E, IDH1, and PTEN mutations in WHO Grade I, II/III, and IV tumors, respectively. In multivariate regression analysis, MGMT methylation and IDH1 mutations were significantly associated with improved OS (HR = 0.44, p = 0.0004 and HR = 0.21, p = 0.007, respectively), while FLT3 and TP53 mutations were significantly associated with poorer OS (HR = 19.46, p < 0.0001 and HR = 1.67, p = 0.014, respectively). MGMT methylation and IDH1 mutations were the only significant alterations associated with improved RFS in the model (HR = 0.42, p < 0.0001 and HR = 0.37, p = 0.002, respectively). These factors were then included in a combined model, which significantly exceeded the predictive value of the base model alone (age, surgery, radiation, chemo, grade) (likelihood ratio test OS p = 1.64 × 10-8 and RFS p = 3.80 × 10-7). This study highlights the genomic landscape of gliomas in a single-institution cohort and identifies a novel association between FLT3 mutation and OS in gliomas.

Aberrant changes in DNA methylation are known to play a major role in the evolution of multiple cancers, but the molecular events responsible for perturbing methylated genomic landscapes have not been completely characterized. In particular, many tumors with dysregulated DNA methylation landscapes do not bear mutations in known regulators of DNA CpG methylation such as IDH1/2, TET2 and DNMT3A. Nevertheless, identification of molecular drivers of aberrant DNA methylation in cancer is essential for directing epigenetic targeted therapy. Thus far, this has proven to be challenging due to various factors including high levels of variation in methylation and multiple co-occurring mutations within a tumor cohort. In order to systematically find mutations that may drive DNA hyper- or hypo- methylation, we analyzed mutation and methylation data from TCGA using a novel computational method based on Boolean implications (if-then rules). The distribution of points in a scatterplot of two variables in a Boolean implication is L-shaped instead of linear, facilitating the discovery of subset (containment) and mutual exclusion relationships between samples with genetic lesions and samples with hypermethylation or hypomethylation of specific CpG sites. We applied the algorithm to recurrent mutations in acute myeloid leukemia (AML), bladder, breast, head &amp; neck, renal clear cell, glioma, lung, ovarian and uterine cancer. Consistent with previous findings, our algorithm identified mutations in IDH2 as a genetic driver of DNA hypermethylation in AML and glioma, and mutation in DNMT3A to be associated with DNA hypomethylation in AML. However, we also found many unexpected associations between somatic mutation and DNA hyper- and hypo- methylation that had not been reported. First, we noted that mutation in the Wilms' Tumor 1 (WT1mut) gene was associated with a high degree of CpG hypermethylation but virtually no hypomethylation. Introduction of a mutant version of WT1 into wildtype AML cells induced consistent DNA hypermethylation in the same set of genes, confirming WT1mut to be causally associated with DNA hypermethylation in AML. We also identified several new associations between somatic mutations and aberrant DNA methylation in several other cancers, including associations between DNA hypomethylation and mutations in (i) STAG2 (a member of the cohesin complex) in bladder cancer and AML, (ii) DNMT3B (an alternative de-novo DNA methyltransferase) in lung adenocarcinomas and (iii) KDM5C (a histone demethylase) clear cell cancer. These data suggest that induction of aberrant DNA methylation by somatic mutation may be an important but relatively under-studied mechanism of cancer evolution. To find potential mutation-specific drug targets, we analysed the genes associated with the hypermethylated CpG sites for each recurrent mutation in AML. Methylated genes in WT1mut samples were highly enriched for polycomb repressor complex 2 (PRC2) targets (q=7.1E-84). Expression of mutant WT1 in normal cord blood stem/progenitor cells induced myeloid differentiation block and treatment of AML primary cells with GSK-126, a small molecule inhibitor of the PRC2/EZH2 complex, promoted myeloid differentiation of WT1mut+ leukemic blasts but not WT1mut- blasts in cell culture. In summary, we present a general method to identify mutation-specific DNA methylation signatures in cancer. This method found a new causal association between WT1mut and DNA hypermethylation in AML that was empirically validated. The WT1mut-specific methylation pattern identified by Boolean implications also revealed a potential treatment strategy for WT1mut AML. The method also found several new candidate genetic drivers of aberrant DNA methylation in other cancers. Our results highlight a strong association between specific mutations and aberrant hyper- or hypo- methylation in cancer and demonstrate that deciphering mutation-specific methylation patterns can lead to therapeutic insights. Citation Format: Subarna Sinha, Daniel Thomas, Ravindra Majeti, David L. Dill. Deciphering the cancer methylome with Boolean implications to find novel drivers of aberrant DNA methylation and actionable drug targets. [abstract]. In: Proceedings of the AACR Special Conference on Computational and Systems Biology of Cancer; Feb 8-11 2015; San Francisco, CA. Philadelphia (PA): AACR; Cancer Res 2015;75(22 Suppl 2):Abstract nr B2-20.

Aberrant changes in DNA methylation are known to play a major role in the evolution of multiple cancers, but the molecular events responsible for perturbing methylated genomic landscapes have not been completely characterized. In particular, many tumors with dysregulated DNA methylation landscapes do not bear mutations in known regulators of DNA CpG methylation such as IDH1/2, TET2 and DNMT3A. Nevertheless, identification of molecular drivers of aberrant DNA methylation in cancer is essential for directing epigenetic targeted therapy. Thus far, this has proven to be challenging due to various factors including high levels of variation in methylation and multiple co-occurring mutations within a tumor cohort. In order to systematically find mutations that may drive DNA hyper- or hypo- methylation, we analyzed mutation and methylation data from TCGA using a novel computational method based on Boolean implications (if-then rules). The distribution of points in a scatterplot of two variables in a Boolean implication is L-shaped instead of linear, facilitating the discovery of subset (containment) and mutual exclusion relationships between samples with genetic lesions and samples with hypermethylation or hypomethylation of specific CpG sites. We applied the algorithm to recurrent mutations in acute myeloid leukemia (AML), bladder, breast, head &amp; neck, renal clear cell, glioma, lung, ovarian and uterine cancer. Consistent with previous findings, our algorithm identified mutations in IDH2 as a genetic driver of DNA hypermethylation in AML and glioma, and mutation in DNMT3A to be associated with DNA hypomethylation in AML. However, we also found many unexpected associations between somatic mutation and DNA hyper- and hypo- methylation that had not been reported. First, we noted that mutation in the Wilms' Tumor 1 (WT1mut) gene was associated with a high degree of CpG hypermethylation but virtually no hypomethylation. Introduction of a mutant version of WT1 into wildtype AML cells induced consistent DNA hypermethylation in the same set of genes, confirming WT1mut to be causally associated with DNA hypermethylation in AML. We also identified several new associations between somatic mutations and aberrant DNA methylation in several other cancers, including associations between DNA hypomethylation and mutations in (i) STAG2 (a member of the cohesin complex) in bladder cancer and AML, (ii) DNMT3B (an alternative de-novo DNA methyltransferase) in lung adenocarcinomas and (iii) KDM5C (a histone demethylase) clear cell cancer. These data suggest that induction of aberrant DNA methylation by somatic mutation may be an important but relatively under-studied mechanism of cancer evolution. To find potential mutation-specific drug targets, we analysed the genes associated with the hypermethylated CpG sites for each recurrent mutation in AML. Methylated genes in WT1mut samples were highly enriched for polycomb repressor complex 2 (PRC2) targets (q=7.1E-84). Expression of mutant WT1 in normal cord blood stem/progenitor cells induced myeloid differentiation block and treatment of AML primary cells with GSK-126, a small molecule inhibitor of the PRC2/EZH2 complex, promoted myeloid differentiation of WT1mut+ leukemic blasts but not WT1mut- blasts in cell culture. In summary, we present a general method to identify mutation-specific DNA methylation signatures in cancer. This method found a new causal association between WT1mut and DNA hypermethylation in AML that was empirically validated. The WT1mut-specific methylation pattern identified by Boolean implications also revealed a potential treatment strategy for WT1mut AML. The method also found several new candidate genetic drivers of aberrant DNA methylation in other cancers. Our results highlight a strong association between specific mutations and aberrant hyper- or hypo- methylation in cancer and demonstrate that deciphering mutation-specific methylation patterns can lead to therapeutic insights. Citation Format: Subarna Sinha, Daniel Thomas, Ravindra Majeti, David L. Dill. Deciphering the cancer methylome with Boolean implications to find novel drivers of aberrant DNA methylation and actionable drug targets. [abstract]. In: Proceedings of the AACR Special Conference on Translation of the Cancer Genome; Feb 7-9, 2015; San Francisco, CA. Philadelphia (PA): AACR; Cancer Res 2015;75(22 Suppl 1):Abstract nr A1-23.

Immunogenomic Landscape of Hematological Malignancies

Lower risk (LR) myelodysplastic syndromes (MDS) are heterogeneous hematopoietic stem and progenitor disorders caused by the accumulation of somatic mutations in various genes including epigenetic regulators that may produce convergent DNA methylation patterns driving specific gene expression profiles. The integration of genomic, epigenomic, and transcriptomic profiling has the potential to spotlight distinct LR-MDS categories on the basis of pathophysiological mechanisms. We performed a comprehensive study of somatic mutations and DNA methylation in a large and clinically well-annotated cohort of treatment-naive patients with LR-MDS at diagnosis from the EUMDS registry (ClinicalTrials.gov.NCT00600860). Unsupervised clustering analyses identified six clusters based on genetic profiling that concentrate into four clusters on the basis of genome-wide methylation profiling with significant overlap between the two clustering modes. The four methylation clusters showed distinct clinical and genetic features and distinct methylation landscape. All clusters shared hypermethylated enhancers enriched in binding motifs for ETS and bZIP (C/EBP) transcription factor families, involved in the regulation of myeloid cell differentiation. By contrast, one cluster gathering patients with early leukemic evolution exhibited a specific pattern of hypermethylated promoters and, distinctly from other clusters, the upregulation of AP-1 complex members FOS/FOSL2 together with the absence of hypermethylation of their binding motif at target gene enhancers, which is of relevance for leukemic initiation. Among MDS patients with lower-risk IPSS-M, this cluster displayed a significantly inferior overall survival (p < 0.0001). Our study showed that genetic and DNA methylation features of LR-MDS at early stages may refine risk stratification, therefore offering the frame for a precocious therapeutic intervention.

Astrocytomas are the most common primary central nervous system tumors in adults. The WHO classification based on histologial characteristics according consider four malignant grades, glioblastoma (GBM) being the most malignant (grade IV). Molecular classification of GBM is based on the integration of somatic mutations, DNA methylation and specific transcript/protein expression and consider four subtypes: proneural, classic, mesenchymal and neural. Lysyl oxidase family is composed of five members (LOX, LOXL1, LOXL2, LOXL3 and LOXL4) which are enzymes responsible for catalyzing lysine-derived cross-links of collagen and elastin. The aim of the present work was to analyze the gene expression of LOXL3 in astrocytomas of different grades and different GBM molecular subtypes, the impact of LOXL3 expression level on overall survival in GBM cases of our cohort. These data were also analyzed in the TCGA database to validate our findings. LOXL3expression increased with astrocytoma malignant grades, and proved to be a prognostic factor, as GBM patients with lower LOXL3 expression presented longer survival time than those with higher expression (p&amp;lt;0.041). Analogous data was obtained in the TCGA database, when considering temozolamide-treated GBM patients (p=0.049). Additionally, LOXL3 expression was higher in mesenchymal GBM subtype than in proneural and classic GBM subtypes in TCGA cohort (p&amp;lt;0.001). To further understand LOXL3 functional role in gliomagenesis, LOXL3 was silenced with two distinct siRNA sequences in U87MG, a mesenchymal subtype of GBM cell line. LOXL3 down regulation was confirmed at gene and protein expression levels. Functional assays with silenced LOXL3 demonstrated a decrease in the cell proliferation (p&amp;lt;0.05), and an increase in the apoptosis associated to temozolamide treatment (p&amp;lt;0.05). Immunofluorescence analysis in U87MG cells showed LOXL3 colocalized with mitochondria. Interestingly, such colocalization decreased when LOXL3 was downregulated, associated to a clear morphological alteration in mitochondria shape. A rough increase in mitochondria volume was observed, suggesting fused mitochondria. In addition, such condition was associated to a 23% decrease in mitochondrial DNA copy number. Altogether, our findings suggest that LOXL3 is upregulated in GBM, with impact in prognosis and might play a role in mitochondria dynamics and/or mitophagy, modulating apoptosis. Transcriptome analysis, currently in progress, will further clarify the associated players in this signaling pathway. Citation Format: Talita de Souza Laurentinho, Roseli da Silva Soares, Suely Kazue Marie, Sueli Mieko Oba-Shinjo. Expression profile and role of LOXL3 in astrocytomas [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2019; 2019 Mar 29-Apr 3; Atlanta, GA. Philadelphia (PA): AACR; Cancer Res 2019;79(13 Suppl):Abstract nr 5175.

BackgroundBladder cancer has numerous genomic features that are potentially actionable by targeted agents. Nevertheless, both pre-clinical and clinical research using molecular targeted agents have been very limited in bladder cancer.ResultsWe created the Genomics of Drug Sensitivity in Bladder Cancer (GDBC) database, an integrated database (DB) to facilitate the genomic understanding of bladder cancer in relation to drug sensitivity, in order to promote potential therapeutic applications of targeted agents in bladder cancer treatment. The GDBC database contains two separate datasets: 1) in-house drug sensitivity data, in which 13 targeted agents were tested against 10 bladder cancer cell lines; 2) data extracted and integrated from public databases, including the Cancer Therapeutics Research Portal, Cancer Cell Line Encyclopedia, Genomics of Drug Sensitivity in Cancer, Kyoto Encyclopedia of Genes and Genomes, and the Cancer Gene Census databases, as well as bladder cancer genomics data and synthetic lethality/synthetic dosage lethality connections.ConclusionsGDBC is an integrated DB of genomics and drug sensitivity data with a specific focus on bladder cancer. With a user-friendly web-interface, GDBC helps users generate genomics-based hypotheses that can be tested experimentally using drugs and cell lines included in GDBC.

BackgroundDifferential distribution of DNA methylation on the parental alleles of imprinted genes distinguishes the alleles from each other and dictates their parent of origin-specific expression patterns. While differential DNA methylation at primary imprinting control regions is inherited via the gametes, additional allele-specific DNA methylation is acquired at secondary sites during embryonic development and plays a role in the maintenance of genomic imprinting. The precise mechanisms by which this somatic DNA methylation is established at secondary sites are not well defined and may vary as methylation acquisition at these sites occurs at different times for genes in different imprinting clusters.ResultsIn this study, we show that there is also variability in the timing of somatic DNA methylation acquisition at multiple sites within a single imprinting cluster. Paternal allele-specific DNA methylation is initially acquired at similar stages of post-implantation development at the linked Dlk1 and Gtl2 differentially methylated regions (DMRs). In contrast, unlike the Gtl2-DMR, the maternal Dlk1-DMR acquires DNA methylation in adult tissues.ConclusionsThese data suggest that the acquisition of DNA methylation across the Dlk1/Gtl2 imprinting cluster is variable. We further found that the Dlk1 differentially methylated region displays low DNA methylation fidelity, as evidenced by the presence of hemimethylation at approximately one-third of the methylated CpG dyads. We hypothesize that the maintenance of DNA methylation may be less efficient at secondary differentially methylated sites than at primary imprinting control regions.

Considering the remarkable heterogeneity of biological features of renal cell carcinoma (RCC), the current clinical classification that only relies on classic clinicopathological features is in urgent need of improvement. Herein, we aimed to conduct DNA methylation modification patterns in RCC. We retrospectively curated multiple RCC cohorts, comprising TCGA-KIRC, TCGA-KICH, TCGA-KIRP, and E-MTAB-1980. DNA methylation modification patterns were proposed with an unsupervised clustering algorithm based on 20 DNA methylation regulators. Immunological features were characterized using tumor-infiltrating immune cells and immunomodulators. Sensitivity to immuno- or targeted therapy was estimated with submap and Genomics of Drug Sensitivity in Cancer (GDSC). DNA methylation score (DMS) was developed with principal component analysis. Three DNA methylation modification patterns were conducted across RCC patients, namely C1, C2 and C3. Among them, C3 displayed the most remarkable survival advantage. The three patterns presented in agreement with immune phenotypes: immune-desert, immune-excluded, and immune-inflamed, respectively. These patterns displayed distinct responses to anti-PD-1 and targeted drugs. DMS enabled the quantification of DNA methylation status individually as an alternative tool for prognostic estimation. The DNA methylation molecular patterns we proposed are an innovative complement to the traditional classification of RCC, which might contribute to precision medicine.

The Genomic of Drug Sensitivity in Cancer (GDSC; www.cancerRxgene.org) resource facilitates development of targeted cancer therapies through pre-clinical identification of therapeutic biomarkers. GDSC is the largest public resource for information on drug sensitivity in cancer cells and links these data to extensive genomic information to identify molecular features that influence anticancer drug response. There is compelling evidence that alterations in cancer genomes strongly influence clinical responses to anticancer therapies. There are several examples where genomic changes are used as molecular biomarkers to stratify patients most likely to benefit from a treatment (e.g. BRAF in melanoma). Despite these successes, the majority of cancer drugs have not been linked to specific molecular features that could be used to direct their clinical use to maximize patient benefit. We are using pharmacogenomic profiling in cancer cell lines as a biomarker discovery platform by systematically linking pharmacological data with genomic information in cancer cells. The GDSC database contains drug sensitivity data generated from high-throughput screening performed by the Cancer Genome Project at the Wellcome Trust Sanger Institute and the Center for Molecular Therapeutics at Massachusetts General Hospital using a collection of &amp;gt;1,200 cancer cell lines. GDSC release v3 (November 2012) contains drug sensitivity data for almost 80,000 experiments, describing response to 142 anticancer drugs across over 700 cancer cell lines. To identify molecular markers of drug response, cell line drug sensitivity data are integrated with large genomic datasets obtained from COSMIC (Catalogue of Somatic Mutations in Cancer), including information on somatic mutations in cancer genes, gene amplification and deletion, tissue type and transcriptional data. Analysis of GDSC data is through a web portal based on queries of specific anticancer drugs or cancer genes. Interactive graphical representations of the data are used throughout with links to related resources, and all datasets are freely available and downloadable. The GDSC database will undergo significant expansion in coming years as drug sensitivity and genomic datasets increase in size and complexity. GDSC provides a unique public resource incorporating large drug sensitivity and genomic datasets to facilitate discovery of new therapeutic biomarkers for cancer therapies. Citation Format: Wanjuan Yang, Jorge Soares, Patricia Greninger, Elena Edelman, Howard Lightfoot, Simon Forbes, Ramaswamy Sridhar, P. Andrew Futreal, Daniel Haber, Michael Stratton, Cyril Benes, Ultan McDermott, Mathew Garnett. Genomics of Drug Sensitivity in Cancer (GDSC): A resource for therapeutic biomarker discovery in cancer cells. [abstract]. In: Proceedings of the 104th Annual Meeting of the American Association for Cancer Research; 2013 Apr 6-10; Washington, DC. Philadelphia (PA): AACR; Cancer Res 2013;73(8 Suppl):Abstract nr 2206. doi:10.1158/1538-7445.AM2013-2206

SummaryCholangiocarcinoma (CCA) is an aggressive malignancy of the bile ducts, with poor prognosis and limited treatment options. Here, we describe the integrated analysis of somatic mutations, RNA expression, copy number, and DNA methylation by The Cancer Genome Atlas of a set of predominantly intrahepatic CCA cases and propose a molecular classification scheme. We identified an IDH mutant-enriched subtype with distinct molecular features including low expression of chromatin modifiers, elevated expression of mitochondrial genes, and increased mitochondrial DNA copy number. Leveraging the multi-platform data, we observed that ARID1A exhibited DNA hypermethylation and decreased expression in the IDH mutant subtype. More broadly, we found that IDH mutations are associated with an expanded histological spectrum of liver tumors with molecular features that stratify with CCA. Our studies reveal insights into the molecular pathogenesis and heterogeneity of cholangiocarcinoma and provide classification information of potential therapeutic significance.

Accurately predicting the response of a cancer patient to a therapeutic agent remains an important challenge in precision medicine. With the rise of data science, researchers have applied computational models to study the drug inhibition effects on cancers based on cancer genomics and transcriptomics. Moreover, a common epigenetic modification, DNA methylation, has been related to the occurrence and development of cancer, as well as drug effectiveness. Therefore, it is helpful for improvement of drug response prediction through exploring the relationship between DNA methylation and drug effectiveness. Here, we proposed a computational model to predict drug responses in cancers through integration of cancer genomics, transcriptomics, epigenomics, and compound chemical properties. Meanwhile, we applied a regularized regression model (Least Absolute Shrinkage and Selection Operator, lasso) to detect the methylation sites that were closely related to drug effectiveness. The prediction models were trained on a well-known pharmacogenomics data resource, Genomics of Drug Sensitivity in Cancer (GDSC). The cross-validation indicates that the performance of the prediction model using DNA methylation is comparable to that of using other cancer omics, including oncogene mutation and gene expression data. It indicates the important role of DNA methylation in prediction of drug responses. Encyclopedia of DNA Elements (ENCODE) and Transcriptional Regulatory Relationships Unraveled by Sentence-based Text mining (TRRUST2) database analyses suggest that the methylation sites associated with drug effectiveness are mainly located in the transcription factor (TF) binding region. Therefore, we hypothesized that the sensitivity of cancer cells to drugs could be regulated by changing the methylation modification of TF binding region. In conclusion, we confirmed the important role of DNA methylation in prediction of drug responses, and provided some methylation sites that closely related to the drug effectiveness, which may be a great regulatory target for improvement of drug treatment effects on cancer patients.

&lt;p&gt;Supplementary Figure 1. Independent validation of C-CIMP A) Supervised clustering of TCGA samples in ccRCC validation dataset assessed by 27k Infinium arrays. The (beta value) level of DNA methylation is represented with a color scale. Each column represents a sample; each row a probe set. The transcriptomic subtype in TCGA, the copy number variation at 9p23.1 locus (CDKN2A), somatic mutation status of four genes (PBRM1, BAP1, SETD2 and VHL) are indicated by red, green and gray squares, with annotations in the legend. B) Kaplan-Meier curves showing distinct outcome of patients according to the three subgroups of DNA methylation classification, with patients belonging to C-CIMP subgroup having the worst outcome. Supplementary Figure 2. Expression vs DNA methylation between C-CIMP and no-CIMP subgroups using volcano and starburst plots. A) All CpG loci analyzed for C-CIMP association; x-axis: β-value difference in DNA methylation between C-CIMP and no-CIMP clusters; y-axis: p-value of the corrected FDR between the two groups; red: probes significantly different between the two subgroups. B) TCGA promoter DNA methylation vs. gene expression analyzed by RNAseq. Log-10 (FDR adjusted p-value) is plotted for DNA methylation (x-axis) and gene expression for each gene (y-axis). Green indicates genes that are downregulated for gene expression and which gain DNA methylation in C-CIMP cluster versus no-CIMP cluster. Red indicates genes that are upregulated for gene expression and gain DNA methylation. Supplementary Figure 3. Association between somatic mutations and DNA methylation subgroups. A) Mutational load is increased in C-CIMP as compared to no-CIMP and low-CIMP subgroups. B) Association between somatic mutations of BAP1 and SETD2 in C-CIMP subgroup as compared to low-CIMP and no-CIMP subgroups. Supplementary Figure 4. Predictive value of VEGF receptor methylation to predict response to sunitinib. Kaplan-Meier curves for progression-free survival of patients treated with sunitinib according to tumor methylation status of (A) FLT4, (B) FLT1, and (C) FLT3 genes in Beuselinck study. Supplementary Figure 5. C-CIMP subgroup in metastatic ccRCC A) Supervised clustering of DNA methylation in Beuselinck study was consistent with the three DNA methylation subgroups C-CIMP, low-CIMP and no-CIMP. B) Kaplan-Meier curves for progression-free survival of patients treated with sunitinib according to their CIMP status. Supplementary Figure 6. Correlation between methylome signature of NSD1 mutations of patients with Sotos syndrome and ccRCC. A) Supervised clustering of DNA methylation in TCGA ccRCC using genome-wide methylome signature of NSD1 mutations in Sotos syndrome. B) Kaplan-Meier curves for overall survival of patients with ccRCCs according to NSD1 mutations in genome-wide methylome signature. Note that the analyzed cohort encompasses 271 ccRCCs assessed by Infinium 450K arrays.&lt;/p&gt;

&lt;p&gt;Supplementary Figure 1. Independent validation of C-CIMP A) Supervised clustering of TCGA samples in ccRCC validation dataset assessed by 27k Infinium arrays. The (beta value) level of DNA methylation is represented with a color scale. Each column represents a sample; each row a probe set. The transcriptomic subtype in TCGA, the copy number variation at 9p23.1 locus (CDKN2A), somatic mutation status of four genes (PBRM1, BAP1, SETD2 and VHL) are indicated by red, green and gray squares, with annotations in the legend. B) Kaplan-Meier curves showing distinct outcome of patients according to the three subgroups of DNA methylation classification, with patients belonging to C-CIMP subgroup having the worst outcome. Supplementary Figure 2. Expression vs DNA methylation between C-CIMP and no-CIMP subgroups using volcano and starburst plots. A) All CpG loci analyzed for C-CIMP association; x-axis: β-value difference in DNA methylation between C-CIMP and no-CIMP clusters; y-axis: p-value of the corrected FDR between the two groups; red: probes significantly different between the two subgroups. B) TCGA promoter DNA methylation vs. gene expression analyzed by RNAseq. Log-10 (FDR adjusted p-value) is plotted for DNA methylation (x-axis) and gene expression for each gene (y-axis). Green indicates genes that are downregulated for gene expression and which gain DNA methylation in C-CIMP cluster versus no-CIMP cluster. Red indicates genes that are upregulated for gene expression and gain DNA methylation. Supplementary Figure 3. Association between somatic mutations and DNA methylation subgroups. A) Mutational load is increased in C-CIMP as compared to no-CIMP and low-CIMP subgroups. B) Association between somatic mutations of BAP1 and SETD2 in C-CIMP subgroup as compared to low-CIMP and no-CIMP subgroups. Supplementary Figure 4. Predictive value of VEGF receptor methylation to predict response to sunitinib. Kaplan-Meier curves for progression-free survival of patients treated with sunitinib according to tumor methylation status of (A) FLT4, (B) FLT1, and (C) FLT3 genes in Beuselinck study. Supplementary Figure 5. C-CIMP subgroup in metastatic ccRCC A) Supervised clustering of DNA methylation in Beuselinck study was consistent with the three DNA methylation subgroups C-CIMP, low-CIMP and no-CIMP. B) Kaplan-Meier curves for progression-free survival of patients treated with sunitinib according to their CIMP status. Supplementary Figure 6. Correlation between methylome signature of NSD1 mutations of patients with Sotos syndrome and ccRCC. A) Supervised clustering of DNA methylation in TCGA ccRCC using genome-wide methylome signature of NSD1 mutations in Sotos syndrome. B) Kaplan-Meier curves for overall survival of patients with ccRCCs according to NSD1 mutations in genome-wide methylome signature. Note that the analyzed cohort encompasses 271 ccRCCs assessed by Infinium 450K arrays.&lt;/p&gt;