IntroductionBrain gliomas are among the most malignant and fatal forms of cancer, known for their highly invasive nature and resistance to standard therapies1,2. These tumors account for a significant proportion of primary brain tumors and pose a substantial clinical challenge due to their poor prognosis and limited treatment options3,4,5. A comprehensive understanding of the molecular and genetic foundations of glioma development and progression is essential for the identification of novel therapeutic targets and the improvement of patient outcomes, a challenge our group has previously addressed3,6.Recent advances in quantitative trait locus (QTL) mapping have revolutionized our understanding of the genetic architecture underlying complex traits, including cancer7,8,9. QTLs are genomic regions associated with variations in quantitative traits, and they encompass several types relevant to cancer research, such as:Expression QTLs (eQTLs): These identify genomic loci that influence gene expression levels10,11. Methylation QTLs (mQTLs): These loci affect DNA methylation patterns, which can regulate gene activity12,13. Splicing QTLs (sQTLs): These loci impact RNA splicing events, potentially altering protein function14,15. Histone modification QTLs (haQTLs): These loci influence histone modifications, affecting chromatin structure and gene expression16,17. Chromatin accessibility QTLs (caQTLs): These loci affect the accessibility of DNA to transcription factors and other regulatory proteins18,19. Integrative analyses of these diverse QTLs provide a holistic view of how genetic variations can influence tumor biology through multiple molecular mechanisms20. Such comprehensive approaches are particularly valuable in uncovering the complex interplay of genetic factors involved in glioma pathogenesis.The 9p21.3 chromosomal region, which harbors critical tumor suppressor genes such as CDKN2A and CDKN2B, alongside the long non-coding RNA (lncRNA) CDKN2B-AS1 (also known as ANRIL), has been robustly implicated in susceptibility to a spectrum of human cancers, including glioma6,21. Genome-wide association studies (GWAS) have consistently identified SNPs within this locus as being significantly associated with glioma risk22,23. However, the precise molecular mechanisms through which these risk-associated SNPs, such as the rs615552_AL359922.1 variant investigated herein, contribute to glioma pathogenesis remain inadequately defined. Specifically, how these variants modulate the expression of adjacent genes(including CDKN2B-AS1), influence local epigenetic landscapes, and ultimately drive oncogenesis in the context of brain cancer requires further clarification24.CDKN2B-AS1 itself is recognized as a multifaceted lncRNA with complex regulatory functions in numerous malignancies, impacting fundamental cellular processes such as cell cycle progression, apoptosis, and immune modulation21,25. Despite these insights, a systematic elucidation of CDKN2B-AS1’s expression dynamics, its interaction with specific risk-associated SNPs (like rs615552_AL359922.1), and its broader regulatory networks in distinct glioma subtypes is still required26.Therefore, to address these knowledge gaps, the present study selected SNP rs615552_AL359922.1 and lncRNA CDKN2B-AS1 as focal points for a detailed investigation. We employed a comprehensive set of quantitative trait locus (QTL) analyses—encompassing expression QTLs (eQTLs), methylation QTLs (mQTLs), splicing QTLs (sQTLs), histone modification QTLs (haQTLs), and chromatin accessibility QTLs (caQTLs)—to systematically investigate the genetic underpinnings of their roles in brain gliomas. Our primary goal was to identify how genetic variation at this locus, exemplified by rs615552_AL359922.1, influences molecular traits related to CDKN2B-AS1 and its neighboring genes, thereby elucidating their potential contributions to glioma pathogenesis. Simultaneously, our genome-wide analytical approach allows for the exploration of other critical glioma-associated loci, such as the well-established risk locus at 5p15.33 harboring the TERT gene, to uncover potentially shared pathogenic mechanisms. Additionally, we aimed to further explore the broader disease associations of the identified genes through phenome-wide association study (pheWAS) colocalization analysis, which can provide insights into shared genetic mechanisms between gliomas and other complex traits.Overall, this study seeks to advance our understanding of the genetic landscape of brain gliomas and identify potential therapeutic targets by integrating multiple QTL approaches and pheWAS colocalization analysis, with a specific focus on the functional implications of rs615552_AL359922.1 and the regulatory role of CDKN2B-AS1. The findings may have important implications for unraveling the complex etiology of gliomas and inform future research directions in this field.MethodData sourceThe data sources utilized in this study encompass various types of quantitative trait loci (QTL) datasets and genome-wide association studies (GWAS).GWAS summary statistics datasets: We have gathered GWAS summary statistics from a variety of prominent databases, including the Neale Lab UKBB v327, the Open GWAS Catalog28, PGC29 and PhenoScanner30. To ensure compatibility with diverse analytical methods, these publicly accessible datasets of GWAS summary association statistics have been meticulously processed and standardized. This involved a series of quality control (QC) measures, including verification of the genome reference (hg19), generation of headers, calculation of any missing statistical data, alignment of alleles with a reference panel, and elimination of non-biallelic SNPs as well as those with a minor allele frequency (MAF) less than 0.01. For brain cancer susceptibility analysis, we utilized the GWAS summary statistics dataset with ID ieu-b-4875 from the Open GWAS Catalog, which included 606 brain cancer cases and 372,016 controls. This dataset was treated as the ‘exposure’ in our colocalization analyses aimed at identifying shared genetic variants with molecular QTLs or other traits.eQTL Data: The GTEx eQTL data was obtained for cis-eQTLs in 21 tissues from GTEx v830,31, including: Adipose Subcutaneous (n = 581), Adipose Visceral Omentum (n = 469), Brain Amygdala (n = 129), Brain Anterior Cingulate Cortex BA24 (n = 147), Brain Caudate Basal Ganglia (n = 194), Brain Cerebellar Hemisphere (n = 175), Brain Cerebellum (n = 209), Brain Cortex (n = 205), Brain Frontal Cortex BA9 (n = 175), Brain Hippocampus (n = 165), Brain Hypothalamus (n = 170), Brain Nucleus Accumbens Basal Ganglia (n = 202), Brain Putamen Basal Ganglia (n = 170), Brain Spinal Cord Cervical C-1 (n = 124), Brain Substantia Nigra (n = 114), Cells Cultured Fibroblasts (n = 483), Cells EBV-Transformed Lymphocytes (n = 147), Muscle Skeletal (n = 706), Nerve Tibial (n = 532), Pituitary (n = 237), Whole Blood (n = 670). EQTLGen data32 was obtained for cis-eQTLs in whole blood (n = 31,684) from eQTLGen. BrainMeta data was obtained for cis-eQTLs in brain cortex samples (n = 2,865) from the SMR website. Additionally, eQTL data from the Religious Orders Study and Memory and Aging Project (ROSMAP), hereafter referred to as the ROSMAP_CMC_eQTL dataset, were also utilized. These data were derived from dorsolateral prefrontal cortex samples (n = 494) as described by Ng et al. (2017, PMID: 28,869,584)33 and accessed via resources like the xQTL server.mQTL Data: Brain-mMeta data was obtained for cis-mQTLs in brain samples (estimated effective n = 1,160) from the SMR website34,35. FB_Brain data (Hannon et al. 2015)36 was obtained for cis-mQTLs in Prefrontal Cortex (PFC), Striatum (STR), and Cerebellum (n = 166). Data from LBC_BSGS37,38 was obtained for cis-mQTLs in whole blood (n = 1,980) from the SMR website. US_Blood data39 was obtained for cis-mQTLs in whole blood (n = 1,175) from the SMR website.haQTL Data: ROSMAP_haQTL data (Ng B et al. 2017, PMID: 28,869,584)33 was obtained for haQTLs in Brain Cortex (n = 561) from Mostafavi Lab.caQTL Data: Bryois_caQTL data (Bryois et al. 2018, PMID: 30,087,329)40 was obtained for cis-caQTLs in Prefrontal Cortex (135 schizophrenia patients and 137 controls) from the SMR website.sQTL Data: GTEx sQTL data was obtained for cis-sQTLs in 21 tissues (the same as the GTEx eQTL data) from GTEx v8.COLOC analysisThe colocalization examination was executed utilizing the COLOC approach41, which applies an approximate Bayes factor evaluation to ascertain whether a pair of traits is associated with shared causal genetic variations within a defined genomic segment. This method assesses the posterior probabilities associated with five potential scenarios: H0, signifying the absence of genetic linkage for both traits; H1/H2, suggesting a genetic link for a single trait exclusively; H3, proposing that each trait is tied to separate causal genetic factors; and H4, suggesting a joint causal variant for both traits. The procedure was carried out using the COLOC R package’s default settings where not otherwise specified. For the analysis, stringent prior probabilities were selected (p1 = 1 × 10–4, p2 = 1 × 10–4, and p12 = 1 × 10–5) to account for a causal SNP in either the primary trait (e.g., GWAS) or a molecular quantitative trait locus (e.g., eQTL, mQTL; referred to as multi-QTL in this context), as well as for a common causal SNP affecting both. The pairwise colocalization posterior probabilities were calculated for an array of pairings, encompassing both GWAS-GWAS comparisons and multi-QTL-GWAS combinations. To identify robust colocalization signals, the established thresholds were strictly adhered to (PP4 > 0.75, and PP4/PP3 > 3). For visual representation of regional associations, the “LocusCompareR” R package42, modified for this purpose, was employed to generate plots.Co-occurrence and co-expression analysisThe examination of co-occurrence and mutual exclusivity was conducted utilizing cBioPortal platform (version 1.3.2, developed by the Memorial Sloan Kettering Cancer Center in New York)43. This analysis leveraged the comprehensive datasets from The Cancer Genome Atlas (TCGA)44 for lower-grade glioma (LGG) and glioblastoma multiforme (GBM), as well as the Cancer Proteome Atlas (CPTAC) Glioma data45. The aim was to explore the concurrent presence and potential interdependence of genetic alterations across these cancer types, providing insights into their molecular interplay.Gene mutation analysisWe obtained somatic mutation data in Mutation Annotation Format (MAF) files for the TCGA-LGG and TCGA-GBM cohorts from the GDC portal. These MAF files, derived from whole-exome sequencing, were analyzed using the ‘maftools’ R package46 to summarize and visualize the mutational profiles, including identifying frequently mutated genes, mutation types, and variant allele frequencies within these glial neoplasms.Copy number variation (CNV) analysisCopy number variation data for CDKN2A, CDKN2B, and TERT in the TCGA-LGG and TCGA-GBM cohorts were obtained from cBioPortal or directly from the TCGA GDC data portal. The CNV data are typically categorized as homozygous deletion, heterozygous deletion, diploid, heterozygous amplification, or homozygous amplification ( based on GISTIC2 scores or similar thresholds provided by TCGA). We analyzed the distribution of these CNV categories for the selected genes in LGG and GBM. Furthermore, the association between CNV status and mRNA expression levels (obtained as described previously) was assessed using appropriate statistical tests (ANOVA or Kruskal–Wallis test), and the prognostic significance of CNVs was evaluated as described in the ‘Survival Analysis’ section.Overexpression functional enrichment analysisIn both the TCGA-LGG and TCGA-GBM cohorts, we conducted differential expression analysis (DEseq247) using the target gene CDKN2B-AS1’s high and low expression levels as the basis for grouping. Genes with significant differential expression (Log2FC > 1, P.adj 0.9, PP4/PP3 > 10). Two of these traits colocalized at the same coloc_snp (rs944801), both representing the same trait, glaucoma (primary open-angle) (GCST006065, GCST90011770), from different cohorts. The other trait, glaucoma (multi-trait analysis) (PMID:31,959,993), colocalized at coloc_snp: rs6475604.In the colocalization analysis between non-epithelial skin cancer (UKBB) and brain cancer, SNP rs10811645, located within the CDKN2B-AS1 gene (chromosome 9p21.3), exhibited a posterior probability of colocalization (PP4) of 0.2184 and a PP4/PP3 ratio of 0.2795. While this PP4 value indicates a low probability and does not meet the stringent criterion (PP4 > 0.75) established for robust colocalization (see Methods ), this observation is reported as an exploratory finding that warrants further investigation.The colocalization analysis between glaucoma (UKBB) and brain cancer revealed that the SNP rs2157719 in the CDKN2B-AS1 gene (chromosome 9p21.3) has a very high colocalization probability (PP4 = 0.9562) and a PP4/PP3 ratio of 21.8249. This strongly suggests that CDKN2B-AS1 may play a crucial role in the pathogenesis of both glaucoma and brain cancer, indicating these diseases may share the same genetic variant.The colocalization analysis between long-term illness, disability, or infirmity (UKBB) and brain cancer identified the SNP rs6475604 in the CDKN2B-AS1 gene (chromosome 9p21.3) with a colocalization probability (PP4) of 0.1143 and a PP4/PP3 ratio of 0.1290. Similarly, the low colocalization probability (PP4 = 0.1143) for long-term illness, disability, or infirmity does not meet the threshold for robust colocalization, and this observation is noted as an exploratory finding requiring further study.The CDKN2B-AS1 gene shows significant colocalization signals across multiple diseases, particularly between glaucoma and brain cancer, with a PP4 of 0.9562 and a PP4/PP3 ratio of 21.8249, suggesting that CDKN2B-AS1 plays a critical role in the pathogenesis of these diseases. Although the colocalization probability is lower in non-epithelial skin cancer and long-term illness, the results still indicate the potential importance of CDKN2B-AS1.We analyzed the loci of the CDKN2B-AS1 and TERT genes and explored the genetic correlations with diseases associated with these loci (Fig. 1I, J). CDKN2B-AS1, CDKN2A, CDKN2B, and AL359922.1 are located in close proximity on chromosome 9p21.3, while TERT is located on chromosome 5p15.33. These genes exhibit high genetic correlations with various diseases. CDKN2B-AS1 shows significant genetic correlations with diseases such as coronary atherosclerosis, ischemic heart disease, and myocardial infarction, which are related to the circulatory system. In contrast, TERT shows significant genetic correlations with diseases such as other non-epithelial skin cancer, hyperplasia of the prostate, and uterine leiomyoma, which are mainly tumor-related.These findings provide important clues for further research into the specific functional mechanisms of these SNPs and genes in brain cancer and other diseases, potentially aiding in the development of new therapeutic strategies and diagnostic methods.Copy number and methylation signatures of CDKN2A, CDKN2B, and TERT in brain cancer prognosisIn the TCGA-LGG/GBM cohorts, we compared the copy number variations (CNVs) and other modal information for the genes CDKN2A, CDKN2B, and TERT (Fig. 2A). Initially, we examined the distribution of five types of CNVs (None, Homo.Del., Hete.Del., Homo.Amp., Hete.Amp.) in both LGG and GBM. We observed similar distribution patterns for CDKN2A and CDKN2B in LGG and GBM. In LGG, the majority of samples had no CNV, followed by Hete.Del., with almost no instances of Homo.Amp. In contrast, in GBM, most samples exhibited Homo.Del., followed by Hete.Del., and similarly, there were no instances of Homo.Amp. For the TERT gene, over 75% of samples in both LGG and GBM did not exhibit CNV. However, Hete.Amp. was significantly more prevalent in GBM compared to LGG (Fig. 2B, C).Fig. 2Analysis of CNVs, Gene Expression, and Methylation in TCGA-LGG/GBM Cohorts. (A) Distribution of CNV Types for CDKN2A, CDKN2B, and TERT genes. (B, C) Prevalence of Heterozygous Amplification (Hete.Amp.) in GBM versus LGG. Panels 2B and 2C highlight the significantly higher prevalence of Hete.Amp. in GBM compared to LGG for the TERT gene, with over 75% of samples in both tumor types lacking CNV. (D) CNV Association with Prognosis in LGG and GBM. This panel illustrates the significant association between CNVs of CDKN2A and CDKN2B and various survival metrics—Disease-Specific Survival (DSS), Overall Survival (OS), and Progression-Free Survival (PFS)—in LGG, with a log-rank P value