Integration of Multi-omics Data Revealed the Orphan CpG Islands and Enhancer-dominated Cis-regulatory Network in Glioma

Integration of Multi-omics Data Revealed the Orphan CpG Islands and Enhancer-dominated Cis-regulatory Network in Glioma | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Research Article Integration of Multi-omics Data Revealed the Orphan CpG Islands and Enhancer-dominated Cis-regulatory Network in Glioma jiawei yao, Penglei Yao, Yang Li, ke he, xinqi ma, Qingsong Yang, and 6 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-3959082/v1 This work is licensed under a CC BY 4.0 License Status: Posted Version 1 posted You are reading this latest preprint version Abstract Background: The complex transcriptional regulatory network leads to the poor prognosis of glioma. The role of orphan CpG islands (oCGIs) in the transcriptional regulatory network has been overlooked. Establishing a sophisticated transcriptional regulatory system is paramount. Methods: We constructed different cis-regulatory models through mutual information and Bayesian networks. We utilized eleven machine learning algorithms to develop classifiers that could effectively integrate multi-omics datasets. we utilized single-cell multi-omics data construct a higher-resolution cis-regulatory network. To investigate the binding interaction between oCGIs and transcription factors, we utilized chromatin immunoprecipitation assay and qRT-PCR. Furthermore, we assessed the proliferative status of various glioma subtypes using the MTT assay and immunohistochemistry. Results: The cis-regulatory network dominated by oCGIs and enhancers was significantly active in the glioma subtypes, mainly characterized by glioblastoma (Cluster 2). Direct regulation of target genes by oCGIs or enhancers is of great importance in the cis-regulatory network. Furthermore, based on single-cell multi-omics data, we found that the highly activated cis-regulatory network in Cluster 2 sustains the high proliferative potential of glioma cells. The upregulation of oCGIs and enhancers related genes in Cluster 2 results in glioma patients exhibiting resistance to radiotherapy and chemotherapy. These findings were further validated through glioma cell line related experiments. Conclusion: Our study systematically elucidated the cis-regulatory role of oCGIs for the first time. The comprehensive characterization of the multi-omics features of the oCGIs- and enhancers-dominated cis-regulatory network offers a novel insight into the pathogenesis of glioma and provides new strategies to treat this challenging disease. oCGIs Enhancer Machine learning Multi-omics Cis-regulatory network Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Figure 6 Introduction Glioma is the most prevalent primary tumor of the brain and spinal cord, with glioblastoma multiforme (GBM) being the most frequent primary malignant tumor of the brain and central nervous system. It accounts for 14.3% of all tumors and 49.1% of malignant tumors. Despite significant efforts from both basic and clinical researchers, the median survival of glioblastoma patients remains only 8 months 1 . The underlying cause of this poor prognosis is the intricate regulatory network of glioma, endowing it with the capacity to adapt to various hostile environments. Elucidating the role of a single target is considerably constrained in this scenario 2,3 . With the advancement of sequencing technologies, high-resolution insights have gradually emerged into the development of glioma, paving the way for understanding transcriptional regulation in glioma 4,5 . Epigenetic modifications, as a regulatory layer, play a key role in both the upstream and downstream components of the transcriptional network 6,7 . Methylation is one of the most prevalent mechanisms regulating transcription 8 . CpG islands (CGIs) are found in various DNA elements involved in the transcriptional regulatory network, with more than half of promoter regions hosting clustered CGIs 9,10 . Furthermore, the presence of CGIs significantly enhances the transcription-activating capacity of enhancers 11 . Enhancers, as distal cis-regulatory elements, are different from proximal regulatory elements such as promoters. They rely on the 3D structure of chromosomes to achieve long-range regulation, thereby playing a pivotal role in the intricate regulatory network of glioma 12,13 . Compared to the high density of CGIs in the promoter regions, CGIs are relatively sparse within enhancers. Nevertheless, the function of both is markedly regulated by the methylation levels within their respective CGIs. Increased methylation reduces chromatin accessibility, consequently affecting the binding of enhancers to transcription factors (TFs) 14,15 . Aberrant epigenetic modifications of CGIs result in transcriptional dysregulation. Abnormal CGIs methylation is closely associated with various diseases, including glioma 16,17 . Excluding CGIs located in the classical regulatory regions, the genome still harbors nearly half of orphan CpG islands (oCGIs) 18 . These isolated oCGIs have long been overlooked, and few studies indicated that some oCGIs play an indispensable role in the positive regulatory effect of enhancers located within the same topologically associating domains (TADs) 19 . Not all oCGIs necessarily serve as bridges for enhancer function, and the independent regulatory potential of oCGIs remains uncertain. These aspects remain unknown in the intricate regulatory network of glioma. In this study, we found that oCGIs act as atypical enhancers, exerting cis-regulatory effects and collaboratively regulating target genes in coordination with enhancers, establishing a complex cis-transcriptional regulatory network in glioma. Furthermore, single-cell multi-omics data revealed that the cis-regulatory role of oCGIs, in conjunction with enhancers, is crucial for maintaining the stemness of glioma cells and is closely associated with various biological behaviors, such as necrosis and invasion. Additionally, it plays a crucial role in treatment resistance, leading to an adverse prognosis for patients. The cis-regulatory role of oCGIs was validated in glioma cell lines. We comprehensively explored potential mechanisms underlying the interaction between oCGIs and enhancers, providing a novel perspective for unraveling the intricate regulatory network in glioma. Methods Data collection and quality control We obtained RNA-seq and DNA methylation data for low-grade glioma (LGG) and GBM from the Cancer Genome Atlas (TCGA) through the UCSC Xena platform ( https://xena.ucsc.edu/ ) and directly downloaded ATAC-seq data from the TCGA ( https://portal.gdc.cancer.gov/ ). The 325 cohort and 693 cohort from the Chinese Glioma Genome Atlas (CGGA, http://www.cgga.org.cn/ ) were utilized for the validation of treatment response. Additionally, for validation purposes, we acquired data from 24 samples, including RNA-seq, DNA methylation-seq, and ChIP-seq, from the Gene Expression Omnibus (GEO) database ( https://www.ncbi.nlm.nih.gov/geo/ , GEO accession: GSE121719, GSE121720, GSE121721, GSE189859, GSE189860, and GSE189857) 20,21 . Data for chromosomal three-dimensional structures (TADs) was retrieved from GSE77565 22 . To validate our findings at the single-cell level, we utilized single-cell RNA sequencing (scRNA-seq) and single-cell reduced representation bisulfite sequencing (scRRBS-seq) data from 8 glioma samples provided by Verhaak et al . 23 . Additionally, spatial transcriptomics sequencing (stRNA-seq) data from 29 glioma samples were obtained from the study of Schnell and GSE194329 dataset 4,24 . Genomic single nucleotide polymorphism (SNP) data were obtained from the 1000 Genomes Project Phase 3 25 . Enhancer data were acquired from The FANTOM5 project 26 . DNA methylation data for 49 glioma cell lines were obtained from GSE68379. Drug sensitivity data for the cell lines were obtained from the Genomics of Drug Sensitivity in Cancer project (GDSC, https://www.cancerrxgene.org/celllines ). All data were analyzed based on the hg19. Identifying glioma subtypes based on DNA methylation levels We applied the following criteria to quality control TCGA DNA methylation data: ( 1 ) Exclusion of probes containing SNPs; ( 2 ) Exclusion of probes expressed in less than 20% of samples; and ( 3 ) Exclusion of samples with less than 20% probe expression. We utilized the k-nearest neighbors (KNN) method for imputing missing values in DNA methylation data. CGIs within a ± 250bp region were merged, and the region extending 1500bp upstream and 500bp downstream of the transcription start site (TSS) was defined as the promoter region. CGIs overlapping with promoters and enhancers were removed, and the remaining CGIs were defined as oCGIs. We conducted consensus clustering of the sample's oCGIs and enhancer methylation data. The number of clusters ranged from 2 to 20. The minimum number of subtypes necessary to effectively differentiate the samples was determined based on the results 27 . Clusters with less than 5 samples were excluded from the analysis. Construction of the oCGIs- and enhancers-dominated cis-regulatory network We measured the strength of the interaction between oCGIs or enhancers and target genes within the same TADs using mutual information (MI) 28 . The following criteria were employed to construct oCGIs-enhancer-gene triplets targeting the same gene: the MI calculated for oCGIs-gene or enhancer-gene pairs had an adjusted p-value of less than 0.05, and any pair with an adjusted p-value of greater than 0.5 in either oCGIs-gene or enhancer-gene relationship was filtered out. The preliminary selection of oCGIs-enhancer-gene triplets was analyzed to identify their regulatory patterns using Bayesian networks ( https://www.bnlearn.com/ ). For each triplet, we computed joint probabilities for potential regulatory patterns. For example, the oCGI direct model represented the direct oCGI regulation of target gene expression, while the oCGI cascade model indicated that oCGI regulates the target gene by modulating enhancers. The joint probabilities were calculated as follows: oCGI-dominated: oCGI direct: P (O, E, G) = P (O) * P (G | O) * P (E) oCGI cascade: P (O, E, G) = P (O) * P (E | O) * P (G | E) oCGI co-responsive: P (O, E, G) = P (O) * P (E | O) * P (G | O) oCGI composite: P (O, E, G) = P(O) * P (E | O) * P (G | O: E) Enhancer-dominated: Enhancer direct: P (O, E, G) = P (O) * P (G | E) * P(E) Enhancer cascade: P (O, E, G) = P (E) * P (O | E) * P (G | O) Enhancer co-responsive: P (O, E, G) = P (E) * P (O | E) * P (G | E) Enhancer composite: P (O, E, G) = P (E) * P (O | E) * P (G | O: E) Co-dominated: P (O, E, G) = P (O) * P (E) * P (G | O: E) Where P(O) and P(E) represent the probability distributions of DNA methylation in oCGIs and enhancers, respectively. P (G|O) indicates the conditional probability of gene expression regulated by oCGIs, and P (G|O: E) represents the conditional probability of gene expression regulated simultaneously by both oCGIs and enhancers. The definitions of the other terms are similar to those mentioned above. We selected the model with the smallest Akaike Information Criterion (AIC) as the regulatory pattern for each triplet. Additionally, we conducted independence testing for each triplet. For example, the p-value of independence testing between oCGI-enhancer and enhancer-gene was less than 0.05 for the oCGI direct model of the triplet to be considered valid. Consistent with conventional understanding, the agreement between DNA methylation level and RNA expression is crucial for model validation. Initially, we defined hypomethylation as β 0.79 using the β distribution of glioma methylation probe values. Hemimethylation was defined as β values falling within the range of 0.32 to 0.79 ( https://github.com/koyelucd/betaclust ). Triplets identified as the CGI direct model in Cluster 2, had an oCGI methylation level lower than that in Cluster 1 and a gene expression level higher than that in Cluster 1. Enrichment analysis was conducted for all upregulated genes in Cluster 2 29 . We employed FIMO to identify potential TF-binding motifs in oCGIs and enhancers 30 . Construction of glioma classifiers based on multi-omics data of oCGIs We constructed the classifier using three different methods, which were cross-validated to enhance the credibility of the model. First, we employed the Partitioning Around Medoids (PAM) algorithm to discern the sample allocations based on distance. Next, we incorporated 11 machine learning algorithms, including cv_glmnet, featureless, kknn, lda, log_reg, naive_bayes, ranger, rpart, svm, xgboost, and debug, to construct the models based on the expression of oCGIs and DNA methylation levels. The machine learning algorithms were implemented using the mlr3 package 31 . TCGA was utilized as the training dataset, while 24 samples from GEO were employed as the validation dataset. We aligned ChIP-seq data from the validation dataset (H3K27ac, H3K4me3, H3K4me1, and H3K27me3) and TCGA ATAC data to the corresponding regions of oCGIs, enhancers, and promoters. scRNA-seq and scRRBS-seq data We conducted quality control on 55,284 cells from the 11 glioma samples in the Verhaak cohort. Cells with a mitochondrial gene percentage exceeding 20% and doublets were filtered out. The samples were integrated using Harmony 32 . Tumor cells were annotated using marker genes in conjunction with copy number variations ( https://github.com/broadinstitute/infercnv ). Cytotrace was employed to assess the differentiation level of tumor cells 33 . Monocle3 was used to identify the differentiation trajectories of glioma cells ( https://github.com/cole-trapnell-lab/monocle3 ). SENIC was utilized to identify significantly activated TFs in glioma subtypes 34 . We applied the same criteria to perform quality control on the scRRBS data. Triplets were identified in tumor cells based on scRNA and scRRBS data. The difference in drug sensitivity between two subtypes of glioma cells was analyzed using the beyondcell package 35 . Spatial transcriptomics data The stRNA-seq data from 29 glioma samples were initially used to identify glioma subtypes using pseudobulk analysis. Subsequently, two pathologists jointly divided the images of stRNA-seq into four regions: vascular, necrotic, cellular, and infiltrating. All stRNA-seq analyses were conducted using SPATA2 package 4 . Copy number variation was assessed using the runCnvAnalysis function. Images representing different features were visualized using the plotSurfaceComparison function. The CellTrek package was employed to integrate scRNA-seq and stRNA-seq data 36 . The communication between tumor cells and other components of the tumor microenvironment in different niches was identified using the nichenetr package 37 . Cell line The glioma cell lines U251, LN229, and A172 were obtained from the Department of Neurosurgery, the Second Affiliated Hospital of Harbin Medical University. The SF126 cell line was acquired from Pricella (Wuhan, China). Cell lines with oCGI (chr11:728884–729383) knockout were established using the CRISPR/Cas9 system. sgRNAs were designed using Benchling’s CRISPR toolkit (sgRNA1: AGCCCCTTGGAAGAAACGGG; sgRNA2: GGAAGCCCCTTGGAAGAAAC). Quantitative PCR with reverse transcription Total RNA was extracted from glioma cell lines using Trizol (Beyotime, China), and reverse transcription was performed using reverse transcription reagents (RNase H-, RNase inhibitor, and dNTP Mix) (Beyotime, China) following the manufacturer's instructions. All primers are listed in Table S1 . Chromatin immunoprecipitation (ChIP) The ChIP assay was conducted following the protocol outlined in the ChIP assay kit (Beyotime, China). Subsequently, PCR was employed to assess the immunoprecipitated DNA level. The primers are listed in Table S1 . MTT Assay In the control group and two oCGI knockout groups (chr11:728884–729383 KO1 and chr11:728884–729383 KO2), 10 µL of MTT (5 mg/mL) (Beyotime, China) was added to each well of a 96-well plate. After 4 hours of incubation, 100 µL of formazan (Beyotime, China) was added, and the absorbance was read at 490 nm. Immunofluorescence In brief, coverslips with confluent cells were fixed with 4% paraformaldehyde. Then, immunostaining was performed using KI67 polyclonal antibody (ProteinTech Group, Inc., USA). Finally, we conducted counterstaining with DAPI. Results Methylation characteristics of oCGIs and enhancers in glioma We initially identified CGIs in different DNA elements. Consistent with previous findings, the majority of CGIs (78.99%) overlapped with genes or promoter regions, but approximately 20% of oCGIs were dispersed throughout the genome (Figure S3A). These dispersed oCGIs have long been ignored. We performed consensus clustering based on the methylation of oCGIs and enhancers to illuminate the methylation characteristics of oCGIs and classical enhancers. We found that glioma can be broadly classified into two subtypes under various clustering scenarios (Figure S1 and Figure S2 ). Furthermore, the clustering results based on oCGIs and enhancers exhibited significant consistency (Figure S3B). Cluster 1 (C1) was predominantly composed of LGG, whereas Cluster 2 (C2) primarily included GBM (Figure S3C). We utilized the clustering results based on oCGIs to further explore the role of oCGIs. Identifying the oCGI- and enhancer-dominated cis-regulatory network To elucidate the regulatory patterns among oCGIs, enhancers, and genes, we initially screened combinations with potential regulatory effects based on interactions between oCGI-gene/enhancer-gene pairs in the same TAD. Furthermore, we proposed 9 regulatory models of oCGI-enhancer-gene triplets (Fig. 1 A). These 9 models elucidated all potential regulatory possibilities, including 4 models where oCGI or enhancer individually played a predominant role (direct, cased, co-responsive, and composite), and 1 model where they were co-dominated (co-dominated). For instance, in the oCGI-dominated direct model, the methylation level of oCGI affected its binding capacity to TFs, thereby affecting the binding of TFs to target gene promoter regions and regulating transcription 14,15 . Subsequently, based on Bayesian networks, we determined the most suitable model for each oCGI-enhancer-gene triplet. In addition, models were selected based on the methylation levels of oCGIs and enhancers in conjunction with gene expression. In the end, only 7 models were validated in our dataset (oCGI_Direct, Enhancer_Direct, oCGI_responsive, Enhancer_responsive, oCGI_Cased, Enhancer_Cased, Coordinate). In both glioma subtypes, the oCGI direct and enhancer direct model predominated, indicating that despite the presence of more complex indirect regulation, oCGIs and enhancers played a dominant role in the direct regulation of target genes. Moreover, the regulation of upregulated triplets in Cluster 2 was the central theme across all models. (Fig. 1 B-C). Furthermore, these upregulated genes were associated with positive regulation of the tumor (positive regulation of cytokine production and cell migration) (Figure S3D). oCGIs exhibited similar regulatory characteristics to classical enhancers. Both oCGIs and enhancers can regulate more than one target gene, but the predominant mode of regulation is one-to-one. Secondly, the regulatory effect of oCGIs or enhancers on target genes within the same TAD is not often limited to adjacent loci. On the contrary, non-adjacent regulation models accounted for the majority of the model count. In a significant number of models, oCGIs or enhancers were separated from their target genes by more than 10 genes. Thirdly, the long-range regulatory characteristics of enhancers were also evident in oCGIs. Similar to enhancers, the regulatory distance of oCGIs typically exceeded 500 kb and even surpassed 1000 kb in some cases. These characteristics were observed in both glioma subtypes (Fig. 2 A-B). A. and B. The number, locations, and distances of target genes were regulated by oCGIs and enhancers in the two glioma subtypes. States of oCGIs and enhancers in the cis-regulatory models Different modes of histone modifications by enhancers serve as markers of their activity. H3K27ac is a marker of active enhancers with positive transcriptional regulatory activity, whereas H3K27me3 is a marker of silenced enhancers. Co-expression of H3K4me1 and H3K27ac indicates enhancers with moderate activity. Promoters also have similar activation and silencing markers. H3K4me3 is significantly enriched in promoter regions as an additional marker 38,39 . Due to the lack of chromatin modification data corresponding to TCGA samples, we used external samples to determine the mode of chromatin modification of the two glioma subtypes. We first determined to which subtype the external samples belonged. Models in 3 dimensions combined with various machine learning algorithms were used to ensure the accuracy of our results. We initially identified the subtypes of 24 glioma samples based on distance metrics (Figure S3E). Next, we constructed classifiers for TCGA glioma samples based on oCGI methylation features and gene RNA expression features using 11 machine-learning algorithms. Except for featureless, log_reg, and debug algorithms, other algorithms exhibited excellent discriminatory capabilities for glioma samples in the TCGA (Figure S4A). We trained 10 models for each of the eight algorithms based on different resampling results. These models were then applied to the test set samples. Model performance was evaluated using classification error rate and ROC, with cv_glmnet showing excellent performance across all 10 models on both oCGI methylation and RNA expression data. It correctly distinguished the subtypes of all samples when utilizing both types of data (Figure S4A-B and Figure S5). Next, we analyzed the distribution of different histone modifications across the genome. Both subtypes exhibited similar distribution patterns, with peaks of different histone modifications significantly enriched in gene-related regions (Figure S4C). Furthermore, we characterized the histone modification features in the 3000 kb regions around all TSS, oCGIs, and enhancers in the two subtypes. Cluster 2 exhibited lower methylation levels, higher chromatin accessibility, and higher levels of active promoter and enhancer feature markers, such as H3K27ac peaks, across all regions. All oCGIs displayed histone modification features similar to enhancers (Figure S6A). These findings support their role as "atypical enhancers" with similar functions. Consistent with the dominance of upregulated models in Cluster 2, Cluster 2 exhibited higher overall transcriptional regulatory activity. Therefore, we focused on genes activated in Cluster 2 and repressed in Cluster 1 in the follow-up analyses. These genes displayed similar histone modification features (Figure S6B-C), confirming our previous findings. Finally, combining the regulatory modes obtained from the triplet analysis, the predominant model characteristics in Cluster 1 were oCGIs and enhancers with lower activity and higher methylation levels, suppressing gene transcription. In contrast, the dominant regulatory mode in Cluster 2 was characterized by active oCGIs and enhancers with lower methylation levels, enabling positive transcriptional regulation (Figure S4D-E). oCGIs- and enhancers-dominated cis-network in glioma cells Although we have constructed a cis-regulatory network based on glioma subtypes, the specific manner in which these regulatory modes function and whether they directly impact tumor cell lifecycle remain elusive. Thus, we investigated regulatory modes dominated by oCGIs and enhancers in tumor cells utilizing scRNA-seq. Our initial approach involved ensuring quality control and cell annotation for scRNA-seq data from 8 samples and filtering cells based on copy number variations, yielding 21,370 tumor cells (Fig. 3 B and S7 A-G). Subsequently, we identified subtypes for these 8 samples using a previously developed RNA classifier (Fig. 3 A) and only retained the tumor cells for subsequent analysis. It should be noted that significant heterogeneity existed among tumor cells from different subtypes (Fig. 3 E and Figure S7G). Although limited coverage of methylation sites in scRRBS-seq prevented identification of regulatory patterns for some triplets, consistent with previous results, oCGI_direct and enhancer_direct modes still primarily regulated both subtypes (Fig. 3 C). To further characterize the heterogeneity between tumor cells from the two subtypes, we assessed the proliferative potential of tumor cells. Tumor cells in Cluster 2 exhibited a significantly lower cytotrace score and a lower degree of differentiation (Fig. 3 D). In line with the results of pseudotime analysis, the tumor cells from Cluster 2 primarily aggregated at the initiation of the differentiation trajectory. As cells progressed along the differentiation trajectory, the composition of tumor cells transitioned from Cluster 2 cells to Cluster 1. Additionally, the upregulated genes in Cluster 2 exhibited a significant negative correlation with differentiation time (Fig. 3 E-F and S8 A-B). The mechanism by which the cis-regulatory network acts in glioma cells We initially focused on the communication between various components in the glioma microenvironment to elucidate the precise mechanisms by which the cis-regulatory network is involved in glioma cells. Therefore, we needed a more detailed identification of tumor cells. All non-tumor cells were classified into oligodendrocytes, microglial, macrophages, T cells, endothelial cells, and pericytes cells (Figure S8C-D). The results of cell communication in the tumor microenvironment showed that interactions between tumor cells and other cells in the tumor microenvironment were stronger in Cluster 2 than in Cluster 1 (Fig. 3 G). Many signals favoring tumor cells were significantly activated in Cluster 2, such as integrin-related signals (COL9A3 − (ITGA2 + ITGB1), JAM3 − (ITGAM + ITGB2), COL6A1 − (ITGA1 + ITGB1)), immune suppression signals (CD99 − PILRA), etc. (Figure S8E). Furthermore, we explored TFs binding to oCGIs or enhancers. Initially, we scanned the DNA sequences of oCGIs or enhancers, preliminarily selecting TFs with binding potential. Subsequently, we identified the activated TFs in the two subtypes of tumor cells (Figure S8F). By combining the two strategies mentioned above, we identified TFs regulating different triplets in different subtypes of glioma cells. Based on univariate cox regression and co-expression analysis between oCGIs/enhancers-related TFs and their target genes, we identified prognosis-related key TFs (Figure S9A-B). The effects of the cis-network regulatory in different niches of gliomas Using stRNA-seq, we explored the regulatory roles of oCGIs and enhancers in different niches. We first identified the subtypes of glioma (Figure S11A). The necrotic and infiltrating niches of the two subtypes of glioma were analyzed and confirmed by pathologists. The necrotic niche surrounding both subtypes exhibited evident hypoxia and differentiation features. Additionally, genes regulated by oCGIs and enhancers in both clusters were significantly downregulated in this region. Differences in the necrotic niche between the two subtypes were reflected in the CNV of spots in Cluster 2, gradually increasing from the necrotic center toward the outer layer. Cluster 1 showed the opposite trend (Fig. 4 A-D and S10 A-B). To elucidate the reasons leading to this difference, we integrated the 8 scRNA-seq datasets from the above studies with the respective stRNA-seq dataset for each subtype. Glioma cells mainly co-localized with microglial and endothelial cells, showing stronger colocalization with microglial in regions closer to the necrotic center (Figure S11B-C). The necrotic niche in Cluster 2 contained a higher proportion of tumor cells, increasing its CNV (Figure S11D). GO analysis showed that MHC-related functions were significantly activated near the necrotic center (Layer 1), while membrane protein activity and extracellular matrix activation were more prominent in the outer layer (Layer 2), with both subtypes exhibiting similar results (Figure S11E). Cell communication in the necrotic niche revealed a significant overactivity of various chemokines and other immune-related receptor-ligand pairs in Cluster 1 (CCL3L3-ACKR2; JAM2-JAM2; CCL2-CCR10, etc.). Similarly, various signals promoting tumor cell activation were more pronounced in Cluster 2 (S100A4-EGFR, SPP1-ITGB1, SPP1-ITGAV, etc.) (Fig. 4 E and S11 F). In the infiltrating niche, the CNV, hypoxia, cytotrace score, and cis score exhibited similar distributions (Fig. 4 F-G and S10 C-D). In the tumor component of the infiltrating niche, tumor cells were predominantly co-localized with oligodendrocytes and endothelial cells, and this distribution pattern was consistent between the two subtypes (Figure S12A-B). The tumor region comprised more tumor cells and fewer immune cells compared with the infiltrating region (Figure S12C). The tumor region in Cluster 1 activated additional immune-related molecular functions compared with Cluster 2 (Figure S12D-E). The communication between glioma cells and components of the tumor microenvironment revealed abundant tumor-promoting signals in the infiltrating niches of both subtypes. Immune-related signals were more abundant in Cluster 1 compared with Cluster 2 (JAM2−−JAM2; CCL3L3−−CCR1; CCL2−−CCR10) (Fig. 4 H and S12 F). The gene POLR2L, regulated by oCGI (Chr11:728884–729383) in our study, played a significant role in cell communication in Cluster 2 glioma cells, although it was not the most significantly activated gene. The upregulation of oCGI/enhancer-related genes in Cluster 2 is associated with treatment resistance in gliomas To elucidate the impact of the cis-regulatory role of oCGI/enhancers on the clinical treatment of glioma patients, we stratified glioma patients from The Cancer Genome Atlas (TCGA) into high and low groups based on the expression levels of oCGI/enhancer-related genes upregulated in Cluster 2 (C2 cis score). The prognosis of patients with higher C2 cis scores was significantly worse under different treatment conditions (untreated, chemotherapy, radiotherapy, chemotherapy + radiotherapy) (Fig. 5 A). This pattern of treatment resistance was further validated in the CGGA cohorts, including the 325 cohort and 693 cohort (Fig. 5 B). Clustering results based on drug sensitivity of tumor cells from two subtypes, as shown by beyondcell, revealed significant heterogeneity between the two subtypes (Fig. 5 C). Ingenol mebutate demonstrated higher sensitivity in Cluster 2 tumor cells, while Marinopyrrole A showed greater sensitivity in Cluster 1 tumor cells (Fig. 5 D-F). Validation of the cis-regulatory role of oCGIs As the POLR2L gene plays a crucial role in cell communication between tumor cells and other components of the tumor microenvironment, we selected oCGI (Chr11:728884–729383) and its target, POLR2L gene, for validation. Additionally, we identified TF E2F7 as a potential regulatory target based on prognosis analysis and co-expression analysis from the above studies. First, we identified the subtypes based on the DNA methylation data of 49 glioma cell lines (Fig. 6 A). Two cell lines from each subtype were chosen for subsequent validation (Cluster 1: A172, SF126; Cluster 2: LN229, U251). The results of ChIP combined with qPCR indicated that E2F7 was significantly enriched in the region of oCGI Chr11:728884–729383 in LN229 and U251 cell lines. Additionally, this enrichment disappeared upon the knockout of oCGI Chr11:728884–729383 (Fig. 6 B). oCGI Chr11:728884–729383 knockout in LN229 and U251 cell lines markedly decreased the expression of POLR2L gene and glioma stem cell-related genes, CD133 and SOX2 . However, such a regulatory relationship was not observed in A172 and SF126 cell lines (Fig. 6 C-D). The effect of oCGI Chr11:728884–729383 on cell proliferation was validated in different glioma subtypes through MTT assay and immunofluorescence assay. After oCGI Chr11:728884–729383 knockout, LN229 and U251 cell viability significantly decreased, and the Ki67 fluorescence signal became noticeably weaker compared with the control group (Fig. 6 E and G-H). The effect of oCGI Chr11:728884–729383 knockout on the viability of A172 and SF126 cells was not significant (Fig. 6 F). To provide better guidance for clinical treatment, we filtered drugs effective for each subtype of glioma. Combining drug sensitivity data from 49 glioma cell lines, we compared the AUC values for the two glioma subtypes. Apart from JNK inhibitor VIII and XAV939, most anti-tumor drugs were more effective on cell lines from Cluster 1 (Fig. 6 I). Discussion Recent studies have increasingly investigated the regulatory effect of classic enhancers on critical targets in cancer. It is now widely acknowledged that enhancers accelerate oncogenesis 40 . As such, many researchers paid attention to the pairing of classic enhancers with critical nodes in cancer. However, for highly heterogeneous tumors like glioma, the complexity of the transcriptional regulatory network provides the potential for a broader definition of enhancers. The coordinated transcription-enhancing function of oCGIs establishes a foundation for their non-classical enhancer function 19 . We constructed the oCGI-enhancer-gene regulatory models to illuminate the importance of oCGIs, which have long been overlooked. We revealed the role of oCGIs as non-classical enhancers in the intricate regulatory network of glioma. First, the comparative genomic analysis highlighted that oCGIs exhibit DNA methylation characteristics similar to classical enhancers, suggesting that oCGIs can exert transcriptional enhancer functions, either through a mechanism akin to that of classical enhancers or by collaborating with classical enhancers. Furthermore, to elucidate the precise regulatory functions of oCGIs across different glioma samples, we applied the 9 possible models of regulation, encompassing oCGIs-enhancer-gene triplets, based on the DNA methylation features of oCGIs in glioma subtypes. Through mutual information analysis and Bayesian networks, we delineated the regulatory patterns of each triplet. As expected, oCGIs directly or indirectly regulated the target genes. Notably, they exhibited regulatory characteristics similar to classical enhancers, enabling them to regulate several target genes and exert long-range regulation within the same TAD. Although oCGIs and enhancers were primarily engaged in the direct regulation of target genes within the cis-regulatory network of glioma, many more complex regulatory models were validated, suggesting that enhancers are potential targets of oCGIs. It enhances our understanding of the cis-regulatory network. The predictive models based on 11 machine-learning algorithms allowed us to accurately identify the glioma subtypes based on their DNA methylation or RNA expression levels. These findings can help integrate data from various sources and comprehensively identify the characteristics of the cis-regulatory network. Chromatin modifications serve as markers for identifying the functional states of enhancers or promoters 38,39 . The results of chromatin modifications in the two glioma subtypes indicated that tumor cells were more active in Cluster 2. In Cluster 2, the oCGIs- or enhancers-dominated regulatory models exerted stronger positive transcriptional control through lower methylation levels and higher chromatin accessibility. Therefore, we focused on the major regulators of transcription in Cluster 2. We employed a high-resolution single-cell atlas to further characterize the details of the cis-regulatory network in tumor cells. Our findings revealed that tumor cells from Cluster 2 typically exhibited higher rates of proliferation. Additionally, the regulatory influence of oCGI- or enhancer-dominated triplets on target genes gradually weakened after differentiation in Cluster 2, suggesting that the active cis-regulatory network in Cluster 2 is a crucial factor for maintaining the heightened activity of tumor cells. Additionally, the communication between tumor cells and the glioma microenvironment was more active in Cluster 2 compared with Cluster 1, exerting a more pronounced pro-tumor effect. To elucidate the precise mechanisms by which the cis-regulatory network and oCGIs were involved in this complex glioma microenvironment, we characterized different niches and conducted stRNA-seq. Tumor cells in the center of the tumor exhibited stronger proliferation and invasive capabilities compared with the infiltrating region 41 . Additionally, due to the vigorous metabolic demands of tumor cells, the tumor center was more susceptible to necrosis. Therefore, we selected the necrotic niche as the tumor center and the infiltrating niche as the tumor periphery. In the necrotic niche of both subtypes, tumor cells were primarily co-localized with microglial and endothelial cells. However, in Cluster 2, this co-localization was associated with significantly weaker immune cell infiltration and tumor-activating signals. An increase in co-localization with oligodendrocytes was observed in the infiltrating niche, similar to the signal network in the necrotic niche. Cluster 2 showed more pronounced activation of pro-tumor signals. The cis activation in Cluster 2 is a crucial factor leading to patients' resistance to various treatment modalities. The cis activation in Cluster 2 renders patients unable to benefit from various clinical treatments and decreased sensitivity to various drugs. To explore the regulatory role of oCGIs in the tumor center and periphery, we focused on the common downstream target in cell communication between the two niches, namely the POLR2L gene, for subsequent validation. The results indicated that oCGI Chr11:728884–729383 regulates glioma proliferation by modulating the regulatory effect of TFs E2F7 on the target gene POLR2L . Clinical drugs were selected based on distinct oCGIs-based glioma subtypes. Furthermore, our study provided a theoretical foundation for the subsequent development of drugs targeting oCGIs-related targets. Although we comprehensively analyzed the oCGIs- and enhancers-dominated cis-regulatory network, limitations inherent to single-cell technologies necessitate the validation of only a fraction of the identified models. With the continued advancement of sequencing technologies, we anticipate a more comprehensive understanding of the cis-regulatory network based on oCGIs. Conclusion For the first time, our study comprehensively explored the role of oCGIs as non-classical enhancers in the transcriptional regulatory network of glioma. Generally, oCGIs directly regulate target genes similar to classical enhancers. In addition, we validated numerous, more intricate models, underscoring the complexity of the regulatory mechanisms of oCGIs in cancer, enabling them to act in a coordinated or regulated manner and significantly affecting glioma progression and treatment resistant. We provided in-depth insights into all potential scenarios within the glioma cis-regulatory network, laying a foundation for unraveling the mechanisms behind glioma development. Abbreviations AIC Akaike Information Criterion CGGA Chinese Glioma Genome Atlas CGIs CpG islands C1 Cluster1 C2 Cluster2 GBM glioblastoma multiforme GEO Gene Expression Omnibus KNN k-nearest neighbors LGG low-grade glioma MI mutual information oCGIs orphan CpG islands scRNA-seq single-cell RNA sequencing scRRBS-seq single-cell reduced representation bisulfite sequencing stRNA-seq spatial transcriptomics sequencing SNP single nucleotide polymorphism TADs topologically associating domains TCGA the Cancer Genome Atlas TFs transcription factors TSS transcription start site Declarations Ethics approval and consent to participate Not applicable. Consent for publication Not applicable. Availability of data and materials The datasets used and/or analyzed during the current study are available from public datasets. Software and resources used for the analyses are described in each method section. Competing Interests The authors have declared that no competing interest exists. Authors' contributions Jiawei Yao: Conceptualization, Data curation, Formal analysis, Methodology, Software, Visualization, Roles/Writing - original draft, Writing - review & editing; Penglei Yao: Data curation, Formal analysis, Methodology, Software, Validation; Yang Li: Methodology, Resources; Ke He: Investigation, Software, Methodology, Validation; Xinqi Ma: Software, Validation; Qingsong Yang: Methodology, Visualization; Junming Jia: Roles/Writing - original draft; Zeren Chen: Investigation, Methodology; Shuqing Gu: Investigation; Weihua Li: Funding acquisition, Supervision; Guangzhi Wang: Resources, Supervision; Mian Guo: Conceptualization, Funding acquisition, Supervision, Project administration, Writing - review & editing. Acknowledgements This work was supported by National Natural Science Foundation of China (82173384 and 81773161) and Shenzhen Science and Technology Innovation Commission (JCYJ20200109120205924). References Ostrom QT, Cioffi G, Waite K, Kruchko C, Barnholtz-Sloan JS (2021) CBTRUS Statistical Report: Primary Brain and Other Central Nervous System Tumors Diagnosed in the United States in 2014–2018. Neuro Oncol 23(12 Suppl 2):iii1–iii105 Sonabend AM, Bansal M, Guarnieri P et al (2014) The transcriptional regulatory network of proneural glioma determines the genetic alterations selected during tumor progression. Cancer Res 74(5):1440–1451 Xiong A, Zhang J, Chen Y, Zhang Y, Yang F (2022) Integrated single-cell transcriptomic analyses reveal that GPNMB-high macrophages promote PN-MES transition and impede T cell activation in GBM. EBioMedicine 83:104239 Ravi VM, Will P, Kueckelhaus J et al (2022) Spatially resolved multi-omics deciphers bidirectional tumor-host interdependence in glioblastoma. Cancer Cell 40(6):639–655e613 Zheng Y, Carrillo-Perez F, Pizurica M, Heiland DH, Gevaert O (2023) Spatial cellular architecture predicts prognosis in glioblastoma. Nat Commun 14(1):4122 Zhao LY, Song J, Liu Y, Song CX, Yi C (2020) Mapping the epigenetic modifications of DNA and RNA. Protein Cell 11(11):792–808 Hogg SJ, Beavis PA, Dawson MA, Johnstone RW (2020) Targeting the epigenetic regulation of antitumour immunity. Nat Rev Drug Discov 19(11):776–800 Hu C, Peng K, Wu Q et al (2021) HDAC1 and 2 regulate endothelial VCAM-1 expression and atherogenesis by suppressing methylation of the GATA6 promoter. Theranostics 11(11):5605–5619 Ferreira HJ, Esteller M (2018) CpG Islands in Cancer: Heads, Tails, and Sides. Methods Mol Biol 1766:49–80 Patsouras MD, Karagianni P, Kogionou P, Vlachoyiannopoulos PG (2019) Differential CpG methylation of the promoter of interleukin 8 and the first intron of tissue factor in Antiphospholipid syndrome. J Autoimmun 102:159–166 Steinhaus R, Gonzalez T, Seelow D, Robinson PN (2020) Pervasive and CpG-dependent promoter-like characteristics of transcribed enhancers. Nucleic Acids Res 48(10):5306–5317 Xiao J, Jin X, Zhang C et al (2021) Systematic analysis of enhancer regulatory circuit perturbation driven by copy number variations in malignant glioma. Theranostics 11(7):3060–3073 Andersson R, Sandelin A (2020) Determinants of enhancer and promoter activities of regulatory elements. Nat Rev Genet 21(2):71–87 Angeloni A, Bogdanovic O (2019) Enhancer DNA methylation: implications for gene regulation. Essays Biochem 63(6):707–715 Thurman RE, Rynes E, Humbert R et al (2012) The accessible chromatin landscape of the human genome. Nature 489(7414):75–82 Malta TM, de Souza CF, Sabedot TS et al (2018) Glioma CpG island methylator phenotype (G-CIMP): biological and clinical implications. Neuro Oncol 20(5):608–620 Nishiyama A, Nakanishi M (2021) Navigating the DNA methylation landscape of cancer. Trends Genet 37(11):1012–1027 Deaton AM, Bird A (2011) CpG islands and the regulation of transcription. Genes Dev 25(10):1010–1022 Pachano T, Sánchez-Gaya V, Ealo T et al (2021) Orphan CpG islands amplify poised enhancer regulatory activity and determine target gene responsiveness. Nat Genet 53(7):1036–1049 Wu Y, Fletcher M, Gu Z et al (2020) Glioblastoma epigenome profiling identifies SOX10 as a master regulator of molecular tumour subtype. Nat Commun 11(1):6434 Lee SD, Song J, LeBlanc VG, Marra MA (2022) Integrative multi-omic analysis reveals neurodevelopmental gene dysregulation in CIC-knockout and IDH1-mutant cells. J Pathol 256(3):297–309 Won H, de la Torre-Ubieta L, Stein JL et al (2016) Chromosome conformation elucidates regulatory relationships in developing human brain. Nature 538(7626):523–527 Johnson KC, Anderson KJ, Courtois ET et al (2021) Single-cell multimodal glioma analyses identify epigenetic regulators of cellular plasticity and environmental stress response. Nat Genet 53(10):1456–1468 Ren Y, Huang Z, Zhou L et al (2023) Spatial transcriptomics reveals niche-specific enrichment and vulnerabilities of radial glial stem-like cells in malignant gliomas. Nat Commun 14(1):1028 Sudmant PH, Rausch T, Gardner EJ et al (2015) An integrated map of structural variation in 2,504 human genomes. Nature 526(7571):75–81 Lizio M, Harshbarger J, Shimoji H et al (2015) Gateways to the FANTOM5 promoter level mammalian expression atlas. Genome Biol 16(1):22 Wilkerson MD, Hayes DN (2010) ConsensusClusterPlus: a class discovery tool with confidence assessments and item tracking. Bioinformatics 26(12):1572–1573 Tong Y, Sun J, Wong CF et al (2018) MICMIC: identification of DNA methylation of distal regulatory regions with causal effects on tumorigenesis. Genome Biol 19(1):73 Zhou Y, Zhou B, Pache L et al (2019) Metascape provides a biologist-oriented resource for the analysis of systems-level datasets. Nat Commun 10(1):1523 Grant CE, Bailey TL, Noble WS (2011) FIMO: scanning for occurrences of a given motif. Bioinformatics 27(7):1017–1018 Lang M, Binder M, Richter J et al (2019) mlr3: A modern object-oriented machine learning framework in R. J Open Source Softw 4:1903 Korsunsky I, Millard N, Fan J et al (2019) Fast, sensitive and accurate integration of single-cell data with Harmony. Nat Methods 16(12):1289–1296 Gulati GS, Sikandar SS, Wesche DJ et al (2020) Single-cell transcriptional diversity is a hallmark of developmental potential. Science 367(6476):405–411 Aibar S, González-Blas CB, Moerman T et al (2017) SCENIC: single-cell regulatory network inference and clustering. Nat Methods 14(11):1083–1086 Fustero-Torre C, Jiménez-Santos MJ, García-Martín S et al (2021) Beyondcell: targeting cancer therapeutic heterogeneity in single-cell RNA-seq data. Genome Med 13(1):187 Wei R, He S, Bai S et al (2022) Spatial charting of single-cell transcriptomes in tissues. Nat Biotechnol 40(8):1190–1199 Browaeys R, Saelens W, Saeys Y (2020) NicheNet: modeling intercellular communication by linking ligands to target genes. Nat Methods 17(2):159–162 Cruz-Molina S, Respuela P, Tebartz C et al (2017) PRC2 Facilitates the Regulatory Topology Required for Poised Enhancer Function during Pluripotent Stem Cell Differentiation. Cell Stem Cell 20(5):689–705e689 Preissl S, Gaulton KJ, Ren B (2023) Characterizing cis-regulatory elements using single-cell epigenomics. Nat Rev Genet 24(1):21–43 Zhao J, Faryabi RB (2023) Spatial promoter-enhancer hubs in cancer: organization, regulation, and function. Trends Cancer Zhao Y, Fu X, Lopez JI et al (2021) Selection of metastasis competent subclones in the tumour interior. Nat Ecol Evol 5(7):1033–1045 Additional Declarations No competing interests reported. Supplementary Files SupplmentaryFigures.docx TableS1.xlsx Cite Share Download PDF Status: Posted Version 1 posted You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-3959082","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":273181407,"identity":"6222b7eb-1139-4a16-ba77-0105bc858869","order_by":0,"name":"jiawei yao","email":"","orcid":"","institution":"The Second Affiliated Hospital of Harbin Medical University","correspondingAuthor":false,"prefix":"","firstName":"jiawei","middleName":"","lastName":"yao","suffix":""},{"id":273181408,"identity":"b9ebc993-a736-4aed-b216-70eea9a63f87","order_by":1,"name":"Penglei Yao","email":"","orcid":"","institution":"The Second Affiliated Hospital of Harbin Medical University","correspondingAuthor":false,"prefix":"","firstName":"Penglei","middleName":"","lastName":"Yao","suffix":""},{"id":273181409,"identity":"576f318b-6918-4fd0-948f-a6bc7fd8bb52","order_by":2,"name":"Yang Li","email":"","orcid":"","institution":"The Second Affiliated Hospital of Harbin Medical University","correspondingAuthor":false,"prefix":"","firstName":"Yang","middleName":"","lastName":"Li","suffix":""},{"id":273181410,"identity":"3d7f3cd0-ed89-4386-bc69-4367e83d3ea7","order_by":3,"name":"ke he","email":"","orcid":"","institution":"The Second Affiliated Hospital of Harbin Medical University","correspondingAuthor":false,"prefix":"","firstName":"ke","middleName":"","lastName":"he","suffix":""},{"id":273181411,"identity":"5dbe527d-fd92-4ced-a4b4-fa5796a4bda3","order_by":4,"name":"xinqi ma","email":"","orcid":"","institution":"The Second Affiliated Hospital of Harbin Medical University","correspondingAuthor":false,"prefix":"","firstName":"xinqi","middleName":"","lastName":"ma","suffix":""},{"id":273181412,"identity":"0c148524-92ee-4a39-84fd-9f4dccaf8754","order_by":5,"name":"Qingsong Yang","email":"","orcid":"","institution":"The Second Affiliated Hospital of Harbin Medical University","correspondingAuthor":false,"prefix":"","firstName":"Qingsong","middleName":"","lastName":"Yang","suffix":""},{"id":273181413,"identity":"f345b567-f8c9-42e0-a36d-5743a2596006","order_by":6,"name":"Junming Jia","email":"","orcid":"","institution":"The Second Affiliated Hospital of Harbin Medical University","correspondingAuthor":false,"prefix":"","firstName":"Junming","middleName":"","lastName":"Jia","suffix":""},{"id":273181414,"identity":"ade34342-6c59-4009-8a7c-a053a32689d4","order_by":7,"name":"Zeren Chen","email":"","orcid":"","institution":"The Second Affiliated Hospital of Harbin Medical University","correspondingAuthor":false,"prefix":"","firstName":"Zeren","middleName":"","lastName":"Chen","suffix":""},{"id":273181415,"identity":"779882c7-b44c-4cb8-9bde-702ef4f4a92c","order_by":8,"name":"Shuqing Gu","email":"","orcid":"","institution":"The First Hospital of Qiqihar","correspondingAuthor":false,"prefix":"","firstName":"Shuqing","middleName":"","lastName":"Gu","suffix":""},{"id":273181416,"identity":"28012d85-33b0-4036-b7ef-5073bf59a7cf","order_by":9,"name":"Weihua Li","email":"","orcid":"","institution":"the First Affiliated Hospital of Shenzhen University Health Science Center","correspondingAuthor":false,"prefix":"","firstName":"Weihua","middleName":"","lastName":"Li","suffix":""},{"id":273181417,"identity":"ca8146ed-75b8-4cd5-86da-7d4154f01e1c","order_by":10,"name":"Guangzhi Wang","email":"","orcid":"","institution":"The Second Affiliated Hospital of Harbin Medical University","correspondingAuthor":false,"prefix":"","firstName":"Guangzhi","middleName":"","lastName":"Wang","suffix":""},{"id":273181418,"identity":"e168e105-d7fb-431f-a69d-5a1ba083fc7f","order_by":11,"name":"Mian Guo","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA00lEQVRIiWNgGAWjYLCCBAYGOQMGHjCbsYFYLcYkagGCxA1Ea5GPSD664UHNnfTt0mePbuZhsJHdcID52QN8WgxvpKXdSDj2LHdnX17abR6GNOMNB9jMDfBqmZFjdiOB7XDuhjM8ZkAthxM3HOBhkyCs5d/hdAOIlv+EtchLALUkth1OgGo5QFiLAc+ztBuJfYcNN5zhS7s5xyDZeOZhNjP8trQnH7v549theYMzvMduvKmwk+073vwMvy0HULlAzIxPPciWBgIKRsEoGAWjYBQwAADxGU7Apg8DeQAAAABJRU5ErkJggg==","orcid":"","institution":"The Second Affiliated Hospital of Harbin Medical University","correspondingAuthor":true,"prefix":"","firstName":"Mian","middleName":"","lastName":"Guo","suffix":""}],"badges":[],"createdAt":"2024-02-15 15:49:47","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-3959082/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-3959082/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":51333004,"identity":"89c301c1-9285-42f6-9ffa-c87fe219e25d","added_by":"auto","created_at":"2024-02-19 18:02:47","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":2560515,"visible":true,"origin":"","legend":"\u003cp\u003eEpigenetic regulatory models in glioma\u003c/p\u003e\n\u003cp\u003eA. Pattern diagrams for 9 oCGI- and enhancer-dominated cis-regulatory models.\u003c/p\u003e\n\u003cp\u003eB. Composition of epigenetic regulatory models for the two glioma subtypes.\u003c/p\u003e\n\u003cp\u003eC. Methylation and gene expression levels of components of the epigenetic regulatory model in the two glioma subtypes. In the bar charts in the upper right corner, red represents Cluster 1, and blue represents Cluster 2.\u003c/p\u003e","description":"","filename":"Picture1.png","url":"https://assets-eu.researchsquare.com/files/rs-3959082/v1/e80ef458ad9b2c69f276cc28.png"},{"id":51331669,"identity":"40b21bbd-429c-44bf-907a-173b8fc45466","added_by":"auto","created_at":"2024-02-19 17:54:47","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":584040,"visible":true,"origin":"","legend":"\u003cp\u003eRegulatory characteristics of oCGIs and enhancers\u003c/p\u003e\n\u003cp\u003eA. and B. The number, locations, and distances of target genes were regulated by oCGIs and enhancers in the two glioma subtypes.\u003c/p\u003e","description":"","filename":"Picture2.png","url":"https://assets-eu.researchsquare.com/files/rs-3959082/v1/25d6b3693c7ebbe8a411e932.png"},{"id":51333005,"identity":"3ff7b027-c126-43fd-8925-a90caa05257f","added_by":"auto","created_at":"2024-02-19 18:02:48","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":4081797,"visible":true,"origin":"","legend":"\u003cp\u003eoCGI- and enhancer-dominated cis-regulatory networks in glioma cells\u003c/p\u003e\n\u003cp\u003eA. Identification of glioma subtypes using the oCGI RNA classifier with 10 models on 8 glioma samples.\u003c/p\u003e\n\u003cp\u003eB. UMAP plot depicting the cell types of 8 glioma samples.\u003c/p\u003e\n\u003cp\u003eC. Composition of oCGI- and enhancer-dominated cis-regulatory models in glioma cells.\u003c/p\u003e\n\u003cp\u003eD. Cytotrace scores in glioma cells, where 0 indicates higher differentiation, and 1 indicates lower differentiation.\u003c/p\u003e\n\u003cp\u003eE. (left) Differentiation trajectory of tumor cells, with 1 representing the starting point of differentiation. (right) The connection between Cluster 1 and Cluster 2 glioma cells with the differentiation trajectory.\u003c/p\u003e\n\u003cp\u003eF. Changes in the expression of target genes of cis-regulatory models in two glioma subtypes along the differentiation trajectory. Top: Cluster 1; bottom: Cluster 2.\u003c/p\u003e\n\u003cp\u003eG. Cell communication in the tumor microenvironment of the two glioma subtypes. Top: Quantity of ligand-receptor pairs between cell types in Cluster 1 compared with Cluster 2. Bottom: Strength of the ligand-receptor pairs between cell types in Cluster 1 compared with Cluster 2.\u003c/p\u003e","description":"","filename":"Picture3.png","url":"https://assets-eu.researchsquare.com/files/rs-3959082/v1/7de1ce999ad1c085898ceb1e.png"},{"id":51331675,"identity":"a012c504-1db6-4d81-a344-5f91c20c85d4","added_by":"auto","created_at":"2024-02-19 17:54:48","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":6269279,"visible":true,"origin":"","legend":"\u003cp\u003eSpatial transcriptomic features of oCGIs- based subtypes\u003c/p\u003e\n\u003cp\u003eA. and C. CNV levels in the necrotic niche of UKF313_T (Cluster 2) and DMG5 (Cluster 1).\u003c/p\u003e\n\u003cp\u003eB. and D. Expression of hypoxia, CNV, cytotrace score, and cis score in the necrotic niche of UKF313_T (Cluster 2) and DMG5 (Cluster 1) based on the distance from the necrotic center.\u003c/p\u003e\n\u003cp\u003eE. Cell communication in the necrotic niche of Cluster 1 and Cluster 2.\u003c/p\u003e\n\u003cp\u003eF. and G. The CNV levels in the infiltrating niche of UKF269_T (Cluster 2) and DMG4 (Cluster 1).\u003c/p\u003e\n\u003cp\u003eH. Cell communication in the infiltrating niche of Cluster 1 and Cluster 2.\u003c/p\u003e","description":"","filename":"Picture4.png","url":"https://assets-eu.researchsquare.com/files/rs-3959082/v1/eb3f300a3957fff73d25a757.png"},{"id":51331671,"identity":"090181a1-e7d6-4e60-9c2d-a3b1c32d165c","added_by":"auto","created_at":"2024-02-19 17:54:48","extension":"png","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":3205258,"visible":true,"origin":"","legend":"\u003cp\u003eoCGIs/Enhancer-related genes contribute to treatment resistance in glioma patients\u003c/p\u003e\n\u003cp\u003eA. and B. Higher C2 cis scores result in shorter overall survival in patients, based on TCGA and CGGA datasets;\u003c/p\u003e\n\u003cp\u003eC. UMAP plot illustrating the bcscore of glioma cells for two subtypes;\u003c/p\u003e\n\u003cp\u003eD. Specific drugs showing sensitivity in the two subtypes.\u003c/p\u003e\n\u003cp\u003eE. UMAP plots of bcscores for Ingenol mebutate and Marinopyrrole A in the two subtypes (top) and the distribution of bcscores in tumor cells (bottom).\u003c/p\u003e","description":"","filename":"Picture5.png","url":"https://assets-eu.researchsquare.com/files/rs-3959082/v1/c9325d05e3ad0e1dea798ae7.png"},{"id":51331676,"identity":"d8db6879-177f-4ab3-b0d5-fcbf15e612b5","added_by":"auto","created_at":"2024-02-19 17:54:48","extension":"png","order_by":6,"title":"Figure 6","display":"","copyAsset":false,"role":"figure","size":2832785,"visible":true,"origin":"","legend":"\u003cp\u003eValidation of oCGIs-dominated cis-regulatory effects\u003c/p\u003e\n\u003cp\u003eA. Application of the oCGI DNA methylation classifier to 49 glioma cell lines.\u003c/p\u003e\n\u003cp\u003eB. ChIP qPCR showing significant enrichment of E2F7 in the oCGI Chr11:728884-729383 region in Cluster 2 cell lines (LN229 and U251). This enrichment disappeared after oCGI knockout. No such enrichment was observed in Cluster 1 cell lines (A172 and SF126).\u003c/p\u003e\n\u003cp\u003eC. In LN229 and U251 cell lines, knockout of oCGI Chr11:728884-729383 decreased the expression of POLR2L, CD133, and SOX2 compared with the control group.\u003c/p\u003e\n\u003cp\u003eD. In A172 and SF126 cell lines, knockout of oCGI Chr11:728884-729383 did not significantly affect the expression of POLR2L, CD133, and SOX2 compared with the control group.\u003c/p\u003e\n\u003cp\u003eE. In LN229 and U251 cell lines, knockout of oCGI Chr11:728884-729383 decreased cell viability compared with the control group.\u003c/p\u003e\n\u003cp\u003eF. In A172 and SF126 cell lines, knockout of oCGI Chr11:728884-729383 did not significantly change cell viability compared with the control group.\u003c/p\u003e\n\u003cp\u003eG. and H. In LN229 and U251 cell lines, knockout of oCGI Chr11:728884-729383 resulted in a noticeable attenuation of KI67 (green) compared to the control group. (Blue: nucleus).\u003c/p\u003e\n\u003cp\u003eI. Drug sensitivity screening of two glioma subtypes cell lines. (*P \u0026lt; 0.05, **P \u0026lt; 0.01, ***P \u0026lt; 0.001, and ****P \u0026lt; 0.0001, P was calculated by unpaired t-test.)\u003c/p\u003e","description":"","filename":"Picture6.png","url":"https://assets-eu.researchsquare.com/files/rs-3959082/v1/2b31a6c1b8a9b9e8f3551a99.png"},{"id":51758894,"identity":"1bf73527-4ad6-4d35-ba9c-fd01a9a2f89b","added_by":"auto","created_at":"2024-02-28 15:28:46","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":3161292,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-3959082/v1/4b1701d1-a4d0-47ff-aa8d-fde2e3f6a2bf.pdf"},{"id":51331673,"identity":"5990370e-db6d-4d2e-9133-6e66d14eba48","added_by":"auto","created_at":"2024-02-19 17:54:48","extension":"docx","order_by":2,"title":"","display":"","copyAsset":false,"role":"supplement","size":8793471,"visible":true,"origin":"","legend":"","description":"","filename":"SupplmentaryFigures.docx","url":"https://assets-eu.researchsquare.com/files/rs-3959082/v1/2e1a5641ef127d16a96ee1b3.docx"},{"id":51331668,"identity":"3cc460cf-1bcf-4f11-a699-3637a41e20ea","added_by":"auto","created_at":"2024-02-19 17:54:47","extension":"xlsx","order_by":2,"title":"","display":"","copyAsset":false,"role":"supplement","size":9479,"visible":true,"origin":"","legend":"","description":"","filename":"TableS1.xlsx","url":"https://assets-eu.researchsquare.com/files/rs-3959082/v1/8cc099c16efa5a700ecd7de5.xlsx"}],"financialInterests":"No competing interests reported.","formattedTitle":"Integration of Multi-omics Data Revealed the Orphan CpG Islands and Enhancer-dominated Cis-regulatory Network in Glioma","fulltext":[{"header":"Introduction","content":"\u003cp\u003eGlioma is the most prevalent primary tumor of the brain and spinal cord, with glioblastoma multiforme (GBM) being the most frequent primary malignant tumor of the brain and central nervous system. It accounts for 14.3% of all tumors and 49.1% of malignant tumors. Despite significant efforts from both basic and clinical researchers, the median survival of glioblastoma patients remains only 8 months\u003csup\u003e1\u003c/sup\u003e. The underlying cause of this poor prognosis is the intricate regulatory network of glioma, endowing it with the capacity to adapt to various hostile environments. Elucidating the role of a single target is considerably constrained in this scenario\u003csup\u003e2,3\u003c/sup\u003e. With the advancement of sequencing technologies, high-resolution insights have gradually emerged into the development of glioma, paving the way for understanding transcriptional regulation in glioma \u003csup\u003e4,5\u003c/sup\u003e.\u003c/p\u003e \u003cp\u003eEpigenetic modifications, as a regulatory layer, play a key role in both the upstream and downstream components of the transcriptional network\u003csup\u003e6,7\u003c/sup\u003e. Methylation is one of the most prevalent mechanisms regulating transcription\u003csup\u003e8\u003c/sup\u003e. CpG islands (CGIs) are found in various DNA elements involved in the transcriptional regulatory network, with more than half of promoter regions hosting clustered CGIs\u003csup\u003e9,10\u003c/sup\u003e. Furthermore, the presence of CGIs significantly enhances the transcription-activating capacity of enhancers\u003csup\u003e11\u003c/sup\u003e. Enhancers, as distal cis-regulatory elements, are different from proximal regulatory elements such as promoters. They rely on the 3D structure of chromosomes to achieve long-range regulation, thereby playing a pivotal role in the intricate regulatory network of glioma\u003csup\u003e12,13\u003c/sup\u003e. Compared to the high density of CGIs in the promoter regions, CGIs are relatively sparse within enhancers. Nevertheless, the function of both is markedly regulated by the methylation levels within their respective CGIs. Increased methylation reduces chromatin accessibility, consequently affecting the binding of enhancers to transcription factors (TFs)\u003csup\u003e14,15\u003c/sup\u003e. Aberrant epigenetic modifications of CGIs result in transcriptional dysregulation. Abnormal CGIs methylation is closely associated with various diseases, including glioma\u003csup\u003e16,17\u003c/sup\u003e. Excluding CGIs located in the classical regulatory regions, the genome still harbors nearly half of orphan CpG islands (oCGIs)\u003csup\u003e18\u003c/sup\u003e. These isolated oCGIs have long been overlooked, and few studies indicated that some oCGIs play an indispensable role in the positive regulatory effect of enhancers located within the same topologically associating domains (TADs)\u003csup\u003e19\u003c/sup\u003e. Not all oCGIs necessarily serve as bridges for enhancer function, and the independent regulatory potential of oCGIs remains uncertain. These aspects remain unknown in the intricate regulatory network of glioma.\u003c/p\u003e \u003cp\u003eIn this study, we found that oCGIs act as atypical enhancers, exerting cis-regulatory effects and collaboratively regulating target genes in coordination with enhancers, establishing a complex cis-transcriptional regulatory network in glioma. Furthermore, single-cell multi-omics data revealed that the cis-regulatory role of oCGIs, in conjunction with enhancers, is crucial for maintaining the stemness of glioma cells and is closely associated with various biological behaviors, such as necrosis and invasion. Additionally, it plays a crucial role in treatment resistance, leading to an adverse prognosis for patients. The cis-regulatory role of oCGIs was validated in glioma cell lines. We comprehensively explored potential mechanisms underlying the interaction between oCGIs and enhancers, providing a novel perspective for unraveling the intricate regulatory network in glioma.\u003c/p\u003e "},{"header":"Methods","content":"\u003cp\u003eData collection and quality control\u003c/p\u003e \u003cp\u003eWe obtained RNA-seq and DNA methylation data for low-grade glioma (LGG) and GBM from the Cancer Genome Atlas (TCGA) through the UCSC Xena platform (\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://xena.ucsc.edu/\u003c/span\u003e\u003cspan address=\"https://xena.ucsc.edu/\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e) and directly downloaded ATAC-seq data from the TCGA (\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://portal.gdc.cancer.gov/\u003c/span\u003e\u003cspan address=\"https://portal.gdc.cancer.gov/\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e). The 325 cohort and 693 cohort from the Chinese Glioma Genome Atlas (CGGA, \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttp://www.cgga.org.cn/\u003c/span\u003e\u003cspan address=\"http://www.cgga.org.cn/\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e) were utilized for the validation of treatment response. Additionally, for validation purposes, we acquired data from 24 samples, including RNA-seq, DNA methylation-seq, and ChIP-seq, from the Gene Expression Omnibus (GEO) database (\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://www.ncbi.nlm.nih.gov/geo/\u003c/span\u003e\u003cspan address=\"https://www.ncbi.nlm.nih.gov/geo/\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e, GEO accession: GSE121719, GSE121720, GSE121721, GSE189859, GSE189860, and GSE189857)\u003csup\u003e20,21\u003c/sup\u003e. Data for chromosomal three-dimensional structures (TADs) was retrieved from GSE77565\u003csup\u003e22\u003c/sup\u003e. To validate our findings at the single-cell level, we utilized single-cell RNA sequencing (scRNA-seq) and single-cell reduced representation bisulfite sequencing (scRRBS-seq) data from 8 glioma samples provided by Verhaak \u003cem\u003eet al\u003c/em\u003e.\u003csup\u003e23\u003c/sup\u003e. Additionally, spatial transcriptomics sequencing (stRNA-seq) data from 29 glioma samples were obtained from the study of Schnell and GSE194329 dataset\u003csup\u003e4,24\u003c/sup\u003e. Genomic single nucleotide polymorphism (SNP) data were obtained from the 1000 Genomes Project Phase 3\u003csup\u003e25\u003c/sup\u003e. Enhancer data were acquired from The FANTOM5 project\u003csup\u003e26\u003c/sup\u003e. DNA methylation data for 49 glioma cell lines were obtained from GSE68379. Drug sensitivity data for the cell lines were obtained from the Genomics of Drug Sensitivity in Cancer project (GDSC, \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://www.cancerrxgene.org/celllines\u003c/span\u003e\u003cspan address=\"https://www.cancerrxgene.org/celllines\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e). All data were analyzed based on the hg19.\u003c/p\u003e \u003cp\u003eIdentifying glioma subtypes based on DNA methylation levels\u003c/p\u003e \u003cp\u003eWe applied the following criteria to quality control TCGA DNA methylation data: (\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e) Exclusion of probes containing SNPs; (\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e) Exclusion of probes expressed in less than 20% of samples; and (\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e) Exclusion of samples with less than 20% probe expression. We utilized the k-nearest neighbors (KNN) method for imputing missing values in DNA methylation data. CGIs within a\u0026thinsp;\u0026plusmn;\u0026thinsp;250bp region were merged, and the region extending 1500bp upstream and 500bp downstream of the transcription start site (TSS) was defined as the promoter region. CGIs overlapping with promoters and enhancers were removed, and the remaining CGIs were defined as oCGIs.\u003c/p\u003e \u003cp\u003eWe conducted consensus clustering of the sample's oCGIs and enhancer methylation data. The number of clusters ranged from 2 to 20. The minimum number of subtypes necessary to effectively differentiate the samples was determined based on the results\u003csup\u003e27\u003c/sup\u003e. Clusters with less than 5 samples were excluded from the analysis.\u003c/p\u003e \u003cp\u003eConstruction of the oCGIs- and enhancers-dominated cis-regulatory network\u003c/p\u003e \u003cp\u003eWe measured the strength of the interaction between oCGIs or enhancers and target genes within the same TADs using mutual information (MI)\u003csup\u003e28\u003c/sup\u003e. The following criteria were employed to construct oCGIs-enhancer-gene triplets targeting the same gene: the MI calculated for oCGIs-gene or enhancer-gene pairs had an adjusted p-value of less than 0.05, and any pair with an adjusted p-value of greater than 0.5 in either oCGIs-gene or enhancer-gene relationship was filtered out. The preliminary selection of oCGIs-enhancer-gene triplets was analyzed to identify their regulatory patterns using Bayesian networks (\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://www.bnlearn.com/\u003c/span\u003e\u003cspan address=\"https://www.bnlearn.com/\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e). For each triplet, we computed joint probabilities for potential regulatory patterns. For example, the oCGI direct model represented the direct oCGI regulation of target gene expression, while the oCGI cascade model indicated that oCGI regulates the target gene by modulating enhancers. The joint probabilities were calculated as follows:\u003c/p\u003e \u003cp\u003eoCGI-dominated:\u003c/p\u003e \u003cp\u003eoCGI direct: P (O, E, G)\u0026thinsp;=\u0026thinsp;P (O) * P (G | O) * P (E)\u003c/p\u003e \u003cp\u003eoCGI cascade: P (O, E, G)\u0026thinsp;=\u0026thinsp;P (O) * P (E | O) * P (G | E)\u003c/p\u003e \u003cp\u003eoCGI co-responsive: P (O, E, G)\u0026thinsp;=\u0026thinsp;P (O) * P (E | O) * P (G | O)\u003c/p\u003e \u003cp\u003eoCGI composite: P (O, E, G)\u0026thinsp;=\u0026thinsp;P(O) * P (E | O) * P (G | O: E)\u003c/p\u003e \u003cp\u003eEnhancer-dominated:\u003c/p\u003e \u003cp\u003eEnhancer direct: P (O, E, G)\u0026thinsp;=\u0026thinsp;P (O) * P (G | E) * P(E)\u003c/p\u003e \u003cp\u003eEnhancer cascade: P (O, E, G)\u0026thinsp;=\u0026thinsp;P (E) * P (O | E) * P (G | O)\u003c/p\u003e \u003cp\u003eEnhancer co-responsive: P (O, E, G)\u0026thinsp;=\u0026thinsp;P (E) * P (O | E) * P (G | E)\u003c/p\u003e \u003cp\u003eEnhancer composite: P (O, E, G)\u0026thinsp;=\u0026thinsp;P (E) * P (O | E) * P (G | O: E)\u003c/p\u003e \u003cp\u003eCo-dominated: P (O, E, G)\u0026thinsp;=\u0026thinsp;P (O) * P (E) * P (G | O: E)\u003c/p\u003e \u003cp\u003eWhere P(O) and P(E) represent the probability distributions of DNA methylation in oCGIs and enhancers, respectively. P (G|O) indicates the conditional probability of gene expression regulated by oCGIs, and P (G|O: E) represents the conditional probability of gene expression regulated simultaneously by both oCGIs and enhancers. The definitions of the other terms are similar to those mentioned above.\u003c/p\u003e \u003cp\u003eWe selected the model with the smallest Akaike Information Criterion (AIC) as the regulatory pattern for each triplet. Additionally, we conducted independence testing for each triplet. For example, the p-value of independence testing between oCGI-enhancer and enhancer-gene was less than 0.05 for the oCGI direct model of the triplet to be considered valid.\u003c/p\u003e \u003cp\u003eConsistent with conventional understanding, the agreement between DNA methylation level and RNA expression is crucial for model validation. Initially, we defined hypomethylation as β\u0026thinsp;\u0026lt;\u0026thinsp;0.32 and hypermethylation as β\u0026thinsp;\u0026gt;\u0026thinsp;0.79 using the β distribution of glioma methylation probe values. Hemimethylation was defined as β values falling within the range of 0.32 to 0.79 (\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://github.com/koyelucd/betaclust\u003c/span\u003e\u003cspan address=\"https://github.com/koyelucd/betaclust\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e). Triplets identified as the CGI direct model in Cluster 2, had an oCGI methylation level lower than that in Cluster 1 and a gene expression level higher than that in Cluster 1. Enrichment analysis was conducted for all upregulated genes in Cluster 2\u003csup\u003e29\u003c/sup\u003e. We employed FIMO to identify potential TF-binding motifs in oCGIs and enhancers\u003csup\u003e30\u003c/sup\u003e.\u003c/p\u003e \u003cp\u003eConstruction of glioma classifiers based on multi-omics data of oCGIs\u003c/p\u003e \u003cp\u003eWe constructed the classifier using three different methods, which were cross-validated to enhance the credibility of the model. First, we employed the Partitioning Around Medoids (PAM) algorithm to discern the sample allocations based on distance. Next, we incorporated 11 machine learning algorithms, including cv_glmnet, featureless, kknn, lda, log_reg, naive_bayes, ranger, rpart, svm, xgboost, and debug, to construct the models based on the expression of oCGIs and DNA methylation levels. The machine learning algorithms were implemented using the mlr3 package\u003csup\u003e31\u003c/sup\u003e. TCGA was utilized as the training dataset, while 24 samples from GEO were employed as the validation dataset. We aligned ChIP-seq data from the validation dataset (H3K27ac, H3K4me3, H3K4me1, and H3K27me3) and TCGA ATAC data to the corresponding regions of oCGIs, enhancers, and promoters.\u003c/p\u003e \u003cp\u003escRNA-seq and scRRBS-seq data\u003c/p\u003e \u003cp\u003eWe conducted quality control on 55,284 cells from the 11 glioma samples in the Verhaak cohort. Cells with a mitochondrial gene percentage exceeding 20% and doublets were filtered out. The samples were integrated using Harmony\u003csup\u003e32\u003c/sup\u003e. Tumor cells were annotated using marker genes in conjunction with copy number variations (\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://github.com/broadinstitute/infercnv\u003c/span\u003e\u003cspan address=\"https://github.com/broadinstitute/infercnv\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e). Cytotrace was employed to assess the differentiation level of tumor cells\u003csup\u003e33\u003c/sup\u003e. Monocle3 was used to identify the differentiation trajectories of glioma cells (\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://github.com/cole-trapnell-lab/monocle3\u003c/span\u003e\u003cspan address=\"https://github.com/cole-trapnell-lab/monocle3\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e). SENIC was utilized to identify significantly activated TFs in glioma subtypes\u003csup\u003e34\u003c/sup\u003e. We applied the same criteria to perform quality control on the scRRBS data. Triplets were identified in tumor cells based on scRNA and scRRBS data. The difference in drug sensitivity between two subtypes of glioma cells was analyzed using the beyondcell package\u003csup\u003e35\u003c/sup\u003e.\u003c/p\u003e \u003cp\u003eSpatial transcriptomics data\u003c/p\u003e \u003cp\u003eThe stRNA-seq data from 29 glioma samples were initially used to identify glioma subtypes using pseudobulk analysis. Subsequently, two pathologists jointly divided the images of stRNA-seq into four regions: vascular, necrotic, cellular, and infiltrating. All stRNA-seq analyses were conducted using SPATA2 package\u003csup\u003e4\u003c/sup\u003e. Copy number variation was assessed using the runCnvAnalysis function. Images representing different features were visualized using the plotSurfaceComparison function. The CellTrek package was employed to integrate scRNA-seq and stRNA-seq data\u003csup\u003e36\u003c/sup\u003e. The communication between tumor cells and other components of the tumor microenvironment in different niches was identified using the nichenetr package\u003csup\u003e37\u003c/sup\u003e.\u003c/p\u003e \u003cp\u003eCell line\u003c/p\u003e \u003cp\u003eThe glioma cell lines U251, LN229, and A172 were obtained from the Department of Neurosurgery, the Second Affiliated Hospital of Harbin Medical University. The SF126 cell line was acquired from Pricella (Wuhan, China). Cell lines with oCGI (chr11:728884\u0026ndash;729383) knockout were established using the CRISPR/Cas9 system. sgRNAs were designed using Benchling\u0026rsquo;s CRISPR toolkit (sgRNA1: AGCCCCTTGGAAGAAACGGG; sgRNA2: GGAAGCCCCTTGGAAGAAAC).\u003c/p\u003e \u003cp\u003eQuantitative PCR with reverse transcription\u003c/p\u003e \u003cp\u003eTotal RNA was extracted from glioma cell lines using Trizol (Beyotime, China), and reverse transcription was performed using reverse transcription reagents (RNase H-, RNase inhibitor, and dNTP Mix) (Beyotime, China) following the manufacturer's instructions. All primers are listed in Table \u003cspan refid=\"MOESM1\" class=\"InternalRef\"\u003eS1\u003c/span\u003e.\u003c/p\u003e \u003cp\u003eChromatin immunoprecipitation (ChIP)\u003c/p\u003e \u003cp\u003eThe ChIP assay was conducted following the protocol outlined in the ChIP assay kit (Beyotime, China). Subsequently, PCR was employed to assess the immunoprecipitated DNA level. The primers are listed in Table \u003cspan refid=\"MOESM1\" class=\"InternalRef\"\u003eS1\u003c/span\u003e.\u003c/p\u003e \u003cp\u003eMTT Assay\u003c/p\u003e \u003cp\u003eIn the control group and two oCGI knockout groups (chr11:728884\u0026ndash;729383 KO1 and chr11:728884\u0026ndash;729383 KO2), 10 \u0026micro;L of MTT (5 mg/mL) (Beyotime, China) was added to each well of a 96-well plate. After 4 hours of incubation, 100 \u0026micro;L of formazan (Beyotime, China) was added, and the absorbance was read at 490 nm.\u003c/p\u003e \u003cp\u003eImmunofluorescence\u003c/p\u003e \u003cp\u003eIn brief, coverslips with confluent cells were fixed with 4% paraformaldehyde. Then, immunostaining was performed using KI67 polyclonal antibody (ProteinTech Group, Inc., USA). Finally, we conducted counterstaining with DAPI.\u003c/p\u003e"},{"header":"Results","content":"\u003cp\u003eMethylation characteristics of oCGIs and enhancers in glioma\u003c/p\u003e\n\u003cp\u003eWe initially identified CGIs in different DNA elements. Consistent with previous findings, the majority of CGIs (78.99%) overlapped with genes or promoter regions, but approximately 20% of oCGIs were dispersed throughout the genome (Figure S3A). These dispersed oCGIs have long been ignored. We performed consensus clustering based on the methylation of oCGIs and enhancers to illuminate the methylation characteristics of oCGIs and classical enhancers. We found that glioma can be broadly classified into two subtypes under various clustering scenarios (Figure \u003cspan class=\"InternalRef\"\u003eS1\u003c/span\u003e and Figure \u003cspan class=\"InternalRef\"\u003eS2\u003c/span\u003e). Furthermore, the clustering results based on oCGIs and enhancers exhibited significant consistency (Figure S3B). Cluster 1 (C1) was predominantly composed of LGG, whereas Cluster 2 (C2) primarily included GBM (Figure S3C). We utilized the clustering results based on oCGIs to further explore the role of oCGIs.\u003c/p\u003e\n\u003cp\u003eIdentifying the oCGI- and enhancer-dominated cis-regulatory network\u003c/p\u003e\n\u003cp\u003eTo elucidate the regulatory patterns among oCGIs, enhancers, and genes, we initially screened combinations with potential regulatory effects based on interactions between oCGI-gene/enhancer-gene pairs in the same TAD. Furthermore, we proposed 9 regulatory models of oCGI-enhancer-gene triplets (Fig.\u0026nbsp;\u003cspan class=\"InternalRef\"\u003e1\u003c/span\u003eA). These 9 models elucidated all potential regulatory possibilities, including 4 models where oCGI or enhancer individually played a predominant role (direct, cased, co-responsive, and composite), and 1 model where they were co-dominated (co-dominated). For instance, in the oCGI-dominated direct model, the methylation level of oCGI affected its binding capacity to TFs, thereby affecting the binding of TFs to target gene promoter regions and regulating transcription\u003csup\u003e14,15\u003c/sup\u003e. Subsequently, based on Bayesian networks, we determined the most suitable model for each oCGI-enhancer-gene triplet. In addition, models were selected based on the methylation levels of oCGIs and enhancers in conjunction with gene expression. In the end, only 7 models were validated in our dataset (oCGI_Direct, Enhancer_Direct, oCGI_responsive, Enhancer_responsive, oCGI_Cased, Enhancer_Cased, Coordinate). In both glioma subtypes, the oCGI direct and enhancer direct model predominated, indicating that despite the presence of more complex indirect regulation, oCGIs and enhancers played a dominant role in the direct regulation of target genes. Moreover, the regulation of upregulated triplets in Cluster 2 was the central theme across all models. (Fig.\u0026nbsp;\u003cspan class=\"InternalRef\"\u003e1\u003c/span\u003eB-C). Furthermore, these upregulated genes were associated with positive regulation of the tumor (positive regulation of cytokine production and cell migration) (Figure S3D).\u003c/p\u003e\n\u003cp\u003eoCGIs exhibited similar regulatory characteristics to classical enhancers. Both oCGIs and enhancers can regulate more than one target gene, but the predominant mode of regulation is one-to-one. Secondly, the regulatory effect of oCGIs or enhancers on target genes within the same TAD is not often limited to adjacent loci. On the contrary, non-adjacent regulation models accounted for the majority of the model count. In a significant number of models, oCGIs or enhancers were separated from their target genes by more than 10 genes. Thirdly, the long-range regulatory characteristics of enhancers were also evident in oCGIs. Similar to enhancers, the regulatory distance of oCGIs typically exceeded 500 kb and even surpassed 1000 kb in some cases. These characteristics were observed in both glioma subtypes (Fig.\u0026nbsp;\u003cspan class=\"InternalRef\"\u003e2\u003c/span\u003eA-B).\u003c/p\u003e\n\u003cp\u003e\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eA. and B. The number, locations, and distances of target genes were regulated by oCGIs and enhancers in the two glioma subtypes.\u003c/p\u003e\n\u003cp\u003eStates of oCGIs and enhancers in the cis-regulatory models\u003c/p\u003e\n\u003cp\u003eDifferent modes of histone modifications by enhancers serve as markers of their activity. H3K27ac is a marker of active enhancers with positive transcriptional regulatory activity, whereas H3K27me3 is a marker of silenced enhancers. Co-expression of H3K4me1 and H3K27ac indicates enhancers with moderate activity. Promoters also have similar activation and silencing markers. H3K4me3 is significantly enriched in promoter regions as an additional marker\u003csup\u003e38,39\u003c/sup\u003e. Due to the lack of chromatin modification data corresponding to TCGA samples, we used external samples to determine the mode of chromatin modification of the two glioma subtypes. We first determined to which subtype the external samples belonged. Models in 3 dimensions combined with various machine learning algorithms were used to ensure the accuracy of our results. We initially identified the subtypes of 24 glioma samples based on distance metrics (Figure S3E). Next, we constructed classifiers for TCGA glioma samples based on oCGI methylation features and gene RNA expression features using 11 machine-learning algorithms. Except for featureless, log_reg, and debug algorithms, other algorithms exhibited excellent discriminatory capabilities for glioma samples in the TCGA (Figure S4A). We trained 10 models for each of the eight algorithms based on different resampling results. These models were then applied to the test set samples. Model performance was evaluated using classification error rate and ROC, with cv_glmnet showing excellent performance across all 10 models on both oCGI methylation and RNA expression data. It correctly distinguished the subtypes of all samples when utilizing both types of data (Figure S4A-B and Figure S5).\u003c/p\u003e\n\u003cp\u003eNext, we analyzed the distribution of different histone modifications across the genome. Both subtypes exhibited similar distribution patterns, with peaks of different histone modifications significantly enriched in gene-related regions (Figure S4C). Furthermore, we characterized the histone modification features in the 3000 kb regions around all TSS, oCGIs, and enhancers in the two subtypes. Cluster 2 exhibited lower methylation levels, higher chromatin accessibility, and higher levels of active promoter and enhancer feature markers, such as H3K27ac peaks, across all regions. All oCGIs displayed histone modification features similar to enhancers (Figure S6A). These findings support their role as \"atypical enhancers\" with similar functions. Consistent with the dominance of upregulated models in Cluster 2, Cluster 2 exhibited higher overall transcriptional regulatory activity. Therefore, we focused on genes activated in Cluster 2 and repressed in Cluster 1 in the follow-up analyses. These genes displayed similar histone modification features (Figure S6B-C), confirming our previous findings. Finally, combining the regulatory modes obtained from the triplet analysis, the predominant model characteristics in Cluster 1 were oCGIs and enhancers with lower activity and higher methylation levels, suppressing gene transcription. In contrast, the dominant regulatory mode in Cluster 2 was characterized by active oCGIs and enhancers with lower methylation levels, enabling positive transcriptional regulation (Figure S4D-E).\u003c/p\u003e\n\u003cp\u003e\u0026nbsp;oCGIs- and enhancers-dominated cis-network in glioma cells\u003c/p\u003e\n\u003cp\u003eAlthough we have constructed a cis-regulatory network based on glioma subtypes, the specific manner in which these regulatory modes function and whether they directly impact tumor cell lifecycle remain elusive. Thus, we investigated regulatory modes dominated by oCGIs and enhancers in tumor cells utilizing scRNA-seq.\u0026nbsp;Our initial approach involved ensuring quality control and cell annotation for scRNA-seq data from 8 samples and filtering cells based on copy number variations, yielding 21,370 tumor cells (Fig.\u0026nbsp;\u003cspan class=\"InternalRef\"\u003e3\u003c/span\u003eB and \u003cspan class=\"InternalRef\"\u003eS7\u003c/span\u003eA-G). Subsequently, we identified subtypes for these 8 samples using a previously developed RNA classifier (Fig.\u0026nbsp;\u003cspan class=\"InternalRef\"\u003e3\u003c/span\u003eA) and only retained the tumor cells for subsequent analysis. It should be noted that significant heterogeneity existed among tumor cells from different subtypes (Fig.\u0026nbsp;\u003cspan class=\"InternalRef\"\u003e3\u003c/span\u003eE and Figure S7G). Although limited coverage of methylation sites in scRRBS-seq prevented identification of regulatory patterns for some triplets, consistent with previous results, oCGI_direct and enhancer_direct modes still primarily regulated both subtypes (Fig.\u0026nbsp;\u003cspan class=\"InternalRef\"\u003e3\u003c/span\u003eC). To further characterize the heterogeneity between tumor cells from the two subtypes, we assessed the proliferative potential of tumor cells. Tumor cells in Cluster 2 exhibited a significantly lower cytotrace score and a lower degree of differentiation (Fig.\u0026nbsp;\u003cspan class=\"InternalRef\"\u003e3\u003c/span\u003eD). In line with the results of pseudotime analysis, the tumor cells from Cluster 2 primarily aggregated at the initiation of the differentiation trajectory. As cells progressed along the differentiation trajectory, the composition of tumor cells transitioned from Cluster 2 cells to Cluster 1. Additionally, the upregulated genes in Cluster 2 exhibited a significant negative correlation with differentiation time (Fig.\u0026nbsp;\u003cspan class=\"InternalRef\"\u003e3\u003c/span\u003eE-F and \u003cspan class=\"InternalRef\"\u003eS8\u003c/span\u003e A-B).\u003c/p\u003e\n\u003cp\u003eThe mechanism by which the cis-regulatory network acts in glioma cells\u003c/p\u003e\n\u003cp\u003eWe initially focused on the communication between various components in the glioma microenvironment to elucidate the precise mechanisms by which the cis-regulatory network is involved in glioma cells. Therefore, we needed a more detailed identification of tumor cells. All non-tumor cells were classified into oligodendrocytes, microglial, macrophages, T cells, endothelial cells, and pericytes cells (Figure S8C-D). The results of cell communication in the tumor microenvironment showed that interactions between tumor cells and other cells in the tumor microenvironment were stronger in Cluster 2 than in Cluster 1 (Fig.\u0026nbsp;\u003cspan class=\"InternalRef\"\u003e3\u003c/span\u003eG). Many signals favoring tumor cells were significantly activated in Cluster 2, such as integrin-related signals (COL9A3 \u0026minus; (ITGA2\u0026thinsp;+\u0026thinsp;ITGB1), JAM3 \u0026minus; (ITGAM\u0026thinsp;+\u0026thinsp;ITGB2), COL6A1 \u0026minus; (ITGA1\u0026thinsp;+\u0026thinsp;ITGB1)), immune suppression signals (CD99\u0026thinsp;\u0026minus;\u0026thinsp;PILRA), etc. (Figure S8E). Furthermore, we explored TFs binding to oCGIs or enhancers. Initially, we scanned the DNA sequences of oCGIs or enhancers, preliminarily selecting TFs with binding potential. Subsequently, we identified the activated TFs in the two subtypes of tumor cells (Figure S8F). By combining the two strategies mentioned above, we identified TFs regulating different triplets in different subtypes of glioma cells. Based on univariate cox regression and co-expression analysis between oCGIs/enhancers-related TFs and their target genes, we identified prognosis-related key TFs (Figure S9A-B).\u003c/p\u003e\n\u003cp\u003eThe effects of the cis-network regulatory in different niches of gliomas\u003c/p\u003e\n\u003cp\u003eUsing stRNA-seq, we explored the regulatory roles of oCGIs and enhancers in different niches. We first identified the subtypes of glioma (Figure S11A). The necrotic and infiltrating niches of the two subtypes of glioma were analyzed and confirmed by pathologists. The necrotic niche surrounding both subtypes exhibited evident hypoxia and differentiation features. Additionally, genes regulated by oCGIs and enhancers in both clusters were significantly downregulated in this region. Differences in the necrotic niche between the two subtypes were reflected in the CNV of spots in Cluster 2, gradually increasing from the necrotic center toward the outer layer. Cluster 1 showed the opposite trend (Fig.\u0026nbsp;\u003cspan class=\"InternalRef\"\u003e4\u003c/span\u003eA-D and \u003cspan class=\"InternalRef\"\u003eS10\u003c/span\u003eA-B). To elucidate the reasons leading to this difference, we integrated the 8 scRNA-seq datasets from the above studies with the respective stRNA-seq dataset for each subtype. Glioma cells mainly co-localized with microglial and endothelial cells, showing stronger colocalization with microglial in regions closer to the necrotic center (Figure S11B-C). The necrotic niche in Cluster 2 contained a higher proportion of tumor cells, increasing its CNV (Figure S11D). GO analysis showed that MHC-related functions were significantly activated near the necrotic center (Layer 1), while membrane protein activity and extracellular matrix activation were more prominent in the outer layer (Layer 2), with both subtypes exhibiting similar results (Figure S11E). Cell communication in the necrotic niche revealed a significant overactivity of various chemokines and other immune-related receptor-ligand pairs in Cluster 1 (CCL3L3-ACKR2; JAM2-JAM2; CCL2-CCR10, etc.). Similarly, various signals promoting tumor cell activation were more pronounced in Cluster 2 (S100A4-EGFR, SPP1-ITGB1, SPP1-ITGAV, etc.) (Fig.\u0026nbsp;\u003cspan class=\"InternalRef\"\u003e4\u003c/span\u003eE and \u003cspan class=\"InternalRef\"\u003eS11\u003c/span\u003eF).\u003c/p\u003e\n\u003cp\u003eIn the infiltrating niche, the CNV, hypoxia, cytotrace score, and cis score exhibited similar distributions (Fig.\u0026nbsp;\u003cspan class=\"InternalRef\"\u003e4\u003c/span\u003eF-G and \u003cspan class=\"InternalRef\"\u003eS10\u003c/span\u003eC-D). In the tumor component of the infiltrating niche, tumor cells were predominantly co-localized with oligodendrocytes and endothelial cells, and this distribution pattern was consistent between the two subtypes (Figure S12A-B). The tumor region comprised more tumor cells and fewer immune cells compared with the infiltrating region (Figure S12C). The tumor region in Cluster 1 activated additional immune-related molecular functions compared with Cluster 2 (Figure S12D-E). The communication between glioma cells and components of the tumor microenvironment revealed abundant tumor-promoting signals in the infiltrating niches of both subtypes. Immune-related signals were more abundant in Cluster 1 compared with Cluster 2 (JAM2\u0026minus;\u0026minus;JAM2; CCL3L3\u0026minus;\u0026minus;CCR1; CCL2\u0026minus;\u0026minus;CCR10) (Fig.\u0026nbsp;\u003cspan class=\"InternalRef\"\u003e4\u003c/span\u003eH and \u003cspan class=\"InternalRef\"\u003eS12\u003c/span\u003eF). The gene POLR2L, regulated by oCGI (Chr11:728884\u0026ndash;729383) in our study, played a significant role in cell communication in Cluster 2 glioma cells, although it was not the most significantly activated gene.\u003c/p\u003e\n\u003cp\u003eThe upregulation of oCGI/enhancer-related genes in Cluster 2 is associated with treatment resistance in gliomas\u003c/p\u003e\n\u003cp\u003eTo elucidate the impact of the cis-regulatory role of oCGI/enhancers on the clinical treatment of glioma patients, we stratified glioma patients from The Cancer Genome Atlas (TCGA) into high and low groups based on the expression levels of oCGI/enhancer-related genes upregulated in Cluster 2 (C2 cis score). The prognosis of patients with higher C2 cis scores was significantly worse under different treatment conditions (untreated, chemotherapy, radiotherapy, chemotherapy\u0026thinsp;+\u0026thinsp;radiotherapy) (Fig.\u0026nbsp;\u003cspan class=\"InternalRef\"\u003e5\u003c/span\u003eA). This pattern of treatment resistance was further validated in the CGGA cohorts, including the 325 cohort and 693 cohort (Fig.\u0026nbsp;\u003cspan class=\"InternalRef\"\u003e5\u003c/span\u003eB). Clustering results based on drug sensitivity of tumor cells from two subtypes, as shown by beyondcell, revealed significant heterogeneity between the two subtypes (Fig.\u0026nbsp;\u003cspan class=\"InternalRef\"\u003e5\u003c/span\u003eC). Ingenol mebutate demonstrated higher sensitivity in Cluster 2 tumor cells, while Marinopyrrole A showed greater sensitivity in Cluster 1 tumor cells (Fig.\u0026nbsp;\u003cspan class=\"InternalRef\"\u003e5\u003c/span\u003eD-F).\u003c/p\u003e\n\u003cp\u003eValidation of the cis-regulatory role of oCGIs\u003c/p\u003e\n\u003cp\u003eAs the \u003cem\u003ePOLR2L\u003c/em\u003e gene plays a crucial role in cell communication between tumor cells and other components of the tumor microenvironment, we selected oCGI (Chr11:728884\u0026ndash;729383) and its target, \u003cem\u003ePOLR2L\u003c/em\u003e gene, for validation. Additionally, we identified TF E2F7 as a potential regulatory target based on prognosis analysis and co-expression analysis from the above studies. First, we identified the subtypes based on the DNA methylation data of 49 glioma cell lines (Fig.\u0026nbsp;\u003cspan class=\"InternalRef\"\u003e6\u003c/span\u003eA). Two cell lines from each subtype were chosen for subsequent validation (Cluster 1: A172, SF126; Cluster 2: LN229, U251). The results of ChIP combined with qPCR indicated that E2F7 was significantly enriched in the region of oCGI Chr11:728884\u0026ndash;729383 in LN229 and U251 cell lines. Additionally, this enrichment disappeared upon the knockout of oCGI Chr11:728884\u0026ndash;729383 (Fig.\u0026nbsp;\u003cspan class=\"InternalRef\"\u003e6\u003c/span\u003eB). oCGI Chr11:728884\u0026ndash;729383 knockout in LN229 and U251 cell lines markedly decreased the expression of \u003cem\u003ePOLR2L\u003c/em\u003e gene and glioma stem cell-related genes, \u003cem\u003eCD133\u003c/em\u003e and \u003cem\u003eSOX2\u003c/em\u003e. However, such a regulatory relationship was not observed in A172 and SF126 cell lines (Fig.\u0026nbsp;\u003cspan class=\"InternalRef\"\u003e6\u003c/span\u003eC-D). The effect of oCGI Chr11:728884\u0026ndash;729383 on cell proliferation was validated in different glioma subtypes through MTT assay and immunofluorescence assay. After oCGI Chr11:728884\u0026ndash;729383 knockout, LN229 and U251 cell viability significantly decreased, and the Ki67 fluorescence signal became noticeably weaker compared with the control group (Fig.\u0026nbsp;\u003cspan class=\"InternalRef\"\u003e6\u003c/span\u003eE and G-H). The effect of oCGI Chr11:728884\u0026ndash;729383 knockout on the viability of A172 and SF126 cells was not significant (Fig.\u0026nbsp;\u003cspan class=\"InternalRef\"\u003e6\u003c/span\u003eF). To provide better guidance for clinical treatment, we filtered drugs effective for each subtype of glioma. Combining drug sensitivity data from 49 glioma cell lines, we compared the AUC values for the two glioma subtypes. Apart from JNK inhibitor VIII and XAV939, most anti-tumor drugs were more effective on cell lines from Cluster 1 (Fig.\u0026nbsp;\u003cspan class=\"InternalRef\"\u003e6\u003c/span\u003eI).\u003c/p\u003e"},{"header":"Discussion","content":"\u003cp\u003eRecent studies have increasingly investigated the regulatory effect of classic enhancers on critical targets in cancer. It is now widely acknowledged that enhancers accelerate oncogenesis \u003csup\u003e40\u003c/sup\u003e. As such, many researchers paid attention to the pairing of classic enhancers with critical nodes in cancer. However, for highly heterogeneous tumors like glioma, the complexity of the transcriptional regulatory network provides the potential for a broader definition of enhancers. The coordinated transcription-enhancing function of oCGIs establishes a foundation for their non-classical enhancer function\u003csup\u003e19\u003c/sup\u003e. We constructed the oCGI-enhancer-gene regulatory models to illuminate the importance of oCGIs, which have long been overlooked. We revealed the role of oCGIs as non-classical enhancers in the intricate regulatory network of glioma.\u003c/p\u003e \u003cp\u003eFirst, the comparative genomic analysis highlighted that oCGIs exhibit DNA methylation characteristics similar to classical enhancers, suggesting that oCGIs can exert transcriptional enhancer functions, either through a mechanism akin to that of classical enhancers or by collaborating with classical enhancers. Furthermore, to elucidate the precise regulatory functions of oCGIs across different glioma samples, we applied the 9 possible models of regulation, encompassing oCGIs-enhancer-gene triplets, based on the DNA methylation features of oCGIs in glioma subtypes. Through mutual information analysis and Bayesian networks, we delineated the regulatory patterns of each triplet. As expected, oCGIs directly or indirectly regulated the target genes. Notably, they exhibited regulatory characteristics similar to classical enhancers, enabling them to regulate several target genes and exert long-range regulation within the same TAD. Although oCGIs and enhancers were primarily engaged in the direct regulation of target genes within the cis-regulatory network of glioma, many more complex regulatory models were validated, suggesting that enhancers are potential targets of oCGIs. It enhances our understanding of the cis-regulatory network. The predictive models based on 11 machine-learning algorithms allowed us to accurately identify the glioma subtypes based on their DNA methylation or RNA expression levels. These findings can help integrate data from various sources and comprehensively identify the characteristics of the cis-regulatory network. Chromatin modifications serve as markers for identifying the functional states of enhancers or promoters\u003csup\u003e38,39\u003c/sup\u003e. The results of chromatin modifications in the two glioma subtypes indicated that tumor cells were more active in Cluster 2. In Cluster 2, the oCGIs- or enhancers-dominated regulatory models exerted stronger positive transcriptional control through lower methylation levels and higher chromatin accessibility. Therefore, we focused on the major regulators of transcription in Cluster 2.\u003c/p\u003e \u003cp\u003eWe employed a high-resolution single-cell atlas to further characterize the details of the cis-regulatory network in tumor cells. Our findings revealed that tumor cells from Cluster 2 typically exhibited higher rates of proliferation. Additionally, the regulatory influence of oCGI- or enhancer-dominated triplets on target genes gradually weakened after differentiation in Cluster 2, suggesting that the active cis-regulatory network in Cluster 2 is a crucial factor for maintaining the heightened activity of tumor cells. Additionally, the communication between tumor cells and the glioma microenvironment was more active in Cluster 2 compared with Cluster 1, exerting a more pronounced pro-tumor effect.\u003c/p\u003e \u003cp\u003eTo elucidate the precise mechanisms by which the cis-regulatory network and oCGIs were involved in this complex glioma microenvironment, we characterized different niches and conducted stRNA-seq.\u0026nbsp;Tumor cells in the center of the tumor exhibited stronger proliferation and invasive capabilities compared with the infiltrating region\u003csup\u003e41\u003c/sup\u003e. Additionally, due to the vigorous metabolic demands of tumor cells, the tumor center was more susceptible to necrosis. Therefore, we selected the necrotic niche as the tumor center and the infiltrating niche as the tumor periphery. In the necrotic niche of both subtypes, tumor cells were primarily co-localized with microglial and endothelial cells. However, in Cluster 2, this co-localization was associated with significantly weaker immune cell infiltration and tumor-activating signals. An increase in co-localization with oligodendrocytes was observed in the infiltrating niche, similar to the signal network in the necrotic niche. Cluster 2 showed more pronounced activation of pro-tumor signals. The cis activation in Cluster 2 is a crucial factor leading to patients' resistance to various treatment modalities. The cis activation in Cluster 2 renders patients unable to benefit from various clinical treatments and decreased sensitivity to various drugs.\u003c/p\u003e \u003cp\u003eTo explore the regulatory role of oCGIs in the tumor center and periphery, we focused on the common downstream target in cell communication between the two niches, namely the \u003cem\u003ePOLR2L\u003c/em\u003e gene, for subsequent validation. The results indicated that oCGI Chr11:728884\u0026ndash;729383 regulates glioma proliferation by modulating the regulatory effect of TFs E2F7 on the target gene \u003cem\u003ePOLR2L\u003c/em\u003e. Clinical drugs were selected based on distinct oCGIs-based glioma subtypes. Furthermore, our study provided a theoretical foundation for the subsequent development of drugs targeting oCGIs-related targets. Although we comprehensively analyzed the oCGIs- and enhancers-dominated cis-regulatory network, limitations inherent to single-cell technologies necessitate the validation of only a fraction of the identified models. With the continued advancement of sequencing technologies, we anticipate a more comprehensive understanding of the cis-regulatory network based on oCGIs.\u003c/p\u003e"},{"header":"Conclusion","content":"\u003cp\u003eFor the first time, our study comprehensively explored the role of oCGIs as non-classical enhancers in the transcriptional regulatory network of glioma. Generally, oCGIs directly regulate target genes similar to classical enhancers. In addition, we validated numerous, more intricate models, underscoring the complexity of the regulatory mechanisms of oCGIs in cancer, enabling them to act in a coordinated or regulated manner and significantly affecting glioma progression and treatment resistant. We provided in-depth insights into all potential scenarios within the glioma cis-regulatory network, laying a foundation for unraveling the mechanisms behind glioma development.\u003c/p\u003e"},{"header":"Abbreviations","content":"\u003cdiv class=\"DefinitionList\"\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eAIC\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eAkaike Information Criterion\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eCGGA\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eChinese Glioma Genome Atlas\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eCGIs\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eCpG islands\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eC1\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eCluster1\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eC2\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eCluster2\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eGBM\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eglioblastoma multiforme\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eGEO\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eGene Expression Omnibus\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eKNN\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003ek-nearest neighbors\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eLGG\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003elow-grade glioma\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eMI\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003emutual information\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eoCGIs\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eorphan CpG islands\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003escRNA-seq\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003esingle-cell RNA sequencing\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003escRRBS-seq\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003esingle-cell reduced representation bisulfite sequencing\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003estRNA-seq\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003espatial transcriptomics sequencing\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eSNP\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003esingle nucleotide polymorphism\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eTADs\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003etopologically associating domains\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eTCGA\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003ethe Cancer Genome Atlas\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eTFs\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003etranscription factors\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eTSS\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003etranscription start site\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003c/div\u003e"},{"header":"Declarations","content":"\u003cp\u003eEthics approval and consent to participate\u003c/p\u003e\n\u003cp\u003eNot applicable.\u003c/p\u003e\n\u003cp\u003eConsent for publication\u003c/p\u003e\n\u003cp\u003eNot applicable.\u003c/p\u003e\n\u003cp\u003eAvailability of data and materials\u003c/p\u003e\n\u003cp\u003eThe datasets used and/or analyzed during the current study are available from public datasets. Software and resources used for the analyses are described in each method section.\u003c/p\u003e\n\u003cp\u003eCompeting Interests\u003c/p\u003e\n\u003cp\u003eThe authors have declared that no competing interest exists.\u003c/p\u003e\n\u003cp\u003eAuthors' contributions\u003c/p\u003e\n\u003cp\u003eJiawei Yao: Conceptualization, Data curation, Formal analysis, Methodology, Software, Visualization, Roles/Writing - original draft, Writing - review \u0026amp; editing; Penglei Yao: Data curation, Formal analysis, Methodology, Software, Validation; Yang Li: Methodology, Resources; Ke He: Investigation, Software, Methodology, Validation; Xinqi Ma: Software, Validation; Qingsong Yang: Methodology, Visualization; Junming Jia: Roles/Writing - original draft; Zeren Chen: Investigation, Methodology; Shuqing Gu: Investigation; Weihua Li: Funding acquisition, Supervision; Guangzhi Wang: Resources, Supervision; Mian Guo: Conceptualization, Funding acquisition, Supervision, Project administration, Writing - review \u0026amp; editing.\u003c/p\u003e\n\u003cp\u003eAcknowledgements\u003c/p\u003e\n\u003cp\u003eThis work was supported by National Natural Science Foundation of China (82173384 and 81773161) and Shenzhen Science and Technology Innovation Commission (JCYJ20200109120205924).\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\u003cli\u003e\u003cspan\u003eOstrom QT, Cioffi G, Waite K, Kruchko C, Barnholtz-Sloan JS (2021) CBTRUS Statistical Report: Primary Brain and Other Central Nervous System Tumors Diagnosed in the United States in 2014\u0026ndash;2018. Neuro Oncol 23(12 Suppl 2):iii1\u0026ndash;iii105\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSonabend AM, Bansal M, Guarnieri P et al (2014) The transcriptional regulatory network of proneural glioma determines the genetic alterations selected during tumor progression. Cancer Res 74(5):1440\u0026ndash;1451\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eXiong A, Zhang J, Chen Y, Zhang Y, Yang F (2022) Integrated single-cell transcriptomic analyses reveal that GPNMB-high macrophages promote PN-MES transition and impede T cell activation in GBM. EBioMedicine 83:104239\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRavi VM, Will P, Kueckelhaus J et al (2022) Spatially resolved multi-omics deciphers bidirectional tumor-host interdependence in glioblastoma. Cancer Cell 40(6):639\u0026ndash;655e613\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eZheng Y, Carrillo-Perez F, Pizurica M, Heiland DH, Gevaert O (2023) Spatial cellular architecture predicts prognosis in glioblastoma. Nat Commun 14(1):4122\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eZhao LY, Song J, Liu Y, Song CX, Yi C (2020) Mapping the epigenetic modifications of DNA and RNA. Protein Cell 11(11):792\u0026ndash;808\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHogg SJ, Beavis PA, Dawson MA, Johnstone RW (2020) Targeting the epigenetic regulation of antitumour immunity. Nat Rev Drug Discov 19(11):776\u0026ndash;800\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHu C, Peng K, Wu Q et al (2021) HDAC1 and 2 regulate endothelial VCAM-1 expression and atherogenesis by suppressing methylation of the GATA6 promoter. Theranostics 11(11):5605\u0026ndash;5619\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eFerreira HJ, Esteller M (2018) CpG Islands in Cancer: Heads, Tails, and Sides. Methods Mol Biol 1766:49\u0026ndash;80\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePatsouras MD, Karagianni P, Kogionou P, Vlachoyiannopoulos PG (2019) Differential CpG methylation of the promoter of interleukin 8 and the first intron of tissue factor in Antiphospholipid syndrome. J Autoimmun 102:159\u0026ndash;166\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSteinhaus R, Gonzalez T, Seelow D, Robinson PN (2020) Pervasive and CpG-dependent promoter-like characteristics of transcribed enhancers. Nucleic Acids Res 48(10):5306\u0026ndash;5317\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eXiao J, Jin X, Zhang C et al (2021) Systematic analysis of enhancer regulatory circuit perturbation driven by copy number variations in malignant glioma. Theranostics 11(7):3060\u0026ndash;3073\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAndersson R, Sandelin A (2020) Determinants of enhancer and promoter activities of regulatory elements. Nat Rev Genet 21(2):71\u0026ndash;87\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAngeloni A, Bogdanovic O (2019) Enhancer DNA methylation: implications for gene regulation. Essays Biochem 63(6):707\u0026ndash;715\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eThurman RE, Rynes E, Humbert R et al (2012) The accessible chromatin landscape of the human genome. Nature 489(7414):75\u0026ndash;82\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMalta TM, de Souza CF, Sabedot TS et al (2018) Glioma CpG island methylator phenotype (G-CIMP): biological and clinical implications. Neuro Oncol 20(5):608\u0026ndash;620\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eNishiyama A, Nakanishi M (2021) Navigating the DNA methylation landscape of cancer. Trends Genet 37(11):1012\u0026ndash;1027\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDeaton AM, Bird A (2011) CpG islands and the regulation of transcription. Genes Dev 25(10):1010\u0026ndash;1022\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePachano T, S\u0026aacute;nchez-Gaya V, Ealo T et al (2021) Orphan CpG islands amplify poised enhancer regulatory activity and determine target gene responsiveness. Nat Genet 53(7):1036\u0026ndash;1049\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWu Y, Fletcher M, Gu Z et al (2020) Glioblastoma epigenome profiling identifies SOX10 as a master regulator of molecular tumour subtype. Nat Commun 11(1):6434\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLee SD, Song J, LeBlanc VG, Marra MA (2022) Integrative multi-omic analysis reveals neurodevelopmental gene dysregulation in CIC-knockout and IDH1-mutant cells. J Pathol 256(3):297\u0026ndash;309\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWon H, de la Torre-Ubieta L, Stein JL et al (2016) Chromosome conformation elucidates regulatory relationships in developing human brain. Nature 538(7626):523\u0026ndash;527\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eJohnson KC, Anderson KJ, Courtois ET et al (2021) Single-cell multimodal glioma analyses identify epigenetic regulators of cellular plasticity and environmental stress response. Nat Genet 53(10):1456\u0026ndash;1468\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRen Y, Huang Z, Zhou L et al (2023) Spatial transcriptomics reveals niche-specific enrichment and vulnerabilities of radial glial stem-like cells in malignant gliomas. Nat Commun 14(1):1028\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSudmant PH, Rausch T, Gardner EJ et al (2015) An integrated map of structural variation in 2,504 human genomes. Nature 526(7571):75\u0026ndash;81\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLizio M, Harshbarger J, Shimoji H et al (2015) Gateways to the FANTOM5 promoter level mammalian expression atlas. Genome Biol 16(1):22\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWilkerson MD, Hayes DN (2010) ConsensusClusterPlus: a class discovery tool with confidence assessments and item tracking. Bioinformatics 26(12):1572\u0026ndash;1573\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eTong Y, Sun J, Wong CF et al (2018) MICMIC: identification of DNA methylation of distal regulatory regions with causal effects on tumorigenesis. Genome Biol 19(1):73\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eZhou Y, Zhou B, Pache L et al (2019) Metascape provides a biologist-oriented resource for the analysis of systems-level datasets. Nat Commun 10(1):1523\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGrant CE, Bailey TL, Noble WS (2011) FIMO: scanning for occurrences of a given motif. Bioinformatics 27(7):1017\u0026ndash;1018\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLang M, Binder M, Richter J et al (2019) mlr3: A modern object-oriented machine learning framework in R. J Open Source Softw 4:1903\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKorsunsky I, Millard N, Fan J et al (2019) Fast, sensitive and accurate integration of single-cell data with Harmony. Nat Methods 16(12):1289\u0026ndash;1296\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGulati GS, Sikandar SS, Wesche DJ et al (2020) Single-cell transcriptional diversity is a hallmark of developmental potential. Science 367(6476):405\u0026ndash;411\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAibar S, Gonz\u0026aacute;lez-Blas CB, Moerman T et al (2017) SCENIC: single-cell regulatory network inference and clustering. Nat Methods 14(11):1083\u0026ndash;1086\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eFustero-Torre C, Jim\u0026eacute;nez-Santos MJ, Garc\u0026iacute;a-Mart\u0026iacute;n S et al (2021) Beyondcell: targeting cancer therapeutic heterogeneity in single-cell RNA-seq data. Genome Med 13(1):187\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWei R, He S, Bai S et al (2022) Spatial charting of single-cell transcriptomes in tissues. Nat Biotechnol 40(8):1190\u0026ndash;1199\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBrowaeys R, Saelens W, Saeys Y (2020) NicheNet: modeling intercellular communication by linking ligands to target genes. Nat Methods 17(2):159\u0026ndash;162\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCruz-Molina S, Respuela P, Tebartz C et al (2017) PRC2 Facilitates the Regulatory Topology Required for Poised Enhancer Function during Pluripotent Stem Cell Differentiation. Cell Stem Cell 20(5):689\u0026ndash;705e689\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePreissl S, Gaulton KJ, Ren B (2023) Characterizing cis-regulatory elements using single-cell epigenomics. Nat Rev Genet 24(1):21\u0026ndash;43\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eZhao J, Faryabi RB (2023) Spatial promoter-enhancer hubs in cancer: organization, regulation, and function. Trends Cancer\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eZhao Y, Fu X, Lopez JI et al (2021) Selection of metastasis competent subclones in the tumour interior. Nat Ecol Evol 5(7):1033\u0026ndash;1045\u003c/span\u003e\u003c/li\u003e\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":true,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true},"keywords":"oCGIs, Enhancer, Machine learning, Multi-omics, Cis-regulatory network","lastPublishedDoi":"10.21203/rs.3.rs-3959082/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-3959082/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003e\u003cb\u003eBackground:\u003c/b\u003e\u003c/p\u003e \u003cp\u003eThe complex transcriptional regulatory network leads to the poor prognosis of glioma. The role of orphan CpG islands (oCGIs) in the transcriptional regulatory network has been overlooked. Establishing a sophisticated transcriptional regulatory system is paramount.\u003c/p\u003e\u003cp\u003e\u003cb\u003eMethods:\u003c/b\u003e\u003c/p\u003e \u003cp\u003eWe constructed different cis-regulatory models through mutual information and Bayesian networks. We utilized eleven machine learning algorithms to develop classifiers that could effectively integrate multi-omics datasets. we utilized single-cell multi-omics data construct a higher-resolution cis-regulatory network. To investigate the binding interaction between oCGIs and transcription factors, we utilized chromatin immunoprecipitation assay and qRT-PCR. Furthermore, we assessed the proliferative status of various glioma subtypes using the MTT assay and immunohistochemistry.\u003c/p\u003e\u003cp\u003e\u003cb\u003eResults:\u003c/b\u003e\u003c/p\u003e \u003cp\u003eThe cis-regulatory network dominated by oCGIs and enhancers was significantly active in the glioma subtypes, mainly characterized by glioblastoma (Cluster 2). Direct regulation of target genes by oCGIs or enhancers is of great importance in the cis-regulatory network. Furthermore, based on single-cell multi-omics data, we found that the highly activated cis-regulatory network in Cluster 2 sustains the high proliferative potential of glioma cells. The upregulation of oCGIs and enhancers related genes in Cluster 2 results in glioma patients exhibiting resistance to radiotherapy and chemotherapy. These findings were further validated through glioma cell line related experiments.\u003c/p\u003e\u003cp\u003e\u003cb\u003eConclusion:\u003c/b\u003e\u003c/p\u003e \u003cp\u003eOur study systematically elucidated the cis-regulatory role of oCGIs for the first time. The comprehensive characterization of the multi-omics features of the oCGIs- and enhancers-dominated cis-regulatory network offers a novel insight into the pathogenesis of glioma and provides new strategies to treat this challenging disease.\u003c/p\u003e","manuscriptTitle":"Integration of Multi-omics Data Revealed the Orphan CpG Islands and Enhancer-dominated Cis-regulatory Network in Glioma","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2024-02-19 17:54:42","doi":"10.21203/rs.3.rs-3959082/v1","editorialEvents":[{"type":"communityComments","content":0}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"ace54faa-038d-4ce0-a2a1-ac22638e136e","owner":[],"postedDate":"February 19th, 2024","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[],"tags":[],"updatedAt":"2024-02-28T15:20:38+00:00","versionOfRecord":[],"versionCreatedAt":"2024-02-19 17:54:42","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-3959082","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-3959082","identity":"rs-3959082","version":["v1"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}