An EGFR co-amplified lncRNA HELDR promotes glioblastoma malignancy through KAT7-driven gene programs.

An EGFR co-amplified lncRNA HELDR promotes glioblastoma malignancy through KAT7-driven gene programs - PMC Skip to main content An official website of the United States government Here's how you know Here's how you know Official websites use .gov A .gov website belongs to an official government organization in the United States. Secure .gov websites use HTTPS A lock ( Lock Locked padlock icon ) or https:// means you've safely connected to the .gov website. Share sensitive information only on official, secure websites. Search Log in Dashboard Publications Account settings Log out Search… Search NCBI Primary site navigation Search Logged in as: Dashboard Publications Account settings Log in Search PMC Full-Text Archive Search in PMC Journal List User Guide PERMALINK Copy As a library, NLM provides access to scientific literature. Inclusion in an NLM database does not imply endorsement of, or agreement with, the contents by NLM or the National Institutes of Health. Learn more: PMC Disclaimer | PMC Copyright Notice Nat Cell Biol . Author manuscript; available in PMC: 2026 Apr 17. Published in final edited form as: Nat Cell Biol. 2026 Mar 27;28(5):972–985. doi: 10.1038/s41556-026-01924-w Search in PMC Search in PubMed View in NLM Catalog Add to search An EGFR co-amplified lncRNA HELDR promotes glioblastoma malignancy through KAT7-driven gene programs Xiaozhou Yu Xiaozhou Yu 1 The Ken & Ruth Davee Department of Neurology, The Lou and Jean Malnati Brain Tumor Institute, The Robert H. Lurie Comprehensive Cancer Center, Simpson Querrey Institute for Epigenetics, Northwestern University Feinberg School of Medicine, Chicago, IL, USA. Find articles by Xiaozhou Yu 1 , Xiao Song Xiao Song 1 The Ken & Ruth Davee Department of Neurology, The Lou and Jean Malnati Brain Tumor Institute, The Robert H. Lurie Comprehensive Cancer Center, Simpson Querrey Institute for Epigenetics, Northwestern University Feinberg School of Medicine, Chicago, IL, USA. Find articles by Xiao Song 1 , Richard A Schäfer Richard A Schäfer 2 Department of Urology, Northwestern University Feinberg School of Medicine, Department of Urology & Robert H. Lurie Comprehensive Cancer Center, Northwestern University Feinberg School of Medicine, Chicago, IL, USA. Find articles by Richard A Schäfer 2 , Qingshu Meng Qingshu Meng 2 Department of Urology, Northwestern University Feinberg School of Medicine, Department of Urology & Robert H. Lurie Comprehensive Cancer Center, Northwestern University Feinberg School of Medicine, Chicago, IL, USA. Find articles by Qingshu Meng 2 , Deanna Tiek Deanna Tiek 1 The Ken & Ruth Davee Department of Neurology, The Lou and Jean Malnati Brain Tumor Institute, The Robert H. Lurie Comprehensive Cancer Center, Simpson Querrey Institute for Epigenetics, Northwestern University Feinberg School of Medicine, Chicago, IL, USA. 3 Present address: Center for Immunotherapy & Precision Immuno-Oncology, Cleveland Clinic, Cleveland, OH, USA. Find articles by Deanna Tiek 1, 3 , Runxin Wu Runxin Wu 1 The Ken & Ruth Davee Department of Neurology, The Lou and Jean Malnati Brain Tumor Institute, The Robert H. Lurie Comprehensive Cancer Center, Simpson Querrey Institute for Epigenetics, Northwestern University Feinberg School of Medicine, Chicago, IL, USA. Find articles by Runxin Wu 1 , Qiu He Qiu He 1 The Ken & Ruth Davee Department of Neurology, The Lou and Jean Malnati Brain Tumor Institute, The Robert H. Lurie Comprehensive Cancer Center, Simpson Querrey Institute for Epigenetics, Northwestern University Feinberg School of Medicine, Chicago, IL, USA. Find articles by Qiu He 1 , Maya Walker Maya Walker 1 The Ken & Ruth Davee Department of Neurology, The Lou and Jean Malnati Brain Tumor Institute, The Robert H. Lurie Comprehensive Cancer Center, Simpson Querrey Institute for Epigenetics, Northwestern University Feinberg School of Medicine, Chicago, IL, USA. Find articles by Maya Walker 1 , Qi Cao Qi Cao 2 Department of Urology, Northwestern University Feinberg School of Medicine, Department of Urology & Robert H. Lurie Comprehensive Cancer Center, Northwestern University Feinberg School of Medicine, Chicago, IL, USA. Find articles by Qi Cao 2 , Rendong Yang Rendong Yang 2 Department of Urology, Northwestern University Feinberg School of Medicine, Department of Urology & Robert H. Lurie Comprehensive Cancer Center, Northwestern University Feinberg School of Medicine, Chicago, IL, USA. Find articles by Rendong Yang 2 , Bo Hu Bo Hu 1 The Ken & Ruth Davee Department of Neurology, The Lou and Jean Malnati Brain Tumor Institute, The Robert H. Lurie Comprehensive Cancer Center, Simpson Querrey Institute for Epigenetics, Northwestern University Feinberg School of Medicine, Chicago, IL, USA. Find articles by Bo Hu 1, ✉ , Shi-Yuan Cheng Shi-Yuan Cheng 1 The Ken & Ruth Davee Department of Neurology, The Lou and Jean Malnati Brain Tumor Institute, The Robert H. Lurie Comprehensive Cancer Center, Simpson Querrey Institute for Epigenetics, Northwestern University Feinberg School of Medicine, Chicago, IL, USA. Find articles by Shi-Yuan Cheng 1, ✉ Author information Article notes Copyright and License information 1 The Ken & Ruth Davee Department of Neurology, The Lou and Jean Malnati Brain Tumor Institute, The Robert H. Lurie Comprehensive Cancer Center, Simpson Querrey Institute for Epigenetics, Northwestern University Feinberg School of Medicine, Chicago, IL, USA. 2 Department of Urology, Northwestern University Feinberg School of Medicine, Department of Urology & Robert H. Lurie Comprehensive Cancer Center, Northwestern University Feinberg School of Medicine, Chicago, IL, USA. 3 Present address: Center for Immunotherapy & Precision Immuno-Oncology, Cleveland Clinic, Cleveland, OH, USA. ✉ Correspondence and requests for materials should be addressed to Bo Hu or Shi-Yuan Cheng. [email protected] ; [email protected] Author contributions Conceptualization, S.-Y.C., B.H. and X.Y.; data production, analysis and investigation, X.Y., X.S., R.A.S., Q.M., R.W., S.-Y.C. and B.H.; writing-original draft: X.Y., B.H. and S.-Y.C., review and editing: X.Y., S.-Y.C., X.S., R.A.S., D.T., R.W., Q.H., M.W., B.H., Q.C. and R.Y. Supervision, S.-Y.C. and B.H.; funding acquisition, S.-Y.C. and B.H. All authors have read and agreed to the published version of the paper. Issue date 2026 May. Reprints and permissions information is available at www.nature.com/reprints . PMC Copyright notice PMCID: PMC13086523  NIHMSID: NIHMS2161547  PMID: 41896311 The publisher's version of this article is available at Nat Cell Biol Previous version available: This article is based on a previously available preprint posted on Research Square on June 24, 2025: " An EGFR Co-Amplified and De Novo Long Noncoding RNA HELDR Promotes Glioblastoma Malignancy through KAT7-Driven Gene Programs ". Abstract EGFR amplification frequently occurs within extrachromosomal DNAs (ecDNAs) and is the most prevalent mutation in glioblastoma (GBM). However, targeting EGFR for GBM treatments has been unsuccessful. Here we show a long non-coding RNA (lncRNA) that is co-amplified with EGFR, which we name hidden EGFR long non-coding downstream RNA (HELDR). HELDR is a GBM-selective lncRNA that promotes tumorigenicity independent of EGFR signalling. HELDR exhibits widespread chromatin association and recruits the transcription co-activator p300 to the KAT7 promoter. p300-induced H3K27ac at the KAT7 promoter enlists other co-transcription factors, activating KAT7 transcription. KAT7 induces H3K14ac and H4K12ac that activate KAT7-driven gene programmes that are critical for GBM malignancy. Targeting KAT7 or HELDR markedly enhances therapeutic effects of anti-EGFR treatments for GBM. These results not only reveal the role of HELDR in EGFR-amplified GBM but also provide a strong rationale to characterize the role of lncRNAs co-amplified with driver oncogenes in human cancers. Glioblastoma (GBM) is a common and the most malignant primary brain tumour in adults with a poor prognosis 1 . In isocitrate dehydrogenase (IDH) wild-type (WT) GBM 1 , >55% tumours have EGFR amplification and/or mutations 2 . EGFR is an oncogenic driver of GBM that is frequently amplified within extrachromosomal DNAs (ecDNAs) 3 , 4 . ecDNAs are megabase-sized, double-stranded circular DNAs in cancer cell nuclei that drive oncogene amplification, dysregulated gene expression and intratumoral heterogeneity, thereby promoting cancer malignancy 5 . EGFR has been designated as a biomarker and therapeutic target for GBM treatments. However, EGFR-targeting therapies remain clinically ineffective for GBM 6 . Epigenetic reprogramming is a hallmark of cancer, including GBM 7 , 8 . Histone acetylation mediated by lysine acetyltransferases (KATs) regulates chromatin accessibility and transcription 9 . The transcriptional co-activator p300 (KAT3B) catalyses histone H3 lysine 27 acetylation (H3K27ac) at active regulatory regions and can be recruited to chromatin by long non-coding RNAs 10 - 12 . Another member of the KAT family, KAT7 (also known as HBO1 or MYST2), primarily acetylates histones H3 and H4 and has been implicated in cancer, although its role in GBM remains unclear 13 , 14 . LncRNAs are another important regulator of epigenetic regulation 15 . LncRNAs modulate cellular functions through RNA–RNA, RNA–DNA and RNA–protein interactions, with their emerging roles in controlling transcriptions 16 , 17 . Recent studies revealed that a large number of lncRNAs, either annotated or unannotated, are co-amplified with oncogene drivers at ecDNAs in human cancers, including GBM 18 , 19 . However, the roles of lncRNAs in EGFR -amplified tumourigenesis and GBM therapy resistance remains unanswered. In this study, we identified a de novo gene that is predominantly transcribed from the antisense strand of the EGFR long non-coding downstream RNA ( ELDR ) located near EGFR within ecDNAs in GBM 3 , 4 , 20 . This undiscovered gene is co-amplified with EGFR in GBM. We named it hidden ELDR ( HELDR ). HELDR binds to promoters, gene bodies and intergenic regions, regulating global gene expression without affecting EGFR signalling. Mechanistically, HELDR binds to and recruits p300 to the promoter of KAT7 , resulting in increased levels of H3K27ac, recruitment of transcription factors and enhanced transcription of KAT7. Last, the HELDR–KAT7 axis facilitates the expression of genes associated with GBM resistance to EGFR inhibitors (EGFRi) and targeting KAT7 or HELDR significantly enhances GBM responses to EGFRi treatments in vivo. Results LncRNA ELDR is co-amplified with EGFR in GBM To study the involvement of lncRNAs in EGFR -amplified GBM malignancy, we performed Spearman correlation analyses using RNA sequencing (RNA-seq) data from The Cancer Genome Atlas (TCGA) and our Northwestern University (NU) cohort 21 ( Supplementary Tables 1 and 2 ). We identified a total of 80 and 118 lncRNAs whose expression significantly correlate with EGFR expression in TCGA and NU datasets, respectively ( Fig. 1a and Extended Data Fig. 1a ). Among these, 14 are positively correlated and 8 lncRNAs are negatively correlated with EGFR expression on both datasets ( Fig. 1b ) where ELDR was the top hit ( Supplementary Table 3 ). We also validated the strong positive correlation between ELDR and EGFR RNA transcript levels in the Clinical Proteomic Tumor Analysis Consortium (CPTAC) cohort ( Extended Data Fig. 1b ). ELDR is an annotated lncRNA 22 transcribed from 7p11, located near EGFR within ecDNAs in GBM 3 , 4 . Whole-genome sequencing data from three typical TCGA patients with GBM revealed that ELDR is situated within the EGFR amplicon and/or contains several other co-amplified genes ( Fig. 1c ) 3 , 23 . ELDR is positively correlated with the EGFR in copy number ( Fig. 1d ), and among patients with EGFR -amplified GBM, more than 89% of cases also contain amplification of the ELDR region ( Extended Data Fig. 1c ). ELDR transcription peaks in three TCGA-GBM tumours are also found in patient-derived xenograft (PDX) GBM models that have EGFR amplification 24 ( Fig. 1e ). Notably, the majority of RNA-seq reads map outside the annotated exons of ELDR , implying that the presently annotated gene may not be a true representation of the transcribed gene in these tumours ( Fig. 1e ). Moreover, ELDR expression is significantly upregulated in GBM relative to normal brain and low-grade glioma ( Extended Data Fig. 1d ) and shows a negative correlation with overall survival in glioma ( Extended Data Fig. 1e ). Last, ELDR expression is markedly elevated in GBM compared with other cancer types in the TCGA datasets ( Extended Data Fig. 1f ). In cancers that include subsets of patients with EGFR amplification (such as head and neck squamous cell carcinoma and lung adenocarcinoma), ELDR expression also correlates with EGFR at the transcript level and is co-amplified at the DNA level ( Extended Data Fig. 1g - i ). These data indicate co-amplification of ELDR with EGFR , accompanied by concordant upregulation of ELDR -locus transcripts. Fig. 1 ∣. The lncRNA ELDR is co-amplified with EGFR in GBM. Open in a new tab a , Heatmap depicts expression of 80 lncRNAs that exhibited significant correlation with EGFR expression in TCGA-GBM samples (∣ r ∣ > 0.4 and P < 0.05). Expression levels of lncRNAs and EGFR and status of EGFR gene amplification and mutation are indicted. b , Venn diagrams for common lncRNAs that significantly correlate with EGFR expression in TCGA and NU GBM samples (∣ r ∣ > 0.4 and P < 0.05). c , Whole-genome sequencing (WGS) data of three representative TCGA-GBM samples for EGFR or EGFRvIII gene copy numbers. EGFR or EGFRvIII and ELDR gene loci are co-amplified. d , The correlation of CNV between EGFR and ELDR gene loci in TCGA ( n = 542) and CPTAC ( n = 99) GBM samples. e , RNA-seq data (IGV) of three representative TCGA samples (same cases in c ) and three Mayo Clinic PDX GBM samples. f , Correlation in RNA transcripts between EGFR and HELDR in TCGA ( n = 169) and CPTAC ( n = 94) GBM samples. Mut, mutation; Amp, amplification; CNV, copy number variation; TCGA, The Cancer Genome Atlas; IGV, Integrative Genomics Viewer. Two-sided Spearman’s rank correlation test ( a , b , d , f ) and one-tailed Fisher’s exact test for overlap significance ( b ). A de novo lncRNA, HELDR is the major transcript at the ELDR locus Lack of the transcribing strand information from where the RNA-seq reads originated results in inaccurate assessment of gene profiling, especially when the gene locus overlaps. We further analysed ELDR transcripts in an RNA-seq dataset 3 and two of our in-house glioma stem-like cell (GSC) RNA-seq datasets with strand direction information and found that RNA transcribed from the ELDR locus primarily comes from the plus strand, whereas the annotated ELDR is transcribed from the negative strand with much less abundance ( Fig. 2a and Extended Data Fig. 2a , b ). These data indicate that a previously unidentified transcript could be the major gene product at the ELDR locus but is transcribed from the antisense strand in relation to ELDR . Long-read RNA-seq on PDX GBM6 cells with EGFR amplification 24 revealed that a group of RNA transcripts with only one exon and slightly varying lengths was the main transcript of this locus ( Fig. 2b ). Analyses of three short-read RNA-seq datasets validated this observation ( Fig. 2a and Extended Data Fig. 2a , b ). Furthermore, rapid amplification of complementary DNA ends (RACE) analyses confirmed the length and the sequence of this unannotated gene ( Fig. 2c , d , Extended Data Fig. 2c - e and Supplementary Table 4 ). Fig. 2 ∣. De novo lncRNA HELDR is the major transcript in the ELDR locus and localized in the nucleus. Open in a new tab a , Short-read RNA-seq of GSC11 in the ELDR locus is shown in IGV. b , Long-read RNA-seq of PDX GBM6 in the ELDR locus is shown in IGV. c , Data of 5′ RACE (left) and 3′ RACE (right) for HELDR in GBM6 cells. Twelve clones of each were collected for Sanger sequencing. d , Maximum length obtained in the RACE and location of primers are shown. e , An online software CPAT ( https://wlcb.oit.uci.edu/cpat/ ) was utilized to assess the coding potential of the newly identified HELDR . Genes of β-actin ( ACTB ), DGCR5 and MALAT1 are controls. f , Left: quantitative PCR with reverse transcription (qRT–PCR) for gene expression in the nucleus. β-actin ( ACTB ) and U1 are controls ( n = 3). Right: immunoblot (IB) for cellular fraction markers. Glyceraldehyde-3-phosphate dehydrogenase (GAPDH) is a cytosolic protein control and H3 is a nuclear protein control. Data are representative of three independent experiments with similar results. g , Single-molecule RNA fluorescence in situ hybridization (smFISH) for HELDR localization in GSC17 cells (representative of 50 cells). HELDR knockout (KO) cells are used as negative control. Arrows indicate HELDR. Ctrl, control. Scale bars, 10 μm. Data were analysed by one-way analysis of variance (ANOVA) with Tukey’s post hoc test and are presented as mean ± s.d. ( f ). ORF Finder 25 and two orthologous algorithms 26 , 27 suggest that HELDR lncRNA lacks the capacity to encode a protein ( Fig. 2e and Extended Data Fig. 2e , f ). Additionally, this transcript is primarily localized in the cell nucleus ( Fig. 2f , g and Extended Data Fig. 3a ). Thus, we identified an undiscovered lncRNA that is transcribed from the plus strand at the ELDR gene locus. We named it hidden ELDR ( HELDR ). HELDR is partially conserved among several mammals ( Extended Data Fig. 2g ). The shared region between HELDR and ELDR in genome potentially overlaps with markers of enhancers, promoters and active regulatory elements ( Extended Data Fig. 2g ). We examined HELDR expression from the Ivy Foundation Glioblastoma Atlas Project (Ivy GAP) 28 to investigate the distribution of HELDR. HELDR was expressed at much lower levels in the leading edge region, correlating with the lower tumour density in this area, whereas its distribution across other regions showed high levels with minor differences ( Extended Data Fig. 2h ). We re-annotated HELDR and, as expected, also observed a significant correlation between HELDR and EGFR expression at the RNA level ( Fig. 1f , Extended Data Fig. 2i and Supplementary Table 3 ). HELDR is critical for the tumorigenicity of GBM with EGFR-HELDR co-amplification Patient-derived GSC and PDX GBM models were used as clinically relevant experimental systems that preserve EGFR amplification or high expression, recapitulate GBM heterogeneity, and are widely used in neuro-oncology research 29 , 30 . We determined the expression levels of HELDR in GSCs and PDX GBM lines with known EGFR amplification 24 , 29 , 31 . The median transcript number of HELDR per cell, measured in several GSCs with higher HELDR expression, ranged from ~22 to 31, corresponding to transcripts per million values of ~19 to 32 ( Fig. 2g and Extended Data Fig. 3a - c ). The expression levels of HELDR are positively correlated with that of EGFR in these GBM models ( Extended Data Fig. 3d ). Moreover, HELDR expression declined moderately during GSCs differentiation into differentiated glioma cells (DGCs) 21 ( Extended Data Fig. 3e ). We next knocked down endogenous HELDR in GSCs with EGFR amplification 29 using shRNAs ( Fig. 3a , b and Extended Data Fig. 3f ). Knockdown (KD) of HELDR markedly impaired proliferation, self-renewal capacity and the expression of several glioma stem-cell markers 32 , 33 in GSCs in vitro ( Fig. 3c - f and Extended Data Fig. 3g - i ). In addition, we also used CRISPR/Cas9-mediated knockout (KO) to deplete HELDR in GSC17 cells ( Extended Data Fig. 3j - m ). Consistently, HELDR KO also significantly impaired GSC17 tumourigenic properties in vitro ( Extended Data Fig. 3n , o ). In vivo, both HELDR KD and depletion markedly suppressed the growth of orthotopic tumour xenografts and improved survival ( Fig. 3g , h and Extended Data Fig. 3p , q ). Moreover, the decreased HELDR levels suppressed tumorigenicity could be rescued by overexpression of an shRNA-resistant form of HELDR 34 ( Fig. 3i - m and Extended Data Fig. 3r - w ). Taken together, these data indicate that HELDR supports EGFR -amplified GBM tumorigenicity. Fig. 3 ∣. HELDR is critical for cell proliferation and tumorigenicity and globally regulates gene expression independent of EGFR in GSCs with EGFR amplification. Open in a new tab a , b , qRT–PCR for shRNA KD of HELDR in GSC17 ( n = 3) ( a ) and GSC11 ( n = 3) ( b ) cells where EGFR is amplified. c , d , Cell proliferation for GSC17 ( n = 3) ( c ) and GSC11 ( n = 3) ( d ) cells with or without HELDR KD. e , f , Glioma sphere formation (self-renewal capacity) of GSC17 ( e ) and GSC11 ( f ) cells with or without HELDR KD. g , Representative bioluminescence imaging (BLI) images of indicated GSC17 tumour xenograft-bearing mice. h , Kaplan–Meier analysis of GSC17 tumour xenograft-bearing mice ( n = 5). i , qRT–PCR for shRNA KD of HELDR and overexpression of shRNA-resistant HELDR in GSC11 cells ( n = 3). j , Cell proliferation for GSC11 cells with HELDR KD and overexpression of shRNA-resistant HELDR ( n = 3). k , Glioma sphere formation of GSC11 cells with HELDR KD and re-expression of a shRNA-resistant HELDR. l , Representative BLI images of indicated GSC11 tumour xenograft-bearing mice. m , Kaplan–Meier analysis of GSC11 tumour xenograft-bearing mice ( n = 5). n , Volcano plot of coding DEGs and HELDR in GSC17 cells after HELDR KD versus control. Two independent shRNAs (two biological replicates per shRNA) were analysed, with each shRNA group compared independently with its matched control. Genes that were significant in both shRNA comparisons (fold change ∣FC∣ > 1.5 and adjusted P < 0.05) are highlighted. Point positions reflect the shRNA-1 comparison. o , KEGG enrichment of significantly downregulated coding DEGs shared by both shRNA comparisons in GSC17 (HELDR KD versus control). FDR, false discovery rate. p , IB for EGFR and EGFR downstream signalling proteins in GSC17 cells with or without HELDR KD. β-actin is a sample processing control. All blots, qRT–PCR and in vitro assays are representative of three independent experiments. Data were analysed by one-way ANOVA with Dunnett’s post hoc test ( a – d ), two-sided likelihood ratio test ( e , f , k ), log-rank test ( h , m ), one-way ANOVA with Tukey’s post hoc test ( i , j ), two-sided Wald test with Benjamini–Hochberg adjustment for multiple comparisons ( n ) and one-sided hypergeometric test with Benjamini–Hochberg adjustment for multiple comparisons ( o ) and are presented as mean ± s.d. ( a – d , i , j ). HELDR regulates global gene expression but does not affect EGFR expression or its downstream signalling To investigate how HELDR regulates EGFR -amplified GBM tumourigenesis, we performed RNA-seq. Across two independent KD groups, we identified more than 900 downregulated and over 400 upregulated protein-coding differentially expressed genes (DEGs) overlapping between groups ( Fig. 3n and Supplementary Table 5 ). As HELDR KD primarily represses gene expression, we focused on the downregulated protein-coding DEGs. Kyoto Encyclopedia of Genes and Genomes (KEGG) analysis revealed enriched pathways, including focal adhesion, axon guidance, MAPK and the RAS pathway ( Fig. 3o ) that are known to promote GBM tumourigenesis and resistance to EGFRi therapy 35 - 38 . LncRNAs can function in cis or trans to regulate the transcription of genes 17 . Thus, we examined the expression of coding genes located near the ELDR/HELDR locus ( Extended Data Fig. 4a ). Most genes, including EGFR , showed no significant changes in expression levels with or without HELDR KD ( Extended Data Fig. 4b ). Using the TransCistor algorithm 39 , HELDR was not predicted to be cis -acting (one-sided Fisher’s exact test, activator P > 0.999, repressor P = 0.0670). We also confirmed that EGFR signalling remains unaltered after HELDR KD ( Fig. 3p and Extended Data Fig. 4c ). Reciprocally, perturbation of EGFR signalling exerted no significant effect on HELDR expression ( Extended Data Fig. 4d , e ). Thus, HELDR that is co-amplified with EGFR globally regulates gene expression but does not influence EGFR expression or its downstream signalling. HELDR exhibits widespread genomic DNA binding One common mechanism by which lncRNAs globally regulate gene expression is to bind genomic DNA and fine-tune gene transcription by regulating the formation of transcription complexes, epigenetic modifications or three-dimensional (3D) genome architecture 16 , 17 . Thus, we employed Chromatin Isolation by RNA Purification (ChIRP) 40 using two independent sets of probes to identify the genomic DNA binding sites of endogenous HELDR ( Fig. 4a ). HELDR shows specific enrichment in both the even and odd probe groups, but not in the negative control groups ( Fig. 4b ). In addition, specific ChIRP-seq peaks were observed in both groups at the DNA locus from which HELDR is transcribed ( Fig. 4c ). These data further confirm the presence of HELDR at its gene locus and underscores the specificity of the ChIRP assay. Fig. 4 ∣. HELDR exhibits widespread genomic DNA binding. Open in a new tab a , Schematic diagram of ChIRP analysis. b , qRT–PCR for RNA levels of HELDR and the negative control U1 retrieved from ChIRP (GSC17). Even and odd, probe pools with even-numbered or odd-numbered probes targeting the HELDR locus sequentially ( n = 3). LacZ, probe pools targeting Escherichia coli -expressed lacZ as a negative control. c , ChIRP-seq shown in IGV in the HELDR locus (GSC17). d , Common peaks between even and odd groups (BED file) in HELDR ChIRP-seq analysis across the whole genome. e , Pie chart shows the distribution of HELDR peaks across indicated genomic features. UTR, untranslated region. f , Venn diagram of the subset of coding DEGs (RNA-seq; shared by both shRNAs) and coding genes with common ChIRP peak in the promoter region (odd and even pools). g , IGV views of KAT7 in HELDR ChIRP-seq and RNA-seq. The peaks of HELDR ChIRP-seq in the promoter region are highlighted. Data were analysed by one-way ANOVA with Tukey’s post hoc test ( b ) and a one-tailed Fisher’s exact test for overlap significance ( f ) and are presented as mean ± s.d. ( b ). Panel a created in BioRender; Yu, X. https://biorender.com/ga6ocvx (2026). A total of 7,095 common peaks between the even and odd groups (one-sided permutation test, P = 0.0009) were identified in the ChIRP-seq experiment ( Supplementary Table 6 ). As a control, we also performed ChIRP-seq targeting two well-characterized nuclear-localized lncRNAs, NEAT1 and MALAT1, both of which are highly expressed and known to bind chromatin 41 , 42 . The enrichment of retrieved DNA at their respective genomic loci confirmed the assay specificity 41 , 42 ( Extended Data Fig. 5a , b ). Peaks identified in ChIRP-seq targeting HELDR showed minimal overlap with those targeting NEAT1 or MALAT1 (Extended Data Fig. 5c , d ). Conversely, as negative controls, only a few peaks were detected in ChIRP-seq experiments performed under RNase-treated conditions, in HELDR-negative GSCs or when using probes complementary to the antisense strand of HELDR ( Extended Data Fig. 5e ). In addition, the majority of HELDR peaks were significantly downregulated following HELDR knockdown ( Extended Data Fig. 5f ). These results further demonstrate the specificity of ChIRP-seq targeting HELDR. The peaks of ChIRP-seq targeting HELDR are distributed across the genome ( Fig. 4d ). We observed that approximately a quarter of the peaks are localized in promoter regions ( Fig. 4e ), which are crucial for transcription regulation. Therefore, we focused on the overlapping protein-coding genes between the DEGs identified in the RNA-seq analysis and those with common peaks in their promoters in the ChIRP-seq data. Ultimately, we identified 58 candidates ( Fig. 4f and Supplementary Table 7 ). There was a trend toward enrichment of HELDR ChIRP promoter occupancy among downregulated compared with upregulated coding DEGs (44 of 958, 4.6% versus 14 of 485, 2.9%; one-tailed Fisher’s exact test, P = 0.0758); together with the greater number of downregulated DEGs, this supports a predominant role for HELDR in transcriptional activation. HELDR recruits p300 to the KAT7 promoter and enhances KAT7 transcription through an epigenetic mechanism Among 58 candidates, we present the localization patterns of several representative genes ( Fig. 4g and Extended Data Fig. 5g ). The enrichment of HELDR at the promoters of these genes was validated ( Extended Data Fig. 5h ). KAT7 has been shown to play an important role in human cancers 43 - 45 , but its role in GBM is largely unknown. Our analysis indicated that HELDR binds to the promoter of KAT7 , and HELDR KD significantly decreased KAT7 expression ( Figs. 4g and 5a and Extended Data Fig. 6a ). In contrast, inhibition of EGFR signalling did not affect KAT7 expression ( Extended Data Fig. 6b ), and HELDR, EGFR and KAT7 were concurrently expressed in EGFR -amplified GSCs and GBM patient samples ( Extended Data Fig. 6c - h ). Additionally, positive correlations were observed between HELDR and KAT7, as well as between EGFR and KAT7 ( Fig. 5b , Extended Data Fig. 6i , j and Supplementary Table 8 ). These data suggest that HELDR plays a role in regulating the expression of KAT7. Fig. 5 ∣. HELDR epigenetically regulates KAT7 expression. Open in a new tab a , IB for KAT7 proteins in GSC17 cells with or without HELDR KD. b , Correlation analysis of RNA-seq data for HELDR RNA in relation to KAT7 in GBM of the NU dataset ( n = 47). c , TF.TARGET.REGNETWORK–based enrichment of the common downregulated coding DEGs identified in both shRNA comparisons following HELDR KD in GSC17 ( http://bioinformatics.sdstate.edu/go/ ). d , RNA pulldown. In vitro-transcribed HELDR binds to p300 in GSC17 cell extract. Antisense (AS) transcript, β-actin and beads are controls. e , RNA immunoprecipitation (RIP). Endogenous HELDR binds to p300 proteins in GSC17 cells. IgG and β-actin ( ACTB ) transcript are controls ( n = 3). f , Images of smFISH for HELDR and immunofluorescence (IF) staining for p300 in GSC17 cells (representative of 30 cells). Arrows show HELDR. g , h , Chromatin immunoprecipitation (ChIP)–qPCR. Enrichment of p300 ( n = 3) or H3K27ac ( n = 3) at KAT7 gene promoters were compared in GSC17 cells with or without HELDR KD. i , Sequence logo plot of GABPA and enrichment in HELDR ChIRP-seq. j , Predicted GABPA binding site in the promoter of KAT7 using online software: Jaspar ( https://jaspar.elixir.no/ ). k , ChIP–qPCR. Enrichment of GABPA at KAT7 gene promoters were compared in GSC17 cells with or without HELDR KD ( n = 3). l , m , Sequential chromatin immunoprecipitation (ChIP–re-ChIP–qPCR). Enrichment of P300/GABPA ( n = 3) ( l ) or H3K27ac/GABPA ( n = 3) ( m ) complexes at KAT7 gene promoters were compared in GSC17 cells with or without HELDR KD. The first ChIP was performed with p300, H3K27ac or IgG, followed by a second ChIP using GABPA. In g , h and k – m , IgG is used as a control. Scale bar, 10 μm. β-actin is a sample processing control in a and d . All blots, qRT–PCR and qPCR are representative of three independent experiments. Data were analysed by two-sided Spearman’s rank correlation test ( b ), one-sided hypergeometric test with Benjamini–Hochberg adjustment for multiple comparisons ( c ), two-sided unpaired t -test ( e , g , h , k – m ) and one-sided cumulative binomial test ( i ) and are presented as means.d. ( e , g , h , k – m ). We hypothesized that the binding of HELDR at gene promoters could either facilitate or impede the binding of transcription-related proteins, thereby regulating transcription. Thus, we conducted gene enrichment analysis and identified potential transcription factors or epigenetic regulators associated with HELDR ( Fig. 5c ). We focused on p300 as p300 is an epigenetic regulator, and two other lncRNAs bind to and recruit p300 to their targeting gene enhancers 12 , 46 , whereas the remaining candidates are general transcription factors that lack well-defined RNA-binding domains 47 . We found that HELDR binds to p300 ( Fig. 5d , e ) and colocalizes with p300 in the nucleus ( Fig. 5f and Extended Data Fig. 6k , l ). p300 induces H3K27ac at targeted gene promoters, enhancers and super enhancers, and recruits transcription factors to activate transcription 12 , 46 , 48 . As anticipated, HELDR plays a role in the enrichment of p300 and H3K27ac at the KAT7 promoter ( Fig. 5g , h and Extended Data Fig. 6m , n ). Further, motif analysis of the ChIRP-seq data revealed that HELDR tends to bind to motifs of several members of the erythroblast transformation-specific (ETS) transcription factor family ( Fig. 5i and Extended Data Fig. 6o ), consistent with the enrichment of ETS-targeted genes in HELDR-regulated DEGs ( Fig. 5c ). We selected GA-binding protein, alpha subunit (GABPA) as a representative target due to its established role in GBM 49 , 50 . HELDR KD markedly reduced GABPA binding at the KAT7 promoter but had no effect on GABPA’s protein expression ( Fig. 5j , k and Extended Data Fig. 7a - c ). Furthermore, HELDR KD also significantly hindered the enrichment of transcription complexes (p300/GABPA or H3K27ac/GABPA) at the KAT7 promoter ( Fig. 5l , m and Extended Data Fig. 7d - g ). KAT7 mediates HELDR-regulated GBM tumourigenic properties To demonstrate that p300 is essential for HELDR’s function and the underlying mechanism, we treated GSCs with a selective p300 inhibitor, A-485 (ref. 51 ). Like the inhibitory effects of HELDR KD, A-485 treatment decreased KAT7 protein expression and attenuated cell proliferation and self-renewal capacity of GSCs ( Fig. 6a - c and Extended Data Fig. 7h - l ). These results suggest that HELDR facilitates the formation of the transcription complex and activates p300-mediated KAT7 transcription. Fig. 6 ∣. KAT7 mediates HELDR-promoted GBM tumourigenesis. Open in a new tab a , IB for KAT7 protein expression in GSC17 cells with indicated modification and treatment (5 μM A-485) for 3 days. b , c , cell proliferation ( n = 3) ( b ) and glioma sphere formation ( c ) of GSC17 cells with indicated modification and treatment. GSC17 cells were treated with 5 μM A-485. d , Volcano plot displaying coding DEGs of RNA-seq data in GSC17 KAT7 KD or control cells with two biological replicates. Coding DEGs (fold change ∣FC∣ > 1.5 and adjusted P < 0.05) are highlighted. e , Pie graph. HELDR KD downregulated coding DEGs that are not regulated by KAT7 KD (KAT7-specific, 62%) or also regulated by KAT7 KD (38%) in GSC17 cells. f , KEGG analysis of significantly downregulated coding DEGs in GSC17 KAT7 KD cells. g – i , Re-expression of KAT7 rescued HELDR KD-suppressed H3K14ac and H412Kac (IB in g ), cell proliferation ( n = 3) ( h ) and glioma sphere formation ( i ) in GSC17 cells. j , Common genes that regulated by both KAT7 KD and HELDR KD in GSC17 cells with two biological replicates. k – m , ChIP–qPCR. Enrichment of KAT7 ( n = 3) ( k ), H3K14ac ( n = 3) ( l ) and H4K12ac ( n = 3) ( m ) at the promoters of indicated genes in GSC17 cells, with or without HELDR KD or rescue by KAT7 overexpression. β-actin is a sample processing control in a and g . All blots, qPCR and in vitro assays are representative of three independent experiments. Data were analysed by one-way ANOVA with Tukey’s post hoc test ( b , h , k–m ), two-sided likelihood ratio test ( c , i ), two-sided Wald test with Benjamini–Hochberg adjustment for multiple comparisons ( d , j ) and a one-sided hypergeometric test with Benjamini–Hochberg adjustment for multiple comparisons ( f ) and are presented as mean ± s.d. ( b , h , k – m ). Next, we assessed the effects of HELDR binding to the KAT7 promoter on GBM tumourigenesis. KAT7 induces H3 and H4 acetylation, primarily H3K14 and H4K12 (ref. 52 ). Indeed, KAT7 KD markedly reduced H3K14ac and H4K12ac ( Extended Data Fig. 8a ), and decreased proliferation and self-renewal capacity ( Extended Data Fig. 8b , c ). These data are consistent with other studies showing that KAT7-induced histone acetylation activates gene transcription and that KAT7 KD decreases proliferation 53 , 54 . KAT7 KD resulted in significantly more downregulated coding DEGs ( Fig. 6d and Supplementary Table 9 ). Of note, among the 958 of downregulated coding DEGs induced by HELDR KD, 364 genes (38%) were also regulated by KAT7 ( Fig. 6e and Supplementary Table 10 ). Pathway analysis revealed that the top enriched pathway caused by HELDR KD shown in Fig. 3o is also enriched in KAT7 KD GSCs ( Fig. 6f ). Furthermore, re-expression of KAT7 in HELDR KD GSCs restored the reduced levels of H3K14ac and H4K12ac, as well as rescued the impaired proliferation and self-renewal capacity caused by HELDR KD ( Fig. 6g - i and Extended Data Fig. 8d - f ). KAT7 activates gene transcription by inducing histone acetylation at gene promoters 53 , 55 . Thus, we investigated whether HELDR and KAT7 regulate histone acetylation at the promoters of key downstream genes in the HELDR–KAT7 axis. We focused on several representative genes with important functions in GBM such as CD44 , cyclin D1 ( CCND1 ), AXL receptor tyrosine kinase ( AXL ), integrin subunit alpha 2 ( ITGA2 ), nerve growth factor receptor ( NGFR ) and caveolin 1 ( CAV1 ) 56 - 60 . The transcription of these genes is downregulated following KD of HELDR or KAT7 ( Fig. 6j , Extended Data Fig. 8g and Supplementary Tables 5 and 9 ). As expected, KD of HELDR or KAT7 markedly decreased, whereas KAT7 overexpression rescued the enrichment of KAT7, H3K14ac and H4K12ac at the promoters of these genes ( Fig. 6k - m and Extended Data Fig. 8h - k ), Together, these results indicate that the KAT7–H3K14ac/H4K12ac axis is a critical downstream mediator for HELDR-regulated GBM tumourigenic properties. Targeting KAT7 enhances the effects of anti-EGFR therapy in preclinical GBM models Next, we tested the effects of targeting KAT7 using a potent inhibitor, WM-3835 (ref. 14 ), alone or in combination with anti-EGFR therapy, in GBM. As expected, EGFR -amplified GSCs/GBM (GBM6, with EGFRvIII (ref. 24 ), GSC34 (ref. 61 ), GSC11 (ref. 29 ) and GSC17 (ref. 29 )) is sensitivity to the EGFR inhibitor erlotinib or KAT7 inhibitor WM-3835, whereas several GSCs (GSC157 (ref. 62 ) and GSC1478 (ref. 31 )) without EGFR amplification are insensitive ( Extended Data Fig. 9a - c ). In vitro, the KAT7 inhibitor WM-3835 exhibited synergistic effects when combined with the EGFR inhibitor erlotinib in suppressing cell proliferation and self-renewal capacity of multiple GSCs or patient-derived GBM6 cells that EGFR is amplified ( Fig. 7a , b and Extended Data Fig. 9b ). Then, we assessed the effect of KAT7 inhibition by WM-3835 on HELDR/KAT7-regulated genes critical for GBM tumourigenesis and resistance to anti-EGFR therapy 37 , 38 , 57 , 63 ( Fig. 6j and Extended Data Fig. 8g ). As expected, WM-3835 treatment decreased protein abundance of CD44, cyclin D1 ( CCND1 ), AXL receptor tyrosine kinase ( AXL ), integrin subunit alpha 2 ( ITGA2 ), nerve growth factor receptor ( NGFR ) and caveolin 1 ( CAV1 ) ( Fig. 7c and Extended Data Fig. 9d ). Fig. 7 ∣. KAT7 inhibitor WM-3835 synergistically enhanced anti-GBM activity of EGFR inhibitor erlotinib. Open in a new tab a , ZIP synergy scores calculated by SynergyFinder ( https://synergyfinder.fimm.fi/ ) for EGFR -amplified GSCs (GSC17, GSC34 and GSC11) and GBM6, which were treated with WM-3835 and erlotinib for 6 days with indicated concentrations. Synergy scores were assessed based on proliferation assays. b , Glioma sphere formation of indicated GSCs that were treated separately or in combination with indicated inhibitors. GSC17: 10 μM WM-3835, 1 μM erlotinib; GSC34: 10 μM WM-3835, 0.6 μM erlotinib; GSC11: 10 μM WM-3835, 0.8 μM erlotinib; and GBM6: 10 μM WM-3835, 0.4 μM erlotinib. c , IB for indicated protein expression in GSC11 or GBM6 cells treated with vehicle or WM-3835 for 72 h. β-actin is a sample processing control. d , The workflow of the animal experiment. I.P., intraperitoneal; P.O., orally. e , BLI images of GSC11 xenograft. f , BLI images of GBM6 xenograft. g , h , Kaplan–Meier analysis for mice bearing indicated GSC ( g ) or GBM ( h ) brain tumour xenografts with indicated treatments (GSC11, n = 5; GBM6, n = 5). i , j , Quantification of IF staining ( Extended Data Fig. 9 ) of brain sections with GBM6 tumour xenografts with indicated treatments (Ki-67, n = 5 ( i ); CD44, n = 5 ( j )). All blots and in vitro assays are representative of three independent experiments. Data were analysed by a two-sided likelihood ratio test ( b ), log-rank test ( g , h ) and one-way ANOVA with Tukey’s post hoc test ( i , j ) and are presented as mean ± s.d. ( i , j ). Panel d created in BioRender; Yu, X. https://biorender.com/ln9rkjo (2026). Last, we evaluated the therapeutic effects on orthotopic tumour xenograft models ( Fig. 7d ). Both WM-3835 and erlotinib monotherapy showed growth suppressive effect on intracranial tumours. Of note, combination treatments showed significantly increased antitumour activity, relative to monotherapy ( Fig. 7e - h ). Intracranial tumours were examined for treatment effect on proliferation, apoptosis, and molecular targets of the KAT7 inhibitor (CD44, H3K14ac and H4K12ac), and the EGFR inhibitor (p-EGFR). We found that tumour xenografts treated with combined WM-3835+erlotinib had significantly lower proliferation indices (by Ki-67 staining) and higher apoptosis levels (by cleaved caspase 3) when compared with WM-3835 or erlotinib treatment ( Fig. 7i , j and Extended Data Fig. 9e , f ). These data show that inhibition of KAT7 enhanced cytotoxicity of EGFRi against tumour xenografts in vivo. Antisense oligonucleotide targeting of HELDR enhances the effects of anti-EGFR therapy for GBM Antisense oligonucleotide (ASO) treatment is considered a promising therapeutic approach for treating multiple neurological diseases 64 . Thus, we explored using ASOs targeting HELDR to treat GBM. ASO treatment significantly inhibited HELDR expression, the KAT7–H3K14ac/H4K12ac axis, and KAT7-regulated signalling pathways and gene expression as described above ( Fig. 8a , b and Extended Data Fig. 10a - c ). Of note, only EGFR -amplified GSC or GBM cells responded to ASO treatment, whereas GSCs without EGFR amplification were unresponsive ( Extended Data Fig. 10d ). ASO treatment enhanced erlotinib suppression of GBM cell proliferation and self-renewal capacity in vitro ( Fig. 8c , d and Extended Data Fig. 10e - j ). Analysis of DEGs following ASO treatment confirmed that HELDR was downregulated ( Supplementary Table 11 ). Comparison of coding DEGs revealed that most of the downregulated genes shared between the two independent HELDR shRNAs also overlapped with those identified in the ASO-treated group, including KAT7 and its downstream target genes ( Extended Data Fig. 10k and Supplementary Tables 11 and 12 ). Last, we evaluated the therapeutic effects of individual or combination therapy by ASO and erlotinib on orthotopic tumour xenograft models ( Fig. 8e ). Both ASO and erlotinib monotherapy suppress tumour growth with combination treatments showing significantly increased antitumour activity ( Fig. 8f - h and Extended Data Fig. 10l ). Treatment with ASOs effectively suppressed HELDR expression in vivo, and the ASO backbone remained detectable in tissues ( Fig. 8i and Extended Data Fig. 10m , n , q , t ), indicating on-target activity by ASO administration. In xenografted tumours, ASO + erlotinib treatment resulted in significantly lower proliferation indices (by Ki-67 staining) and higher apoptosis level (by cleaved caspase 3) comparing with monotherapy ( Extended Data Fig. 10n , o , r ). In addition, both ASO and erlotinib suppressed their known targets as well as targets identified in this study, whereas EGFR expression was unaffected ( Extended Data Fig. 10n - t ). Taken together, these data indicate that ASO-targeting HELDR markedly enhanced the anti-GBM activity of erlotinib in orthotopic tumour xenografts. Fig. 8 ∣. ASO-targeting HELDR synergistically enhanced anti-GBM activity of EGFR inhibitor erlotinib. Open in a new tab a , qRT–PCR analysis of GBM6 cells 24 h after transfection with 100 nM control or HELDR-targeting ASO in vitro ( n = 3). b , IB for indicated proteins in GBM6 cells 72 h after transfection with 100 nM control or HELDR-targeting ASO in vitro. β-actin is a sample processing control. c , d , Cell proliferation ( n = 3) ( c ) and glioma sphere formation ( d ) of GBM6 cells with indicated treatment. 100 nM ASO; 0.4 μM erlotinib. e , The workflow of the animal experiment. f , Kaplan–Meier analysis of GBM6 xenografts ( n = 5). g , BLI images for mice bearing GBM6 brain tumour xenografts with indicated treatments. h , Kaplan–Meier analysis for mice bearing GSC11 brain tumour xenografts with indicated treatments ( Extended Data Fig. 10 ) ( n = 5). i , qRT–PCR analysis of HELDR expression in tumour area from brain sections ( n = 4). j , Graphical representation of the mechanism by which HELDR regulates EGFR treatment resistance and the translation strategy. All blots, qRT–PCR and in vitro assays are representative of three independent experiments. Data were analysed by a one-way ANOVA with Dunnett’s post hoc test ( a ), one-way ANOVA with Tukey’s post hoc test ( c , i ), two-sided likelihood ratio test ( d ) and log-rank test ( f , h ) and are presented as mean ± s.d. ( a , c , i ). Panels created in BioRender: e , Yu, X. https://biorender.com/ln9rkjo (2026); j , Yu, X. https://biorender.com/ph2lb90 (2026). Discussion In this study, we identified and characterized a previously unannotated lncRNA, HELDR, which is co-amplified with EGFR in GBM and promotes tumourigenesis independently of EGFR expression and signalling ( Fig. 8j ). HELDR broadly associates with chromatin and activates oncogenic transcriptional programmes, in part through upregulation of the lysine acetyltransferase KAT7. Functionally, the HELDR–KAT7 axis contributes to GBM malignancy and modulates responses to EGFR inhibitor. HELDR was an unanticipated finding from our search for EGFR-correlated, upregulated lncRNAs in GBM. We found that the previously characterized ELDR is the top lncRNA with the highest correlation with EGFR. Significantly, ELDR co-amplifies with EGFR in the same ecDNAs in GBM tumours 3 , 4 , 20 . Moreover, our in-depth RNA analyses revealed an undiscovered lncRNA, HELDR , in the ELDR locus that is transcribed from its antisense strand with high efficiency. Functionally, HELDR promotes EGFR -amplified GBM tumourigenesis by inducing genes in oncogenic pathways critical for GBM malignancy independent of EGFR expression and signalling. These are the second (ELDR) and the third (HELDR) lncRNAs that co-amplify with the driver oncogene EGFR at ecDNAs in GBM. We recently reported that the lncRNA LINC02283 co-amplifies with PDGFRA and enhances PDGFRA-mediated signalling and promotes GBM tumourigenesis 65 . ecDNA amplifications that harbour driver oncogenes such as EGFR , MYC , MDM2 and CDK4 were detected in 17.1% of 39 human tumour subtypes, including GBM 18 . ecDNAs are also enriched with regulatory elements including promoters, enhancers and lncRNAs 18 , 19 . Thus, our discovery of the unannotated HELDR that co-amplifies with EGFR in GBM suggests that additional lncRNAs could be found in tumour ecDNAs and in the antisense strand of known gene loci. As amplified lncRNAs can promote driver oncogene-promoted tumourigenesis dependent 65 or independent (this study) of the driver oncogenic signalling, investigation of these co-amplified annotated and unannotated lncRNAs 18 , 19 such as LINC02283 and HELDR in tumour ecDNAs would further advance our understanding of their roles in driver oncogene-promoted tumourigenesis and therapy resistance of human cancers. LncRNAs can exhibit their cis - or trans -regulatory functions through RNA–DNA interactions that activate or silence gene transcription 17 . In this study, we show that HELDR, a GBM-selective nuclear lncRNA, interacts with genomic DNA at multiple regulatory regions. HELDR engages promoter-associated chromatin to promote KAT7 expression relevant to GBM tumourigenesis and influence responses to anti-EGFR therapies. Moreover, our data reveal that HELDR is globally associated with the promoters of more than 600 coding genes, including 58 genes with altered expression, which may be critical for GBM tumourigenesis. Thus, the effects of HELDR-regulated transcription through binding to their gene promoters in GBM tumourigenesis and therapy responses warrant further investigation. EGFR is a driver oncogene 2 and a therapeutic target for treating GBM 6 . Thus far, anti-EGFR treatments have been disappointing 37 . In this study, we reveal an uncharacterized HELDR–KAT7 axis that promotes EGFR -amplified GBM tumourigenesis and therapy resistance independent of EGFR expression and signalling. Consistently, targeting HELDR-activated KAT7 or targeting HELDR by ASOs augments anti-GBM activity by erlotinib, supporting a functional role for the HELDR–KAT7 axis in EGFR-independent GBM malignancy. In summary, our study identifies HELDR as a previously undiscovered lncRNA co-amplified with EGFR that drives GBM malignancy independently of EGFR signalling. By uncovering a HELDR–KAT7 regulatory axis, this work highlights a functional contribution of ecDNA-associated lncRNAs to oncogene-driven tumour biology. These findings suggest that targeting lncRNA-driven epigenetic programmes may offer therapeutic opportunities to overcome resistance to EGFR-targeted therapies in GBM. Methods This research complies with all relevant ethical guidelines. All patient studies were conducted in accordance with Declaration of Helsinki. Research involving human tissue was performed under protocols approved by the NU Institutional Review Board under protocol number STU00217476. All participants provided written informed consent for the use of their tissues for research purposes. No specific participant compensation was provided for the donation of these tissues. Detailed patient information is provided in the ‘ Reporting summary ’ section. All animal experiments were conducted under the Institutional Animal Care and Use Committee approved protocols (IS00027857) at NU, in accordance with National Institutes of Health (NIH) and institutional guidelines. All animals were euthanized upon reaching predefined humane end points consistent with Institutional Animal Care and Use Committee (IACUC) guidelines (for example, neurological deficits, >15–20% weight loss or impaired ambulation). This rule was fully enforced in all animal experiments. Cell culture Human HEK293T cells (ATCC, CRL-3216), U87 cells (ATCC, HTB-14) and GBM6 (ref. 24 ) cells were cultured in DMEM (Thermo Fisher Scientific, 11995-065) supplemented with 10% FBS (Thermo Fisher Scientific, 10437028) and 1% penicillin–streptomycin (Thermo Fisher Scientific, 15140122). Patient-derived GSCs, including GSC157 (ref. 65 ), GSC7-2, GSC17, GSC23, GSC11, GSC34 (ref. 66 ), and GSC1478 (ref. 31 ), as well as GBM6 (used in the sphere formation assay), were cultured in GSC medium. GSC7-2, GSC17, GSC23, GSC11 and GSC34 were kindly provided by E. Sulman (Duke University); GBM6 was a gift from J. Sarkaria (Mayo Clinic); GSC157 and GSC1478 were from our laboratory collection and were previously characterized 31 , 62 . The GSC medium consists of DMEM/F12 medium (Thermo Fisher Scientific, 11320-033), 2% B27 supplement (Thermo Fisher Scientific, 17504-044), 1× antibiotic-antimycotic (Thermo Fisher Scientific, 15240062), 5 mg ml −1 heparin (Sigma-Aldrich, 9041-08-1), 20 ng ml −1 EGF (Peprotech, 100-15 R) and 20 ng ml −1 bFGF (Peprotech, 100-18B). All cells were authenticated by short tandem repeat analysis at IDEXX BioAnalytics, Texas Tech University Health Sciences Center or NU’s NUSeq core facility. All cell lines tested negative for Mycoplasma using the VenorGeM Mycoplasma Detection Kit (Sigma-Aldrich, MP0025). The latest authentication and Mycoplasma testing were performed in December 2022. All cell lines were cultured for fewer than 20 passages before use. RNA extraction and purification RNA extraction and purification were conducted using the RNeasy Mini kit (QIAGEN, 74104) according to the manufacturer’s instructions. Under some situation, we combine the TRIzol (Invitrogen, 15596026) method with the RNeasy Mini kit to purify RNA. PCR, qPCR and qRT–PCR cDNA was generated via reverse transcription using the iScript cDNA Synthesis kit (Bio-Rad, 1708891) according to the manufacturer’s instructions. PCR was performed using the Q5 High-Fidelity PCR kit (NEB, E0555S) according to the manufacturer’s instructions. EvaGreen qPCR MasterMix (Bullseye, BEQPCR-R) on an Applied Biosystems StepOne Plus Real-Time Thermal Cycling Block was used to conduct qPCR and qRT–PCR. The relative level of gene expression was calculated using the ΔΔCt method. Indicated primers are shown in Supplementary Table 13 . Subcellular fractionation Cells were lysed in ice-cold PBS/0.1% NP-40. After centrifugation at 720 g for 5 min, the supernatant was collected as the cytoplasmic fraction. The pellet (nuclear fraction) was washed and dissolved in either SDS lysis buffer (for protein) or TRIzol (for RNA), depending on the intended application. Rapid amplification of cDNA ends RACE was conducted using the SMARTer RACE 5′/3′ kit (Takara, 634859) according to the manufacturer’s instructions. In brief, 1 μg of freshly isolated RNA from GSC11 or GBM6 cells was used to generate first-strand cDNA. 5′ and 3′ RACE PCRs were performed with universal and gene-specific primers listed in Supplementary Table 13 . For 5′ RACE PCR, the following conditions were used: 94 °C for 30 s, 68 °C for 30 s and 72 °C for 3 min, for a total of 30 cycles. For 3′ RACE PCR, a nested PCR strategy was employed. Specifically, the first round of PCR was performed with the following conditions: 94 °C for 30 s, 68 °C for 30 s and 72 °C for 3 min, for 30 cycles. The second round of PCR used these conditions: 94 °C for 30 s, 70 °C for 30 s and 72 °C for 3 min, for 25 cycles. PCR products were visualized by agarose gel electrophoresis and cloned into a linearized pRACE vector. A total of 12 clone each were randomly selected from the 5′ RACE and 3′ RACE products and subjected to Sanger sequencing to confirm their DNA sequences. Immunofluorescence staining and single-molecule RNA fluorescence in situ hybridization of cells Cells were grown on coverslips. Cells or frozen xenograft tissue sections were fixed with 3.7% formaldehyde in 1× PBS at room temperature for 10 min. For IF staining for cells, fixed cells were washed and permeabilized with 0.1% Tween-20 at room temperature for 15 min. Then, the cells were incubated with the indicated antibody ( Supplementary Table 14 ) at room temperature for 1 h. For fixed frozen xenograft tissue sections, the tissues were incubated with the indicated antibodies at 4 °C overnight. After washing with 1× PBS, coverslips or slides were incubated with the appropriate fluorophore-conjugated antibody ( Supplementary Table 14 ) and 4,6-diamidino-2-phenylindole (DAPI) (Invitrogen, 2615842) at room temperature for 20 min. After washing, coverslips or slides were mounted with Fluoro-Gel (Electron Microscopy Sciences, 17983-20). For smFISH, fixed cells were washed with 1× PBS and permeabilized in 70% ethanol at 4 °C overnight. For sequential IF followed by smFISH, after staining with fluorophore-conjugated antibodies, cells were re-fixed in 3.7% formaldehyde in 1× PBS for 10 min at room temperature and then subjected to the smFISH procedure. Stellaris Quasar 570–labelled probes against HELDR ( Supplementary Table 15 ) were designed with the Stellaris Probe Designer ( https://oligos.biose-archtech.com/products/rna-fish/stellaris-probe-designer ) and synthesized by Biosearch Technologies. smFISH was performed according to the manufacturer’s protocol. In brief, cells were hybridized overnight at 37 °C in a humidified chamber with hybridization buffer (Biosearch, SMF-HB1–10) containing 10% formamide and 1 μl of a 12.5 μM probe pool. Cells were then washed twice at 37 °C for 30 min each with diluted Wash Buffer A (Biosearch, SMF-WA1-60) containing 10% formamide, followed by a 5-min wash at room temperature with diluted Wash Buffer B (Biosearch, SMF-WB1-20). Nuclei were counterstained with DAPI and coverslips were mounted with Fluoro-Gel. HELDR KO cells and RNase-treated cells were used as negative controls. In RNase treatment control, cells were treated with RNase A (NEB, T3018L) at 37 °C for 30 min before the hybridization step. The images were acquired using an Olympus BX53 microscope equipped with a DP72 camera (Olympus cellSens Entry, v.1.7) or a Nikon Ti2 widefield microscope equipped with a CMOS camera (Nikon NIS-Elements, v.5.21.00). Image analyses were performed using FIJI (v.2.16.0), followed by statistical analysis. IF staining and smFISH of patient-derived GBM specimens Formalin-fixed, paraffin-embedded human GBM specimens were de-identified and obtained from the Northwestern Nervous System Tumor Bank. Sections underwent standard de-waxing and rehydration as described previously 67 , followed by routine IF staining. For sequential smFISH followed by IF staining, deparaffinized and rehydrated tissue sections were treated with 10 μg ml −1 proteinase K for 20 min at 37 °C and then subjected to the smFISH procedure. Afterward, the sections were re-fixed in 3.7% formaldehyde in 1× PBS for 10 min at room temperature, followed by routine IF staining. Cell viability assay Cells were seeded at a density of 1,000 cells per well with triplicates in a 96-well plate. The CellTiter-Glo Luminescent Cell Viability Assay kit (Promega, G9241) was used to determine cell viability following the manufacturer’s instructions. Luminescence was measured using the SpectraMax M3 Multi-Mode Microplate Reader (Molecular Devices). Limiting dilution assays A single-cell suspension of GSCs was seeded onto 96-well plates with varying cell numbers. After 1 week of culture, the wells with tumour sphere formation were counted. RNA pulldown assay DNA templates for in vitro transcription were acquired by PCR reaction. Biotinylated RNAs were generated using an in vitro transcription kit (Thermo, A57622) according to the instructions. DNA templates were then removed by digestion with DNase I (NEB, M0303L) at 37 °C for 30 min. RNAs were purified using the RNeasy Mini kit (QIAGEN, 74104). In vitro-transcribed RNA was heated to 65 °C for 10 min, then slowly cooled to room temperature to allow proper secondary structure formation. GSC cells were lysed in RIPA buffer (50 mM Tris, pH 7.4, 150 mM NaCl, 1 mM EDTA, 0.1% SDS, 1% NP-40, 0.5% sodium deoxycholate, 0.5 mM dithiothreitol (DTT), 1× PIC and 1× PhIC). The cell lysate was mixed with RNA and incubated at 4 °C for 4 h on a rotary platform. The reaction mixture was then incubated with streptavidin-conjugated magnetic beads (Thermo, 65001) at 4 °C for 1 h on a rotary platform. The beads were washed five times with RIPA buffer. RNA-associated proteins were eluted by boiling in SDS buffer and detected by IB. RNA immunoprecipitation Cells were crosslinked with 0.3% formaldehyde at room temperature for 10 min. To neutralize the formaldehyde, 1.25 M glycine was added to a final concentration of 0.125 M and the mixture was incubated at room temperature for 5 min. Cells were then washed twice with PBS and lysed in RIPA buffer. The cell lysate was mixed with a primary antibody ( Supplementary Table 14 ) and incubated at 4 °C for 4 h on a rotor. The reaction mixture was subsequently incubated with protein A agarose beads (CST, 9863) at 4 °C for 1 h on a rotor. The beads were washed ten times with RIPA buffer. Proteinase K (NEB, P8107S) and reaction buffer were added, and the mixture was incubated for at 55 °C for 30 min. The RNA was retrieved as described in the RNA extraction protocol. CRISPR/Cas9-mediated HELDR deletion The gRNAs were designed using the SYNTHGO CRISPR Design Tool for knockouts ( https://design.synthego.com ). Two gRNAs were cloned into the lentiCRISPRv2GFP plasmid (Addgene, 82416), and two gRNAs were cloned into the lentiCRISPR v2 plasmid (Addgene, 52961) ( Supplementary Table 16 ). Plasmids for sequence-verified clones were co-transfected into HEK293T cells with the packaging vector psPAX2 (Addgene, 12260) and pVSV-G (Addgene, 138479) to produce lentiviral particles. GSC cells were infected with the virus overnight and then selected using puromycin (Thermo Fisher Scientific, A1113803) and the FACSMelody 3-Laser Sorter (BD Biosciences). Single clones were selected and genotyped by PCR and Sanger sequencing. Gene overexpression and knockdown All relevant plasmids were acquired from VectorBuilder. Sequences were incorporated into PiggyBac expression vectors. The PiggyBac plasmid was co-electroporated with the PiggyBac Hypase Plasmid using the Neon Transfection System (Invitrogen, MPK10096) as per the manufacturer’s instructions. The information of indicated plasmids is shown in Supplementary Table 16 . Immunoblotting Cells were lysed using an SDS lysis buffer consisting of 2% SDS, 50 mM Tris-HCl, 10 mM EDTA and 10% glycerol (pH 8), supplemented with 1× protease inhibitor cocktail (Sigma-Aldrich, P8340) and 1× phosphatase inhibitor cocktail (Roche, 4906845001). The protein samples were then separated via SDS–PAGE and transferred to nitrocellulose (NC) membranes. After the transfer, the membranes were blocked with 5% non-fat milk in TBS-T at room temperature for 1 h. They were then incubated with the indicated antibodies ( Supplementary Table 14 ) at 4 °C overnight. On the second day, following washing with TBS-T, the membranes were incubated with host-specific secondary antibodies conjugated with horseradish peroxidase (HRP; Supplementary Table 14 ). The signal was developed using an enhanced chemiluminescence (ECL) (Amersham Bioscience, RPN2109) reaction according to the manufacturer’s instructions. The image of ECL signals on IB membranes were captured by using an iBright CL1500 Western Blot Imaging System (Thermo Fisher). Short-read RNA sequencing Libraries were prepared using the NEBNext RNA Ultra Library Preparation Kit (E7770) with the poly(A) enrichment module for total RNA. The libraries were sequenced on a NOVAX-02_NovaSeq X Plus (Illumina) at NU’s NUSeq core facility, resulting in an average coverage of 100 million 150-bp paired-end reads per sample. Sequencing data were processed using Quest, NU’s High-Performance Computing Cluster. Reads were aligned to the human genome reference hg38 using HISAT2 (v.2.0.4). Subsequently, the number of reads was quantified using HTSeq-count (v.0.13.5). Differential gene expression analysis was performed using the DESeq2 package (v.1.38.3) in R software with a two-sided Wald test with Benjamini–Hochberg adjustment for multiple comparisons. Gene enrichment analysis was conducted using the online software ShinyGO (v.0.85.1) ( http://bioinformatics.sdstate.edu/go/ ) with a one-sided hypergeometric test with Benjamini–Hochberg adjustment for multiple comparisons. Long-read RNA sequencing The long-read RNA-seq was performed using the platform from Oxford Nanopore Technologies (ONT). Poly(A) RNA was purified from the total RNA sample using the DynaBeads mRNA Purification kit (Invitrogen, 61006). The SQK-DCS109 kit (ONT) was used to prepare the library according to the manufacturer’s instructions. Sequencing was conducted on a Nanopore MinION platform at NU’s NUSeq core facility. ONT reads were base-called and demultiplexed using Guppy (v.6.5.7). The sequencing data was analysed using Porechop (v.0.2.4) and FLAIR (v.2.1) according to the tools’ standard workflows. Chromatin isolation by RNA purification sequencing (ChIRP and ChIRP-seq) Twenty million cells per group were crosslinked with 1% glutaraldehyde in PBS at room temperature for 10 min. The crosslinking reaction was then quenched with 125 mM glycine at room temperature for 5 min. After centrifugation, the cell pellet was washed with cold PBS. Cells were lysed in lysis buffer (50 mM Tris-HCl, pH 7.0, 10 mM EDTA and 1% SDS) at a ratio of 1 ml per 100 mg of cell mass/pellets, with freshly added phenylmethyl sulfonyl fluoride (PMSF) (Sigma-Aldrich, 10837091001), protease inhibitor cocktail (Sigma-Aldrich, P8340) and RNase inhibitor (Applied Biosystems, N8080119). The samples were then fragmented using a Q500 sonicator (Thermo). We designed biotinylated antisense oligonucleotide tiling probes targeting HELDR, antisense strand of HELDR, MALAT1 and LacZ (control), and used commercial NEAT1 probe sets ( Supplementary Table 17 ). Probes were numbered and then mixed in equimolar amounts into two independent pools: one containing even-numbered probes and the other containing odd-numbered probes. The hybridization reaction was conducted at 37 °C for 12 h by adding two volumes of hybridization buffer (750 mM NaCl, 1% SDS, 50 mM Tris-HCl, pH 7.0, 1 mM EDTA and 15% formamide) with freshly added PMSF, protease inhibitor cocktail and RNase inhibitor, to one volume of samples and 100 pmol of biotinylated DNA probe pools. The complex was pulled down using 100 μl of streptavidin-magnetic C1 beads (Thermo, 65001) at 37 °C for 30 min and washed five times with washing buffer (2× SSC and 0.5% SDS). For RNA isolation, 5 μl of Proteinase K (NEB, P8107S) and 95 μl of Proteinase K buffer (100 mM NaCl, 10 mM Tris-HCl, pH 7.5, 1 mM EDTA and 0.5% SDS) were added and incubated at 50 °C for 45 min with moderate shaking. RNA was retrieved by combining TRIzol and the RNeasy Mini kit. qRT–PCR was used to validate the specificity of the reaction. For DNA isolation, DNA was eluted using 1 ml of elution buffer (50 mM NaHCO 3 , 1% SDS) containing 10 μl of RNase A (10 mg ml −1 ) (NEB, T3018L) and 10 μl of RNase H (10 U μl −1 ) (NEB, M0297L) at 37 °C for 1 h with moderate shaking. Next, 15 μl of Proteinase K (NEB, P8107S) was added to each sample, followed by incubation at 50 °C for 45 min with shaking. The supernatant was then collected after separation by magnetic stand. The DNA was purified using the ChIP DNA Clean and Concentrator kit (Zymo Research, D5205) according to the manufacturer’s protocol and used for sequencing or qPCR. The library preparation was performed using the KAPA HyperPrep kit (Roche, 07962347001) and xGen UDI-UMI adaptors (IDT, 10006914) following the manufacturer’s instructions. Subsequently, 150-bp pair-end sequencing was conducted using the Element AVITI sequencer (Element Biosciences) at NU’s NUSeq core facility. Analysis of ChIRP-seq For ChIRP-seq data analysis, the paired-end sequencing reads were trimmed using fastp (v.0.22.0) and filtered using RNAnue (v.0.2.0). Window trimming (–wsize 3) was applied with a minimum Phred score of 20. Quality control of the reads before and after preprocessing was conducted using FastQC (v.0.12.1). The pre-processed reads were aligned by using Bowtie2 (v.2.5.1) with the very-sensitive parameter against the GRCh38/hg38 reference genome. Subsequently, the aligned reads were deduplicated using UMI tools (v.1.1.4). The bamCompare from deep-Tools (v.3.5.1) was used to normalize the odd/even samples against the input samples. Peaks were called for each replicate individually using MACS3 (v.3.0.1) with parameters –qvalue 0.05 –broad –broad-cutoff 0.05. Overlapping peaks were acquired by using BEDTools (v.2.30.1). Motif analysis was conducted using HOMER (v.4.11) with a one-sided cumulative binomial test. To assess the statistical significance of the overlap between peaks retrieved from the even and odd probe pools targeting HELDR, a one-sided permutation test was performed using the R package regioneR 68 (with 1,000 iterations). Differential binding analysis of ChIRP-seq data at HELDR binding sites was performed using the R package DiffBind 69 , treating even and odd probe sets targeting HELDR as replicates (two-sided Wald test with Benjamini–Hochberg adjustment for multiple comparisons). Chromatin immunoprecipitation and sequential ChIP ChIP assays were performed by using the SimpleChIP Enzymatic Chromatin IP kit (CST, 9003) according to the manufacturer’s instructions. In brief, cells were crosslinked with 1% formaldehyde at room temperature for 10 min and then quenched with a glycine solution provided by the kit. The crosslinked chromatin was digested with micrococcal nuclease and then sonicated using a Bioruptor (diagenode). The reaction system was incubated with the indicated antibody overnight at 4 °C. After a 2-h incubation with Protein G beads at 4 °C, the complex was decrosslinked. The DNA was purified and subsequently used for qPCR. For ChIP–re-ChIP, following the first round of immunoprecipitation with P300, H3K27ac or IgG, the complex was eluted using 100 μl of elution buffer containing 10 mM DTT at 37 °C for 30 min. The elution was diluted 25-fold with a buffer containing 20 mM Tris-HCl (pH 8.0), 150 mM NaCl, 2 mM EDTA and 1% Triton X-100, followed by re-immunoprecipitation using the GABPA antibody. Animal experiments Athymic nude mice (outbred, CrTac:NCr-Foxn1 nu , Taconic Biosciences), aged 6–8 weeks, were used for all in vivo studies ( n = 5 mice per group for all cohorts). Mice were housed under a 12-h light–dark cycle, with controlled temperature (20–26 °C) and humidity (40–60%). Food and water were available at all times. Both male and female mice were used in this study. Specifically, male mice were used for the HELDR KD and HELDR KO experiments involving GSC17 cells. For all other experiments, female mice were used. GSCs or GBM6 cells expressing luciferase were injected intracranially with 2 × 10 5 cells. Bioluminescence imaging (BLI) was conducted to monitor tumour growth using the SII Lago imaging system (Aura, v.3.2). In the therapy experiment targeting KAT7, 1 week post-implantation, the mice were randomly assigned to four groups ensuring that the mean BLI signal was comparable across groups: (1) control; (2) WM-3835 (MCE, HY-134901) (100 mg kg −1 d −1 , Monday to Friday until moribund, intraperitoneal injection); (3) erlotinib (MCE, HY-50896) (50 mg kg −1 d −1 , Monday to Friday until moribund, oral gavage); and (4) a combination of WM-3835 and erlotinib. In the HELDR-targeting therapy experiment, the ASOs ( Supplementary Table 18 ) were designed by using an online tool provided and synthesized by IDT. ASOs were injected intratumorally mixed with in vivo-jetPEI reagent (Polyplus, 0.4 μl per mouse), as described in our previous publication 31 . Specifically, 1 week post-implantation, mice were randomly assigned to four groups based on similar mean BLI signals: (1) control (vehicle, oral gavage Monday–Friday until moribund; ASO control 4 μg per mouse per day, intratumoral injection Monday and Thursday until moribund); (2) ASO-targeting HELDR (4 μg per mouse per day, intratumoral injection Monday and Thursday until moribund; vehicle, oral gavage Monday–Friday until moribund); (3) erlotinib (MCE, HY-50896) (50 mg kg −1 d −1 , oral gavage Monday–Friday until moribund; ASO control 4 μg per mouse per day, intratumoral injection Monday and Thursday until moribund); and (4) a combination of ASO-targeting HELDR and erlotinib. In all animal experiments, BLI monitoring was commenced on brain glioma xenograft-bearing animals weekly from day 7 post-implantation. Animals were killed upon reaching predefined humane end points consistent with IACUC guidelines (for example, neurological deficits, >15–20% weight loss or impaired ambulation). For Kaplan–Meier survival analyses, mice were killed when one of these humane end points was reached. Bioinformatics analysis of public datasets To study the EGFR -correlated lncRNA expression profile in GBM, two-sided Spearman’s rank correlation analysis between the expression of EGFR and each lncRNA was performed. lncRNAs with an absolute correlation coefficient (∣ r ∣) > 0.4 and a P < 0.05 were included. A heatmap showing the expression of lncRNAs was generated with R software. Whole-genome sequencing (WGS) data were used for copy number variation (CNV) analysis. Binned coverage on chromosome 7, encompassing the EGFR and HELDR genes, was calculated using the WGD suite ( https://github.com/RCollins13/WGD ). Three representative samples were visualized using the Sushi R package, v.1.34.0. Visualization of sequencing coverage and genomic regions IGV (v.2.16.1) was used to visualize coverage and genomic regions in short-read RNA-seq, long-read RNA-seq and ChIRP-seq datasets. Statistics and reproducibility Data are presented as mean ± s.d., unless otherwise indicated. For box plots, the centre line represents the median, the box spans the 25th–75th percentiles, and the whiskers extend to the minimum and maximum values. All replicates used in this study are biological replicates. For limiting dilution assays, the results were calculated using the Extreme Limiting Dilution Analysis online tool ( http://bioinf.wehi.edu.au/software/elda/ ) with a two-sided likelihood ratio test. Cis-acting prediction of lncRNA is calculated using the TransCistor online tool ( https://shiny.bioinformatics.unibe.ch/apps/transcistor/ ), which applies a one-sided Fisher’s exact test. All other statistical analyses were performed using GraphPad Prism (v.10), SPSS (v.27) or R (v.4), unless otherwise specified. Statistical significance was assessed using one-way ANOVA followed by Tukey’s or Dunnett’s post hoc test; Kruskal–Wallis test followed by Dunn’s multiple comparisons test; two-sided unpaired or paired Student’s t -tests; two-sided Spearman’s rank correlation test; one-tailed Fisher’s exact test; or the log-rank test, as indicated in the figure legends. No statistical methods were used to predetermine sample sizes but our sample sizes are similar to those reported in previous publications (PMID: 38321204, 2024; PMID: 39025928, 2024). The data distribution was assumed to be normal unless stated otherwise; however, this was not formally tested. No data were excluded. Unless stated otherwise, the experiments were not randomized. Data collection and analysis were not performed blind to the conditions of the experiments. The investigators were not blinded to allocation during experiments and outcome assessment. Materials availability All newly generated materials, including oligonucleotides and plasmids described in this study, are available from the corresponding author upon reasonable request. Reporting summary Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article. Extended Data Extended Data Fig. 1 ∣. The lncRNA ELDR is co-amplified with EGFR in GBM. Open in a new tab a , Heatmap depicts expressions of 118 lncRNAs that exhibited significant correlation with EGFR expression in NU GBM samples (∣r ∣ > 0.4 and p < 0.05). Expression levels of lncRNAs and EGFR are indicated. b , The correlation in RNA transcripts between EGFR and ELDR in CPTAC ( n = 94) GBM samples. c , Percentage of ELDR amplification in GBM patients with or without EGFR amplification in TCGA ( n = 284/319 vs. 2/223) and CPTAC ( n = 61/64 vs. 0/35). d , Expression of ELDR RNA in normal brain ( n = 5), LGG ( n = 530), and GBM ( n = 169) in TCGA dataset. e , Kaplan–Meier analysis of ELDR expression in TCGA glioma samples. f , TCGA Pan-Cancer analysis of expression of RNA transcript at the ELDR locus via the UALCAN portal ( https://ualcan.path.uab.edu/ ). In box-and-whisker plots, the centre line indicates the median; the box boundaries represent the interquartile range (25th–75th percentiles); and the whiskers show the minimum and maximum values. BLCA, bladder urothelial carcinoma ( n = 414); BRCA, breast invasive carcinoma ( n = 1102); CESC, cervical squamous cell carcinoma and endocervical adenocarcinoma ( n = 304); COAD, colon adenocarcinoma ( n = 478); CHOL, cholangiocarcinoma ( n = 36); ESCA, oesophageal carcinoma ( n = 161); GBM, glioblastoma ( n = 156); HNSC, head and neck squamous cell carcinoma ( n = 500); KICH, kidney chromophobe ( n = 65); KIRC, kidney renal clear cell carcinoma ( n = 538); KIRP, kidney renal papillary cell carcinoma ( n = 288); LIHC, liver hepatocellular carcinoma ( n = 371); LUAD, lung adenocarcinoma ( n = 533); LUSC, lung squamous cell carcinoma ( n = 502); PAAD, pancreatic adenocarcinoma ( n = 177); PCPG, pheochromocytoma and paraganglioma ( n = 178); PRAD, prostate adenocarcinoma ( n = 498); READ, rectum adenocarcinoma ( n = 166); SARC, sarcoma ( n = 259); STAD, stomach adenocarcinoma ( n = 375); THCA, thyroid carcinoma ( n = 502); THYM, thymoma ( n = 119); UCEC, uterine corpus endometrial carcinoma ( n = 551); ACC, adrenocortical carcinoma ( n = 79); DLBC, diffuse large B-cell lymphoma ( n = 48); LGG, low-grade glioma ( n = 511); UCS, uterine serous carcinoma ( n = 56); UVM, uveal melanoma ( n = 80); OV, ovarian serous cystadenocarcinoma ( n = 374); LAML, acute myeloid leukaemia ( n = 151); MESO, mesothelioma ( n = 86); TGCT, testicular germ cell tumour ( n = 150); SKCM, skin cutaneous melanoma ( n = 103); Mets, metastasis ( n = 367). g , Correlation of EGFR and ELDR RNA expression in head and neck squamous cell carcinoma (HNSC; n = 381) and lung adenocarcinoma (LUAD; n = 389) samples. h , The correlation of CNV between EGFR and ELDR gene loci in HNSC ( n = 399) and LUAD ( n = 562) samples. i , Proportion of ELDR amplification in HNSC (with EGFR amplification: 11/17; without: 6/382) and LUAD (with: 15/22; without: 9/540). Two-sided Spearman’s rank correlation test in a , b , g , h , one-way ANOVA with Tukey’s post hoc test in d , log-rank test in e . Data are presented as mean ± s.d. ( d ). Extended Data Fig. 2 ∣. De novo lncRNA HELDR is the major transcript in the ELDR locus and localized in the nucleus. Open in a new tab a, b , Integrative genomics viewer (IGV) showed short-read RNA-seq of GSC34 and GBM39 at the ELDR locus. c , Agarose gel electrophoresis. 5’ rapid amplification of cDNA ends (RACE) and 3’ RACE for HELDR transcripts in GBM6. Data are representative of three independent experiments with similar results. d , Representative Sanger sequencing for 5’ RACE and 3’ RACE products. e , Open reading frame (ORF) assessment using the ORF Finder ( https://www.ncbi.nlm.nih.gov/orffinder ). No ORFs > 300 nt were detected. f , Coding potential of indicated genes predicted by an online software CPC ( https://cpc.gao-lab.org/ ). g , UCSC Genome Browser tracks showing HELDR conservation (PhyloP 100, vertebrates) and ENCODE 70 ChIP–seq marks for enhancers (H3K4me1), promoters (H3K4me3), and active regulatory elements (H3K27ac). h , RNA-seq data from the Ivy Glioblastoma Atlas Project showing HELDR expression across GBM anatomical regions (LE, leading edge, n = 7; IT, infiltrating tumour, n = 9; CT, cellular tumour, n = 36; PZ, perinecrotic zone, n = 12; PS, pseudopalisading cells around necrosis, n = 14; HPV, hyperplastic blood vessels in the cellular tumour, n = 10; MP, microvascular proliferation, n = 11). i , Correlation of EGFR and HELDR RNA expression in Ivy GAP dataset ( n = 269). Kruskal–Wallis test followed by Dunn’s multiple comparisons test in h , two-sided Spearman’s rank correlation test in i , the centre line marks the median, the box spans the 25th–75th percentiles, and the whiskers extend to the minimum and maximum in h . Extended Data Fig. 3 ∣. HELDR is critical for cell proliferation and tumorigenicity. Open in a new tab a , smFISH for HELDR in GSC34 and GSC11 cells (representative of 50 cells). Cells treated with RNase served as a negative control. Arrow: HELDR . b , smFISH quantification of HELDR loci per cell (n = 50 cells). c , HELDR expression (TPM) from RNA-seq. d , qRT–PCR for HELDR expression in U87, patient-derived GSCs, and patient-derived xenograft (PDX) GBM6 cell lines with indicated levels of EGFR expression (n = 3). HELDR was presented as the log 2 of its relative expression level. e , qRT–PCR of HELDR in GSCs and matched differentiated glioblastoma cells (DGCs) (n = 3). f , qRT–PCR for shRNA KD of HELDR in GSC34 (n = 3) cells that EGFR is amplified. g , Cell proliferation for GSC34 (n = 3) cells with or without HELDR KD. h , Glioma sphere formation (self-renewal capacity) of GSC34 cells with or without HELDR KD. i , IB for representative GBM stem-cell markers in GSCs with or without HELDR KD. β-actin is a sample processing control. j , CRISPR-Cas9-mediated HELDR deletion (Del) in GSCs. k, l , PCR ( k ) and Sanger sequencing ( l ) for two single clones of CRISPR/Cas9-mediated HELDR deletion GSC17 cells. m , qRT–PCR results confirmed deletion (Del) of HELDR in GSC17 cells (n = 3). n , Cell proliferation for GSC17 cells with or without HELDR deletion (n = 6). o , Self-renewal capacity of GSC17 cells with or without HELDR deletion. p , Representative BLI images of indicated GSC17 tumour xenograft-bearing mice. q , Kaplan–Meier analysis of GSC17 tumour xenograft-bearing mice (n = 5). r, s , qRT–PCR for shRNA KD of HELDR and overexpression of shRNA-resistant HELDR in GSC17 (n = 3) and GSC34 cells (n = 3). t, u , Cell proliferation for GSC17 (n = 3) and GSC34 (n = 3) cells with HELDR KD and overexpression of shRNA-resistant HELDR. v, w , Glioma sphere formation of GSC17 and GSC34 cells with HELDR KD and overexpression of shRNA-resistant HELDR. Scale bar, 10 μm. All blots, qRT–PCR, and in vitro assays are representative of three independent experiments. Two-sided unpaired t -test in e , one-way ANOVA with Dunnett’s post hoc test in f, g, m, n , two-sided likelihood ratio test in h, o, v, w , log-rank test in q , one-way ANOVA with Tukey’s post hoc test in r-u . Data are shown as mean ± s.d. ( b, d-g, m, n , and r-u ). Extended Data Fig. 4 ∣. EGFR pathway activity is unaffected by HELDR, and EGFR does not affect HELDR expression. Open in a new tab a , Diagram of EGFR, ELDR/HELDR gene locus and nearby genes in Chromosome (Chr) 7 genomic DNA 71 or ecDNAs 3 , 4 in GBM. b , Dot plot showing expression changes of the indicated genes at the EGFR and ELDR/HELDR loci in GSC17 cells after HELDR knockdown with two independent shRNAs. The box highlights SEC61G, the only gene significantly regulated by HELDR KD in both shRNAs. c , IB for EGFR and EGFR downstream signalling proteins in GSC34 or GSC11 cells with or without HELDR KD. d, e , IB of EGFR-pathway proteins ( d ) and qRT–PCR (n = 3) of HELDR ( e ) in GSCs with or without EGFR knockdown or treatment with the EGFR inhibitor erlotinib. β-actin is a sample processing control in c and d . All blots and qRT–PCR are representative of three independent experiments. Two-sided Wald test with Benjamini–Hochberg adjustment for multiple comparisons in b , one-way ANOVA with Dunnett’s post hoc test in e , Data are shown as mean ± s.d. ( e ). Extended Data Fig. 5 ∣. HELDR exhibits widespread genomic DNA binding. Open in a new tab a, b , ChIRP-seq targeting NEAT1 ( a ) and MALAT1 ( b ) visualized in IGV at the NEAT1 and MALAT1 loci (GSC17). c, d , Venn diagrams illustrating the shared and distinct peaks identified by ChIRP-seq targeting NEAT1 ( c ) or MALAT1 ( d ) in comparison with HELDR (GSC17). e , Bar plot showing common peak number of ChIRP-seq in probe targeting HELDR and negative controls including RNase-treated sample before hybridization, probes targeting antisense strand of HELDR, and HELDR negative cell (GSC1478). f , Volcano plot showing log 2 fold changes and adjusted p-values (padj) for HELDR occupancy (ChIRP-seq) differences between control and HELDR-KD GSC17 cells across 7,095 HELDR binding sites ( Supplementary Table 5 ). Peaks with ∣log 2 FC ∣ > 1 and padj < 0.05 were considered significant. g , IGV of two representative genes, UBASH3B and ATP2B4 that were identified in HELDR ChIRP-seq and RNA-seq analysis. The peaks of HELDR ChIRP-seq in the promoter region are shaded. h , ChIRP–qPCR: Enrichment of HELDR at the promoters of KAT7 (n = 3), UBASH3B (n = 3), and ATP2B4 (n = 3) in GSC17 cells. qPCR assays are representative of three independent experiments. Two-sided Wald test with Benjamini–Hochberg adjustment for multiple comparisons in f , two-sided unpaired t -test in h , Data are presented as mean ± s.d. ( h ). Extended Data Fig. 6 ∣. HELDR facilitates KAT7 gene expression through recruiting p300 to the KAT7 gene promoter. Open in a new tab a , IB for KAT7 protein expression in GSC34 and GSC11 cells with or without HELDR KD. b , IB of KAT7 in GSCs with or without EGFR KD or treatment with the EGFR inhibitor erlotinib. c-e , Representative images of HELDR smFISH with IF for KAT7 or EGFR in GSCs (representative of 50 cells). Arrow: HELDR. f-h , Representative HELDR smFISH with IF for KAT7 or EGFR in human GBM tissue (representative of 7 patients). Arrow: HELDR. i , Gene correlation analysis of HELDR and KAT7 genes in TCGA (n = 169) and CPTAC (n = 91) GBM samples. j , Correlation analysis of RNA-seq data for EGFR RNA in relation to KAT7 in GBM of indicated datasets (TCGA, n = 169; CPTAC, n = 91; NU, n = 47). k , Images of smFISH (HELDR) and IF (p300) in GSC34 and GSC11 cells (representative of 30 cells). Arrow: HELDR. l , HELDR localization compared between top-50% and bottom-50% p300 intensity regions within each nucleus (n = 30 cells). m, n , ChIP-qPCR. Enrichment of p300 (left) and H3K27ac (right) at the KAT7 gene promoter in GSC34 (n = 3) and GSC11 (n = 3) cells with or without HELDR KD. IgG is a control. o , Sequence logo plot and enrichment of ETS family member-binding motif in HELDR ChIRP-seq. Scale bar, 10 μm. β-actin is a sample processing control in a and b . All blots and qPCR are representative of three independent experiments. Two-sided Spearman’s rank correlation test in i , j , two-sided paired t -test in l , two-sided unpaired t -test in m, n , one-sided cumulative binomial test in o , Data are presented as mean ± s.d. ( l-n ). Extended Data Fig. 7 ∣. HELDR recruits p300 and ETS transcription factors to the KAT7 promoter to promote KAT7 transcription. Open in a new tab a, b , ChIP-qPCR. Enrichment of GABPA at KAT7 gene promoters were compared in GSC34 (n = 3) and GSC11 (n = 3) cells with or without HELDR KD. IgG is a control. c, IB for GABPA protein expression in GSCs with or without HELDR KD. d-g , ChIP-re-ChIP-qPCR. Enrichment of P300/GABPA (n = 3) or H3K27ac/GABPA (n = 3) complexes at KAT7 gene promoters were compared in GSC34 or GSC11 cells with or without HELDR KD. The first ChIP was performed with p300, H3K27ac, or IgG, followed by a second ChIP using GABPA. IgG is a control in ( a, b , and d-g ). h-l , Effects of ± treatment of 5 μM p300 inhibitor A-485 on KAT7 protein expression ( h , IB, treatment for three days), cell proliferation (n = 3) ( i , j ) or self-renewal (glioma sphere formation, k , l ) in GSC34 or GSC11 cells with or without HELDR KD. β-actin is a sample processing control in c and h . All blots, qPCR, and in vitro assays are representative of three independent experiments. Two-sided unpaired t -test in a, b , and d-g , one-way ANOVA with Tukey’s post hoc test in i and j , two-sided likelihood ratio test in k and l . Data are presented as mean ± s.d. ( a, b, d-g, i , and j ). Extended Data Fig. 8 ∣. KAT7 mediates HELDR-promoted GBM tumorigenesis by inducing transcriptions of genes that are critical for GBM malignancy. Open in a new tab a , IB for KAT7 proteins and its substrate H3K14ac and H4K12ac in GSCs with or without KAT7 KD. b, c , KAT7 KD inhibits cell proliferation (n = 3) ( b ) and glioma sphere formation ( c ) in GSCs. d-f , Re-expression of KAT7 rescued HELDR KD-suppressed KAT7 proteins and its substrate H3K14ac and H4K12ac ( d , IB), cell proliferation (n = 3) ( e ), and glioma sphere formation ( f ) of GSCs. g , Co-regulated genes shared between KAT7 knockdown and HELDR knockdown in GSC17 cells (two biological replicates). The dot plot shows gene-expression changes after HELDR knockdown (shRNA-2). h , ChIP-qPCR. Enrichment of KAT7 (see Fig. 6k ) and IgG (negative control) at the promoters of indicated genes in GSC17 cells, with or without HELDR KD or rescue by KAT7 OE (n = 3). i-k , ChIP-qPCR. Enrichment of KAT7 (n = 3) ( i ), H3K14ac (n = 3) ( j ), and H4K12ac (n = 3) ( k ) at the promoters of indicated genes in GSC17 cells with or without KAT7 KD. β-actin is a sample processing control in a and d . All blots, qPCR, and in vitro assays are representative of three independent experiments. One-way ANOVA with Dunnett’s post hoc test in b , two-sided likelihood ratio test in c and f , one-way ANOVA with Tukey’s post hoc test in e and h , two-sided Wald test with Benjamini–Hochberg adjustment for multiple comparisons in g , two-sided unpaired t -test in i-k . Data are presented as mean ± s.d. ( b, e , and h-k ). Extended Data Fig. 9 ∣. KAT7 inhibitor WM-3835 synergistically enhanced anti-GBM activity of EGFR inhibitor erlotinib. Open in a new tab a , Cell viability of EGFR -amplified (GSC11, GBM6) and non-amplified (GSC1478, GSC157) GSC/GBM cells after 6 days of erlotinib treatment at indicated concentrations (n = 3). b , Cell inhibition rates for combinations of indicated concentrations of KAT7 inhibitor (WM-3835) and EGFR inhibitor (Erlotinib) are presented in the matrix. Data were obtained by cell viability assays. c , Cell viability of EGFR -amplified (GSC11, GBM6) and non-amplified (GSC1478, GSC157) GSC/GBM cells after 6 days of WM-3835 treatment at indicated concentrations (n = 3). d , IB for indicated protein expressions in GSC11 or GBM6 cells treated with vehicle or WM-3835 for 72 h. β-actin is a sample processing control. e , Representative images of IF staining for brain sections with GBM6 xenograft tumours with indicated treatments. f , Quantification of IF staining for brain sections with GBM6 (n = 5) and GSC11 (n = 5) xenograft tumours. Scale bar, 100 μm. All blots and in vitro assays are representative of three independent experiments. By one-way ANOVA with Tukey’s post hoc test. Data are presented as mean ± s.d. ( a , c , and f ). Extended Data Fig. 10 ∣. ASO-mediated inhibition of HELDR synergistically enhanced anti-GBM activity of EGFR inhibitor erlotinib. Open in a new tab a , qRT–PCR analysis of GSC11 cells 24 h after transfection with 100 nM HELDR-targeting ASOs or a control ASO in vitro (n = 3). b, c , IB for indicated proteins in GBM6 and GSC11 cells 72 h after transfection with 100 nM HELDR-targeting ASOs or a control ASO in vitro. d , Cell viability of EGFR -amplified (GSC11, GBM6) and non-amplified (GSC1478, GSC157) GSC/GBM cells after 6 days of indicated ASO treatment (n = 3). e-j , Cell proliferation (n = 3) ( e-g ), and glioma sphere formation ( h-j ) with indicated treatment. In d - j , ASO, 100 nM; Erlotinib, 0.8 μM. k , Analysis of coding DEGs from RNA-seq of GSC17 cells 24 h after transfection with 100 nM control or HELDR-targeting ASO in vitro (two biological replicates; threshold: ∣fold change∣> 1.5 and adjusted p < 0.05). Venn diagrams showing common coding DEGs between HELDR knockdown by shRNA (shared by both shRNAs; see Fig. 3n ) and ASO treatment. l , BLI images for mice bearing GSC11 brain tumour xenografts with indicated treatments (n = 5 in each group). Treatments: ASO, 4 μg/mouse, twice/week, erlotinib, 50 mg/kg, 5 days on/2 days off; until morbibund. m , qRT–PCR analysis of HELDR expression in tumour area from brain sections (n = 4). n , Representative images of IF staining for brain sections with GBM6 xenograft tumours with indicated treatments. o-q , Quantification of IF staining shown in n (n = 5). r-t , Quantification of IF staining for brain sections with GSC11 xenograft tumours with indicated treatments (n = 5). Scale bar, 100 μm. β-actin is a sample processing control in b and c . All blots, qRT-PCR, and in vitro assays are representative of three independent experiments. One-way ANOVA with Dunnett’s post hoc test in a , two-sided unpaired t -test in d , one-way ANOVA with Tukey’s post hoc test in e-g, m , and o - t , two-sided likelihood ratio test in h-j , two-sided Wald test with Benjamini–Hochberg adjustment for multiple comparisons in k , Data are presented as mean ± s.d. ( a, d, e-g, m , and o-t ). Supplementary Material Supplementary Tables-2 NIHMS2161547-supplement-Supplementary_Tables-2.xlsx (298.3KB, xlsx) Supplementary Tables-1 NIHMS2161547-supplement-Supplementary_Tables-1.xlsx (2.2MB, xlsx) Raw_data_of_IB_and_DNA_gels NIHMS2161547-supplement-Raw_data_of_IB_and_DNA_gels.pdf (3.9MB, pdf) Supplementary information The online version contains supplementary material available at https://doi.org/10.1038/s41556-026-01924-w . Acknowledgements All shared resources at RHLCC and NU Feinberg School of Medicine were supported by National Cancer Institute Cancer Center grant P30CA060553. We thank E. Sulman and J. Sarkaria for providing the GSC cell lines and GBM6 cells. We thank the NU Feinberg School of Medicine Center for Advanced Microscopy/Nikon Imaging Center and NUSeq core for help with experiments. This work was supported by US NIH grants NS113160, NS126810, NS115403, NS122375 and NS125318 (S.-Y.C.); NIH CA234799, United States Army Medical Research Acquisition Activity W81XWH-22-10373 (D.T.); W81XWH-22-1-0374 and HT9425-24-1-0573 (X.S.); NIH CA259388 and GM142441 (R.Y.); and CA285684, CA278832, CA256741 and P50CA180995 Development Research Program, the Polsky Urologic Cancer Institute of the Robert H. Lurie Comprehensive Cancer Center of NU at Northwestern Memorial Hospital (Q.C.). The funders had no role in study design, data collection and analysis, decision to publish or preparation of the manuscript. Footnotes Competing interests The authors declare no competing interests. Extended data is available for this paper at https://doi.org/10.1038/s41556-026-01924-w . Online content Any methods, additional references, Nature Portfolio reporting summaries, source data, extended data, supplementary information, acknowledgements, peer review information; details of author contributions and competing interests; and statements of data and code availability are available at https://doi.org/10.1038/s41556-026-01924-w . Data availability The short-read RNA-seq, long-read RNA-seq and ChIRP-seq data that support the findings of this study have been deposited in the Gene Expression Omnibus under accession codes GSE286179 , GSE286318 , GSE286319 , GSE286320 and GSE286432 . Previously published RNA-seq datasets for Mayo Clinic PDX models and NU cohort that were re-analysed here are available under accession codes PRJNA548556 (ref. 24 ), SRR8236743 (ref. 3 ) and GSE147352 (ref. 21 ). The Ivy GAP dataset was acquired from its portal ( https://glioblastoma.alleninstitute.org/ ) 28 . Publicly available datasets from the TCGA and CPTAC cohorts were obtained from the GDC Data Portal ( https://portal.gdc.cancer.gov/ ). All other data supporting the findings of this study are available from the corresponding author on reasonable request. Source data are provided with this paper. References 1. Horbinski C, Berger T, Packer RJ & Wen PY Clinical implications of the 2021 edition of the WHO classification of central nervous system tumours. Nat. Rev. Neurol 18, 515–529 (2022). [ DOI ] [ PubMed ] [ Google Scholar ] 2. Brennan CW et al. The somatic genomic landscape of glioblastoma. Cell 155, 462–477 (2013). [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 3. Wu S. et al. Circular ecDNA promotes accessible chromatin and high oncogene expression. Nature 575, 699–703 (2019). [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 4. Morton AR et al. Functional enhancers shape extrachromosomal oncogene amplifications. Cell 179, 1330–1341.e13 (2019). [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 5. Yan X, Mischel P & Chang H Extrachromosomal DNA in cancer. Nat. Rev. Cancer 24, 261–273 (2024). [ DOI ] [ PubMed ] [ Google Scholar ] 6. Westphal M, Maire CL & Lamszus K EGFR as a target for glioblastoma treatment: an unfulfilled promise. CNS Drugs 31, 723–735 (2017). [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 7. McClellan BL et al. Impact of epigenetic reprogramming on antitumor immune responses in glioma. J. Clin. Invest 133, e163450 (2023). [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 8. Esteller M. et al. The epigenetic hallmarks of cancer. Cancer Discov. 14, 1783–1809 (2024). [ DOI ] [ PubMed ] [ Google Scholar ] 9. Shvedunova M & Akhtar A Modulation of cellular processes by histone and non-histone protein acetylation. Nat. Rev. Mol. Cell Biol 23, 329–349 (2022). [ DOI ] [ PubMed ] [ Google Scholar ] 10. Roadmap Epigenomics Consortium et al. Integrative analysis of 111 reference human epigenomes. Nature 518, 317–330 (2015). [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 11. Ogryzko VV, Schiltz RL, Russanova V, Howard BH & Nakatani Y The transcriptional coactivators p300 and CBP are histone acetyltransferases. Cell 87, 953–959 (1996). [ DOI ] [ PubMed ] [ Google Scholar ] 12. Maldotti M. et al. The acetyltransferase p300 is recruited in trans to multiple enhancer sites by lncSmad7. Nucleic Acids Res. 50, 2587–2602 (2022). [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 13. Yokoyama A, Niida H, Kutateladze TG & Cote J HBO1, a MYSTerious KAT and its links to cancer. Biochim. Biophys. Acta 1867, 195045 (2024). [ Google Scholar ] 14. White J. et al. Histone lysine acetyltransferase inhibitors: an emerging class of drugs for cancer therapy. Trends Pharmacol. Sci 45, 243–254 (2024). [ DOI ] [ PubMed ] [ Google Scholar ] 15. Mattick JS et al. Long non-coding RNAs: definitions, functions, challenges and recommendations. Nat. Rev. Mol. Cell Biol 24, 430–447 (2023). [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 16. Statello L, Guo CJ, Chen LL & Huarte M Gene regulation by long non-coding RNAs and its biological functions. Nat. Rev. Mol. Cell Biol 22, 96–118 (2021). [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 17. Ferrer J & Dimitrova N Transcription regulation by long non-coding RNAs: mechanisms and disease relevance. Nat. Rev. Mol. Cell Biol 25, 396–415 (2024). [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 18. Bailey C. et al. Origins and impact of extrachromosomal DNA. Nature 635, 193–200 (2024). [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 19. Tang J. et al. Enhancing transcription-replication conflict targets ecDNA-positive cancers. Nature 635, 210–218 (2024). [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 20. Noorani I. et al. Extrachromosomal DNA-driven oncogene spatial heterogeneity and evolution in glioblastoma. Cancer Discov. 15, 2078–2095 (2025). [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 21. Huang T. et al. PRMT6 methylation of RCC1 regulates mitosis, tumorigenicity, and radiation response of glioblastoma stem cells. Mol. Cell 81, 1276–1291.e9 (2021). [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 22. Sur S & Ray RB Emerging role of lncRNA ELDR in development and cancer. FEBS J. 289, 3011–3023 (2022). [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 23. Rossi MR et al. Novel amplicons on the short arm of chromosome 7 identified using high resolution array CGH contain over expressed genes in addition to EGFR in glioblastoma multiforme. Genes Chromosomes Cancer 44, 392–404 (2005). [ DOI ] [ PubMed ] [ Google Scholar ] 24. Vaubel RA et al. Genomic and phenotypic characterization of a broad panel of patient-derived xenografts reflects the diversity of glioblastoma. Clin. Cancer Res 26, 1094–1104 (2020). [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 25. Niazi F & Valadkhan S Computational analysis of functional long noncoding RNAs reveals lack of peptide-coding capacity and parallels with 3′ UTRs. RNA 18, 825–843 (2012). [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 26. Wang L. et al. CPAT: coding-potential assessment tool using an alignment-free logistic regression model. Nucleic Acids Res. 41, e74 (2013). [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 27. Kong L. et al. CPC: assess the protein-coding potential of transcripts using sequence features and support vector machine. Nucleic Acids Res. 35, W345–W349 (2007). [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 28. Puchalski RB et al. An anatomic transcriptional atlas of human glioblastoma. Science 360, 660–663 (2018). [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 29. Wu S. et al. EGFR amplification induces increased DNA damage response and renders selective sensitivity to talazoparib (PARP inhibitor) in glioblastoma. Clin. Cancer Res 26, 1395–1407 (2020). [ DOI ] [ PubMed ] [ Google Scholar ] 30. Guo G. et al. EGFR ligand shifts the role of EGFR from oncogene to tumour suppressor in EGFR-amplified glioblastoma by suppressing invasion through BIN3 upregulation. Nat. Cell Biol 24, 1291–1305 (2022). [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 31. Song X. et al. RNA splicing analysis deciphers developmental hierarchies and reveals therapeutic targets in adult glioma. J. Clin. Invest 134, e173789 (2024). [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 32. Lathia JD et al. Integrin α6 regulates glioblastoma stem cells. Cell Stem Cell 6, 421–432 (2010). [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 33. Anido J. et al. TGF-β receptor inhibitors target the CD44(high)/Id1(high) glioma-initiating cell population in human glioblastoma. Cancer Cell 18, 655–668 (2010). [ DOI ] [ PubMed ] [ Google Scholar ] 34. Zhu Y. et al. The long noncoding RNA glycoLINC assembles a lower glycolytic metabolon to promote glycolysis. Mol. Cell 82, 542–554.e6 (2022). [ DOI ] [ PubMed ] [ Google Scholar ] 35. Haderk F. et al. Focal adhesion kinase-YAP signaling axis drives drug-tolerant persister cells and residual disease in lung cancer. Nat. Commun 15, 3741 (2024). [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 36. Chokshi CR et al. Targeting axonal guidance dependencies in glioblastoma with ROBO1 CAR T cells. Nat. Med 30, 2936–2946 (2024). [ DOI ] [ PubMed ] [ Google Scholar ] 37. Lin B. et al. EGFR, the Lazarus target for precision oncology in glioblastoma. Neuro Oncol. 24, 2035–2062 (2022). [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 38. Huang L & Fu L Mechanisms of resistance to EGFR tyrosine kinase inhibitors. Acta Pharm. Sin. B 5, 390–401 (2015). [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 39. Dhaka B. et al. Functional identification of cis-regulatory long noncoding RNAs at controlled false discovery rates. Nucleic Acids Res. 52, 2821–2835 (2024). [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 40. Luo H. et al. HOTTIP-dependent R-loop formation regulates CTCF boundary activity and TAD integrity in leukemia. Mol. Cell 82, 833–851.e11 (2022). [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 41. Fierro C. et al. The long non-coding RNA NEAT1 is a deltaNp63 target gene modulating epidermal differentiation. Nat. Commun 14, 3795 (2023). [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 42. West JA et al. The long noncoding RNAs NEAT1 and MALAT1 bind active chromatin sites. Mol. Cell 55, 791–802 (2014). [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 43. Li L. et al. Transcriptional regulation of the Warburg effect in cancer by SIX1. Cancer Cell 33, 368–385.e7 (2018). [ DOI ] [ PubMed ] [ Google Scholar ] 44. Niu Z. et al. HBO1 catalyzes lysine lactylation and mediates histone H3K9la to regulate gene transcription. Nat. Commun 15, 3561 (2024). [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 45. Au YZ et al. KAT7 is a genetic vulnerability of acute myeloid leukemias driven by MLL rearrangements. Leukemia 35, 1012–1022 (2021). [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 46. Lv W. et al. Long noncoding RNA lncMREF promotes myogenic differentiation and muscle regeneration by interacting with the Smarca5/p300 complex. Nucleic Acids Res. 50, 10733–10755 (2022). [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 47. Lunde BM, Moore C & Varani G RNA-binding proteins: modular design for efficient function. Nat. Rev. Mol. Cell Biol 8, 479–490 (2007). [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 48. Chen Q. et al. Histone acetyltransferases CBP/p300 in tumorigenesis and CBP/p300 inhibitors as promising novel anticancer agents. Theranostics 12, 4935–4948 (2022). [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 49. Bell RJ et al. Cancer. The transcription factor GABP selectively binds and activates the mutant TERT promoter in cancer. Science 348, 1036–1039 (2015). [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 50. Mancini A. et al. Disruption of the β1L isoform of GABP reverses glioblastoma replicative immortality in a TERT promoter mutation-dependent manner. Cancer Cell 34, 513–528.e8 (2018). [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 51. Lasko LM et al. Discovery of a selective catalytic p300/CBP inhibitor that targets lineage-specific tumours. Nature 550, 128–132 (2017). [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 52. Su Z, Zhang Y, Tang J, Zhou Y & Long C Multifunctional acyltransferase HBO1: a key regulatory factor for cellular functions. Cell. Mol. Biol. Lett 29, 141 (2024). [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 53. Fang Z. et al. NOGOB receptor deficiency increases cerebrovascular permeability and hemorrhage via impairing histone acetylation-mediated CCM1/2 expression. J. Clin. Invest 132, e151382 (2022). [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 54. Kueh AJ et al. Stem cell plasticity, acetylation of H3K14, and de novo gene activation rely on KAT7. Cell Rep. 42, 111980 (2023). [ DOI ] [ PubMed ] [ Google Scholar ] 55. Xiao Y. et al. HBO1 is a versatile histone acyltransferase critical for promoter histone acylations. Nucleic Acids Res. 49, 8037–8059 (2021). [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 56. Reifenberger G, Wirsching HG, Knobbe-Thomsen CB & Weller M Advances in the molecular genetics of gliomas - implications for classification and therapy. Nat. Rev. Clin. Oncol 14, 434–452 (2017). [ DOI ] [ PubMed ] [ Google Scholar ] 57. Guo G. et al. A TNF-JNK-Axl-ERK signaling axis mediates primary resistance to EGFR inhibition in glioblastoma. Nat. Neurosci 20, 1074–1084 (2017). [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 58. Korovina I, Vehlow A, Temme A & Cordes N Targeting integrin α2 as potential strategy for radiochemosensitization of glioblastoma. Neuro Oncol. 25, 648–661 (2023). [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 59. Ahn BY et al. Glioma invasion mediated by the p75 neurotrophin receptor (p75(NTR)/CD271) requires regulated interaction with PDLIM1. Oncogene 35, 1411–1422 (2016). [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 60. Li Y. et al. TRAF4 maintains deubiquitination of caveolin-1 to drive glioblastoma stemness and temozolomide resistance. Cancer Res. 82, 3573–3587 (2022). [ DOI ] [ PubMed ] [ Google Scholar ] 61. Ding J. et al. EGFR suppresses p53 function by promoting p53 binding to DNA-PKcs: a noncanonical regulatory axis between EGFR and wild-type p53 in glioblastoma. Neuro Oncol. 24, 1712–1725 (2022). [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 62. Mao P. et al. Mesenchymal glioma stem cells are maintained by activated glycolytic metabolism involving aldehyde dehydrogenase 1A3. Proc. Natl Acad. Sci. USA 110, 8644–8649 (2013). [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 63. Park HS et al. Targeting YAP-p62 signaling axis suppresses the EGFR-TKI-resistant lung adenocarcinoma. Cancer Med. 10, 1405–1417 (2021). [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 64. Rinaldi C & Wood MJA Antisense oligonucleotides: the next frontier for treatment of neurological disorders. Nat. Rev. Neurol 14, 9–21 (2018). [ DOI ] [ PubMed ] [ Google Scholar ] 65. Goenka A. et al. Oncogenic long noncoding RNA LINC02283 enhances PDGF receptor A-mediated signaling and drives glioblastoma tumorigenesis. Neuro Oncol. 25, 1592–1604 (2023). [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 66. Bhat KPL et al. Mesenchymal differentiation mediated by NF-κB promotes radiation resistance in glioblastoma. Cancer Cell 24, 331–346 (2013). [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 67. Huang T. et al. A regulatory circuit of miR-125b/miR-20b and Wnt signalling controls glioblastoma phenotypes through FZD6-modulated pathways. Nat. Commun 7, 12885 (2016). [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 68. Gel B. et al. regioneR: an R/Bioconductor package for the association analysis of genomic regions based on permutation tests. Bioinformatics 32, 289–291 (2016). [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 69. Stark R & Brown G DiffBind: differential binding analysis of ChIP-seq peak data. R package version 100, 2–21 (2011). [ Google Scholar ] 70. Rosenbloom KR et al. ENCODE data in the UCSC Genome Browser: year 5 update. Nucleic Acids Res. 41, D56–D63 (2013). [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 71. Eley GD et al. A chromosomal region 7p11.2 transcript map: its development and application to the study of EGFR amplicons in glioblastoma. Neuro Oncol. 4, 86–94 (2002). [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] Associated Data This section collects any data citations, data availability statements, or supplementary materials included in this article. Supplementary Materials Supplementary Tables-2 NIHMS2161547-supplement-Supplementary_Tables-2.xlsx (298.3KB, xlsx) Supplementary Tables-1 NIHMS2161547-supplement-Supplementary_Tables-1.xlsx (2.2MB, xlsx) Raw_data_of_IB_and_DNA_gels NIHMS2161547-supplement-Raw_data_of_IB_and_DNA_gels.pdf (3.9MB, pdf) Data Availability Statement The short-read RNA-seq, long-read RNA-seq and ChIRP-seq data that support the findings of this study have been deposited in the Gene Expression Omnibus under accession codes GSE286179 , GSE286318 , GSE286319 , GSE286320 and GSE286432 . Previously published RNA-seq datasets for Mayo Clinic PDX models and NU cohort that were re-analysed here are available under accession codes PRJNA548556 (ref. 24 ), SRR8236743 (ref. 3 ) and GSE147352 (ref. 21 ). The Ivy GAP dataset was acquired from its portal ( https://glioblastoma.alleninstitute.org/ ) 28 . Publicly available datasets from the TCGA and CPTAC cohorts were obtained from the GDC Data Portal ( https://portal.gdc.cancer.gov/ ). All other data supporting the findings of this study are available from the corresponding author on reasonable request. Source data are provided with this paper. ACTIONS View on publisher site PDF (7.9 MB) Cite Collections Permalink PERMALINK Copy RESOURCES Similar articles Cited by other articles Links to NCBI Databases Cite Copy Download .nbib .nbib Format: AMA APA MLA NLM Add to Collections Create a new collection Add to an existing collection Name your collection * Choose a collection Unable to load your collection due to an error Please try again Add Cancel Follow NCBI NCBI on X (formerly known as Twitter) NCBI on Facebook NCBI on LinkedIn NCBI on GitHub NCBI RSS feed Connect with NLM NLM on X (formerly known as Twitter) NLM on Facebook NLM on YouTube National Library of Medicine 8600 Rockville Pike Bethesda, MD 20894 Web Policies FOIA HHS Vulnerability Disclosure Help Accessibility Careers NLM NIH HHS USA.gov Back to Top