Transcriptomic profiles and matched clinical information for GBM were retrieved from The Cancer Genome Atlas (TCGA; https://portal.gdc.cancer.gov), the Gene Expression Omnibus (GEO; https://www.ncbi.nlm.nih.gov/geo), the Chinese Glioma Genome Atlas (CGGA; http://www.cgga.org.cn) and the Ivy Glioblastoma Atlas Project (Ivy GAP; https://glioblastoma.alleninstitute.org). Bulk-RNA-seq datasets (TCGA-GBM, CGGA-325, CGGA-693, GSE43378, GSE13041, GSE74187 and GSE83300) and a single-cell RNA-sequencing (scRNA-seq) dataset (GSE182109) were downloaded. Ensembl gene identifiers were converted to gene symbols, clinical variables were matched to each sample, and all datasets were merged in R (v 4.1.2). ECHDC2 was identified by intersecting genes with significant prognostic value in Kaplan–Meier analysis across the GSE83300, GSE43378, GSE74187, and GSE13041 datasets, followed by subsequent analyses (Supplementary Figure S1).

scRNA-seq data were processed with Seurat package (v 4.3.0.1) (Hao et al., 2021). Quality-control filters were applied to retain cells with 200 < nFeature_RNA <4 000, nCount_RNA <30 000 and percent. mt < 20%. Batch effects were corrected with Harmony package (v 0.1.1). Neighbour graphs and clusters were generated with FindNeighbors and FindClusters, respectively, and visualised by Uniform Manifold Approximation and Projection (UMAP). Cell types were annotated on the basis of canonical marker genes.

Time-dependent receiver-operating-characteristic (ROC) curves were generated with “timeROC” package (v 0.4). Patients were stratified into high- and low-ECHDC2 groups according to the median ECHDC2 mRNA level. Overall survival was compared by Kaplan–Meier analysis using “survival” package (v 3.7–0).

Relative immune-cell infiltration was estimated by single-sample gene-set enrichment analysis (ssGSEA) implemented in “GSVA” package (v 1.48.3) (Hänzelmann et al., 2013). Absolute immune-cell fractions were inferred with CIBERSORT (Newman et al., 2015), which distinguishes 22 immune-cell subsets on the basis of 547 marker genes. Stromal, immune and ESTIMATE scores of the tumor micro-environment (TME) were calculated with “ESTIMATE” package (v 1.0.13) (Yoshihara et al., 2013). Correlations between ECHDC2 expression and cytokines, chemokines or immune checkpoints were evaluated by Spearman’s test and Wilcoxon rank-sum test (two-tailed; p < 0.05). Intersections across datasets were depicted with Venn diagrams.

A protein-protein interaction (PPI) network centered on ECHDC2 was constructed using the STRING database (Szklarczyk et al., 2021) (http://www.stringdb.org/). Based on the CGGA-325 RNA-seq dataset, we identified differentially expressed genes (DEGs) between high and low ECHDC2 expression groups using the limma package (v 3.56.2) in R (v 4.1.2), with the median expression level of ECHDC2 as the cutoff. The specific screening criteria were: |log2FoldChange| > 1 and adjusted P-value < 0.05. The upregulated and downregulated DEGs were merged into a single gene set for subsequent analyses (Supplementary Table S1). Gene Ontology (GO) enrichment analysis, Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment analysis, and Gene Set Enrichment Analysis (GSEA) were performed using the “clusterProfiler” package (v 4.8.3) (Wu et al., 2021).

ECHDC2 mRNA expression across cancer types was obtained from the Cancer Cell Line Encyclopedia (CCLE) (https://sites.broadinstitute.org/ccle/) and the TIMER web server (https://cistrome.shinyapps.io/timer) (Li T. et al., 2017). Normal- and tumor-tissue RNA-seq data were downloaded from TCGA. Kaplan-Meier curves and univariate Cox regression were generated with “survival” package (v 3.7–0) and visualised with “survminer” package (v 0.5.0) to evaluate overall survival (OS), disease-specific survival (DSS), disease-free interval (DFI) and progression-free interval (PFI).

Human GBM U251 and A172 cells (Cell Bank, Chinese Academy of Sciences, Shanghai, China) were cultured in DMEM (C119955008BT, Gibco) supplemented with 10% fetal bovine serum (A5256701, Gibco) at 37 °C in a humidified 5% CO₂ incubator. Three small interfering RNAs (siRNAs) targeting ECHDC2 (si-ECHDC2-1: 5′-GGC​CGA​CGA​CUG​AGU​GGA​A-3′; si-ECHDC2-2: 5′-GGA​AUG​UCU​UCG​UCA​GUG​A-3′; si-ECHDC2-3: 5′-CGU​GUU​CUG​UGC​AGG​UGC​A-3′) were synthesized by OBiO (Shanghai, China). Transfections were performed with Lipofectamine 3000 (L3000-015, Invitrogen) in Opti-MEM (31985070, Gibco).

Total RNA was extracted with the SteadyPure Rapid RNA kit (AG21023, Accurate Biology) and reverse-transcribed using EvoScript RT premix (AG11706, Accurate Biology). The qRT-PCR was performed with a SYBR Green Pro Taq HS kit (AG11718, Accurate Biology). Relative expression was calculated by the 2^-ΔΔCT method, normalizing to β-actin. The sequences of primers were as follows: ECHDC2, Forward primer 5′-AGT​GCG​TGT​CCT​GCT​CTT​C-3′, Reverse primer 5′-CAT​CTG​TTC​CCG​CTC​CTT​CA-3’; β-actin, Forward primer 5′-CCT​TTG​CCG​ATC​CGC​CG-3′, Reverse primer 5′-AAT​CCT​TCT​GAC​CCA​TGC​CC-3’. AKT, Forward primer 5′-CTG​CAC​AAA​CGA​GGG​GAG​TA-3′, Reverse primer 5′-GCG​CCA​CAG​AGA​AGT​TGT​TG-3’.

Cells were lysed in SDS buffer containing protease and phosphatase inhibitors (HY-K0010 and HY-K0021, MCE). Proteins were separated on 10% SDS-PAGE gels and transferred to PVDF membranes (IPVH00010, Merck). After blocking (PS108P, Yamei), membranes were incubated overnight at 4 °C with anti-ECHDC2 (26126-1-AP, Proteintech; 1:1000), phosphorylated AKT (p-AKT, Ser473 phosphorylation) antibody (AF0016, Affinity; 1:1000), AKT antibody (AF6261, Affinity; 1:1000), or anti-β-actin (4970S, Cell Signaling Technology; 1:2000), followed by HRP-conjugated secondary antibody (7074S, Cell Signaling Technology; 1:2000) and chemiluminescence detection.

The CCK-8 Cell Counting Kit (C6005, NCM Biotech) was used to measure the proliferation ability of GBM cells. For CCK-8 assays, 100 ul of medium containing 1*10³ cells was added to each well of a 96-well plate. Before measuring, the medium in the wells to be tested was replaced with fresh medium containing 10% CCK-8, and then placed in an incubator for 2 h. Then a microplate reader was used to measure the OD of the wells to be tested at a wavelength of 450 nm.

Edu assay was performed using Edu kit (C0075S, Beyotime), according to the instructions. The EdU solution was prepared as a 1:1000 concentration of medium and added to the pre-prepared 12-well plates with 500 ul per well, which were incubated at 37 °C for 2 h and fixed in 4% paraformaldehyde for 15 min. The fixed cells were incubated with 100 ul of reaction solution for 30 min sheltered from light, and then the nuclei were stained for 10 min by using DAPI. Cell proliferation was imaged by utilization of a 20x fluorescence microscope.

Cell migration and invasion assays were performed using 6.5 mm Transwell chambers with 8.0 μm pore polycarbonate membrane inserts (3422, Corning). For the migration assay, 650 μL of medium containing 10% FBS was added to the lower chamber, and then 200 μL of serum-free cell suspension (2*10⁴ cells) was seeded into the upper chamber. The cells were incubated at 37 °C for 24 h, followed by fixation with 4% paraformaldehyde and staining with crystal violet solution. The stained cells in the upper chamber were gently removed, and the cells that migrated to the underside of the membrane were imaged and counted under a light microscope.

For the invasion assay, before seeding the cells (4*10⁴ cells) into the upper chamber, the membrane of the upper chamber was coated with 60 μL of Matrigel (diluted 1:8, 354234, Corning). The remaining steps were identical to those of the migration assay.

Confluent monolayers were scratched with a sterile 200 µL pipette tip held perpendicular to the plate. After washing away debris, cells were incubated in 2% FBS medium, and wound closure was photographed at 0 h and 24 h. Migratory distance was quantified in ImageJ (v 1.53q).

Data are presented as mean ± standard deviation. Two-group comparisons were performed with unpaired two-tailed Student’s t-tests; multiple-group comparisons used one-way ANOVA followed by Tukey’s post hoc test. p < 0.05 was considered statistically significant. Statistical analyses were conducted in R (v 4.1.2) or GraphPad Prism (v 9.0.0).

Across CGGA, GEO, TCGA and Ivy GAP cohorts, ECHDC2 mRNA was markedly elevated in WHO grade III and grade IV gliomas (Figures 1A,G,K). Higher ECHDC2 expression correlated with older age (Figure 1B), primary rather than recurrent tumors (Figure 1C), an unmethylated MGMT promoter (Figure 1D) and 1p/19q non-codeletion (Figure 1H). Expression was significantly greater in IDH-wild-type GBM than in IDH-mutant tumors (Figures 1E,I,M) and highest in GBM relative to lower-grade gliomas (Figures 1F,J,L). Within molecular or anatomical subclasses, ECHDC2 was enriched in the TCGA Classical subtype (Figure 1N) and in pseudopalisading necrosis zones (PNZ; Figure 1O). These observations indicate that ECHDC2 expression increases with tumor grade and is prevalent in key GBM niches. Notably, Single-cell RNA-seq data (GSE182109) identified six principal cell clusters (Figure 2A); ECHDC2 was broadly expressed in malignant cells, astrocytes and myeloid cells (Figure 2B).

Patients in CGGA-325, GSE13041, GSE43378, GSE74187 and GSE83300 were stratified by the median ECHDC2 level. Kaplan-Meier analysis demonstrated that high ECHDC2 expression conferred shorter OS in CGGA-325 (p = 0.003; Figure 2C) and in the four GEO cohorts (p = 0.009–0.049; Figures 2D–F,H). Progression-free survival (PFS) was likewise reduced (p = 0.006; Figure 2G). Time-dependent ROC curves yielded area-under-the-curve (AUC) values of 0.688, 0.640 and 0.628 at 1, 2 and 3 years in CGGA-325 (Figure 2I), with comparable AUCs (0.600–0.922) in GEO datasets (Figures 2J–N). Thus, elevated ECHDC2 is a robust adverse prognostic marker in GBM.

Across five bulk RNA-seq datasets, Stromal, Immune, and ESTIMATE scores were each positively correlated with ECHDC2 expression (Figures 3A–C; Supplementary Figures 2A, B). These findings indicate that patients with high ECHDC2 expression exhibit a highly active TME, increased immune infiltration, and thus may be more sensitive to immunotherapy. Accordingly, we stratified analyses of cytokines (Figure 3D; Supplementary Figures 2C–F), immune checkpoints (Figure 3E; Supplementary Figures 2G–J), and infiltrating immune cells by ECHDC2 expression levels (Figures 3F,G; Supplementary Figures 2K–R).

Across the five bulk datasets, ECHDC2 expression was positively correlated with the cytokine leukemia inhibitory factor (LIF) (Figures 4A,E) and the immune checkpoint pair TNFSF14/TNFRSF14 (Figures 4B,F,G), suggesting an immunomodulatory role for ECHDC2. CIBERSORT analysis identified neutrophils as the most frequently intersecting immune cell population among 22 immune subsets (Figure 4C), although neutrophil abundance was not significantly correlated with ECHDC2 expression (Figure 4H). ssGSEA results revealed higher enrichment levels of effector-memory CD8⁺ T cells (TEM) and plasmacytoid dendritic cells (pDCs) in tumors with high ECHDC2 expression (Figure 4D), and both of these immune cell populations were positively correlated with ECHDC2 (Figures 4I,J).

A protein–protein interaction (PPI) network constructed using STRING localized ECHDC2 to a metabolic hub, with ACAA1, HADH, HSDL2, SCP2, and HMGCL as its core interacting partners (Figure 5A). KEGG and GSEA enrichment analyses indicated that ECHDC2 may contribute to the promotion of GBM progression via the PI3K/Akt signaling pathway (Figures 5B,C). GO enrichment analyses emphasized the potential involvement of ECHDC2 in the biological processes of gliogenesis, glial cell differentiation, and glial cell development (Supplementary Figure S3A).

To elucidate the functional role of ECHDC2 in GBM cells, we knocked down or overexpressed ECHDC2 in GBM cells by transfecting ECHDC2-targeting siRNAs or ECHDC2-overexpressing plasmids, respectively. The knockdown and overexpression efficiencies of ECHDC2 were detected via Western blotting and qRT-PCR, respectively, in both U251 and A172 cells (Figures 5D,E). The ECHDC2-targeting siRNA #2, which exhibited the highest knockdown efficiency, was selected for subsequent functional experiments. The level of p-AKT was significantly decreased in ECHDC2-knockdown cells, while it was increased in ECHDC2-overexpressing cells, with no obvious alterations in total AKT expression (Figures 5F,G; Supplementary Figure S3B). Compared with cells in the negative control (NC) group, GBM cells with ECHDC2 knockdown showed significantly inhibited migration and invasion capacities, whereas ECHDC2-overexpressing cells exhibited the opposite effects. These findings were validated by both wound-healing and Transwell assays (Figures 6A,B). Furthermore, ECHDC2 knockdown suppressed the proliferative capacity of GBM cells, while ECHDC2 overexpression exerted the opposite effect (Figure 7A). Consistent results were obtained in the EdU assay (Figure 7B).

CCLE data indicated ubiquitous ECHDC2 transcription across normal tissues (Supplementary Figure S4A). TCGA analysis showed reduced ECHDC2 in 14 cancers—including BRCA, CHOL and COAD—and increased expression in PRAD (Supplementary Figure S4B).

Univariate Cox regression across 32 tumors linked ECHDC2 to OS and DSS in nine cancers (Supplementary Figure S4C). ECHDC2 acted as a high-risk gene in THYM and LGG, but as a protective factor in bladder, head-and-neck, kidney, liver, lung, mesothelioma and pancreatic cancers. Kaplan–Meier curves mirrored these trends (Supplementary Figures S4F–N). For DFI, ECHDC2 conferred higher risk in COAD and PRAD (Supplementary Figures S4D,E). PFI analysis identified favorable associations in BLCA, HNSC, KIRC, KIRP, LUAD, MESO and PAAD, but an adverse impact in LGG and PRAD (Supplementary Figures S5A–S).

Pan-cancer immune deconvolution showed positive correlations between ECHDC2 and B cells, follicular helper T (TFH) cells and natural-killer (NK) cells, and negative correlations with macrophages, myeloid-derived suppressor cells (MDSC) and cancer-associated fibroblasts (CAF) (Supplementary Figure S6A).