B7-H3 interaction with cMet enhances lipolysis to drive malignant progression of glioma via pSTAT3-ATGL axis

Abstract

Background

Although metabolic dysregulation serves as a pivotal hallmark in glioma, the detailed information about the lipid metabolism regulation is largely unknown. In the context of this investigation, we have identified B7-H3 — an immune checkpoint molecule of the B7 family — as a regulator of lipid metabolism in glioma.

Methods

A series of enrichment analyses were applied to explore the interaction between metabolic reprogramming and prognosis in glioma. The expression and prognostic value of B7-H3 were detected in both public databases and clinical samples. Furthermore, the effect of B7-H3 on Glioma cells malignance and lipid metabolism was evaluated through in vivo and in vitro functional experiments. Finally, the molecular mechanism of B7-H3 was investigated by the study of transcriptome sequencing and key interacting proteins.

Results

The metabolism associated genes (MAG) were firstly screened from the prognosis related differentially expressed genes (DEG) using the TCGA and verified CGGA derived primary and recurrent cohorts of glioma patients, and the immune checkpoint molecule CD276 (B7-H3) was selected due its significant poor prognosis correlation in both primary and recurrent glioma. Furthermore, both RNA sequencing and untargeted metabolomics profiling confirmed a strong correlation between B7-H3 and lipid metabolism in glioma. Notably, B7-H3 showed a significant positive correlation with lipolytic enzyme ATGL (PNPLA2) according to the above transcriptomic data and CGGA database of glioma. However, the receptors of B7-H3 on cancer cells remained poorly understood. Here, we reported that B7-H3 modulated abnormal lipolysis during glioma progression by forming a complex with cMet, and subsequent simultaneous promoting the phosphorylation of STAT3 and expression of ATGL. More importantly, B7-H3 protein level was associated with poor prognosis and correlated with ATGL status in glioma.

Conclusions

Our results define B7-H3 as a lipolysis related factor that acts by driving glioma malignancy via the cMet-pSTAT3-ATGL pathway, thereby nominating both B7-H3 and ATGL as potential therapeutic targets for glioma.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12967-026-08151-9.

Introduction

Glioma, particularly glioblastoma (GBM), exhibits high heterogeneity and an invasive nature, leading to poor prognosis and high recurrence rates. Except previously confirmed mechanisms, including enhanced tumor heterogeneity [1], activation of tumor stem cells [2], immune microenvironment remodeling, and abnormal epigenetic regulation [3], involved in the regulation of glioma progression and recurrence, metabolic abnormality is also an important feature of poor prognosis in glioma. The cholesterol metabolism dysregulation resulted by the m6A-regulatory protein, hnRNPA2B1, could contribute to glioma stemness and malignancy [4]. Moreover, a recent study uncovered that glioma cells could generate lactate through aerobic glycolysis, thereby forming an acidic microenvironment resulted in CD8⁺T cells inhibition and enhanced activity of immunosuppressive cells (such as MDSCs and Tregs) [5]. It is considered that the synergistic interaction between the high-throughput glycolysis and glutaminolysis pathways maintain ATP content via the aerobic fermentation of both glucose and glutamine and promote malignant glioma growth [6]. Consequently, the identification of critical genes governing metabolic reprogramming may constitute a novel research avenue with direct implications for glioma treatment and patient prognosis.

B7-H3 is a B7 family derived immune checkpoint molecule, which is aberrantly overexpressed in various tumors, including glioma. Previous studies have shown that B7-H3 is associated with overall survival (OS) in glioma patients [7] and can drive tumor progression through the dual axis of immune suppression-metabolic reprogramming. It is reported that B7-H3 can mediate glucose competition between cancer cells and CD8⁺T cells through the PI3K/Akt/mTOR signaling pathway [8]. Moreover, B7-H3 can stimulate glycolysis and mitochondrial activity in non-tumor adipocyte progenitor cells, while B7-H3 knockout mice exhibit spontaneous obesity, metabolic dysfunction, and adipose tissue inflammation. Therefore, targeting B7-H3 and related pathway has opened up potential new avenues for metabolic diseases treatment [9]. However, the role of B7-H3 in metabolic regulation within glioma remains unexplored. Given the acknowledged critical impact of metabolic dysregulation in glioma progression, targeting B7-H3 may present a novel perspective for therapeutic intervention.

In the present study, we systematically investigated the role of B7-H3 through bioinformatics-based screening, experimental validation, and mechanistic exploration. This integrated approach revealed a novel regulatory mechanism of B7-H3, providing new insights into its metabolic regulation in gliomas.

Materials and methods

Patients and clinical samples

Expression levels of B7-H3 and ATGL were analyzed in 80 glioma patients who were through surgical procedure and confirmed by pathology department at the First Affiliated Hospital of Soochow University (Suzhou, China) between January 2013 and December 2019. Additionally, the clinic-pathologic characteristics of the patients and the follow-up details with different B7-H3 or ATGL expression were described in Table S1. All clinical samples were obtained with patients’ informed consent and approved by the Ethics Committee of The First Affiliated Hospital of Soochow University.

Cell lines

The glioma cell line DBTRG-05MG (05MG) was obtained from Cell Bank of Shanghai Institute for Biological Sciences, Chinese Academy of Sciences (Shanghai, China). The glioma cell line U87MG was purchased from the Procell Life Science & Technology Co., Ltd., which verified the authenticity of cells by STR profiling. All cell lines were tested negative for Mycoplasma contamination and cultured according to the supplier’s instructions.

Lentivirus and small interfering RNA

The glioma cell U87MG was transfected with the lentiviral vector GV492-puro encoding the human B7-H3 mRNA sequence (NM_001024736.2) and designated as B7-H3, while the empty vector was used as a negative control and designated as EV. Furthermore, to establish stable cell lines with target gene knockout, sgRNAs against B7-H3 were synthesized by Applied Biological Materials Inc (abm, Jiangsu, China) and designated as sgB7-H3. The empty vector was used as a negative control and designated as sgCtrl. Details were shown in Table S2. Subsequently, U87MG sgB7-H3 cells with ATGL overexpression was achieved via the GV358-puro lentiviral vectors encoding the human PNPLA2 gene and designated as ATGL. All the lentiviral vectors used in this study were synthesized by Genechem (Shanghai, China). Moreover, the small interfering RNA (siRNA) targeting specific human B7-H3 sites (siB7-H3#1 and siB7-H3#2), and negative control siRNA (siNC) were synthesized by GenePharma (Shanghai, China; Table S3).

Western blotting (WB) analysis

The glioma cells were homogenized in RIPA lysis buffer (#P0013C, Beyotime Biotechnology, Shanghai, P.R.C) with protease inhibitor cocktail (#P1005, Beyotime Biotechnology, ×100), and protein concentrations were determined by BCA assay (#P0010, Beyotime Biotechnology). Furthermore, the protein extracted from glioma cells were resolved by SDS-polyacrylamide gels and transferred to PVDF membranes (#IPVH00010, 0.45 μm, Millipore, Billerica, MA, USA). After that, the PVDF membranes were incubated overnight at 4 °C with the primary antibodies against B7-H3 (#14453-1-AP, 1:1,000, Proteintech Group, Inc), cMet (#8198, 1:1,000, Cell Signaling Technology, Danvers, MA, USA), phospho-Met (#3077, Tyr1234/1235, 1:1,000, Cell Signaling Technology), STAT3 (#9139, 1:1,000, Cell Signaling Technology), phospho-STAT3 (#9145, Tyr705, 1:1,000, Cell Signaling Technology), ATGL (#ab207799, 1:500, Abcam, Cambridge, MA, USA), and GAPDH (#10494-1-AP,1:10,000, Proteintech Group, Inc), followed by HRP-conjugated secondary antibodies (#7074, Anti-rabbit IgG, HRP-linked Antibody; #7076, Anti mouse IgG, HRP-linked Antibody, Cell Signaling Technology) for 1 h at RT. Finally, the bands were visualized by the enhanced Pierce ECL Western blotting substrate kit (Thermo Scientific/Pierce, Rockford, IL, USA), and quantified via ImageJ.

Chemo-resistance assays

Firstly, 5,000 glioma cells were seeded into 96-well plates and incubated overnight at 37 °C. The cells were then treated with 5-Fluorouracil (5-FU, #HY-90006, MCE) for 48 h. Thereafter, the glioma cells were incubated with Cell Counting Kit-8 (CCK-8, Dojindo, Japan) for 4 h at 37 °C and the cell viability was determined by enzyme immunoassay analyzer (Bio-Rad, CA, USA) according to the manufacturer’s instructions. Five replicate wells were included in each analysis, and at least three independent experiments were conducted.

Dual-color immunofluorescence

The immunofluorescence imaging of glioma cells was fulfilled with Tyramide signal amplification system (#RS0035, Immunoway, TX, US). Briefly, cells were fixed in 4% paraformaldehyde, permeabilized with 0.1% Triton X-100 and blocked with 5% BSA. After blocking, cells were incubated with primary antibodies specific for rabbit anti-B7-H3(#14453-1-AP, Proteintech Group, Inc), and mouse anti-cMet (#8741, Cell Signaling Technology) overnight at 4 °C. Furthermore, incubation of HRP-conjugated secondary antibodies and Tyramide signal amplification system were separately carried out for 30 min and 10 min at room temperature. Finally, DAPI (#D212, Dojindo Laboratories, Kumamoto, Japan) was then used for staining the nuclei, and images were acquired with Pannoramic MIDI II (3DHISTECH Ltd).

Lipid droplets (LDs) visualization

For Lipid droplets visualization, the fixed glioma cells were rinsed 3 times with PBS and stained with BODIPY 493/503 (#D3922, Invitrogen) for 30 min at RT. After nuclei staining with DAPI, images were acquired with a confocal microscope (OLYMPUS IX83, Tokyo, Japan).

Immunohistochemistry

Hematoxylin and eosin (HE) staining and immunohistochemistry (IHC) were performed on paraffin-embedded tissues using standard protocols as we previously described [10]. In brief, a pressure cooker was used for antigen retrieval using 0.01 M citrate buffer (pH 6.0). Antibodies specific to B7-H3 (#14453-1-AP, Proteintech Group, Inc) and ATGL (#ab207799, Abcam) were used in this study. Moreover, the immunodetection was performed on the following day using DAB (Dako, Carpinteria, CA) according to the manufacturer’s instructions. Finally, immunostaining scores were evaluated as described previously [11]. Slides were scanned using Pannoramic MIDI II (3DHISTECH Ltd) and images were shown through Case Viewer software.

Co-immunoprecipitation analysis

The rProtein A/G Magnetic IP/Co-IP Kit (#abs9649, Absin Bioscience Inc, PRC) was employed to assess the interaction between B7-H3 and cMet in glioma cells. In brief, U87MG cells were harvested using 1×Lysis/Wash Buffer (Enhanced, Absin Bioscience Inc). Furthermore, the cell lysate was immunoprecipitated with anti-B7-H3 antibody (#66481-1-Ig, Proteintech Group, Inc), separated by SDS-PAGE and subjected to WB analysis with anti-cMet antibody (#240085, ZEN-BIOSCIENCE). Meanwhile, the cell lysate was also immunoprecipitated with anti-cMet antibody (#YM8428, Immunoway), and subjected to WB analysis with anti-B7-H3 antibody (#14453-1-AP, Proteintech Group, Inc).

Cleavage under target and release using Nuclease (CUT&RUN) assays

CUT&RUN assays were conducted using the pSTAT3 (#9145, Tyr705, Cell Signaling Technology) antibody in accordance with the manufacturer’s protocol based on the pA/G-MNase (Hyperactive pA/G-MNase CUT&RUN Assay Kit for PCR/qPCR, HD103, Vazyme). Co-immunoprecipitated DNA was detected by quantitative PCR using primers specific to the PNPLA2 (ATGL) promoter region (Table S4).

Animal experiments

The 6-week male athymic BALB/c nude mice were purchased from the Shanghai Lingchang Biotechnology Co.,Ltd, and were housed in a specific-pathogen free (SPF) facility under a12-hour light/dark cycle at 21–23^◦C, with sterile food and water provided. All animal experiments were approved by the Animal Care and Use Committee of Soochow University (SUDA20250507A05). To establish xenograft tumors, 5 × 10⁶ sgB7-H3 or sgCtrl glioma cells U87MG were suspended at a 1:1 ratio in 100µL of complete medium and Matrigel (#356234, BD Biosciences), and delivered via subcutaneous injections into BALB/c nude mice. The tumor formation was monitored for 3 weeks, and 5 mice were in each experimental condition. For establishing intracranial xenograft tumor models, nude mice were anesthetized by intraperitoneal injection of ketamine/xylazine according to previous description [12]. Briefly, 5 × 10⁵ glioma cells in 5 µL culture medium were stereotactically injected into specific brain regions, followed by monitoring via bioluminescence imaging (BLI) to track growth and progression. At 3 weeks post orthotopic injection, the brains were removed and fixed in formalin for subsequent analysis. Here, at least 3 mice were in each experimental condition, and three independent experiments.

RNA-seq and data analysis

Total RNA of U87MG sgB7-H3 and sgCtrl cells were extracted using Trizol (Invitrogen). Library construction, quality control, and sequencing procedures were performed by CapitalBio Technology Co., Ltd. RNA-seq raw data had been submitted to the ‌Genome Sequence Archive (GSA)‌ database (HRA014343). Gene expression was quantified feature Counts software in a Linux Server, and DEGs were identified via DESeq2 (version 1.36.0) with stringent criteria: ∣log2FC∣>1, FDR < 0.05. Moreover, the functional GO/KEGG enrichment analysis cluster Profiler of DEGs were also performed. For a deeper understanding of the biological pathways linked to these differentially expressed genes, gene set enrichment analysis (GSEA) was performed utilizing the GseaVis package [13]. Furthermore, the Immuno-Oncology Biological Research (IOBR) package was applied to decode the tumor microenvironment and signatures [14].

Untargeted metabolomics

Briefly, U87MG sgB7-H3 and sgCtrl cells were snap-frozen in liquid nitrogen, homogenized, and extracted with cold methanol-water (volume ratio: 80:20). Untargeted metabolomics was performed using Liquid Chromatograph Mass Spectrometer (LC-MS) by CapitalBio Technology Co., Ltd. MS data had been acquired in both positive and negative ESI modes via Q-TOF m/z50–1000. Metabolites were annotated via HMDB and METLIN databases masstolerance < 5ppm. Statistical analyses PCA, OPLS − DA and pathway mapping were conducted in MetaboAnalyst and KEGG with stringent criteria: VIP value > 1.0, p-value < 0.05 (t-test or non-parametric test), fold change > 2.

Gene screening and verification analysis

A total of 695 cases of lower grade glioma (LGG) and glioblastoma (GBM) derived RNA-seq data were downloaded from TCGA for DEGs analysis. Then, these DEG were further employed for prognosis analysis via combining the survival data. Prognosis related DEGs from LGG and GBM samples were selected through intersection. Top 20 of prognosis related DEG were further selected using machine learning method random forest, followed by intersection with Metabolism associated genes (MAGs) obtained from Genecards (https://www.genecards.org/) for candidate genes screening. Moreover, the prognosis validation of these candidate genes was also performed using two cohorts with 325 and 693 GBM patients from CGGA (https://www.cgga.org.cn/analyse/RNA-data.jsp). All the analyses were performed using R (version 4.3.1).

Statistical analysis

All statistical values were measured using SPSS 26.0 software (Chicago, IL, USA) or GraphPad Prism 9.0. Data were presented as mean ± standard deviation (SD). For parametric data, the two-tailed Student’s t-test was used to determine statistical significance between two groups, while the one-way ANOVA test was used in multiple comparisons. For nonparametric data, the Wilcox test was used to determine statistical significance between two groups. The analysis of co-expression of B7-H3 and ATGL in glioma patients with different tumor grades or survival status was performed using the Chi-squared test. In addition, the Pearson correlation test and Kaplan–Meier analysis were separately employed for correlation analyses and survival curves. p values were designated as follows: *, p < 0.05; **, p < 0.01; ***, p < 0.001; ****, p < 0.0001.

Results

B7-H3 was confirmed as prognosis and metabolism-related DEG in glioma

To explore the genes involved in metabolic reprogramming and prognosis in glioma, the following roadmap analyses were performed. As shown in Fig. 1A, firstly, 4490 DEGs was identified through differential analyses using LGG and GBM datasets from the TCGA database (Fig. S1A). Subsequently, prognosis-associated genes were further identified (3091 genes for LGG and 407 genes for GBM) via combining analyses with survival data. Finally, a total of 306 shared prognosis-associated candidate genes were obtained and the top 20 clinical related genes were confirmed using machine learning method random forest (Fig. S1B, S1C). Among these 20 genes, 7 genes were confirmed as the metabolism-related genes via intersection analysis with metabolism gene set, and they were IGFBP2, PLAUR, PTX3, CD276(B7-H3), HEXB, PLOD1, and TGFBI. (Fig. 1A). To validate their prognostic relevance in primary and recurrent gliomas, we analyzed these seven genes using the CGGA derived datasets. According to the results, all 7 genes showed prognosis significance using the 325-cohort dataset (Fig. S1D). Moreover, only B7-H3 exhibited a significant correlation with poor prognosis in recurrent patients (Fig. 1B). Further immunohistochemical analysis of B7-H3 expression in 80 glioma patients revealed an ascending trend in tumor tissues with increasing malignancy grade (Fig. 1C and D). Meanwhile, elevated B7-H3 levels strongly correlated with survival status and unfavorable prognosis (Fig. 1E and F, Table S1). Taken together, B7-H3 could serve as metabolism related prognosis genes in glioma, and its aberrantly overexpressed in glioma tissues with positive correlated with tumor grade, death risk and poor prognosis.

Fig. 1.

Identification and validation of prognosis and metabolism-related DEG in glioma. (A) Flow chart of DEG screening from prognosis and metabolism angles. Briefly, TCGA derived LGG and GBM data were employed prognosis related DEG analysis. Machine learning method random forest were used for metabolism related screening using above prognosis related DEGs. Finally, seven genes (IGFBP2, PLAUR, PTX3, CD276(B7-H3), HEXB, PLOD1, TGFBI) were identified. (B) According to the survival information from CGGA derived datasets, B7-H3 was the only prognosis related DEGs in patients with glioma recurrence. Kaplan-Meier curves were used for prognosis evaluation and the p values calculated via log-rank test were shown. (C) IHC analysis of the expression of B7-H3 in glioma tissues from different Tumor grades. Scale bars 20 μm. (D) Violin plots of B7-H3 expression in glioma patients with different Tumor grade: Grade 1/2, green plot; Grade 3, orange plot; Grade 4, purple plot. p value was calculated via Wilcox test. (E) Violin plots of glioma patient survival in the different state: Live, green plot; Dead, orange plot. p value was calculated via Wilcox test. (F) Kaplan–Meier plots of overall survival of 80 glioma patients in this research. Among them, 12 patients exhibit abnormally high expression of B7-H3, while 68 patients have low expression of B7-H3. X-axis represented the timeline of patient survival, Y-axis represented the survival fraction (the hazard ratio (HR): 3.09, 95% CI: 1.29–7.42), and P value was calculated by log-rank test

B7-H3 promotes glioma malignant transformation by regulating lipid metabolism

Chemoresistance in glioma is a key factor contributing to poor prognosis. Our study revealed that knockout or silencing of B7-H3 expression in glioma cells significantly inhibited their chemoresistance, and vice versa (Fig. 2A and C). Additionally, B7-H3 expression levels markedly influenced glioma cell migration, invasion, and colony formation capabilities (Fig. S2). Both orthotopic and subcutaneous xenograft models demonstrated that B7-H3 knockout substantially suppressed tumor formation in glioma-bearing mice (Fig. 2D and E). RNA sequencing and KEGG pathway enrichment analysis of U87MG sgCtrl and U87MG sgB7-H3 cells highlighted prominent alterations in metabolic pathways (Fig. 2F). Importantly, we subsequently confirmed that the genes associated with the lipid metabolism pathway were suppressed after B7-H3 knockout via GSEA (Fig. 2G). Consistently, untargeted metabolomics profiling further confirmed a strong correlation between B7-H3 and lipid metabolism in glioma models (Fig. 2H and I). These findings collectively suggest that dysregulated lipid metabolism mediates B7-H3-driven malignant transformation in glioma.

Fig. 2.

B7-H3 promoted the malignancy of Glioma via regulating abnormal lipid metabolism. (A) Immunoblotting validated the downregulation or overexpression of B7-H3 in glioma cells. U87MG cells were transduced with CRISPR-Cas9 or overexpression vector of B7-H3, while 05MG cells were transduced with siRNA for B7-H3. (B, C) B7-H3 enhanced drug sensitivity of glioma. U87MG cells transduced with CRISPR-Cas9 or overexpression vector of B7-H3 were applied for analyzing the sensitivity of 5-FU (B). 05MG cells transduced with siRNA for B7-H3 were applied for analyzing the sensitivity of 5-FU (C). Data represent means ± SD. ***P < 0.001, ****P < 0.0001; compared with sgCtrl, LVcon or siNC group by two-tailed Student’s t-test or multiple t-tests. (D) B7-H3 knockout significantly inhibited the tumor-forming capacity in orthotopically inoculated glioma model mice. We established an orthotopic tumor model by lateral ventricular injection of sgCtrl and sgB7-H3 U87MG cells in Nude mice.5 × 10⁵ cells per injection, n = 3 mice per group. The tumor relative area was calculated via ImageJ. Data represent means ± SD. **P < 0.01; compared with sgCtrl group by two-tailed Student’s t-test. (E) Representative images of tumor. Subcutaneous growth of sgCtrl and sgB7-H3 U87MG cells in Nude mice. 5 × 10⁶ cells per injection, n ≥ 5 mice per group. The primary tumor growth rate in different experimental groups was measured once per week. Experiments were performed in triplicate. ****P < 0.0001; P-values are determined by two-tailed Student’s t-test. (F) The top 20 biological process (BP) terms, cellular component (CC) terms, and molecular function (MF) terms in the enrichment analysis between sgCtrl and sgB7-H3 U87MG cells. (G) Gene set enrichment analysis (GSEA) was conducted on the differentially expressed genes between sgCtrl and sgB7-H3 as detected by RNA-Seq (H) The enriched Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway terms of DEGs for metabolism between sgCtrl and sgB7-H3 U87MG cells. (I) Clustering analysis based on untargeted metabolomic sequencing revealed that the metabolic differences were predominantly concentrated in lipid metabolism

B7-H3 modulates lipid metabolism and promote glioma progression via regulating the lipolytic enzyme ATGL

To elucidate how B7-H3 regulates lipid metabolism in glioma cells, we performed enrichment analysis of lipid metabolism-related enzymes in transcriptomic data from U87MG sgB7-H3 and U87MG sgCtrl cells. Key enzymes in lipid catabolism PNPLA2, lipid synthesis ACLY, and β-oxidation CPT1A exhibited significant alterations upon B7-H3 knockout (Fig. 3A). In primary glioma datasets, high expression of PNPLA2 or ACLY correlated with poor patient prognosis, while in recurrent glioma datasets, only PNPLA2 overexpression was strongly associated with adverse outcomes (Fig. 3B). Notably, B7-H3 showed a significant positive correlation with PNPLA2 in both primary and recurrent glioma cohorts (Fig. 3C). Analysis of ATGL (PNPLA2) expression in 80 glioma tissues revealed progressively elevated levels with increasing tumor grade (Fig. 3D and E). Genetic knockout or silencing of B7-H3 in glioma cells markedly downregulated ATGL expression, and vice versa (Fig. 3F). Furthermore, lipid droplet staining confirmed that B7-H3 downregulation could cause increased BODIPY signal, suggesting the presence of lipid droplets accumulation in B7-H3 downregulated cells (Fig. 3G and H).

Fig. 3.

Abnormal lipolysis regulated by B7-H3-ATGL axis in Glioma cells in vitro. (A) Lipid metabolism gene related heatmap was generated using B7-H3 knockout cells derived RNA-seq data and was further classified according to lipid metabolism processes: lipolysis, synthesis, β-oxidation. Among these genes, three key enzymes PNPLA2, ACLY, CPT1A underwent significant changes. *P < 0.05, **P < 0.01; P-values are determined by two-tailed Student’s t-test. (B) Overall Survival information and three key genes expression data of primary and recurrent glioma patients were extracted from the CGGA dataset. To describe the contribution of genes to patient survival time, we used Kaplan–Meier analysis. 95% CI and p value were calculated via log-rank test and shown at the bottom of each plot. (C) The statistical correlation between B7-H3 expression and PNPLA2 in patients with primary or recurrent glioma. Pearson’s correlation coefficient test was used to correlation analysis. (D) IHC analysis of the expression of ATGL (PNPLA2) in glioma tissues from different Tumor grades. Scale bars 50 μm. (E) Violin plots of ATGL expression in glioma patients with different Tumor grade: Grade 1, blue plot; Grade 2, orange plot; Grade 3, purple plot; Grade 4, green plot. p value was calculated via Wilcox test. (F) The level of ATGL protein altered correspondingly with changes in B7-H3 expression levels. The protein levels of B7-H3, ATGL, and GAPDH in U87MG and 05MG cells were determined by immunoblotting. (G) Confocal micrographs displaying sgCtrl and sgB7-H3 U87MG cells, with LDs stained with BODIPY 493/503, and nuclei were marked with blue using DAPI. Scale bars 50 μm. (H) Confocal images of LDs stanning in 05MG cells transduced with siRNA for B7-H3. Scale bars 50 μm

Based on these facts, upregulating the expression of ATGL was observed to further promote the decomposition of lipid droplets that abnormally accumulated due to B7-H3 knockout (Fig. 4A and B). Additionally, incorporation of ATGL could also abolish cell proliferation inhibition and colony-forming and chemoresistance in B7-H3 knockout glioma cells (Fig. 4C and F). Similarly, downregulation of ATGL effectively suppressed the B7-H3 overexpression-induced proliferative effects (Fig. S3). Furthermore, we found that B7-H3 knockout significantly impaired the tumor-forming ability of glioma cells, but this tumor-suppressive effect was reversed upon ATGL upregulation (Fig. 4G and H, S4). Collectively, our findings indicate that B7-H3 could regulate lipid metabolism via the lipolytic enzyme ATGL, thereby facilitating the malignant transformation of glioma.

Fig. 4.

B7-H3 regulated glioma cells lipid metabolism, proliferation and colony formation in vitro and tumor growth in vivo in an ATGL dependent manner. (A) B7-H3 downregulates the expression of ATGL in glioma cells, but interfering with ATGL expression reverses the inhibition. U87MG cells transduced with CRISPR-Cas9 lentivirus for B7-H3 and overexpression lentivirus for ATGL were applied for WB assay. (B) Confocal micrographs displaying U87MG cells transduced with CRISPR-Cas9 lentivirus for B7-H3 and overexpression lentivirus for ATGL, with LDs stained with BODIPY 493/503, Scale bars 50 μm. (C, D) Downregulation of B7-H3 suppressed the motility of glioma cells, but overexpression of ATGL reversed the inhibitory effect. U87MG cells transduced with CRISPR-Cas9 lentivirus for B7-H3 and overexpression lentivirus for ATGL were applied for proliferation assay and cloning assay. sgCtrl was used as control. The bar chart represents the quantification of proliferation assay (C) and cloning assay (D). Data represent means ± SD. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001; compared with sgCtrl group or sgB7-H3 group by multiple t-tests or two-tailed Student’s t-test. (E, F) Downregulation of B7-H3 suppressed the drug sensitivity of glioma cells, but overexpression of ATGL reversed the inhibitory effect. U87MG cells transduced with CRISPR-Cas9 lentivirus for B7-H3 and overexpression lentivirus for ATGL were treated with different concentrations of 5-FU (1.25 µg/mL, 10 µg/mL). Data represent means ± SD. **P < 0.01, ***P < 0.001, ****P < 0.0001; compared with sgCtrl group or sgB7-H3 group by two-tailed Student’s t-test or multiple t-tests. (G, H) The volume of the glioma U87MG tumors and representative tumor images. Statistical analysis was performed using one-way ANOVA. **P < 0.01, ****P < 0.0001, n = 5 mice per group. Error bars denote mean ± SD

B7-H3 promotes glioma progression by regulating ATGL through the cMet-pSTAT3 signaling pathway to participate in lipid metabolism

Our previous studies had demonstrated the direct interaction between B7-H3 and the proto-oncoprotein cMet in colorectal cancer and glioma cells [15]. Here, confocal microscopy confirmed the co-localization of B7-H3 and cMet in glioma cell lines U87MG and 05MG (Fig. 5A). Co-immunoprecipitation assays further validated their physical interaction in U87MG cells (Fig. 5B). Notably, transcriptomic datasets of both primary and recurrent gliomas revealed a significantly positive correlation between B7-H3 and cMet expression (Fig. 5C). To investigate downstream signaling pathways regulated by the B7-H3/cMet interaction, we performed KEGG enrichment analysis on RNA-seq data from U87MG sgCtrl and U87MG sgB7-H3 cells, identifying the JAK-STAT pathway as the most prominently altered pathway (Fig. 5D). Meanwhile, GSEA was carried out and the genes associated with the IL-6/Jak/Stat3 signaling were found to be markedly suppressed following B7-H3 depletion (Fig. 5E). Subsequent validation demonstrated that B7-H3 modulated ATGL expression by regulating phosphorylation levels of both pMet and pSTAT3 (Fig. 5F). Similarly, the expression levels of pSTAT3 and ATGL in tumor tissues of the B7-H3 knockout group were significantly downregulated (Fig. S5), consistent with the in vitro experimental results. Furthermore, we observed that p-STAT3(Tyr705) bound to the promoter of PNPLA2 (ATGL) in U87MG sgCtrl cells, while B7-H3 knockout significantly impaired the binding (Fig. 5G). In line with this observation, a STAT3 inhibitor cryptotanshinone suppressed the Tyr705 phosphorylation of STAT3 and the transcriptional level of PNPLA2 (ATGL), which was triggered by the exogenous expression of B7-H3 (Fig. S6). Collectively, these findings indicate that the B7-H3/cMet interaction activates downstream pSTAT3 signaling to upregulate ATGL expression, thereby driving glioma malignant transformation (Fig. 5H).

Fig. 5.

B7-H3 regulated ATGL through cMet-pSTAT3 signaling pathway during lipolysis. (A) Endogenous B7-H3 (green) and cMet (red) immunofluorescence staining of glioma cells. B7-H3 co-localization with cMet was both in the cell edge and cytoplasm regions. Scale bars, 20 μm. (B) The interaction of B7-H3 and cMet in U87MG was validated via co-immunoprecipitation assays. Cell lysates from U87MG were IP against B7-H3 and IB against cMet. Furthermore, Cell lysates from U87MG were IP against cMet and IB against B7-H3. mIgG or RIgG was used as a control. (C) The statistical correlation between B7-H3 and cMet in primary or recurrent glioma patients. Pearson’s correlation coefficient test was used to correlation analysis. (D) The top 10 enriched KEGG pathway terms of DEGs in U87MG after B7-H3-knockout. (E) GSEA was conducted on the differentially expressed genes between sgCtrl and sgB7-H3 as detected by RNA-Seq. (F) The phosphorylation of STAT3 was inhibited after the downregulation of B7-H3. B7-H3 was silenced by either siB7-H3#1 or siB7-H3#2 in glioma cells, and siNC was used as non-targeting control. The protein levels of B7-H3, cMet, STAT3, and phosphorylation of cMet, STAT3 in U87MG and 05MG cells were determined by immunoblotting. (G) CUT&RUN-qPCR assays for U87MG (sgCtrl and sgB7-H3) were performed using anti-p-STAT3(Tyr705) antibody and control IgG, and this was followed by real-time qPCR using the designed primer specific for the PNPLA2 (ATGL) promoter. (H) Model of B7-H3 regulation of tumorigenesis via lipolysis. Schematic illustration depicting the roles of B7-H3 as a lipolysis related factor to drive the malignancy of glioma via the cMet-pSTAT3-ATGL pathway

Patients with B7-H3 and ATGL co-expression had the worst prognosis

To investigate the clinical relevance of the B7-H3/ATGL signaling axis in patient survival, we first retrieved glioma patients’ data from the CGGA dataset. We found that high co-expression of B7-H3/ATGL predicted poor survival in both primary glioma (Fig. 6A) and recurrent glioma patients (Fig. 6B). More importantly, we performed immunostaining to examine the co-expression of B7-H3 and ATGL in glioma specimens from the cohort enrolled in our study (Fig. 6C). Further investigation revealed that the co-expression of both B7-H3 and ATGL in tumor cells predicted the shorter survival in glioma patients (Fig. 6D). Remarkably, a positive correlation was observed between the co-expression levels of B7-H3/ATGL and tumor grade (Fig. 6E). In addition to intrinsic factors, the malignant transformation of tumor cells is closely linked to the tumor microenvironment (TME). Our study compared the TME and immune profiles of patients with primary and recurrent glioma patients (Fig. S7), revealing that aberrant expression of B7-H3 and ATGL correlated with immune characteristics of the TME (Fig. S8). Especially, co-expression of B7-H3 and ATGL aligned with TME features that promoted tumor malignancy (Fig. 6F and G). Together, these results reveal that co-expression of B7-H3 and ATGL predicts a poorer prognosis in glioma patients.

Fig. 6.

ATGL protein levels positively correlated with B7-H3 expression in glioma patients. (A, B) Overall Survival information and gene expression data of glioma patients (primary glioma (A), recurrent glioma (B)) were extracted from the CGGA dataset. To describe the contribution of B7-H3/ATGL to patient survival time, we used Kaplan-Meier analysis. We used the geometric mean of B7-H3/PNPLA2 value to represent the weighted co-expression of B7-H3/ATGL. 95% CI and P value were calculated via log-rank test and shown at the bottom of each plot. (C) Representative image of IHC staining in primary tumors from glioma patients. Scale bars, 50 μm, magnified picture 20 μm. (D) The relationship between B7-H3 and ATGL co-expression and the survival status of glioma patients. p value was calculated via Chi-squared test. (E) The relationship between B7-H3 and ATGL co-expression and different Tumor grade of glioma patients. p value was calculated via chi-square test. (F, G) the Immuno-Oncology Biological Research (IOBR) analyses were applied to decode the relationship between tumor microenvironment signatures score and prognosis phenotype according to B7H3 and ATGL grouping using glioma patients (primary glioma(F), recurrent glioma (G))

Discussion

Glioma, as the most common malignant tumor in the central nervous system, exhibits a high degree of heterogeneity and an aggressive phenotype closely associated with metabolic reprogramming. In this study, we conducted comprehensive enrichment analyses to explore the interaction between metabolic reprogramming and prognosis in glioma. Subsequently, we found that a top co-associated gene, CD276 (B7-H3), exhibited a significant correlation with poor prognosis in both primary and recurrent glioma patients. Notably, through extensive in vitro and in vivo experiments, we validated that B7-H3 orchestrates abnormal lipolysis during glioma progression via complex formation with cMet. Importantly, we further uncovered its underlying novel signaling axis of B7-H3-cMet-ATGL in glioma pathogenesis.

Metabolic abnormalities often promote malignant progression by reshaping the intrinsic signaling network and microenvironmental ecosystem of glioma cells. Glioma cells preferentially utilized glucose to generate lactate through the Warburg effect, even under oxygen rich conditions, maintaining this metabolic pattern. Not only did it rapidly supply energy, but it also produced a large number of intermediate metabolites to support biosynthesis and redox homeostasis, driving cell proliferation and therapeutic resistance [16]. Glutamate dehydrogenase 1 (GDH1) was a key enzyme in glutaminolysis that converted glutamate to α-ketoglutarate (α-KG). GDH1-produced α-KG directly bonded to and activated IKKβ and NF-κB signaling, which promoted glucose uptake and tumor cell survival by upregulating GLUT1, thereby accelerating gliomagenesis [17]. In addition, metabolic abnormalities in the glioma microenvironment could significantly suppress anti-tumor immune responses. It was reported that monocyte-derived macrophages (MDM), but not microglia, were major contributors to the immunosuppression associated with tumor associated macrophages in GBM, and demonstrated that PERK-perturbed glucose metabolism drove MDM immunosuppressive programs via histone lactylation [18]. The excessive activation of Indoleamine 2,3-dioxygenase-1 (IDO-1) in the tryptophan metabolism pathway consumed local tryptophan, generating metabolites such as kynurenine, which directly inhibited T cell proliferation and induced apoptosis [19]. Moreover, glioma cells produced sPD-L1 through the Wnt/β-catenin signaling pathway, which interacted with the PD-1 receptor on CD8⁺ T cells, inhibiting their function by reducing IFN-γ levels [20]. In general, the relationship between metabolic abnormalities and glioma progression involves a complex multi-level regulatory network. Consequently, targeting critical metabolic pathways combined with immune checkpoint inhibitors or epigenetic regulators may offer novel therapeutic strategies to improve glioma prognosis.

In this study, we applied a series of enrichment analyses to explore the interaction between metabolic reprogramming and prognosis in glioma, aiming to elucidate potential underlying mechanisms. As a result, we found that B7-H3 exhibited a significant correlation with poor prognosis in both primary and recurrent glioma patients. B7-H3 was expressed on a number of cells, and it acted as a driver of immune evasion through hampering T cell infiltration and promoting exhaustion of CD8⁺ T cells [21]. B7-H3 expression in tumor cells contributed to CCL2-CCR2-M2 macrophage axis-mediated immunosuppression and tumor progression in high-grade serous ovarian cancer [22]. In addition, although PD-L1/PD-1 blockade therapy had been approved to treat many types of cancers, the majority of patients with solid tumors did not respond well. A recent study demonstrated that tumor-expressed B7-H3 inhibited the function of CD8⁺ T cells and suggested that B7-H3 might be a target in patients who were not responsive to PD-L1/PD-1 inhibition [23]. Here, we demonstrated that B7-H3 was aberrantly overexpressed in glioma tissues and significantly associated with poor prognosis in both primary and recurrent patients. More importantly, dysregulated lipid metabolism mediated B7-H3-driven malignant transformation in glioma.

B7-H3 affected the metabolism of triple negative breast cancer cells. Once the expression of B7H3 in tumor cells was downregulated, it leads to a decrease in glycolytic capacity of tumor cells [24]. Indeed, the glycolytic pathway was highly enriched in immune-evading patients, where lactate derived from tumor cells induced histone lactylation, leading to upregulation of B7-H3 gene expression. This elevated B7-H3 expression subsequently suppressed the infiltration rate and cytotoxicity of CD8⁺ T cells within the tumor microenvironment, thereby promoting tumor progression [25]. Furthermore, our previous findings suggested that B7-H3 could be a novel regulator of glucose metabolism and chemoresistance by modulating HK2 expression in colorectal cancer cells [26]. Here, we found significant changes in key enzymes involved in lipid catabolism PNPLA2, lipid synthesis ACLY, and β-oxidation CPT1A following B7-H3 knockout. In primary glioma datasets, high expression of PNPLA2 or ACLY correlated with poor patient prognosis, while in recurrent glioma datasets, only PNPLA2 overexpression was strongly associated with adverse clinical outcomes. Notably, B7-H3 showed a significant positive correlation with PNPLA2 in both primary and recurrent glioma cohorts.

ATGL(PNPLA2), as a key enzyme in lipid metabolism, it has attracted much attention in tumor research in recent years by regulating triglyceride hydrolysis and fatty acid release. ATGL correlated with worse overall survival in men with advanced, castration-resistant prostate cancer (CRPC). ATGL inhibition induced metabolic plasticity, causing a glycolytic shift that could be exploited therapeutically by co-targeting both metabolic pathways [27]. Subsequent studies demonstrated that GRP78 secreted within the hypoxic tumor microenvironment mediates the reprogramming of tumor cells toward macrophages, and sustained the immunosuppressive tumor microenvironment by maintaining the protein stability of adipose triglyceride lipase ATGL to promote lipolysis [28]. Besides, the ubiquitin-specific peptidase 22 facilitated EPI-induced breast cancer progression and metastasis by enhancing ATGL-mediated lipolysis [29]. In glioma, upregulating the expression of ATGL was observed to further promote the degradation of lipid droplets that abnormally accumulated due to B7-H3 knockout. Additionally, we experimentally validated that B7-H3 knockout suppressed the proliferation and colony-forming capacity of glioma cells, whereas upregulation of ATGL expression counteracted this inhibitory effect. Similarly, we found that B7-H3 knockout significantly impaired the tumor-forming ability of glioma cells, but this tumor-suppressive effect was reversed upon ATGL upregulation. In light of this evidence, our findings indicated that B7-H3 primarily facilitated the malignant transformation of glioma by regulating lipid metabolism through the lipolytic enzyme ATGL. Thus, targeting ATGL or combining it with immune checkpoint inhibitors may represent a novel therapeutic strategy for glioma.

Our study revealed that B7-H3 directly interacted with cMet, triggering cMet/STAT3 phosphorylation and enhancing the lipolysis of glioma cell. As the receptor tyrosine kinase for hepatocyte growth factor, cMet directly modulates the metabolic network of tumor cells by activating the PI3K-AKT/mTOR and RAS-MAPK signaling axes [30]. EGFRvIII-mediated cross-activation of cMet, regulated GSCs maintenance and promote tumor recurrence. And then, GBM tumorigenicity could be inhibited by combination of pan-EGFR inhibitor afatinib and Temozolomide [31]. Moreover, co-activation of AKT and cMet induced hepatocellular carcinoma (HCC) development by the mTORC1/ Fatty Acid Synthase (FASN) pathway [32]. Notably, the clinical trials had shown that cMet inhibitor crizotinib had a significant antitumor activity in cMet amplified lung cancer [33]. However, tumor cells could escape therapeutic targeting by activating bypass signaling pathways or downstream effector molecules to acquire resistance mutations (D1228N, Y1230C). Additionally, cMet abnormalities exhibited significant heterogeneity across tumor types and individuals, necessitating precise identification of patient populations likely to benefit from targeted therapies. Notably, excessively high expression of B7-H3 has been described in various human malignancies and rarely in normal healthy tissues [34]. Therefore, the combination of cMet inhibitors with immune checkpoint inhibitors (such as PD − 1/PD−L1, B7-H3 antibodies) could achieve synergistic mechanisms to jointly target tumors, and enhance anti-tumor efficacy. In this study, we also reveal that the interaction between B7-H3 and cMet regulated ATGL to promote malignant transformation of tumors. ATGL inhibitors exhibited unique therapeutic potential in the prevention of metabolic diseases and heart failure [35], but had not been reported in the field of tumor treatment. Metabolic enzyme inhibitors targeted tumor-specific metabolic pathways, disrupting their energy supply and biosynthesis. Their combination with chemotherapy, targeted therapy, and immunotherapy held promising potential for synergistic anti-tumor strategies. Therefore, combining cMet inhibitor with immune checkpoint inhibitors or metabolic targets might achieve dual metabolic-immune modulation, thereby improving therapeutic outcomes.

In summary, our in silico and experimental studies revealed that B7-H3 regulates abnormal lipolysis by targeting the pSTAT3/ATGL axis signaling, mediated through its novel interaction with cMet. Future studies should focus on further delineating the relationship between metabolic heterogeneity and the tumor immune microenvironment to enhance the synergistic therapeutic potential of combining metabolic modulation with immune activation approaches.

Supplementary Information

Below is the link to the electronic supplementary material.

Author contributions

YX, FL and ZW conceived and designed the research. XH, LZ, YL, and JW performed most of the experiments and analyzed the data. JA and XL provided human samples. XL and YY analyzed tissue specimens by immunohistochemistry and interpreted clinical significance. ZW and WQ performed the computational analysis and interpreted the results. YX, FL, and ZW supervised the experiments. YX, ZW and HX wrote the paper.

Data availability

Our sequencing and processed data files were submitted to the GSA (GSA-Human; https://ngdc.cncb.ac.cn/gsa-human) repository HRA014343 (RNA-seq). In addition, the datasets used and/or analyzed during the current study are available from the corresponding author upon reasonable request. Original western blots and original images of immunofluorescence are attached as Supplemental materials.

Declarations

Ethical approval

This study was carried out in accordance with guidelines outlined in the Declaration of Helsinki and approved by the First Hospital Affiliated to Soochow University for the Biomedical Research Ethics Committee. And all animal experiments were done in the animal laboratory center as per the study protocol according to the NIH Guide for the Care and Use of Laboratory Animals, approved by the Animal Care and Use Committee of Soochow University (SUDA20250507A05).

Competing interests

The authors declare no competing interests.