Background
Glioblastoma (GBM) is the most common and lethal malignant tumor of the central nervous system. Owing to its highly invasive nature and lack of effective therapeutic methods, its prognosis remains poor. The median overall survival (OS) of patients with GBM is < 2 years [
1]. Currently, maximum tumor resection combined with radiochemotherapy remains the standard treatment for GBM. Recently, genome-wide molecular profiling studies have identified many target genes that have advanced our understanding of GBM tumorigenesis and chemoresistance. Based on these studies, several individualised therapies and novel therapeutic strategies have been developed. For example, patients with O6-methylguanine-DNA-methyltransferase (MGMT) promoter methylation may have a more effective treatment response and better prognosis [
2]. However, none of these individualised targeted therapies have been shown to improve patient prognosis owing to the considerable heterogeneity between different glioma subtypes. Therefore, there is an urgent need to identify novel molecular targets and develop effective therapies for GBM.
Mitochondria perform multifaceted roles in normal physiology, including energy conversion, apoptosis regulation, biosynthetic metabolism, and cellular proliferation [
3,
4]. They are vital for stress sensing, environmental adaptation, and tumorigenesis [
5]. They are also involved in tumor development, progression, and treatment resistance by overproducing reactive oxygen species (ROS), which induce genomic instability, and regulate gene expression and signaling pathways [
6‐
10]. An increasing number of studies have demonstrated that mitochondria play an important role in gliomas, including GBM. Mitochondrial DNA (mtDNA) alterations are associated with cellular and metabolic consequences, diagnosis, prognosis, and treatment of GBM [
11]. Mitochondrial dynamics are reportedly essential for the development of gliomas. Dynamin-related protein 1 (
DRP1), a key mediator of mitochondrial fission, upregulation is correlated with poor prognosis in GBM, and
DRP1 knockdown decreases glioma cell proliferation, migration, and invasiveness [
12,
13]. In addition, mitochondria related ROS are involved in the oncogenesis of gliomas at various phases, such as tumor initiation and progression [
14]. Excessive ROS production induced by mitochondrial damage simulataneously activates autophagy and apoptosis in GBM cells [
15]. According to recent studies, anticancer agents directly targeting mitochondria related genes bypass drug resistance and improve the prognosis of patients with GBM [
16,
17]. Therefore, a comprehensively analysis of mitochondria-related genes (MRGs) and exploration of their function in GBM might be useful for identifying novel prognostic biomarkers and developing effective therapeutic strategies for GBM.
In this study, we identified prognostic MRGs and constructed and validated a novel prognostic model for GBM base on 12 differentially-expressed (DE)-MRGs. Our model presented remarkable performance in predicting the prognosis of GBM patients and was confirmed to be an independent risk predictive factor. Furthermore, our risk score was enriched in inflammatory response, extracellular matrix, and pro-cancer-related and immune related pathways and closely associated with gene mutation and immune cell infiltration. To confirm the importance of these 12 DE-MRGs, we selected single-stranded DNA-binding protein 1 (SSBP1) for further in vitro studies and found that it was upregulated in GBM tissues. Furthermore, we demonstrated that SSBP1 knockdown significantly inhibits GBM cells proliferation and migration by disturbing mitochondrial function. SSBP1 knockdown also enhances temozolomide (TMZ) sensitivity by enhancing ROS induced ferroptosis.
Materials and methods
Data acquisition
The gene expression profiles of The Cancer Genome Atlas (TCGA)-GBM cohort (count and tpm) and clinical information were downloaded from the GDC Data Portal (
https://portal.gdc.cancer.gov/). In addition, the single cell data of GSE84465 and the external independent GBM databases, GSE147352 and GSE16011, were gained from the Gene Expression Omnibus (GEO,
https://www.ncbi.nlm.nih.gov/gds). The samples with complete survival information were retained for analysis. The gene expression data of Chinese Glioma Genome Atlas (CGGA) GBM cohort were downloaded from
http://www.cgga.org.cn/.
Identification of prognostic mitochondrial-related genes (MRGs) in GBM
First, we analysed the single-cell data of GSE84465 to identify differentially expressed genes (DEGs) between neoplastic and non-neoplastic cells by using “Seurat” package. The threshold for data filtering included minimum cells = 3, 200 < nFeature RNA < 7500, and percentage of ribosome RNA < 15. FindMarkers function was used to DEGs. based on the following criterions: logFC > 0.25 and adjusted p value < 0.05. The 686 MRGs (Additional file
6: Table S1) were obtained from the uniport database (
https://www.uniprot.org/). By taking the intersection of DEGs and MRGs, we finally got 201 differentially-expressed (DE)-MRGs, of which 197 DE-MRGs can be matched in the TCGA GBM database. Subsequently, univariate Cox analysis was performed to identify the prognostic DE-MRGs based on the TCGA GBM cohort. The FeaturePlot function was used to present gene expression on a dimensional reduction plot between cell clusters.
Construction and validation of prognostic risk score model based on the DE-MRGs
Based on the TCGA cohort, the Least absolute shrinkage and selection operator (LASSO) Cox regression analysis was used for further dimensionality reduction of prognostic DE-MRGs. Finally, 12 DE-MRGs were selected to the construct a prognostic risk model. The formula for risk score was as follows: Risk score = 0.2132 * expression level of PLAUR + 0.0261 * expression level of RBP1 + 0.0048 * expression level of ABCB8 + 0.2553 * expression level of TOMM7 − 0.1868 * expression level of MFF + 0.0714 * expression level of SSBP1 + 0.1927 * expression level of MRPL36 + 0.1686 * expression level of AGK + 0.07 * expression level of HK1—0.1339 * expression level of APEX1 + 0.2896 * expression level of NUDT1 − 0.3503 * expression level of PHB2. The Kaplan–Meier (K-M) analysis, univariate cox and multivariate cox analysis, and time dependent Receiver Operation Characteristic (ROC) curve were used to reveal the prognostic value of our model. The GSE16011 and GSE147352 GBM cohorts were used to validate this prognostic model.
Gene Oncology (GO) and pathway enrichment analysis of our risk model
Firstly, The R package (Deseq2) was used to identify the DEGs between high- and low-risk groups based on the TCGA GBM cohort. The standards for DEGs were logFC > 1 and adjusted p value < 0.05. Then, the GO analysis of biological process (BP), cellular component (CC), and molecular function (MF) was performed by using DAVID (
https://david.ncifcrf.gov/) based on the DEGs between risk groups. The pathway enrichment analysis of KEGG and HALLMARK gene sets was performed via GSAE (gene set enrichment analysis) method. An adjusted p-value < 0.05, q-value < 0.05, and absolute normalised enrichment score (NES) > 1 were used as the threshold for determination of significance.
Ferroptosis score
Ferroptosis-related gene sets, including driver and suppressor genes, were obtained from FerrDb (
http://www.zhounan.org/ferrdb/current/), the first database dedicated to ferroptosis regulators and ferroptosis-disease associations. The ferroptosis score of each sample was calculated using the single-sample GSEA (ssGSEA) method.
Immune infiltration analysis
The infiltration levels of 28 immune cells in each sample were assessed by using single sampleGSEA based on the R package (GSVA). Additionally, the ESTIMATE algorithm was also used for calculated the ImmuneScore, StromalScore, and ESTIMATEScore of each sample based on the “estimate” package in R.
Cell culture and transfection
All glioma cell lines (U87, U251, and SHG-44) were obtained from the Shanghai Life Academy of Sciences Cell Library (Shanghai, China). Glial cells (HEB) were obtained from the Sun Yat-Sen University Cancer Center. U87 cells were grown at 37 °C and 5% CO2 in Dulbecco’s modified Eagle’s medium (DMEM; HyClone, United States) supplemented with 10% foetal bovine serum. TMZ was obtained from Sigma-Aldrich Corporation.
All small interference RNA (siRNA) against the target genes and negative control siRNA were synthesised by GenPharma (Suzhou, China). U87 cells were transfected using Lipofectamine® RNAiMAX Transfection Reagent (Invitrogen, Carlsbad, California, United States), according to the manufacturer’s instructions. The sequences of the SSBP1 siRNA were as follows: siRNA #1 F 5’-CAACAAUCAUAGCUGAUAAUA, 3’-UUAUCAGCUAUGAUUGUUGUU; siRNA #2 F 5’- UAAUACAGGUCUUCGAAACAU, 3’-GUUUCGAAGACCUGUAUUACA.
RNA extraction and quantitative PCR (qPCR)
Total RNA was extracted from U87 cells using TRIzol reagent (Invitrogen, Carlsbad, CA, United States) and reverse transcription was performed using PrimeScript™ RT Reagent Kit (Takara, Dalian, China). Real-time PCR was performed using SYBR Green Real-Time PCR Kit (Takara, Dalian, China). β-Actin was used as a normalising control. The relative expression levels were evaluated using the 2-ΔΔCT method. The primer sequences used in this study are included in Additional file
7: Table S2.
Cell proliferation and migration assay
U87 cells were seeded in 96-well plates at a density of 5000 cells per well, Cell viability was measured using the Cell Counting Kit-8(Sigma-Aldrich, Shanghai, China) according to the manufacturer’s instructions. After knocking down SSBP1 for 72 h, U87 cells were seeded into the upper chambers at a density of 5.0 × 104 cells in 300 µl of serum-free cell culture medium, while 500 µl of medium containing 20% FBS was added into the lower chambers. The cell migration assay was performed using 24-well transwell chambers (Corning, NY, United States) according to the manufacturer’s instructions.
Measurement of ROS level, iron content, and GSH level
Mitochondrial ROS and mitochondrial membrane potential (MMP) were detected using MitoSOX™ Red Mitochondrial Superoxide Indicator (Invitrogen, United States) and MitoTracker™ Red CMXRos (Invitrogen, United States), respectively. For 24 h, 1.0 × 105 U87 cells were cultured in a Nunc™ Glass Bottom Dish (Invitrogen, United States). Cells were washed with warm DPBS before incubation with 2.5 µM MitoSOX™ Red or 100 nM MitoTracker™ Red. Fluorescence intensity was analysed using a Zeiss LSM 800 confocal microscope and measured using ImageJ software.
U87 cells were transfected with siRNA and incubated with and without 400 μM TMZ for 48 h and collected. Iron content (Abcam, Cambridge, UK) and intracellular GSH levels (Abcam, Cambridge, UK) were measured according to the manufacturer’s instructions.
Western blotting
Glioma cells and HEB were lysed in 300 µL SDS sample buffer (Sangon Biotech, China) containing 1 mM phosphatase inhibitor and 1 mM PMSF, and denatured proteins (20 µg) were resolved on 15% SDS PAGE gels and transferred to PVDF membranes. After blocking with 5% milk at room temperature, the membranes were incubated with the primary antibody overnight at 4 °C, followed by incubation with a horseradish peroxidase–conjugated secondary antibody for 2 h at room temperature. After washing the membranes thrice with TBST, ECL Reagent (Sangon Biotech, China) was added for chemiluminescent detection.
Immunofluorescence
U87 cells were seeded in a six-well plate with three coverslips per well and cultured for 72 h at 37 °C. Further, 100 nM MitoTracker™ Red CMXRos (Invitrogen, United States) were incubated with U87 cells at 37 °C for 30 min before fixing in 4% paraformaldehyde for 10 min at room temperature. Permeabilization was performed using 0.5% Triton X-100 for 10 min at room temperature. The slides were washed and blocked with 1% BSA for 1 h. The primary antibodies were then incubated with U87 cells overnight at 4 °C. Next morning, the slides were incubated with Alexa Fluor 488 (Abcam, Shanghai, China) for 1 h at room temperature. Cell nuclei were stained with DAPI (Boyetime, Wuhan, China). Immunofluorescence images were observed using a Zeiss LSM 800 confocal microscope and analysed using ImageJ software.
Statistical analysis
All results were analysed using GraphPad Prism 9 and R software with R packages. The experimental data were presented as means ± SDs (standard deviations) and unpaired Student’s t-test was used for continuous variables between groups. The log-rank test was utilized for K-M survival analysis. The Pearson’s correlation coefficient was applied to calculate the correlation between the expression of mtDNA-encoded genes and SSBP1. A two-side p < 0.05 was considered to indicate a statistically significant result.
Discussion
Multiple studies have revealed that MRGs are essential for tumorigenesis and tumor progression and targeting MRGs may be a promising therapeutic approaches for cancer treatment [
4]. However, few studies have focused on MRGs and elucidated their role in GBM. In the present study, we constructed and validated a prognostic risk score model for GBM based on 12 DE-MRGs through differential gene expression, univariate Cox, and LASSO-Cox analyse. Our risk model presented excellent performance in predicting the GBM prognosis; the risk score was an independent prognostic factor associated with the clinicopathological and molecular features of GBM. In addition, stratified analysis demonstrated that our risk model could well distinguish the prognosis of patients with radiotherapy or TMZ chemotherapy, which indicated that our model has potential clinical application value and that the 12 DE-MRGs might be involved in chemoradiotherapy resistance in GBM. Indeed, some of these genes have been shown to be related to chemoradiotherapy sensitivity in GBM. Cho et al. reported that the relative expression of APEX1 (APE1) was significantly increased in TMZ-resistant cell lines [
35]. Furthermore, functional enrichment and immune cell infiltration analyses revealed that our risk score was significantly associated with inflammatory response, extracellular matrix, pro-cancer-related and immune related pathways, tumor mutation burden and immune cell infiltration in GBM. Therefore, these 12 genes are potential therapeutic targets for GBM treatment.
To the best of our knowledge, our study is the first to identify DE-MRGs in neoplastic and non-neoplastic cells based on single cell data for the first time. Single cell RNA sequencing presents great advantages over bulk RNA sequencing in dissecting heterogeneity in cell populations [
36]. Therefore, single cell analysis is more conducive to finding differentially expressed genes among cell types. Based on single cell analysis we screened 201 DE-MRGs in GBM tissues. To further screen out important DE-MRGs in GBM, univariate Cox and LASSO Cox analyses were performed and 12 DE-MRGs were identified, which were used to construct a prognostic risk score model for GBM patients. In addition, our risk score excellently predicted the prognosis of patients with GBM, which in turn indicates that our screening method is feasible. The LASSO is a penalization method used to shrink and select variates for regression [
37]. Moreover, our risk model also associated with the prognosis of other cancers and the predictive efficiency of our model was better than that of several reported signatures, including immune-related gene signature, pyroptosis-related gene signature, and autophagy-related gene signature, in GBM. A possible reason behind the difference of predictive power might be due to the involvement of mitochondria in various biological processes, including immune response, hypoxia, and cell death [
38]; thus, the screened 12 DE-MRGs are likely to play an important role in GBM and might be therapeutic candidates. Indeed, some of these 12 genes have been reported to be involved in glioma tumorigenesis or to be significant in predicting OS. For example, the expression of RBP1 is significantly elevated in gliomas, and the overexpression of RBP1 enhances the growth, self-renewal ability, invasion, and migration of glioma cells [
22]. AGK is markedly overexpressed in glioma and might play an important role in glioma development and progression [
23]. LINC00470 inhibits the ubiquitination of HK1, the first key enzyme in the glycolysis pathway, thereby affecting glycolysis, and inhibiting cell autophagy in gliomas [
39]. APEX1 (APE1) activity is elevated in gliomas and induces resistance to chemoradiotherapy [
35].
NUDT1, also known as
MTH1, is overexpressed in GBM, and its silencing significantly alters glioma cell viability [
40]. These studies support the important role of the risk score. Therefore, the members of the 12 DE-MRGs that have not been studied in glioma are worth for further study.
Recently, an increasing number of studies have demonstrated that SSBP1 is significantly correlated with poor patient prognosis and is involved in tumorigenesis, proliferation, and drug sensitivity in certain human cancers [
9,
17,
30]. The upregulation of SSBP1 is associated with the aggressiveness of osteosarcoma cells [
27]; whereas, its depletion triggers cell death in colorectal cancer cells by affecting the mitochondrial proteome [
28]. SSBP1 is a suppressor of triple-negative breast cancer metastasis [
29]. SSBP1 participates in mtDNA repair in cancer cells during oxidative stress by interacting with p53 [
30]. However, as one of the 12 DE-MRGs, the role of SSBP1 in GBM remains unclear. Therefore, we investigeted the role of SSBP1 in GBM. Based on bioinformatic analysis, we found that SSBP1 is aberrantly upregulated in GBM tissue and significantly related to the poor prognosis of primary GBM patients. Previous studies have shown that SSBP1 is essential for mtDNA maintenance and replication. mtDNA can lead to devastating, heritable, and multisystem diseases that have different tissue-specific presentations and are important in the initiation and maintenance of tumorigenesis in GBM [
41‐
43]. Therefore, we focused on exploring the role of SSBP1 in mitochondrial function. Our study demonstrates that silencing SSBP1 expression inhibits GBM cell proliferation and migration. SSBP1 enhances mtDNA replication under physiological conditions, resulting in ATP generation through oxidative phosphorylation [
42]. Therefore, we speculated that SSBP1 knockdown might affect mitochondrial ROS production by regulating oxidative phosphorylation.
In recent years, many biomaterials with new technologies have been developed for cancer treatment and have shown promising application prospects. Nanotechnology-based approaches exhibit higher efficacy, higher target specificity, and great potential to bypass the limitations of traditional therapies [
44]. As a factory of energy involved in the proliferation of cancer cells, mitochondria are naturally regarded as an important target for cancer therapeutics. Th mitochondria in cancer cells are characterized by ROS overproduction, which promotes cancer development. Recently, multiple novel agents specific for ROS targets have been shown to efficiently maximize chemotherapy efficacy and minimize side effects [
45,
46]. ROS-responsive micro- and nano-particles specifically release their drug cargo guided by ROS concentration, which is enhanced in the cellular environment within specific tumors, and thus, show marked cytotoxicity for cancer cells compared to non-ROS responsive molecules [
14]. Hence, a better understanding of the role of mitochondrial ROS in GBM will help identify novel therapeutic targets. In this study, we demonstrated that
SSBP1 knockdown promotes ROS production and alters MMP in GBM cells, which is consistent with the results of
SSBP1 knockdown in other cancers [
28‐
30]. Ubiquinol-cytochrome c reductase hinge protein (UQCRH) regulates electron transfer from cytochrome c1 to cytochrome c, and its upregulation enhances ROS production [
47]. UQCRC1 is a subunit of UQCRH, and its upregulation can result in enhanced ROS production. ROS production may also be elevated due to the upregulation of mitochondrial ETC function, as implied by upregulation of other ETC components. Our results indicated that SSBP1 might regulates ROS by regulating the expression of UQCRC1, however, the exact mechanisms underlying these processes require further study.
We found that mitochondria of U87 cells with SSBP1 knockdown were fragmented and aggregated. These changes seem to correlate with the expression of mitochondrial morphology mediators, such as the upregulation of OMA1 and DRP1 and downregulation of OPA1. Since we demonstrated that SSBP1 is a potential mitochondrial biomarker of GBM, we further investigated whether it could also be a therapeutic target for GBM. Although TMZ is the first-line chemotherapy for GBM, its efficacy is limited by acquired chemoresistance. Oliva et al. found that TMZ-dependent acquired chemoresistance might be due to a mitochondrial adaptive response to TMZ genotoxic stress with a major contribution from cytochrome c oxidase [
48]. Lomeli et al. have also reported that TMZ can lead to mitochondrial dysfunction, oxidative stress, and apoptosis [
49]. A recent study demonstrated that TMZ can suppress tumor cell proliferation by inducing ferroptosis, which might be a result of ROS accumulation [
50]. Thus, targeting mitochondrial ROS may overcome the TMZ resistance and improve TMZ efficacy. Recently, ferroptosis has become a hot research topic, and several studies have investigated ferroptosis-related biomarkers through bioinformatics analysis or experiments [
51,
52]. Here, we combined bioinformatics and experimental methods to confirm that
SSBP1 knockdown enhances ROS production to trigger ferroptosis in U87 cells. Furthermore, upon TMZ treatment,
SSBP1 knockdown activated the AMPK pathway and inhibited the NF-κB pathway in U87 cells. AMPK pathway activation and NF-κB pathway inhibition have been reported to enhance the anti-cancer effects of chemotherapy [
53,
54]. Therefore, we can conclude that
SSBP1 knockdown increases TMZ sensitivity by promoting mitochondrial ROS to trigger ferroptosis and regulate the AMPK and NF-κB pathways. This result implied that the strategy of combining an SSBP1 inhibitor with TMZ would benefit tumour treatment by enhancing TMZ sensitivity, however, additional efforts are required to translate this strategy into the clinical setting to benefit GBM patients. Further studies on the mechanisms of MRGs and ferroptosis in TMZ resistance would provide new ideas for the clinical reversal of TMZ resistance and improve the efficacy of chemotherapy.
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.