Abstract Background Glutathione plays critical roles in detoxifying xenobiotics, cell signaling, cell death and the antioxidant defence in an emerging body of evidence, the most abundant intracellular low molecular weight thiol in tissues. However, all glutathione metabolism pertinent genes (GMPGs) expression and their diagnostic/prognostic/micropeptide potential analyses have not been investigated to perform in pan-cancers. Methods We gained GMPGs from the MsigDB 7.2, 12,123 samples were used to reveal the differentially expressed genes (DEGs) and the survival analysis in 32 types of cancers from TCGA, GTEx, and GEO datasets for the first time. All statistical analyses were performed by R for bioinformatics, such as DEGs, prognostic, diagnostic analysis, ceRNA, micropeptide prediction and immune infiltration. In addition, we utilized siRNA technology to target knockdown the expression of the G6PD gene in Huh7 hepatocellular carcinoma cells. Results G6PD was significantly expressed and poor prognosis in liver hepatocellular carcinoma (LIHC) and predicted RBM26-AS1 encoded micropeptide might target G6PD in LIHC. In vitro experiments show that G6PD knockout in Huh7 cells reduces their proliferation, migration, and invasion capabilities. Conclusions We confirmed that G6PD played a crucial role in the occurrence and progression of LIHC. G6PD is positively associated with Th2 cells in LIHC, regulating immune responses in the immune system. We considered that micropeptide RBM26-AS1 might be a new player involved in LIHC by interacting with G6PD, might perform a key function in liver cancer. Supplementary Information The online version contains supplementary material available at 10.1007/s12672-025-01945-1. Keywords: Glutathione, G6PD, Biomarkers, LIHC, Micropeptide Introduction Mitochondrial metabolism are the primary source of reactive oxygen species, mitochondrial glutathione appears as the main defense for keeping the befitting mitochondrial redox environment [[36]1, [37]2]. Glutathione is transported into mitochondria, and could provide a valuable approach to hampering mitochondrial dysfunction diseases [[38]3, [39]4]. The abundance of glutathione in tissues and cells was the predominant low molecular weight thiol, playing critical roles in detoxifying xenobiotics, cell signaling, cell death and antioxidant defence [[40]5]. More research has revealed that glutathione metabolism pertinent genes (GMPGs) played an essential role in cancer metabolism [[41]6]. Pharmacologic inhibition of the Nrf2/GSH pathway could be valuable for Isocitrate dehydrogenase (NADP( +)) 1 (IDH1)-mutated cancer treatment [[42]7]. Alanine-glyoxylate and serine-pyruvate aminotransferase (AGXT) expression is downregulated in Kidney renal clear cell carcinoma (KIRC) and associated with poorer prognosis [[43]8]. Glutamate decarboxylase 1 (GAD1) expression changes could alter glutamate metabolism in brain cancer [[44]9]. Asparagine synthetase (glutamine-hydrolyzing) (ASNS) is also marked as a survival-related protein in Clinical Proteomic Tumor Analysis Consortium (CPTAC) and The Cancer Proteome Atlas (TCPA) database [[45]10]. 4-Aminobutyrate aminotransferase (ABAT) was positively regulated by Hepatocyte nuclear factor 4 alpha (HNNF4A) and suppressed tumorigenic capability in KIRC [[46]11]. Glucose-6-phosphate dehydrogenase (G6PD) is a key gene in the context of GMPGs and plays a critical role in maintaining cellular redox balance and protecting cells from oxidative damage [[47]12]. G6PD deficiency is a common genetic disorder implicated in hematological diseases, cancer metabolic reprogramming, and redox signaling [[48]13, [49]14]. In hepatocellular carcinoma, G6PD is found to inhibit ferroptosis by targeting cytochrome P450 oxidoreductase [[50]15]. miR-206/miR-122 are negatively correlated with hepatocellular carcinoma by targeting G6PD [[51]16, [52]17]. The ceRNA competition mechanism is widely used in the study of tumors. We found that there will be new micropeptides closely related to tumor target genes. Micropeptide is a small peptide encoded by a small open reading frame [[53]18], can also be named by its genomic location, which has been proven to be effective in maintaining cell homeostasis [[54]19], and the length is less than 100–150 amino acids (AAs). Micropeptides play a crucial role in cancer development and progression, extensively participating in various signaling pathways and gene expression regulation. The high expression in liver cancers of LINC00998 encoded micropeptide Small integral membrane protein 3 (SMIM30) was also related to poor prognostic, SMIM30 could activate the tyrosine kinase SRC/YES1, which further activated the MAPK signaling pathway in promoting the proliferation and migration of liver cancer development [[55]20]. The expression level of circ-AKT3 in glioblastoma tissue is lower than in normal tissue [[56]21]. Overexpression of micropeptide AKT3-174AA played a negative regulatory role in the PI3K/AKT signaling pathway to reduce the proliferation of glioblastoma cells [[57]22]. However, GMPGs expression and their diagnostic / prognostic / micropeptide potential have not been investigated from a multi-cancer survey yet. By investigating the expression profiles and clinical features of GMPGs in 32 tumor types, The Cancer Genome Atlas (TCGA) four-letter abbreviation was listed in Nonstandard Abbreviations, and the 32 types of tumors and normal tissues from TCGA, Gene Expression Omnibus (GEO), Genotype-Tissue Expression (GTEx) were listed in Table S1. In addition, the extent to which different tumor types express these genes associated with glutathione has not yet to be determined. A comprehensive feature of GMPGs expression might help to reveal the glutathione metabolism across human cancers. To improve our understanding of GMPGs in the present study, we use the data from TCGA to comprehensively compute the differences in GMPGs expression. We seek to (1) identify patterns of GMPGs dysregulation; (2) associate the survival rates of differentially expressed GMPGs (DE-GMPGs) in all 32 cancer types screening; T (3) integrate the TCGA data with previously published data to gain related insight regarding GMPGs on tumorigenesis and prognosis, and provide a new research idea to deeply explore the potential effects of lncRNA-encoded micropeptides on targeting GMPGs (Fig. [58]1A); (4) selecting one particular cancer type for further analysis. Fig. 1. [59]Fig. 1 [60]Open in a new tab A The workflow of GMPGs and potential micropeptides in humans. B Summary of the correlation between expression of glutathione metabolism pertinent genes and patient survival. Differential gene expression of 54 glutathione metabolism-related genes in 32 different cancer types. Fold change and q-value shown were obtained through comparison of tumor tissue to control tissue; Summary of the correlation between expression of GMPGs and patient survival, the orange box represents a higher expression of GMPGs associated with worse survival, and the green box represents an association with better survival, only p-value < 0.05 are shown. C Sankey diagram for selected GMPGs in cancers, a green box filled with blue means down-regulated DEGs with better survival, an orange box filled with blue means down-regulated DEGs with worse survival, a green box filled with red means up-regulated DEGs with better survival, and an orange box filled with red means up-regulated DEGs with worse survival. Next, we used MiPepid to find potential micropeptide in these 9 lncRNAs; LINC00869 and [61]AC012146.7 did not exist micropeptide, so we marked it as a grey triangle; smORFunction was selected to predict potential micropeptide, AC145207.5, PROSER2-AS1, CCDC18-AS1, and NEAT1 were not indicated as functional micropeptide, so we marked them as a blue triangle; G6PD / hsa-miR-1-3p / RBM26-AS1, GLS / hsa-miR-23a-3p / STAG3L5P-PVRIG2P-PILRB, GLS / hsa-miR-23a-3p / VPS9D1-AS1, GLS / hsa-miR-23a-3p / [62]AC078846.1 axis was validated by MultiMiR, and these three lncRNAs (STAG3L5P-PVRIG2P-PILRB, VPS9D1-AS1, [63]AC078846.1) were predicted as functional micropeptides by smORFunction, especially in LIHC, which we marked as a green triangle. Validated ceRNA. Gene was marked as a green parallelogram, miRNA was marked as green diamond, and lncRNAs were formed as a triangle, grey means no micropeptide prediction, blue means no functional micropeptide prediction, and green means useful prediction. D The visualization of chromosome location among the data for lncRNA-micropeptide-smORFunction Materials and methods Data collection and differentially expressed genes and prognosis analysis We gained the glutathione metabolism pertinent genes from the Molecular Signatures Database (MsigDB) v7.2 ([64]http://www.gsea-msigdb.org/gsea/downloads.jsp), 12,123 samples were used to illustrate the differentially expressed genes (DEGs) (|log fold change| (|logFC|) > 1 and q-value < 0.05) and survival analysis (p-value < 0.05) in 32 types of cancers from TCGA ([65]https://portal.gdc.cancer.gov), GTEx ([66]https://gtexportal.org/home/) and GEO dataset ([67]https://www.ncbi.nlm.nih.gov/geo/), using limma package [[68]23]. Then, we selected the GMPGs from pan-cancer DEGs for further analysis. Functional analysis was also performed by the Kyoto Encyclopedia of Genes and Genomes (KEGG) [[69]24]. Subsequently, we extracted the mRNA expression of GMPGs from pan-cancer expression data to compute the patients’ overall survival analysis (using the cox method). Sankey plot was used to display the upregulated and risky DE-GMPGs in pan-cancers [[70]25]. CeRNA discovery MultiMiR v2.3 [[71]26] and LnCeVar [[72]27] were used for the upregulated and risky DE-GMPGs related to ceRNA discovery. The strategy is that (1) multiMiR was selected to find ceRNAs which the validated experiment by “Luciferase reporter assay//Real-time quantitative reverse transcription PCR (qRT-PCR)//Western blot”; (2) Then we used the LnCeVar website to discover the ceRNAs as well; (3) We took the collection to visual by Cytoscape v3.8.2 [[73]28]. iLoc-LncRNA [[74]29], iLoc-mRNA [[75]30], RNALocate v2.0 [[76]31] and DM3Loc [[77]32] were chosen to predict the location of the key genes, lncRNAs. The cutoffs values: nucleus = 0.68, exosome = 0.98, cytosol = 0.2, ribosome = 0.39, membrane = 0.24, endoplasmic reticulum = 0.22 [[78]33]. Principal component analysis (PCA) We combined TCGA-LIHC mRNA expression data ([79]https://tcga-xena-hub.s3.us-east-1.amazonaws.com/download/TCGA.LIH C.sampleMap%2FHiSeqV2.gz) with clinical data and divided the data into tumor VS normal group and the size of the primary tumor stage T1 VS T4, using the R package ggplot2 [[80]34] to draw PCA scatter plot. Screening key GMPGs in LIHC Screening of common differentially expressed genes (tumor VS normal tissue; T1 VS T4), using the limma package, threshold |logFC|> 1 and p-value < 0.05 [[81]23]. We used the Venn diagram to display the common GMPGs. We selected these common GMPGs for subsequent analysis. Univariable Cox analysis was used to screen out the critical prognostic factors (p-value < 0.05). Next, we selected 5 GMPGs with clinical features using randomForestSRC [[82]35] to rank the importance of prognostic-related GMPGs (the random survival forest algorithm was nrep = 100 and nstep = 5, the final signature was set as the genes with relative importance > 0.2). The receiver operating characteristic (ROC) curve is an excellent tool to select possible optimal cut-point for a given diagnostic test. Area's diagnostic efficacy was reasonable under the ROC curve (AUC) > 0.7 and was poor by an AUC between 0.5–0.7. Finally, an AUC of no more than 0.5 indicated the marker's lack of a diagnostic value [[83]36, [84]37]. The immune landscape of LIHC cancer was drawn by a pheatmap [[85]38]. The lollipop chart identified the relationship of G6PD and significantly immune gene sets. A scatter plot visualized the correlation between G6PD and immune cells. Correlation analysis between the key GMPGs and clinical characteristics To understand the correlation between the key GMPGs and the clinical characteristics of LIHC, we divided the expression levels of the critical GMPGs into high and low groups according to the best cut-off value. We analyzed the expression levels of each gene and the clinical characteristics of LIHC by using the mosaic package to draw a mosaic map for visualization [[86]39]. Using the human protein atlas (HPA) database ([87]https://www.proteinatlas.org) to obtain immunohistochemical detection of gene protein expression of the key GMPGs. We calculated the correlation between G6PD and GMPGs by “cor” (Pearson method) in R language. Micropeptide prediction and functional analysis Using MiPepid [[88]40] to predict micropeptides encoded by key lncRNAs (length ≥ 27nt, classification = coding, probability > 0.990). The predicted micropeptide and related lncRNAs were located in the human chromosome and revealed by R package “RIdeogram” [[89]41]. Next, we used smORFunction to identify the functional enrichment of the predicted micropeptides [[90]42]. Cell culture and cell transfection The human hepatocellular carcinoma cell line Huh7 was obtained from the cell bank of the central laboratory of Yijishan Hospital, Wannan Medical College (Wuhu, China). cells were cultured in Dulbecco's Modified Eagle Medium (DMEM) (Gibco BRL, USA) with 100 g/L fetal bovine serum (FBS) and 10 g/L double antibiotics in an incubator at 37 ℃ and 5% CO[2] concentration. The cells were divided into a negative control (NC) group and a si-G6PD group for subsequent experimental studies. The siRNA targeting G6PD was synthesized by Emosa Biotechnology Co., Ltd (Anhui, China), and transfection experiments were carried out using cells in logarithmic growth phase, with the number of cells equal to one-third of the amount of conventionally cultured cells. siRNA fragments were transfected into Huh7 cells by using siRNA transfection reagent (SignaGen, USA), and then the cells were placed in a cell culture incubator for further cultivation, and the targeted G6PD siRNA sequences are shown in Table S2. Quantitative real-time PCR (qRT-PCR) analysis After 48 h of transfection, cells were lysed to extract total RNA using TRIzol reagent (Servicebio, G3013-100ML) for measuring G6PD mRNA expression levels across different groups. Following RNA extraction, it was reverse-transcribed to cDNA using the PrimeScript™ RT Reagent Kit (Vazyme, R232-01). Fluorescent quantitative PCR was then performed using SYBR Green PCR Master Mix on the CFX96 real-time PCR system, and transfection efficiency was assessed via qRT-PCR. Specific primer sequences were obtained from Shenger Biotechnology Co., Ltd (Shanghai, China), as detailed in Table S2. RNA isolation, cDNA synthesis, and qPCR assays were conducted according to previously established methods [[91]43]. Data analysis employed the 2^−ΔΔCt method to determine the relative expression levels, normalized against GAPDH. Cell-cunting-kit-8 (CCK8) assay and transwell Cells in the logarithmic growth phase were seeded at a density of 1 × 10^3 cells per well in a 96-well plate. After a 24-h incubation, cell viability was assessed using a CCK-8 assay kit (SB-CCK8, Share-Bio). The optical density (OD) values of each well at 24, 48, 72, and 96 h were measured at a wavelength of 450 nm using an Epoch2 microplate reader (BioTek, USA). Transwell chambers were utilized for cell migration and invasion assays. Cells in the logarithmic growth phase were counted and adjusted to a density of 1 × 10^5 cells/mL in serum-free DMEM. These cells were then seeded into the upper chamber of a Transwell setup, while serum-enriched medium was added to the lower chamber. To evaluate cell migration and invasion, various treatment conditions were applied: Matrigel Matrix Gel (BD Biocoat, USA) was used to assess invasion, whereas no treatment was used to evaluate migration. After a defined period, cells in the upper chamber were fixed with 4% paraformaldehyde (PFA) for 15 min and stained with 0.1% crystal violet (Solarbio, China) for another 15 min. The migrated or invaded cells were then photographed under a 20X objective lens using an optical microscope and subsequently counted. Statistical analysis All experiments were conducted at least three times. Data were analyzed using R software. Comparisons between groups were performed using the t-test, with a p-value of less than 0.05 considered statistically significant. Results Differentially expressed genes and prognosis analysis in multi-cancers Glutathione metabolism-related gene expression differences for the 54 genes were calculated from 32 cancers in the TCGA, GTEx, and GEO databases. UVM comparatively lacked a further analysis of glutathione metabolism compared to adjacent normal eye tissue. We summarized the KEGG pathway enrichment analysis of the 54 GMPGs (Fig. S1), “Metabolism” pathways were the main components in GMPGs. Still, there were four pathways (k000330, k000270, k000360, and k000350) on level 3 that belonged to “amino acid metabolism.” We also found that “alanine, aspartate, and glutamate metabolism” (k000250) and “butanoate metabolism” (k000650), which belonged to “carbohydrate metabolism” contained 36 GMPGs. From gene expression, we found that all glutathione metabolism-related genes were differentially expressed in different types of cancers (Fig. [92]1B and Fig. S2); Aspartoacylase (ASPA) was down-regulated in 22 types of cancers, whereas Adenylosuccinate lyase (ADSL), Carbamoyl-phosphate synthetase 2, aspartate (CAD), and Phosphoribosyl pyrophosphate amidotransferase (PPAT) were up-regulated in multi-cancers. ABAT was up-regulated and protective in ACC, whereas down-regulated and protective in KIRC, KIRP, and LIHC. Interleukin 4 induced 1 (IL4I1) was up-regulated and risky in GBM, KIRC, KIRP, and LIHC, whereas down-regulated and risky in LAML. Primarily, we found several GMPGs were especially highly up-regulated and risky in only one tumor, such as GCLM in BLCA, GLS in LIHC, and IDH1 in LGG (Fig. [93]1B). It’s worth noting that differentially expressed genes highly expressed in the tumor will lead to high risk or low risk. Subsequently, we further combined the expression data with poor prognosis analysis together, then formed the Sankey plot (Fig. [94]1C and Table S3), selected the data from left heatmap, we distinctly detected that G6PD was significantly upregulated in BLCA (logFC = 1.06, p-value = 1.00E-03), KIRP (logFC = 1.58, p-value = 1.20E-12) and LIHC (logFC = 2.60, p-value = 3.59E-32) compared cancer to normal tissues. LIHC contained six GMPGs (IL4I1, Glutaminase (GLS), Glutamate decarboxylase 1 (GAD1), G6PD, CAD, and ASNS) which were up-regulated and risky in cancer (Fig. [95]1C). However, IL4I1 was up-regulated and poor prognosis in four cancers (LIHC, KIRP, KIRC, and GBM). CeRNA discovery We used MultiMiR v2.3 and LnCeVar to discover validated ceRNAs in these 18 GMPGs; only G6PD and GLS found 9 pairs of ceRNAs. Then we used MiPepid to find potential micropeptide in these 9 lncRNAs; LINC00869 and [96]AC012146.7 did not exist micropeptide, so we marked it as a grey triangle. Next, smORFunction was selected to predict potential micropeptides, AC145207.5, PROSER2-AS1, CCDC18-AS1, and NEAT1 were not predicted as functional micropeptides, so we marked them as a blue triangle. G6PD / hsa-miR-1-3p / RBM26-AS1, GLS / hsa-miR-23a-3p / STAG3L5P-PVRIG2P-PILRB, GLS / hsa-miR-23a-3p / VPS9D1-AS1, GLS / hsa-miR-23a-3p / [97]AC078846.1 axis validated by MultiMiR, we visualized the human chromosome to lncRNAs by “RIdeogram” (Fig. [98]1C). [99]AC078846.1, STAG3L5P-PVRIG2P-PILRB, were located in chromosome 7. We chose iLoc-mRNA to predict the subcellular location of GLS and G6PD, GLS was located in cytoplasm and cytosol, and G6PD was located in the ribosome. We validated the expected site of these two genes by RNALocate. GLS was validated to find in the cytoplasm and cytosol of HeLa cells by 44 k oligoarray, and the cytoplasm and cytosol of HepG2 and K562 cells by CSCD. These three lncRNAs (STAG3L5P-PVRIG2P-PILRB, RBM26-AS1, [100]AC078846.1) were located in the cytoplasm, and cytosol VPS9D1-AS1 was found in the ribosome. By next-generation sequencing, VPS9D1-AS1 was located in the ribosome of the colon cancer cell. CSCD validated the localization of STAG3L5P-PVRIG2P-PILRB in cytoplasm and cytosol of HeLa-S3 (Table S4). Micropeptide prediction and chromosome location MiPepid was used to predict 27 micropeptides STAG3L5P-PVRIG2P-PILRB, 20 micropeptides [101]AC078846.1, 7 micropeptides VPS9D1-AS1 and 4 micropeptides RBM26-AS1, which range from 8 to 100 amino acids. Subsequently, smORFunction was selected to identify the functional enrichment of the predicted micropeptides; we filtered 23 micropeptides mappings with sORFunction prediction ID in LIHC (Table S5). Then we integrated the enriched KEGG pathway (Table S6) to assess the potential function of micropeptides in LIHC. For example, micropeptide STAG3L5P-PVRIG2P-PILRB was related to cell adhesion molecules cams and allograft rejection, micropeptide [102]AC078846.1 and VPS9D1-AS1 were connected with neuroactive ligand-receptor interaction, micropeptide RBM26-AS1 was related to the spliceosome. DM3Loc was chosen to predict the location of lncRNA encoded with micropeptides. STAG3L5P-PVRIG2P-PILRB-57aa, VPS9D1-AS1-26aa, [103]AC078846.1-25aa and [104]AC078846.1-36aa were located in the ribosome, [105]AC078846.1-47aa and [106]AC078846.1-64aa were located in ribosome or exosome (Table S4). We chose iLoc-mRNA to predict the subcellular location of GLS and G6PD, GLS was located in cytoplasm and cytosol, and G6PD was located in the ribosome. We validated the expected site of these two genes by RNALocate. GLS was validated to find in the cytoplasm and cytosol of HeLa cells by 44 k oligoarray, and the cytoplasm and cytosol of HepG2 and K562 cells by CSCD. These three lncRNAs (STAG3L5P-PVRIG2P-PILRB, RBM26-AS1, [107]AC078846.1) were located in the cytoplasm, and cytosol VPS9D1-AS1 was found in the ribosome. By next-generation sequencing, VPS9D1-AS1 was located in the ribosome of the colon cancer cell. CSCD validated the localization of STAG3L5P-PVRIG2P-PILRB in cytoplasm and cytosol of HeLa-S3 (Table S4). Based on the visualization of chromosome location among the data for lncRNA-micropeptide-smORFunction (Table S7), we visualized the human chromosome to the predicted micropeptides and related lncRNAs by “RIdeogram” (Fig. [108]1D). [109]AC078846.1, STAG3L5P-PVRIG2P-PILRB, and their encoded micropeptides were located in chromosome 7. Their results of smORFunction prediction in LIHC were also found in the same chromosome 7, RBM26-AS1, and its encoded micropeptide was located in chromosome 13, VPS9D1-AS1 encoded micropeptide were located in chromosome 16. Therefore, we speculated that these micropeptides encoded by lncRNAs might play essential roles in liver cancer. The analysis of GMPGs in LIHC We have discovered the particularity of GMPGs in liver cancer, so subsequent, and we will analyse the role of GMPGs in liver cancer. To compare the transcriptomes of LIHC and normal tissues, T1 and T4, we performed the PCA of LIHC and normal tissues (Fig. [110]2A) of T1 and T4 (Fig. [111]2B). Based on the volcano plots (|logFC|> 1, p-value < 0.05), we found eight DE-GMPGs (GAD1, G6PD, CAD, ASNS, GLUL, IL4I1, ALDH4A1, and GLS) were down-regulated in normal tissue (Fig. [112]2C), five DE-GMPGs (IL4I1, GCLM, G6PD, PPAT, and GSR) were down-regulated in T1 (Fig. [113]2D). Then, we used the Venn diagram to integrate the common DE-GMPGs in LIHC, and 10 DE-GMPGs were detected in normal VS tumors and T1 VS T4 (Fig. [114]2E). The forest plot displayed the common GMPGs in LIHC. Through Uni-Cox regression, G6PD, ALDH5A1, GAD1, GPT, and IL4I1 were confirmed as survival-related genes (p-value < 0.05) (Fig. [115]2F), among the G6PD and GAD1, were high-risk GMPGs with a hazard ratio (HR) from 1.048 to 1.543. Fig. 2. [116]Fig. 2 [117]Open in a new tab The selection of the critical GMPGs in LIHC. A PCA plot of the transcriptomes of LIHC and normal tissues. B PCA plot of the transcriptomes of T1 and T4. C The volcano plot of DE-GMPGs in normal tissues VS tumor. D The volcano plot of DE-GMPGs in T1 VS T2. E Venn diagram of common GMPGs in normal tissue VS tumor and LIHC T1 VS T4. F The forest plot of common GMPGs in LIHC Subsequently, we applied the randomForestSRC to feature selection with a relative importance > 0.2, and drew the correction of the number of trees and the error rate (Fig. [118]3A), and then, ranked the order of the out-of-bag importance of the five GMPGs (Fig. [119]3B). As expected, tumors with higher expressed G6PD demonstrated poorer prognosis than lower expressed G6PD (p-value < 0.001, HR = 2.64, 95% confidence interval (CI) = [1.74–4.01]) (Fig. [120]3C); whereas tumors with lower expressed Glutamic-pyruvic transaminase (GPT) demonstrated poorer prognosis than higher expressed GPT (p-value = 0.002, HR = 0.57, 95%CI = [0.39–0.85]) (Fig. [121]3D). Fig. 3. [122]Fig. 3 [123]Open in a new tab Random forest and survival analysis. A The error rate of random forest. B The rank of the out-of-bag importance for the five GMPGs. Association between G6PD (C) and GPT D in LIHC for overall survival Eight clinical factors (grade, M, N, T, age, gender, stage, and risk score) with two key GMPGs were included in the analysis. Based on the univariate cox (unicox) and survival analysis, it revealed that T, stage and risk score were significant independent factors to overall survival (OS) of LIHC patients (p-value < 0.001) (Fig. S3A). Then, through multivariate cox (multicox) analysis, the only risk score was statistically significant (p-value < 0.001) (Fig. S3B). HR risk scores are 1.953 (95%CI [1.535–2.485]), p-value < 0.001). The prognostic accuracy of the risk score was confirmed by the ROC curve (AUC = 0.799) (Fig. S3C). G6PD expression was elevated as the risk score increased, while GPT expressed decreasingly as the risk score increased (Fig. S3D). ROC curves analyses indicating the discriminatory potential of G6PD and GPT (Fig. S3E), we finally confirmed G6PD (AUC = 0.763) as the key GMPG in LIHC. We systematically used the single-sample gene set enrichment analysis scores to visualize the immune landscape of TCGA-LIHC (371 patients) from 24 immune cell types by heatmap (Fig. [124]4A). The annotated below introduces the immune infiltration, relative family cancer history, gender, neoplasm histologic grade, and survival. The relationship of G6PD and significantly immune gene sets were identified and visualized within a lollipop chart in LIHC (p-value < 0.05) (Fig. [125]4B). We found the correlation of G6PD and T helper 2 (Th2) cells was significantly positive (r = 0.48, p-value < 2.71E-22) (Fig. [126]4C). G6PD and plasmacytoid dendritic cell (pDC) were negatively correlated (r = -0.46, p-value < 1.73E-20) (Fig. [127]4D). Fig. 4. [128]Fig. 4 [129]Open in a new tab The immune landscape of TCGA-LIHC. A The single-sample gene set enrichment analysis scores to visualize the immune landscape of TCGA-LIHC (371 patients) from 24 immune cell types by heatmap. B The lollipop chart in LIHC for the relationship of G6PD and significantly immune gene sets (p-value < 0.05). C Correlation between G6PD and Th2.cells. D Correlation between G6PD and pDC We found the highly expressed G6PD was positively related to females and males (blue) and negatively associatedwith the LIHC stage 1 and stage 2 (red) (Fig. [130]5A). The immunohistochemical detection of G6PD expression in LIHC patients was downloaded from the HPA database (Fig. [131]5B). We further calculated the correlation between G6PD and GMPGs, GCLM, ADSS, ASNS, CAD, GSS, GSR, IL4I1, GLS, PPAT and ADSL, were significantly positively correlated with G6PD, and G6PD, GCLM, GSS and GSR were also enriched in glutathione metabolism (hsa00480) (Fig. S1); whereas CPS1, ALDH5A1, GPT, AGXT and ABAT were significantly negatively correlated with G6PD (Fig. [132]5C). Fig. 5. [133]Fig. 5 [134]Open in a new tab Correlation analysis between G6PD and clinical characteristics. (A) The mosaic map displays the correlation of G6PD and clinical features. (B) The immunohistochemical detection of G6PD expression in LIHC patients. (C) The correlation between G6PD and other genes Knocking down G6PD expression effectively inhibits the proliferation, migration, and invasion capabilities of hepatocellular carcinoma cells To investigate the role of G6PD in hepatocellular carcinoma, we initially performed a transfection knockdown of the G6PD gene and verified the efficiency of this transfection via qRT-PCR (Fig. [135]6A). The results indicated a significant reduction in G6PD mRNA expression in Huh7 cells within the knockdown group post-transfection. Subsequently, we employed a CCK-8 assay to assess the effects of G6PD depletion on the proliferative capabilities of hepatocellular carcinoma cells (Fig. [136]6B), revealing a markedly lower proliferative rate in the si-G6PD group compared to the NC group. Additional transwell assays corroborated these findings, demonstrating that G6PD knockdown reduced the migration and invasion abilities of these cells (Fig. [137]6C–E). Collectively, these data suggest that G6PD knockdown impairs the proliferation, migration, and invasion of hepatocellular carcinoma cells. Fig. 6. [138]Fig. 6 [139]Open in a new tab Knockdown of G6PD can inhibit the proliferation, migration, and invasion of hepatocellular carcinoma cells. A RT-qPCR experiments confirmed the inhibitory effect of knocking down G6PD on mRNA expression in Huh7 cells. B CCK-8 assay were conducted on Huh7 cells from NC group and si-G6PD group, indicating that knocking down G6PD inhibits the proliferation of hepatocellular carcinoma cells. C–E Transwell experiments showed that compared to the NC group, the migration ability of Huh7 cells in the si-G6PD group decreased, and the number of invading cells decreased. Compared with the NC group, *p < 0.05, **P < 0.01, ***P < 0.001, and **** p < 0.0001 Discussion The multi-cancer omics studies integrated different landscapes of molecular data to detect the similarities and differences in single or multiple tumors. We hypothesized that GMPGs were potential tumor-specific mediators and biomarkers of clinical features. To test these, we first combined transcriptome data from TCGA and GTEx to identify the DE-GMPGs with their prognostic value to predict patient survival after comparing the GMPGs in 32 tumors, which elucidate potential biomarkers of glutathione within the tumor microenvironment. We found several DE-GMPGs in multi-tumors, such as IL4I1 was up-regulated and risky in GBM, KIRC, KIRP, and LIHC. IL4I1 was a metabolic immune checkpoint and up-regulated in GBM [[140]44], which could promote tumor progression in GBM [[141]45], and IL4I1-driven AHR was a new signature in GBM as well [[142]46]. GCLM was involved in (anti)oxidant metabolism and potential therapeutic targets and prognostic markers for BLCA, which played a vital role in the pathogenesis [[143]47], we detected that GCLM was significantly highly up-regulated and risky only in BLCA in this study. We specifically and systematically studied the transcriptome and clinical data and found that the critical GMPG in LIHC was G6PD. G6PD is the key rate-controlling enzyme performed as an oncogene in bladder, liver, and breast cancer [[144]48, [145]49]. In rodent models, elevated G6PD expression correlates with an increased prevalence and severity of hepatic precancerous conditions, highlighting its pivotal role in the pathogenesis of liver cancer [[146]50, [147]51]. Th2 cells were reported to focus on humoral immunity [[148]52], and inhibit Th1 cells' activity from attenuating the macrophage response [[149]53]. G6PD was positively connected with Th2 cells in LIHC, which might affect the effect of Th2 cells on macrophages and humoral immunity. G6PD is prognostic, high expression is unfavorable in liver cancer in the HPA database ([150]https://www.proteinatlas.org/ENSG00000160211-G6PD/pathology), G6PD protein expression was also highly expressed in LIHC [[151]54]. The expression of G6PD was upregulated in LIHC patients (compared with normal patients) and T4 (compared with T1). We also detected that G6PD was a high-risk gene in LIHC. We identified and validated that G6PD played a crucial role in the occurrence and progression of LIHC [[152]55, [153]56]. Non-coding RNAs (circRNA, lncRNA, and pri-miRNA) have potential short open reading frames and can encode micropeptides. Short peptides encoded by non-coding RNAs are closely related to tumors and can be used as potential prognostic markers and therapeutic targets for tumors [[154]57]. G6PD was marked as a hub gene for metabolic reprogramming and redox signaling in cancer [[155]58], and could be inhibited by aldolase B to suppress LIHC cells [[156]59]. LIHC contained G6PD, which was up-regulated and risky for cancer. By analyzing the potential micropeptides, we predicted RBM26-AS1 encoded a micropeptide to target G6PD and hoped that micropeptide RBM26-AS1 could become a new small molecule drug for the treatment of LIHC. Numerous studies have revealed evidence of G6PD and endogenous microRNA sponges. LINC00242 / miR-1-3p / G6PD axis affected the proliferation and apoptosis of gastric cancer cells by regulating the Warburg effect [[157]60]. LncRNA SNHG14 / miR-206 / G6PD pathway contributes to the progression of NSCLC in lung cancer [[158]61]. This study also discovered that G6PD / hsa-miR-1-3p / RBM26-AS1 axis and predicted micropeptide RBM26-AS1 might be a potential treatment to LIHC micropeptide RBM26-AS1 was related to the spliceosome, RBM26-AS1 and its encoded micropeptide were located in chromosome 13, which need more research. Therefore, we speculated that these micropeptides encoded by lncRNAs might play essential roles in liver cancer. G6PD was marked as a hub gene for metabolic reprogramming and redox signaling in cancer [[159]58], and could be inhibited by aldolase B to suppress LIHC cells [[160]59]. LIHC contained G6PD, which was up-regulated and risky for cancer. G6PD exhibits oncogene-like properties in the hepatocellular carcinoma cell line Huh7 by promoting cell proliferation, migration, and invasion in this study. The observed inhibitory effects of G6PD knockdown on these cellular behaviors suggest that targeting G6PD could represent a promising therapeutic strategy. In summary, our findings provide new insights and clues for the treatment and prognosis of LIHC; however, comprehensive studies are necessary to elucidate G6PD's mechanisms fully. This should include detailed investigations into G6PD's specific roles in the development and progression of LIHC and its interactions with other related molecules. Additionally, further experimental and clinical studies are essential to confirm the viability and effectiveness of G6PD targeting as a therapeutic option, potentially leading to significant advancements in cancer treatment and meriting continued research. Conclusions Based on the systematic overview of GMPGs in pan-cancer for the first time, our study highlights the potential of multi-cancer GMPGs as biomarkers for detecting the molecular mechanisms in cancer progression and developing more potential immunotherapy strategies. No mutation in G6PD was detected in LIHC patients, which revealed that G6PD gene status seemed to be highly stable during the evolution of LIHC. G6PD was positively connected with Th2 cells in LIHC, which might affect the effect of Th2 cells on macrophages and humoral immunity. Knockdown of G6PD inhibits the proliferation, migration, and invasion of hepatocellular carcinoma cells. Our findings offer valuable insights for the detection, treatment, and mechanistic understanding of LIHC. We predicted that micropeptide RBM26-AS1 was a new player involved in LIHC and interacted with G6PD in LIHC, regulating immune responses in the immune system. Supplementary Information [161]Additional file1 (JPG 905 KB)^ (904.6KB, jpg) [162]Additional file2 (JPG 3955 KB)^ (3.9MB, jpg) [163]Additional file3 (JPG 2595 KB)^ (2.5MB, jpg) [164]Additional file4 (XLSX 47 KB)^ (47KB, xlsx) Acknowledgements