Transcriptomics combined with in vitro experimental validation probes polyamine metabolic profiles in low-grade gliomas

Introduction

Glioma originates in the glial cells within the brain (1), with an average global incidence of 5–8 cases per 100,000 people. Based on histological criteria and clinical behavior, gliomas are predominantly stratified into two major subtypes: low-grade gliomas (LGG) and glioblastomas (GBM). Although LGGs are slow-growing and have a better prognosis, they have a high recurrence rate and are frequently progress to develop into high-grade gliomas due to the infiltrative growth of glioma cells, which makes complete surgical removal of the tumor tissue impossible (2,3). LGGs have a very high recurrence rate and are at high risk of developing into high-grade gliomas. Although considerable therapeutic advances have been achieved in managing LGG over the past decade, clinical outcomes remain remarkably heterogeneous among patients, attributable to intertumoral genomic variance even when patients are managed under standardized protocols (4-9).

Emerging evidence from multi-omics investigations demonstrates that oncogenesis involves systematic metabolic reprogramming (10). For example, lipid metabolism has been shown to promote the growth and metastasis of prostate cancer (11); excessive activation of the serine/glycine biosynthetic pathway is closely linked to tumorigenesis, highlighting its critical role in tumor development (12). Polyamines, including putrescine, spermidine, and spermine, are promoters of cell regeneration and proliferation, present in both normal and cancerous cells, and play an essential part in diverse physiological activities (13). Biogenic polyamines (putrescine, spermidine, spermine) critically regulate cellular proliferation and oncogenic transformation by modulating cell cycle progression, and elevated polyamine levels are associated with immunosuppression. Ornithine decarboxylase (ODC) is a key enzyme in polyamine metabolism, and one study has shown that it is transcriptionally regulated by the myelocytomatosis oncogene (MYC) (14). Increased MYC expression is demonstrated in most cancer types, and positively correlates with enhanced polyamine biosynthesis via ODC (15). ODC is a transcriptional trigger for MYC cancer generators. Polyamines are also engaged in the regulation of cellular functionality, such as genetic expression and programmed cell death, and possess anti-inflammatory and antioxidant properties, which are closely associated with cancer development (16). In addition, spermine plays an indispensable role in shielding DNA from reactive oxygen species (ROS) damage, reducing subsequent mutations, and preventing potential carcinogenesis. In carcinoma, polyamine metabolism is often aberrantly regulated, principally by enhancing the activity of the enzyme polyamine biosynthesis. This leads to an elevation of polyamine levels, which consequently supports both malignant transformation and tumor progression. Therefore, polyamine metabolic pathways are an attractive target for anticancer therapy and are closely linked to cancer progression. In the most recent studies, it has been shown that spermine synthase (SMS) is associated with the tumorigenesis in multiple cancers, and that high levels of SMS expression correlate with poor prognosis (17-19). However, the possible functions of SMS in LGG progression and its potential regulatory mechanisms remain incompletely explored.

In this research, we first classified the LGG samples into two distinct subtypes using cluster analysis of genes associated with polyamine metabolism. Next, we compared the two subtypes in terms of immune cell immersion, gene mutations, and drug responsiveness. Nine genes linked to LGG patient prognosis were identified via least absolute shrinkage and selection operator (LASSO) and Cox regression analyses, leading to the development of polyamine metabolism-related gene risk score (PRGRS). Additionally, we examined the role of SMS in LGG oncogenesis progression through in vitro experiments. SMS could serve as a prospective prognostic biological marker and treatment candidate for LGG patients. We present this article in accordance with the TRIPOD reporting checklist (available at https://tcr.amegroups.com/article/view/10.21037/tcr-2025-776/rc).

Methods

Data collection and processing

Transcriptomic and clinical data for LGG patients were obtained from The Cancer Genome Atlas (TCGA) website (https://portal.gdc.cancer.gov/). After excluding samples lacking survival data, a cohort of 506 LGG patients was retained for further analysis (table available at https://cdn.amegroups.cn/static/public/tcr-2025-776-1.xlsx). RNA-seq and clinical data for the Chinese Glioma Genome Atlas (CGGA)-693 and CGGA-325 cohorts were obtained from the CGGA database (http://www.cgga.org.cn) as a validation cohort (20). We obtained from the Molecular Signatures Database (MSigDB) (https://www.gsea-msigdb.org/gsea/msigdb) database (“Human Gene Set: REACTOME_METABOLISM_OF_POLYAMINES”) and included a total of 59 polyamine metabolism-related genes (Table S1). On the basis of the patients’ overall survival (OS), we first conducted a single-variable Cox regression analysis and screened 36 prognostically relevant polyamine metabolism-related genes for subsequent analysis based on P<0.05. The study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments.

Consistency cluster analysis

Using the “ConsensusClusterPlus” R package (reps =500, pItem =0.8, pFeature =1), we grouped the TCGA-LGG cohort based on the expression of 36 prognosis-linked genes. After integrating the consistency matrix, cumulative distribution function (CDF), and proportion of ambiguous clusters (PAC), evaluating k=2 and k=3, we established that k=2 is the preferred number of clusters. To conclude, we used principal component analysis (PCA) to assess the robustness of the clustering and the “pheatmap” R package to illustrate the pattern of expression of genes concerned with polyamine metabolism between the two molecular subtypes.

Development of prognostic characteristics related to polyamine metabolism

We utilized the LASSO regression and progressive Cox regression for our analyses. Ultimately, nine crucial genes were picked as model genes, and the PRGRS was derived for each LGG patient on the basis of risk coefficients and gene expression. The risk score was calculated as:

PRGRS=SMS∗0.67+PSMC2∗0.19+PSMD12∗0.67+PSMB9∗0.2−PSMB5∗0.47+PSMD5∗0.34+PSMF1∗0.15+PSMD14∗0.04−OAZ3∗0.52[1]

With the threshold value of the intermediate risk score, patients were stratified into high- versus low-risk groups.

Tumor microenvironment (TME) assessment of molecular subtypes

Integrative computational deconvolution of tumor-infiltrating lymphocyte profiling enabled systematic characterization of immunophenotypic divergence across molecularly defined glioma subtypes. The ESTIMATE algorithm was employed to systematically quantify immune infiltration differentials between molecular subtypes through bulk transcriptomic data. Subsequently, single-sample gene set enrichment analysis (ssGSEA) was employed to evaluate the distribution of immune cell subsets between the two subtypes. In addition, we used Cell Type Identification By Estimating Relative Subsets Of RNA Transcripts (CIBERSORT), MicroCell Population counter (MCP-counter), Estimating the Proportions of Immune and Cancer cells (EPIC), Tumor IMmune Estimation Resource (TIMER), and Xcell to comprehensively assess the TME profile (21). Finally, we concentrated on the expression of major regulatory factors in the TME across the two subtypes, and we collected genes closely related to anti-tumor immunity from the previous research (22) and assessed the expression differences between the two metabolic isoforms.

Tumor mutation status and genomic variant analysis

Further single-nucleotide variant (SNV) data were obtained for the TCGA-LGG cohort, and the top 20 genes with the highest mutation frequency differences between the two subtypes were identified and were visualized with the “maftools” R software package. One study has found that tumor mutation burden (TMB) can be utilized as a biological marker for immunotherapy (23). We prognostically stratified the TCGA-LGG cohort based on median TMB combined with polyamine metabolism-related molecular subtypes, respectively, and the prognostic discrepancy between the respective groups was also tested by utilizing the Kaplan-Meier survival curves.

Drug sensitivity prediction

We used the “oncoPredict” R package to predict the half maximal inhibitory concentration (IC50) of common clinical chemotherapeutic agents in patients with different molecular subtypes of LGG (24). The drug information was obtained from Cancer Drug Sensitivity Genomics (GDSC2). Temozolomide (TMZ) is still the first-line drug commonly used clinically for the treatment of LGG patients, and potentially effective small-molecule drugs need to be further screened. We obtained drug information based on the Cancer Treatment Response Portal (CTRP) and Profiling Relative Inhibition Simultaneously in Mixtures (PRISM) count and screened potential chemotherapeutic agents based on correlation analysis.

Functional enrichment analysis and exploration of potential molecular mechanisms

We utilized the “limma” package to recognize differentially expressed genes (DEGs) between subtypes and risk groups, with thresholds of |log₂fold change (FC)| >1 and P<0.05 (25). Transcriptomic stratification analysis employing the limma algorithm identified subtype-specific and risk-associated differentially expressed transcripts. Subsequent functional annotation via clusterProfiler revealed enriched biological processes through Gene Ontology (GO) analysis, while Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway and gene set enrichment analysis (GSEA) were conducted using curated gene sets (c2.cp.kegg.v7.5.1) from the MSigDB (26). Additionally, we leveraged the “h.all.v2023.1.Hs.symbols.gmt” gene set from the MSigDB database to conduct gene set variation analysis (GSVA) on differential genes and assess the correlation among key modeling genes and the 50 marker pathways to simulate the potential oncogenic mechanism of the modeling genes. oncogenic mechanism of action of the modeled genes.

Clinical characterization and nomogram construction

To refine the clinical orientation of the molecular typing and risk grouping results, we selected age, OS, isocitric dehydrogenase (IDH) mutation status, and chromosomal deletion status as important clinical features to be compared among different subgroups and subtypes. We determined independent prognostic factors linked to OS in LGG patients using univariate and multivariate Cox regression analyses. In addition, we analyzed the clinical significance of risk subgroups using survival curves in different clinical characteristic subgroups. Finally, we used the “rms” R package to complete the construction of nomograms and evaluate the accuracy of predicting 1-, 3- and 5-year survival. Calibration curves and clinical decision curve analysis (DCA) were used to refine and evaluate the clinical efficacy of the nomogram.

Immunotherapy response assessment

IMvigor210 is a clinical trial of a programmed death ligand 1 (PD-L1) blocker (atezolizumab) in the treatment of locally advanced or metastatic uroepithelial cancer. The cohort has been used to forecast the efficacy of immunotherapy for a number of types of cancer, such as gastric and bladder cancers (27). We used key modeling genes in the immunotherapy cohort to simulate risk groupings in the TCGA-LGG cohort. Survival curves were used to validate prognostic differences between risk subgroups and predict response outcomes to PD-L1 blocker regimens. To further assess the predictive power of PRGRS for immunotherapy efficacy, we also performed the same analysis in the anti-programmed cell death protein 1 (PD-1) (pembrolizumab)-treated melanoma cohort, GSE78220, and the T-cell therapy [adoptive cell therapy (ACT)]-treated melanoma cohort, GSE100797 (28,29).

Cell cultivation and transfection

Neuroglioma cells U251 and LN229 were obtained from the Cell Bank of the Chinese Academy of Sciences and employed for in vitro experiments. Cells were maintained in Dulbecco’s Modified Eagle’s Medium containing 10% fetal bovine serum (FBS) and 1% penicillin-streptomycin. SMS-targeting small interfering RNA (siRNA) and control siRNA were procured from Gemma Genetics. U251 and LN229 cells were transfected with siRNA using Lipofectamine 2000 for 12 hours and then tested for function.

Real-time quantitative polymerase chain reaction (RT-qPCR) and Western blotting

RNA was extracted from si-SMS and NC-transfected glioma cells (LN229, U251). cDNA was synthesized using Synergetic Binding Reagent (SYBR) Green qPCR mixture for PCR. For protein analysis, cells were lysed with radioimmunoprecipitation assay buffer (RIPA) containing phenylmethanesulfonyl fluoride (PMSF). Proteins are detached using 10% Sodium dodecyl sulfate polyacrylamide gel electrophoresis (SDS-PAGE) and transported to a PolyVinylideneFluoride (PVDF) membrane. The membrane was incubated overnight at 4 °C with a primary antibody for 15 minutes after membrane blocking and then incubated with a secondary antibody for 2 hours.

Proliferation and colony formation experiments

Proliferation kinetic profiling of SMS-silenced glioma cells (U251/LN229) was performed through standardized Cell Counting Kit-8 (CCK-8) assays. Cells transfected with SMS-targeting siRNA for 24 h prior to 96-well plate seeding (3×10³/well) underwent optical density (OD) measurements (OD450, BioTek, California, USA) at 4–72 h intervals post-treatment using 10% CCK-8 reagent. Clonogenic assays involved 14-day culture of 2×10³ cells followed by 4% paraformaldehyde fixation and 1% crystal violet quantification of viable colonies (>50 cells/cluster).

Transwell migration, invasion and wound healing assays

U251 and LN229 were transfected with SMS siRNA for 24 h and then cultivated in 24-well plates with 8-µm well plugs to assess both migration and invasion. For Transwell assays, 4×10⁴ cells were seeded into the superior chamber with 200 µL of supernatant-free medium and 800 µL of medium containing 10% FBS was applied to the lower chamber. After 48 hours, migrated cells were fixed, stained, and counted under a light microscope (200×). For scratch assays, cells were cultured in 24-well plates, scratched using a 200-µL pipette tip, and cultured in FBS-free Dulbecco’s Modified Eagle Medium (DMEM). Wound closure was imaged at 0 and 24 h and quantified using ImageJ (40×).

Statistical analysis

R software (version 4.0.2) and GraphPad were employed for data analysis and visualization. The Wilcoxon test was used for comparison of the variance between the two groups. Kaplan-Meier curves and clinical characteristics were characterized by the log-rank test. Materiality of cell line assays was ascertained by t-test in GraphPad Prism 9. Significance levels were set at P<0.05.

Results

Identification of polyamine metabolism-related isoforms

From 59 polyamine-response genes curated in MSigDB, univariate Cox regression identified 36 prognostic biomarkers, with a forest plot highlighting top 15 significant candidates (Figure 1A). Consensus clustering based on these 36-polyamine metabolism-related genes separated LGG samples into C1 and C2 subgroups (Figure 1B). CDF plots showed that the slopes of the CDF curves in the range of the consensus indices (0.1–0.9) were relatively small at k=2 (Figure 1C). t-stochastic neighbor embedding (t-SNE) analysis showed uniform distribution of LGG patients between the two clusters (Figure 1D). A heatmap revealed differential expression of polyamine metabolism-related genes, including downregulation in C1 and upregulation in C2 (Figure 1E). Next, we analyzed by GSEA enrichment in the C2 subgroup, where we enriched the polyamine metabolism pathway (Figure 1F). Kaplan-Meier analysis demonstrated significantly better prognosis in C1 (Figure 1G).

Figure 1 Identification of polyamine metabolism-related isoforms. (A) Forest plot of univariate regression analysis for LGG patients in the TCGA-LGG dataset. (B,C) Consistent heat map matrix and CDF plots for two clusters. (D) The t-SNE downscaling plot between subtypes. (E) Comparison of polyamine metabolism-related gene expression levels across the two subtypes. (F) GSEA analysis demonstrating the polyamine metabolic pathway in subtype C2. (G) Survival analysis revealed a more favorable prognosis for C1. CDF, cumulative distribution function; CI, confidence interval; GSEA, gene set enrichment analysis; HR, hazard ratio; LGG, low-grade gliomas; t-SNE, t-stochastic neighbor embedding; TCGA, The Cancer Genome Atlas.

TME characterization of polyamine metabolism-related subtypes

We assessed the aggregate immune immersion between the two subgroups using the ESTIMATE algorithm, which evaluates stromal scores, ESTIMATE scores, immune scores, and tumor purity. The C2 subgroup had greater stromal scores, greater immune scores, and lower tumor purity. This evidence indicated that the C2 subgroup had a greater immune infiltration than the C1 subgroup (Figure 2A-2D). We had gained insight into the pattern of immune infiltration in the different subtypes by analyzing the relationship between the two subtypes and 28 immune cell subpopulations using the ssGSEA method. Significantly increased infiltration of activated CD4 T cells, activated CD8 T cells, and effector memory CD8 T cells, macrophages, myeloid-derived suppressor cells (MDSC), and regulatory T cells were observed in the C2 subgroup (Figure 2E). Further immune immersion was analyzed and showed that the C2 subtype had a more advanced immune cell infiltration level, with macrophages and their M1 and M2 subpopulations significantly elevated compared to C1 (Figure 2F). In addition, C2 exhibited elevated expression of antigen presentation and immune checkpoint genes (Figure 2G).

Figure 2 Immunologic microenvironment of distinct molecular subtypes. (A-D) Immunity scores, ESTIMATE scores, stromal scores and tumor purity were used to characterize the different immune status between subtypes. (C) ssGSEA-quantified immune landscape stratification reveals subtype-specific leukocyte infiltration gradients (F) Immune infiltrating cells were assessed via multiple algorithms across two distinct subgroups. (G) Differences in the expression of immunomodulator molecules between two different subgroups. ns, P>0.05; ***, P<0.001; ****, P<0.001. CIBERSORT, Cell Type Identification By Estimating Relative Subsets Of RNA Transcripts; EPIC, Estimating the Proportions of Immune and Cancer cells; MCP-counter, MicroCell Population counter; ssGSEA, single-sample gene set enrichment analysis; TIMER, Tumor IMmune Estimation Resource.

Mutational landscape differences between subtypes

In the TCGA-LGG cohort, somatic mutation distributions between the low-risk and high-risk groups were analyzed (Figure 3A,3B) to investigate their association with polyamine metabolism and LGG development. Compared with patients in the C1 subgroup, patients in the C2 subgroup had a significantly higher overall mutation frequency (95.03% vs. 96.2%), with notable differences in IDH1 (85% vs. 64%), TP53 (50% vs. 43%), ATRX (39% vs. 32%), and CIC (25% vs. 18%). A recent study had found Tumor mutation burden (TMB) to be a promising biomarker in gliomas and may be associated with predicting immunotherapy response (30). C2 subtype patients exhibited elevated TMB levels (Figure 3C), and LGG patients with low TMB exhibited a significant prognostic advantage (Figure 3D). In addition, combining molecular subtypes with TMB could effectively stratify patient prognosis (Figure 3E).

Figure 3 Differences in mutational landscapes between subtypes. (A,B) Changes in somatic mutation distribution between subtypes. (C) Differences in TMB between subtypes. (D) Survival prognosis comparison between high and low TMB groups. (E) Survival differences combining TMB and subtype. TMB, tumor mutation burden.

Pathway enrichment and drug sensitivity analysis

We conducted functional enrichment analyses of the subgroup DEGs to explore altered biological processes associated with polyamine metabolism-related genes. The results showed that the C1 subgroup’s biological processes exhibited enrichment in synaptic transmission and vesicle trafficking. Cellular components were primarily linked to synaptic membranes, neuronal synapses, and projections. Molecular functions (MFs) were predominantly enriched in calcium-dependent binding proteins, SNARE binding (Figure 4A). Biological process (BP) demonstrated enrichment of C2 isoforms for antigen processing and presentation. Cellular components (CC) were shown to be associated with major histocompatibility complex (MHC) protein complexes, endoplasmic reticulum (ER)-to-Golgi transport vesicles, and others. MFs were mainly enriched for peptide binding, antigen binding, and MHC protein complex binding (Figure 4B). GSVA analysis based on the Hallmark gene set revealed significant enrichment of immune-related pathways in C2 such as interferon-α/γ response, allograft rejection, IL2/STAT5 signaling pathway, IL6/JAK/STAT3 signaling pathway, angiogenesis, and inflammatory response. C1 was significantly enriched in pancreatic β-cells, the Hedgehog signaling pathway, KRAS signaling, and spermatogenesis (Figure 4C). To explore the differences in drug sensitivity between subtypes to guide clinical dosing, we used drug information from the GDSC database to assess the IC50 of common chemotherapeutic agents. Among them, drugs, such as gemcitabine, temozolomide, cisplatin, and dasatinib were more sensitive to patients in the C1 subgroup, lapatinib and carmustine are more sensitive to LGG patients in subgroup C2, and may have better efficacy (Figure 4D-4I).

Figure 4 Pathway enrichment and drug sensitivity assessment. (A,B) KEGG analysis of two molecular subtypes. (C) GSVA analysis two molecular subtype, red bar means positive correlation, blue bar means minus correlation. (D-I) Drug sensitivity analysis between C1 subgroup and C2 subgroup. *, P<0.05; **, P<0.01; ***, P<0.001. BP, biological process; CC, cellular component; GSVA, gene set variation analysis; KEGG, Kyoto Encyclopedia of Genes and Genomes; MF, molecular function.

Construction and validation of a prognostic signature related to polyamine metabolism

A machine learning-driven pipeline integrating LASSO regularization with stepwise Cox modeling identified 9 polyamine-responsive biomarkers for LGG risk stratification. Prognostic models were developed based on the expression characteristics of these 9 polyamine metabolism-related genes, of which 2 polyamine metabolism-related genes were used as protective factors and 7 polyamine metabolism-related genes were used as risk factors (Figure 5A-5C). Receiver operating characteristic (ROC) curve analysis showed that the area under the curve (AUC) of the TCGA training set exceeds 0.8 (Figure 5D), indicating good predictive performance results. According to the median risk score, Patients were stratified into high or low risk groups. Kaplan-Meier analysis revealed that OS was worse for the high-risk group (Figure 5E). The 5-year AUC values for the CGGA-325 cohort were 0.76, 0.83, 0.84, 0.85, and 0.84 (Figure 5F), and Kaplan-Meier analysis of the CGGA-325 cohort revealed that high-risk patients exhibited poorer OS (Figure 5G). We further validated themodel’s accuracy and reliability using the CGGA-693 cohort, achieving AUC values of 0.62, 0.66, 0.67, 0.7, and 0.7 (Figure 5H), in which the CGGA-693 Kaplan-Meier analysis showed that it had a similar result (Figure 5I). The above results establish the polyamine metabolism-related genes related prognostic model’s prognostic utility. In addition, we plotted Risk distribution plots. Patients in the high-risk group exhibited a higher mortality rate and shorter survival time, with differential gene expression between the risk subgroups (Figure 5J-5L).

Figure 5 Development and validation of prognostic features linked to polyamine metabolism. (A) 10-fold cross-validated LASSO regression for parameter optimization. (B) Coefficient screening under LASSO analysis. (C) Demonstration of the 9 key polyamine metabolism-related genes coef. (D) Time-independent ROC curve profiles in the TCGA-LGG cohort were analyzed. (E) The comparison of Kaplan-Meyer curves of LGG patients in the TCGA-LGG high and low risk groups. (F) Time-dependent ROC curve analysis in the CGGA-325 cohort. (G) Kaplan-Meyer curves of LGG patients in the CGGA-325 high-risk and low-risk groups. (H) Time-dependent ROC curve analysis in the CGGA-693 cohort. (I) Kaplan-Meyer curves of LGG patients in the CGGA-693 high- and low-risk groups. (J-L) Risk factor cascade diagram. AUC, area under the curve; CI, confidence interval; LASSO, least absolute shrinkage and selection operator; LGG, low-grade gliomas; TCGA, The Cancer Genome Atlas; CGGA, Chinese Glioma Genome Atlas; ROC, receiver operating characteristic.

Pathway alterations and drug sensitivity differences in risk groups associated with polyamine metabolism

Correlation evaluation demonstrated marked positive correlations of the modeled gene set with biological processes, including programmed cell death, immunological signaling, and cell cycle progression (Figure 6A,6B), with the highest correlations for SMS, PSMC2, and PSMB9, underscoring their pathogenic contributions to glioma evolution. GSEA analysis indicated that pathways associated with the cell cycle and DNA replication were significantly enriched in the high-risk group (Figure 6C). Risk scores were significantly positively correlated with immune-related pathways (T-cell receptor signaling pathway and Toll-like receptor signaling pathway), whereas risk scores were significantly inversely correlated with multiple metabolic pathways (lysine degradation, unsaturated fatty acid synthesis, glycerolipid metabolism, and taurine metabolism) (Figure 6D). In addition, we used CTRP-derived and PRISM-derived drug response data. PRISM-derived analyses identified 23 commonly used chemotherapeutic agents (Figure 6E,6F), and CTRP-derived compound analyses identified 5 therapeutic agents (Figure 6G,6H), all of which are likely to have better efficacy in the high-risk group. Atorvastatin and Fluvastatin had the highest correlation.

Figure 6 Differences in pathway enrichment and drug sensitivity between risk subgroups. (A-D) Pathway alterations in key modeling genes and inter-risk group pairs. (E-H) Lollipop and box-and-line plots demonstrating differences in drug sensitivity predicted by the CTRP and PRISM databases. ns, P>0.05; *, P<0.05; **, P<0.01; ***, P<0.001. AUC, area under the curve; CTRP, Cancer Treatment Response Portal; ECM, extracellular matrix; NES, normalized enrichment score; PRISM, Profiling Relative Inhibition Simultaneously in Mixtures; SMS, spermine synthase.

Integrated analysis of polyamine metabolic profiles with clinical features

Initial prognostic risk stratification revealed significant covariation with clinicopathological parameters, particularly age distribution, World Health Organization (WHO) grade, and molecular classification (IDH-wildtype prevalence) in the TCGA-LGG cohort (Figure 7A). The distribution of clinical features in the C1 and C2 subgroups had similar manifestations (Figure 7B). To assess the combined prognostic value of polyamine metabolic-related features and clinicopathologic characteristics, we conducted univariate and multivariate Cox regression analyses. The results demonstrated that the risk score served as an independent prognostic factor (Figure 7C,7D). Survival analyses demonstrated a more significant predictive effect of applying risk scores in patients with LGG G3-stage and age ≤65 years (Figure 7E-7G). To make the risk profile more suitable for clinical application, we integrated tumor grade and PRGRS to create a predictive nomogram (Figure 7H). Calibration plots demonstrated that the nomogram accurately predicted survival outcomes. (Figure 7I). The nomogram was more capable of identifying high-risk patients compared with other clinical features (Figure 7J).

Figure 7 Correlation between low and high-risk groups and clinical characteristics. (A,B) Correlations of risk groups as well as subtypes with clinical characteristics (0, alive; 1, dead). (C) Univariate and (D) multivariate Cox regression analysis for different clinical characteristics. (EG) Survival curves of LGG patients in the high- and low-risk groups with pathologic staging and age subgroups. (H) Column line graphs combining age, sex, grading, and risk score. (I) Calibration curves for the 1-, 3-, and 5-year nomograms constructed. (J) DCA decision curve analysis. ***, P<0.001. CI, confidence interval; CL, classical; DCA, decision curve analysis; IDH, isocitric dehydrogenase; LGG, low-grade gliomas; ME, mesenchymal; NE, neural; OS, overall survival; PN, proneural.

Immunotherapy response prediction

Multi-cohort validation revealed the prognostic signature’s clinical utility: high-risk stratification consistently predicted reduced OS across immunotherapy datasets. PRGRS was elevated in patients with stable disease (SD) or progressive disease (PD) compared to those with complete response (CR) or partial response (PR). Additionally, the proportion of SD/PD patients was higher in the high-risk group (Figure 8A). Survival differences between risk groups were more pronounced in stage III + IV patients than in stage I + II patients (Figure 8B,8C). Similar results were observed in the GSE78220 and GSE100797 cohorts (Figure 8D,8E). These findings suggest that PRGRS effectively predicts patient prognosis and immunotherapy response.

Figure 8 PRGRS predicts response to immunotherapy. (A) Comparison of Kaplan-Meyer curves for the high- and low-risk groups of the IMvigor210 cohort. Box plots show differences in risk scores and proportion of responses to immunotherapy in the IMvigor210 cohort. (B,C) Comparison of Kaplan-Meyer curves for early (I + II) and late (III + IV) patients in the IMvigor210 cohort. (D) Kaplan-Meier curves comparing high- and low-risk groups and the proportion of CR/PR or SD/PD patients receiving immunotherapy in the GSE78220 cohort. (E) Kaplan-Meier curves comparing high- and low-risk groups in the GSE100797 cohort, boxplots showing risk score differences between response and non-response patients, and percentage of patients with CR/PR or SD/PD undergoing immunotherapy in the high- and low-risk groups. *, P<0.05; ***, P<0.001; ns, P>0.05. CR, complete response; PD, progressive disease; PR, partial response; PRGRS, polyamine metabolism-related gene risk score; SD, stable disease.

SMS promoting malignant progression in LGG

In the above analysis, we identified SMS as a significant risk factor with the highest coefficient (Figure 5C). Comparative analysis revealed significantly elevated SMS mRNA expression in LGG tumor tissues versus normal tissues (Figure S1A). We subsequently discovered that LGG patients with high expression of SMS had a poorer prognosis (Figure S1B). Time-dependent ROC curves demonstrated AUC >0.8 for 1-/3-year survival prediction, which had a superior performance (Figure S1C). The immunohistochemical outcomes of the HPA data provided further evidence that SMS protein expression was remarkably up-regulated in patients with LGG (Figure S1D). These results suggest that SMS is a significant oncogene and has not been adequately studied in LGG. As a result, we performed in vitro experiments on SMS. Results showed that SMS mRNA and protein levels were higher in glioma cells (U251 and LN229) than in HMC3 cells (Figure 9A,9B). CCK-8 and colony formation assays indicated that SMS knockdown suppressed LGG cell proliferation (Figure 9C,9D). Transforaminal and wound healing assays indicate that SMS knockdown decreases glioma cell migration and invasion (Figure 9E,9F). Flow cytometry revealed that SMS knockdown caused G2/M phase arrest, reducing proliferation (Figure 10A,10B). Additionally, SMS knockdown decreased CDK1, CDK2, and Cyclin D1 protein levels (Figure 10C,10D). In summary, SMS facilitates glioma cell multiplication, migration, invasion, and cell cycling progression.

Figure 9 SMS facilitates LGG progression. (A) RT-qPCR to verify SMS expression. (B) Western blot to verify the protein expression of SMS. (C) CCK-8 assay of U251 and LN229. (D) Clonal spotting of U251 and LN229 with crystalline violet staining (1×). (E) Transwell migration and invasion assay of U251 and LN229 with crystalline violet staining under a light microscope (200×). (F) Scratch test of U251 and LN229 using ImageJ (40×). **, P<0.01; ***, P<0.001. CCK-8, Cell Counting Kit-8; LGG, low-grade gliomas; NC, negative control; OD, optical density; RT-qPCR, real-time quantitative polymerase chain reaction; SMS, spermine synthase.

Figure 10 SMS facilitates LGG progression. (A) Flow cytometry to analyze the cell cycle. (B) Proportion of cell cycle distribution (C) Western blot analysis confirmed the effect of SMS knockdown on cell cycle markers in U251 cells at the protein level. (D) Western blot confirmed the impact of SMS knockdown on LN229 cell cycle markers at the protein level. **, P<0.01; ***, P<0.001. LGG, low-grade gliomas; NC, negative control; SMS, spermine synthase.

Discussion

Glioma is the most common of the primary malignant brain tumors in adults and is a highly heterogeneous disease (31). The molecular heterogeneity of LGG contributes to their adverse prognosis. The diffuse and aggressive nature of glioma cells not only hampers complete surgical resection but also enhances resistance to chemotherapy and radiotherapy. LGG may not only develop into high-grade gliomas but also cause other disorders such as epilepsy (32-34). Despite the many studies on LGG today, the mortality rate of LGG patients is still high, so there is an imminent clinical need for a new prognostic model for further exploration of new targeted sites of LGG and to offer new insights for clinical treatment.

A growing body of evidence suggests that polyamine metabolism has a critical role in tumor biology and tumorigenesis and therapy. Elevated polyamine levels stimulate cell proliferation and angiogenesis in tumors (35), which promotes tumorigenesis and progression, thus dysregulation of polyamine metabolism is directly related to cancer development, and it is common in various types of cancer (14,36). For example, elevated levels of putrescine, spermidine, and spermine have been used as biological markers for early detection in patients with ovarian cancer (37,38). Spermine oxidase (SMOX), a key enzyme in polyamine metabolism, promotes gastric cancer by stimulating DNA damage through hydrogen peroxide produced by polyamine metabolism (39-42). Recent data suggest that polyamine metabolism also plays a crucial role in regulating antitumor immunotherapy. For example, Lee MS found that pancreatic cancer cells promote polyamine synthesis through ornithine aminotransferase (OAT), which is an important mechanism for tumor growth, and that the use of OAT inhibitors was effective in alleviating the proliferation of pancreatic cancer cells. Consequently, polyamine metabolism has appeared as a promising therapeutic target in cancer therapy (43).

In this study, we performed a comprehensive bioinformatics analysis to elucidate the role of polyamine metabolism-related genes in predicting molecular subtypes and prognosis of LGG patients. Thirty-six genes significantly associated with OS in LGG were first screened from polyamine metabolism-related genes, and further classified into two subtypes by a consensus clustering algorithm for LGG patients exhibiting different survival prognosis, pathway alterations, immune profiles, and drug sensitivity. The C2 subtype exhibited poorer OS and was associated with elevated dendritic cell infiltration, reflecting a highly proliferative and malignant tumor phenotype. In contrast, the C1 subtype demonstrated better OS and robust immune infiltration, with activated immune response pathways. We identified nine PRG-related prognostic features (OAZ3, PSMB5, PSMD14, PSMF1, PSMD5, PSMB9, PSMD12, PSMC2, and SMS) and constructed a stable PRGRS. This signature not only strongly predicted prognosis but also showed potential for predicting immunotherapy response in LGG patients. The C2 subtype’s poor prognosis and immune characteristics, such as elevated immune checkpoint expression, suggest limited benefit from immunotherapy, while the C1 subtype’s immune activation may enhance treatment efficacy. Among the nine key polyamine metabolism-related genes, SMS was a significant poor prognostic factor, and according to the results of TCGA and Human Protein Atlas (HPA) databases, the study also revealed that the mRNA and protein levels of SMS were notably elevated in LGG patients compared to normal tissues. SMS encodes proteins belonging to the spermine/spermidine synthase family and catalyzes the production of spermine from spermidine. It plays an important role in polyamine metabolism and spermine metabolism. By knocking down the expression of SMS in U251 and LN229 cells in vitro, we observed that the proliferation, migration, invasive ability, and cell cycle progression of tumor cells were significantly inhibited. More recently, studies have demonstrated that elevated expression levels in carcinomas such as hepatocellular carcinoma and squamous cell carcinoma of the head and neck are associated with poor prognosis. Specifically, metabolic reprogramming in hepatocellular carcinoma, characterized by increased glycolysis and lactylation, contributes to tumor progression and poor prognosis (44). In head and neck squamous cell carcinoma, lactylation of histone H3K9 activates interleukin-11 (IL-11), which suppresses CD8⁺ T cell proliferation and promotes tumor development through the JAK2/STAT3 pathway (45). These findings highlight the potential of targeting metabolic and immune-related pathways to improve therapeutic outcomes in these cancers. In lung adenocarcinoma, SMS has been proven to play an instrumental role in arginine metabolism and may be engaged in immune escape from lung adenocarcinoma (17).

Our study has several limitations. First, the reliance on public data sources introduces potential selection bias, necessitating validation in larger prospective trials to assess the clinical utility of polyamine metabolism-based prognostic models. Second, the limited sample size underscores the need for large-scale cohort studies to evaluate model robustness. Third, while PRGRS-predicted immunotherapy responses were analyzed using cohorts like IMvigor210, their applicability to LGG patients requires further validation in clinical trials. Finally, additional in vivo experiments are needed to clarify the biological functions and molecular mechanisms of the genes included in the model, particularly their roles in tumor progression and immune evasion. Addressing these limitations will enhance the translational potential of our findings.

Conclusions

In summary, this study conducted an in-depth analysis of polyamine metabolism-related genes in LGG patients using comprehensive bioinformatics analysis and identified SMS as a significant risk factor for LGG prognosis. Experimental analysis revealed that the knockdown of SMS inhibited the malignant biological behavior of LGG cells. Moreover, the knockdown of SMS affected cell cycle progression and the expression of several key cycling proteins. PRGRS has the potential to provide a novel and reliable aid for the diagnosis and prognosis of LGG patients, and SMS, a key gene for polyamine metabolism, maybe a potential prognostic biomarker and therapeutic target for future clinical treatment.

Acknowledgments

None.

Footnote

Reporting Checklist: The authors have completed the TRIPOD reporting checklist. Available at https://tcr.amegroups.com/article/view/10.21037/tcr-2025-776/rc

Data Sharing Statement: Available at https://tcr.amegroups.com/article/view/10.21037/tcr-2025-776/dss

Peer Review File: Available at https://tcr.amegroups.com/article/view/10.21037/tcr-2025-776/prf

Funding: This study was supported by Special Program for Priority Support and Cultivation of Key Disciplinary Areas at Bengbu Medical University (No. 2024bypy016), Bengbu Municipal Science and Technology Innovation Guidance Projects (No. 2024ZD0004), General Program for Natural Sciences Research of Bengbu Medical University (No. 2024byzd026), Scientific Research Projects of Anhui Provincial Health Commission (No. AHWJ2023A20289), Major Science and Technology Projects of Anhui Provincial Science and Technology Innovation Platforms (No. S202305a12020038) and Research Funds of Joint Research Center for Regional Diseases of IHM (No. 2024bydjk007).

Conflicts of Interest: All authors have completed the ICMJE uniform disclosure form (available at https://tcr.amegroups.com/article/view/10.21037/tcr-2025-776/coif). The authors have no conflicts of interest to declare.

Ethical Statement: The authors are accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved. The study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments.

Open Access Statement: This is an Open Access article distributed in accordance with the Creative Commons Attribution-NonCommercial-NoDerivs 4.0 International License (CC BY-NC-ND 4.0), which permits the non-commercial replication and distribution of the article with the strict proviso that no changes or edits are made and the original work is properly cited (including links to both the formal publication through the relevant DOI and the license). See: https://creativecommons.org/licenses/by-nc-nd/4.0/.

References

1. Liu J, Cao S, Imbach KJ, et al. Multi-scale signaling and tumor evolution in high-grade gliomas. Cancer Cell 2024;42:1217-1238.e19. [Crossref] [PubMed]
2. Choi S, Yu Y, Grimmer MR, et al. Temozolomide-associated hypermutation in gliomas. Neuro Oncol 2018;20:1300-9. [Crossref] [PubMed]
3. Lin H, Liu C, Hu A, et al. Understanding the immunosuppressive microenvironment of glioma: mechanistic insights and clinical perspectives. J Hematol Oncol 2024;17:31. [Crossref] [PubMed]
4. Pace A, Dirven L, Koekkoek JAF, et al. European Association for Neuro-Oncology (EANO) guidelines for palliative care in adults with glioma. Lancet Oncol 2017;18:e330-40. [Crossref] [PubMed]
5. Tan AC, Ashley DM, López GY, et al. Management of glioblastoma: State of the art and future directions. CA Cancer J Clin 2020;70:299-312. [Crossref] [PubMed]
6. Mijiti M, Maimaiti A, Chen X, et al. CRISPR-cas9 screening identified lethal genes enriched in Hippo kinase pathway and of predictive significance in primary low-grade glioma. Mol Med 2023;29:64. [Crossref] [PubMed]
7. Dunn GP, Cloughesy TF, Maus MV, et al. Emerging immunotherapies for malignant glioma: from immunogenomics to cell therapy. Neuro Oncol 2020;22:1425-38. [Crossref] [PubMed]
8. Boussiotis VA, Charest A. Immunotherapies for malignant glioma. Oncogene 2018;37:1121-41. [Crossref] [PubMed]
9. Maxwell R, Luksik AS, Garzon-Muvdi T, et al. The Potential of Cellular- and Viral-Based Immunotherapies for Malignant Glioma-Dendritic Cell Vaccines, Adoptive Cell Transfer, and Oncolytic Viruses. Curr Neurol Neurosci Rep 2017;17:50. [Crossref] [PubMed]
10. Dey P, Kimmelman AC, DePinho RA. Metabolic Codependencies in the Tumor Microenvironment. Cancer Discov 2021;11:1067-81. [Crossref] [PubMed]
11. Shrestha RK, Nassar ZD, Hanson AR, et al. ACSM1 and ACSM3 Regulate Fatty Acid Metabolism to Support Prostate Cancer Growth and Constrain Ferroptosis. Cancer Res 2024;84:2313-32. [Crossref] [PubMed]
12. Tong H, Jiang Z, Song L, et al. Dual impacts of serine/glycine-free diet in enhancing antitumor immunity and promoting evasion via PD-L1 lactylation. Cell Metab 2024;36:2493-2510.e9. [Crossref] [PubMed]
13. Zhu Y, Zhou Z, Du X, et al. Cancer cell-derived arginine fuels polyamine biosynthesis in tumor-associated macrophages to promote immune evasion. Cancer Cell 2025;43:1045-1060.e7. [Crossref] [PubMed]
14. Casero RA Jr, Murray Stewart T, Pegg AE. Polyamine metabolism and cancer: treatments, challenges and opportunities. Nat Rev Cancer 2018;18:681-95. [Crossref] [PubMed]
15. Coni S, Bordone R, Ivy DM, et al. Combined inhibition of polyamine metabolism and eIF5A hypusination suppresses colorectal cancer growth through a converging effect on MYC translation. Cancer Lett 2023;559:216120. [Crossref] [PubMed]
16. Flynn AT, Hogarty MD. Myc, Oncogenic Protein Translation, and the Role of Polyamines. Med Sci (Basel) 2018;6:41. [Crossref] [PubMed]
17. Wang Z, Zhang J, Shi S, et al. Predicting lung adenocarcinoma prognosis, immune escape, and pharmacomic profile from arginine and proline-related genes. Sci Rep 2023;13:15198. [Crossref] [PubMed]
18. Xiang L, Piao L, Wang D, et al. Overexpression of SMS in the tumor microenvironment is associated with immunosuppression in hepatocellular carcinoma. Front Immunol 2022;13:974241. [Crossref] [PubMed]
19. Fahrmann JF, Vykoukal J, Fleury A, et al. Association Between Plasma Diacetylspermine and Tumor Spermine Synthase With Outcome in Triple-Negative Breast Cancer. J Natl Cancer Inst 2020;112:607-16. [Crossref] [PubMed]
20. Zhao Z, Zhang KN, Wang Q, et al. Chinese Glioma Genome Atlas (CGGA): A Comprehensive Resource with Functional Genomic Data from Chinese Glioma Patients. Genomics Proteomics Bioinformatics 2021;19:1-12. [Crossref] [PubMed]
21. Zeng D, Ye Z, Shen R, et al. IOBR: Multi-Omics Immuno-Oncology Biological Research to Decode Tumor Microenvironment and Signatures. Front Immunol 2021;12:687975. [Crossref] [PubMed]
22. Thorsson V, Gibbs DL, Brown SD, et al. The Immune Landscape of Cancer. Immunity 2018;48:812-830.e14. [Crossref] [PubMed]
23. Chan TA, Yarchoan M, Jaffee E, et al. Development of tumor mutation burden as an immunotherapy biomarker: utility for the oncology clinic. Ann Oncol 2019;30:44-56. [Crossref] [PubMed]
24. Maeser D, Gruener RF, Huang RS. oncoPredict: an R package for predicting in vivo or cancer patient drug response and biomarkers from cell line screening data. Brief Bioinform 2021;22:bbab260. [Crossref] [PubMed]
25. Ritchie ME, Phipson B, Wu D, et al. limma powers differential expression analyses for RNA-sequencing and microarray studies. Nucleic Acids Res 2015;43:e47. [Crossref] [PubMed]
26. Liberzon A, Birger C, Thorvaldsdóttir H, et al. The Molecular Signatures Database (MSigDB) hallmark gene set collection. Cell Syst 2015;1:417-25. [Crossref] [PubMed]
27. Mariathasan S, Turley SJ, Nickles D, et al. TGFβ attenuates tumour response to PD-L1 blockade by contributing to exclusion of T cells. Nature 2018;554:544-8. [Crossref] [PubMed]
28. Hugo W, Zaretsky JM, Sun L, et al. Genomic and Transcriptomic Features of Response to Anti-PD-1 Therapy in Metastatic Melanoma. Cell 2016;165:35-44. [Crossref] [PubMed]
29. Lauss M, Donia M, Harbst K, et al. Mutational and putative neoantigen load predict clinical benefit of adoptive T cell therapy in melanoma. Nat Commun 2017;8:1738. [Crossref] [PubMed]
30. Budczies J, Kazdal D, Menzel M, et al. Tumour mutational burden: clinical utility, challenges and emerging improvements. Nat Rev Clin Oncol 2024;21:725-42. [Crossref] [PubMed]
31. Zheng H, Zhao Y, Zhou H, et al. The Comprehensive Analysis of m6A-Associated Anoikis Genes in Low-Grade Gliomas. Brain Sci 2023;13:1311. [Crossref] [PubMed]
32. Cancer Genome Atlas Research Network. Comprehensive, Integrative Genomic Analysis of Diffuse Lower-Grade Gliomas. N Engl J Med 2015;372:2481-98. [Crossref] [PubMed]
33. Rudà R, Bello L, Duffau H, et al. Seizures in low-grade gliomas: natural history, pathogenesis, and outcome after treatments. Neuro Oncol 2012;14:iv55-64. [Crossref] [PubMed]
34. Haddad AF, Young JS, Oh JY, et al. The immunology of low-grade gliomas. Neurosurg Focus 2022;52:E2. [Crossref] [PubMed]
35. Gerner EW, Meyskens FL Jr. Polyamines and cancer: old molecules, new understanding. Nat Rev Cancer 2004;4:781-92. [Crossref] [PubMed]
36. Holbert CE, Cullen MT, Casero RA Jr, et al. Polyamines in cancer: integrating organismal metabolism and antitumour immunity. Nat Rev Cancer 2022;22:467-80. [Crossref] [PubMed]
37. Akinjiyan FA, Ibitoye Z, Zhao P, et al. DDR2-regulated arginase activity in ovarian cancer-associated fibroblasts promotes collagen production and tumor progression. Oncogene 2024;43:189-201. [Crossref] [PubMed]
38. Waalkes TP, Rosenshein NB, Shaper JH, et al. A feasibility study in the development of biological markers for ovarian cancer. J Surg Oncol 1982;21:207-14. [Crossref] [PubMed]
39. McNamara KM, Sierra JC, Latour YL, et al. Spermine oxidase promotes Helicobacter pylori-mediated gastric carcinogenesis through acrolein production. Oncogene 2025;44:296-306. [Crossref] [PubMed]
40. Gobert AP, Wilson KT. Polyamine- and NADPH-dependent generation of ROS during Helicobacter pylori infection: A blessing in disguise. Free Radic Biol Med 2017;105:16-27. [Crossref] [PubMed]
41. Chaturvedi R, de Sablet T, Peek RM, et al. Spermine oxidase, a polyamine catabolic enzyme that links Helicobacter pylori CagA and gastric cancer risk. Gut Microbes 2012;3:48-56. [Crossref] [PubMed]
42. Xuan M, Gu X, Li J, et al. Polyamines: their significance for maintaining health and contributing to diseases. Cell Commun Signal 2023;21:348. [Crossref] [PubMed]
43. Lee MS, Dennis C, Naqvi I, et al. Ornithine aminotransferase supports polyamine synthesis in pancreatic cancer. Nature 2023;616:339-47. [Crossref] [PubMed]
44. Yang Z, Yan C, Ma J, et al. Lactylome analysis suggests lactylation-dependent mechanisms of metabolic adaptation in hepatocellular carcinoma. Nat Metab 2023;5:61-79. [Crossref] [PubMed]
45. Wang R, Li C, Cheng Z, et al. H3K9 lactylation in malignant cells facilitates CD8(+) T cell dysfunction and poor immunotherapy response. Cell Rep 2024;43:114686. [Crossref] [PubMed]