Development of a fatty acid metabolism (FAM)-related gene signature for prognosis prediction and personalized therapy in lower grade gliomas

Introduction

According to the latest global statistics, there were a total of 321,476 cases of central nervous system (CNS) tumors and 248,305 deaths in 2022 (1). Gliomas, which arise from glial cells, represent the most common primary CNS tumor encountered in clinical practice, posing a major risk to human health (2,3). Based on the World Health Organization (WHO) glioma classification guidelines, these tumors are divided into four distinct grades. Lower grade gliomas (LGGs) encompass grade 2 and 3 gliomas (4). While LGGs are generally less aggressive than grade 4 gliomas, the 5-year survival rates for grade 2 glioma and 3 glioma are only 47.3–79.1% and 26.5–50.7%, respectively, and some cases will progress to higher grade gliomas (5). Current therapeutic modalities for LGGs encompass observation, surgical resection, radiation therapy, and chemotherapy, each accompanied by distinct risks and benefits, including cognitive and neurological side effects (6). Surgical resection typically serves as the primary approach, supported by evidence indicating that a more extensive resection correlates with enhanced progression-free survival and overall survival (OS) rates (7,8). Nevertheless, the efficacy of radiation and chemotherapy remains a subject of ongoing research and discussion. Recent clinical trials have demonstrated that the integration of radiotherapy with chemotherapy, particularly the PCV regimen (comprising procarbazine, lomustine, and vincristine), yields superior survival outcomes compared to radiotherapy alone (9). Nonetheless, the enduring cognitive and neurological side effects associated with these treatments underscore the importance of appropriate patient selection and timing of interventions. Therefore, it is necessary to choose individualized treatment methods based on accurate risk stratification of the disease.

Combination of histopathology and genetic profiling can more accurately grade and stratify the risk of gliomas. Mutations in isocitrate dehydrogenase (IDH) genes are commonly observed in LGGs and are linked to a more favorable clinical outcome (10). The integration of IDH mutations, clinicopathologic features, and 1p/19q co-deletion has enhanced patient stratification and facilitated the development of tailored therapeutic approaches (11). However, the broad spectrum of prognoses reflects the heterogeneity of LGGs and highlights the limitations of the prognostic factors to some extent. Thus, the identification of additional prognostic biomarkers is imperative to refine predictive accuracy and enable personalized treatment strategies.

Fatty acid (FA) metabolism is crucial for various cellular processes, including energy production, membrane structure, and signaling molecules synthesis (12). FAs are essential for various CNS functions, and dysregulation of FA metabolism (FAM) has been implicated in severe CNS disorders like Alzheimer’s (13) and Parkinson’s disease (14). Emerging evidence highlights the relevance of FAM in solid tumors (15,16), notably in high-grade gliomas relating to CNS pathogenesis, such as glioblastoma (GBM) (17). Qi et al. demonstrated that genes associated with FA catabolic metabolism have significant prognostic value and are closely correlated with immune cells in gliomas (18). Nevertheless, there is a lack of comprehensive studies on developing prognostic models for LGGs based on FAM. This knowledge gap impedes the clinical application of FAM-related genes (FAMRGs) in risk stratification and tailored therapeutic approaches for LGGs.

Objective

This study initially conducted a bioinformatics analysis to identify FAMRGs from The Cancer Genome Atlas (TCGA) dataset and subsequently developed a prognostic model for LGGs. The model underwent validation using data from the TCGA test cohort and the Chinese Glioma Genome Atlas (CGGA) database. Functional enrichment analyses, including Gene Ontology (GO), immune infiltration analysis, Gene Set Variation Analysis (GSVA), Kyoto Encyclopedia of Genes and Genomes (KEGG), and drug sensitivity analysis, were performed to unravel the biological significance and mechanisms of these genes in the immune microenvironment. The results provide valuable insights into the prognostic and personalized therapeutic potential of these genes 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-2025-1320/rc).

Methods

Data acquisition

The RNA sequencing (RNA-seq) data and associated clinical details for 529 LGG samples were sourced from University of California, Santa Cruz Xena (UCSC Xena), and the data for 255 normal cerebral cortex samples were obtained from the Genotype-Tissue Expression (GTEx) project. Moreover, a testing cohort comprising 420 LGG patients with RNA-seq data and clinical information was established using the CGGA database. The TCGA dataset was stratified into training (n=262) and testing cohorts (n=262). Furthermore, 162 FAMRGs were identified using the Human Molecular Signatures Database via the R package “msigdbr”. The study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments.

Differential analysis

Differentially expressed genes (DEGs) between TCGA lower-grade glioma samples and normal cerebral cortex samples were identified using the R package “edgeR”. The screening criteria for DEGs included a false discovery rate (FDR) less than 0.05 and an absolute log₂ fold change greater than 1. The intersection of DEGs and FAMRGs was illustrated using the R package “VennDiagram”.

Construction and evaluation of the prognostic risk model

To develop the prognostic risk model, we utilized differentially expressed FAMRGs in conjunction with sample survival data. Initially, samples were randomly allocated into training and test datasets. Prognostic genes were identified through univariate Cox analysis, multivariate Cox analysis, and least absolute shrinkage and selection operator (LASSO) regression analysis via the R packages “survival” and “glmnet”. The risk model was formulated based on the selected genes, with individual risk scores calculated using the following equation:

Riskscore=∑i=1nexprgenei×coefficientgenei[1]

In this formula, coefficient_genei represents the regression coefficient for each FAMRGs, while expr_genei denotes its expression level. Patients were then classified into high and low-risk groups based on the median risk score. To compare OS between the two subgroups, Kaplan-Meier curves were generated using the R package “survminer”. The prognostic model’s performance was evaluated through receiver operating characteristic (ROC) curve analysis, implemented via “timeROC” package.

Verification of the prognostic risk model

To confirm the robustness of the developed prognostic risk model, we employed two independent datasets: the CGGA dataset, encompassing 264 samples, and the TCGA testing dataset. The validation process involved several steps: we first applied multivariate Cox analysis to reassess prognostic significance of the identified genes in these new datasets. Subsequently, OS was then compared between patients classified as high-risk and low-risk groups using Kaplan-Meier survival curves with the R package. Additionally, the model’s predictive performance was evaluated by conducting ROC curve analysis.

Construction of nomogram

To enhance the clinical applicability of our findings, we created an integrated nomogram that combines key clinicopathological characteristics with the derived risk scores. This tool was designed using the “rms” package in R, aiming to provide a more holistic approach to prognosis prediction in LGGs. Both univariate and multivariate Cox regression analyses were utilized to pinpoint independent risk factors. Calibration curves at 1-, 3-, 5-, and 8-year were drawn for visual assessment. Furthermore, the accuracy of the nomogram was gauged using ROC curves.

Functional enrichment analysis

The biological functions of FAM-related DEGs were further explored through GO analysis, encompassing cellular component (CC), biological process (BP), and KEGG analysis, as well as molecular function (MF) using the “ClusterProfilter” package. Furthermore, to compare the two groups, Reactome pathway analysis was executed utilizing the “GSVA” package. A significance level of P<0.05 was set to determine meaningful enrichment in pathways and processes.

Immune infiltration analysis

The stromal, immune, and estimated scores for LGGs were evaluated using the “ESTIMATE” package. To determine the distribution of 22 different immune cell types between the high- and low-risk groups, the CIBERSORT algorithm was applied through the “IOBR” package.

Drug sensitivity

Sensitivity scores were computed for the prediction of the half-maximal inhibitory concentration (IC50) of prevalent chemotherapy drugs in LGG patients using the “oncoPredict” package and the Genomics of Drug Sensitivity in Cancer database.

Validation of prognostic genes

The immunohistochemistry image depicting protein expression of prognostic genes in LGGs and cerebral cortex samples was obtained from the Human Protein Atlas (HPA) database.

Statistical analysis

All statistical analyses were performed with R software (version 4.3.2). To analyze the differences in OS among the subgroups, Kaplan-Meier curves were generated and assessed using the log-rank test, were employed to evaluate differences in OS between subgroups. Regression analysis facilitated the identification of FAMRGs for the risk model. Comparisons of clinicopathological data between the groups were accomplished using the Wilcoxon test, with statistical significance defined as a P value of less than 0.05.

Results

Identification of prognostic FAM-related DEGs in the TCGA-LGG cohort and functional enrichment analyses

The study conducted DEG analysis to investigate FAMRGs differentially expressed in TCGA LGG samples compared to normal cerebral cortex tissue. This analysis identified a total of 8,733 DEGs. By overlapping the 8,733 DEGs from the TCGA database with 162 FAMRGs, 37 FAM-related DEGs were identified (Figure 1A,1B). Functional enrichment analysis was subsequently conducted to examine molecular characteristics and biological functions. The results highlighted enrichment of these genes in BPs such as lipid metabolic process, FAM process, and nicotinamide adenine dinucleotide phosphate (NADP) metabolic process. Moreover, the genes were found to be highly enriched in CCs like cytosol, cytoplasm, and mitochondrion, and in MFs including protein homodimerization activity, oxidoreductase activity, and nicotinamide adenine dinucleotide (NAD) binding (Figure 2A-2C). KEGG analysis demonstrated enrichment in pathways such as metabolic pathways, peroxisome proliferator-activated receptor (PPAR) signaling pathway, and tryptophan metabolism (Figure 2D). Reactome pathways analysis comparing two groups revealed upregulation of the innate immune system in the high-risk group, whereas pathways related to G2/M DNA damage checkpoint, arachidonic acid metabolism, and biological oxidations were upregulated in the low-risk cohort (Figure 2E).

Figure 1 Identification of prognostic FA related DEGs. (A) The Venn diagram presented FA related DEGs; (B) the volcano plot showed FA related DEGs (the dots of the red and blue represent the downregulated and upregulated DEGs, respectively). DEGs, differentially expressed genes; FA, fatty acid; FC, fold change; FDR, false discovery rate; TCGA, The Cancer Genome Atlas.

Figure 2 GO annotations and pathway enrichment analyses. (A) Biological process. (B) Cellular component. (C) Molecular function. (D) KEGG analysis. (E) Reactome analysis of high and low-risk groups. GO, Gene Ontology; KEGG, Kyoto Encyclopedia of Genes and Genomes; NAD, nicotinamide adenine dinucleotide; NADH, reduced nicotinamide adenine dinucleotide; NADP, nicotinamide adenine dinucleotide phosphate.

For the development of a prognostic risk model, 524 LGG patients were equally split into training and test groups. Among 37 genes, univariate Cox regression analysis identified 10 potential prognostic markers. Subsequent LASSO regression analysis selected seven genes (Figure 3A,3B), which were further refined to four genes (CPT2, UROS, GPD1, HSD17B10) through multivariate Cox regression analysis for the final model (Figure 3C). The risk scores were computed using the expression profiles and corresponding regression coefficients of the identified genes. Utilizing the median risk score as a threshold, the training dataset was categorized into two groups: high- and low-risk. Figure 3D presents a comparative analysis of these groups, highlighting differences in survival status distribution, survival times, and expression patterns of the selected genes. Within the TCGA training cohort, patients classified as high-risk exhibited increased mortality rates. Notably, three genes (CPT2, GPD1, HSD17B10) showed elevated expression levels in the high-risk group relative to their low-risk counterparts.

Figure 3 Construction and validation of FAM-related prognostic risk model. (A) By LASSO regression analysis, seven genes were screened out. (B) Coefficients in LASSO regression model. (C) Four prognostic genes were screened out by multivariate Cox regression analysis. Risk score and survival time distributions, and expression levels of the prognostic genes in the TCGA training (D) and CGGA set (G). ROC curves of the prognostic model for predicting the 1-, 3-, 5-, and 8-year OS in the TCGA training (E) and CGGA set (H). Kaplan-Meier survival curves of the TCGA training (F) and CGGA set (I). AUC, area under the curve; CGGA, Chinese Glioma Genome Atlas; CI, confidence interval; FAM, fatty acid metabolism; HR, hazard ratio; LASSO, least absolute shrinkage and selection operator; OS, overall survival; ROC, receiver operating characteristic; TCGA, The Cancer Genome Atlas.

To evaluate the predictive power of the prognostic model for LGG patient outcomes, ROC curve analysis was employed. The training dataset yielded area under the curve (AUC) values of 0.8314, 0.7972, 0.7337, and 0.7337 for 1-, 3-, 5-, and 8-year survival predictions, respectively (Figure 3E). The TCGA training cohort revealed a statistically significant reduction in OS among high-risk patients compared to those in the low-risk category (Figure 3F).

Corroboration of the prognostic risk model

The prognostic gene signature, comprising CPT2, UROS, GPD1, and HSD17B10, exhibited consistent patterns in survival status, survival duration, and gene expression profiles in both the TCGA test cohort (Figure S1) and the CGGA cohort (Figure 3G), similar to those observed in the TCGA training cohort. Multivariate Cox regression analysis reaffirmed the significant impact of these four genes on OS within the test cohort. ROC analysis in the TCGA test cohort yielded AUC values of 0.7369, 0.7684, 0.6973, 0.6446 for 1-, 3-, 5-, 8-year survival predictions, respectively (Figure S2). Similarly, in the CGGA cohort, the AUC values were 0.6788, 0.6983, 0.7194, and 0.6941 for the same time points (Figure 3H). Moreover, Kaplan-Meier analyses in both testing groups demonstrated significantly worse prognosis in the high-risk groups compared to the low-risk groups (P<0.05) (Figure S3 and Figure 3I). These consistent findings across multiple cohorts substantiate the prognostic risk model’s efficacy in predicting outcomes for LGG patients.

Development and validation of the nomogram

A comprehensive nomogram was constructed using the TCGA dataset, incorporating the risk score with various key clinicopathological factors, including age, gender, WHO grade, 1p19q codeletion status, IDH status, and O6-methylguanine-DNA methyl-transferase (MGMT) promoter status. The cumulative score derived from these parameters inversely correlates with the 1-, 3-, 5-, and 8-year survival probabilities (Figure 4A). Internal and external calibration analyses were performed to assess the nomogram’s predictive accuracy. In the TCGA cohort, the nomogram demonstrated AUC values of 0.895, 0.908, 0.849, 0.831 for 1-, 3-, 5-, 8-year OS rates, respectively (Figure 4B). Calibration plots exhibited a strong concordance between observed and predicted OS outcome (Figure 4C). In the CGGA cohort, the AUCs were 0.724, 0.779, 0.782, 0.836 for the corresponding time points (Figure 4D), with calibration plots similarly showing high agreement between observed and predicted OS rates (Figure 4E). These findings validate the nomogram’s efficacy in predicting outcomes for LGGs and suggest its potential utility in clinical decision-making.

Figure 4 Construction and validation of the prognostic nomogram. The nomogram was constructed by risk level and clinicopathological features. Gender: 0=Female, 1=Male (A). ROC curves of the nomogram in the TCGA (B) and CCGA set (C). Calibration curves of the nomogram for predicting 1-, 3-, 5-, and 8-year OS in the TCGA (D) and CGGA set (E). CGGA, Chinese Glioma Genome Atlas; IDH, isocitrate dehydrogenase; OS, overall survival; ROC, receiver operating characteristic; TCGA, The Cancer Genome Atlas; WT, wild-type.

To explore the risk score’s validity as an autonomous prognostic indicator in LGGs, we conducted univariate and multivariate Cox regression analyses, incorporating various clinicopathological characteristics. As depicted in Table 1, these analyses identified the risk score as a standalone risk factor within the TCGA cohort, a finding that was corroborated by results from the CGGA cohort (Table S1). In the TCGA cohort, multivariate Cox regression analysis indicated that patients older than 40 years, with higher WHO grades, lacking 1p19q codeletion, and possessing wild-type IDH1, experienced significantly reduced OS, with all P values being less than 0.05. Conversely, the findings from the CGGA cohort showed that age did not hold prognostic significance. Therefore, the risk model can be considered a dependable and novel prognostic biomarker. Besides, WHO grade, 1p19q codeletion status, IDH status were also identified as independent prognosis factors.

We conducted a comparison of risk scores distributions across various clinicopathological subgroups (Figure 5). The results indicated that patients exhibiting characteristics such as higher WHO grade, wild-type IDH1, non-codeletion of 1p19q, and unmethylated MGMT promoter tended to have higher risk scores (all P<0.05). In contrast, age and gender subgroups did not show significant differences in risk score distribution. These findings substantiate the risk model’s capacity to reflect key clinicopathological features in both the TCGA and CGGA cohorts.

Figure 5 The distribution of risk scores among different clinicopathological subgroups. ns, not significant; ***, P<0.001; ****, P<0.0001. IDH, isocitrate dehydrogenase; MGMT, O6-methylguanine-DNA methyl-transferase; WT, wild-type.

Immune landscape analysis of different risk subgroups

In our study, we utilized CIBERSORT analysis to evaluate the immune responses of 22 tumor-infiltrating immune cells (TIICs) across risk groups. Significant variances were observed in the infiltration levels of various immune cell types, such as naive B cells, CD8 T cells, resting memory CD4 T cells, follicular helper T cells, NK cells (both resting and activated), M2 macrophages, activated dendritic cells, and mast cells (both resting and activated) (Figure 6A). The risk score showed a significant positive correlation with the infiltration levels of M2 macrophages, activated dendritic cells, and resting mast cells (r =0.19, 0.16, and 0.20, respectively; all P<0.05). Conversely, it was negatively correlated with naive B cells, resting dendritic cells, and activated mast cells (r =−0.18, −0.13, and −0.19, respectively; all P<0.05) (Figure 6B). Additionally, ESTIMATE analysis indicated a positive association between the risk score and Immune Score, ESTIMATE Score, and Stromal Score (Figure 6C-6E).

Figure 6 Immune characteristics of different risk subgroups. The fraction of 22 types of immune cells in high- and low-risk groups based CIBERSORT (A). Correlation analysis between immune cell infiltration and risk score (B). Levels of immune score, estimate score and stromal score in high- and low-risk groups (C-E). The expression levels of immune checkpoints in high- and low-risk groups (F). *, P<0.05. **, P<0.01. ***, P<0.001. ****, P<0.0001. ns, not significant.

Recognizing the importance of immune checkpoints in modulating immunological tolerance and protecting cells from indiscriminate attacks, we investigated the expression profiles of immune checkpoint-related genes across different risk categories using TCGA dataset. The gene signature was derived from a previous study (19). Our findings revealed significant upregulation of several key immune checkpoint genes within the high-risk group, including programmed cell death ligand 1 (CD274, PD-L1), programmed cell death 1 ligand 2 (PDCD1LG2), programmed cell death 1 (PDCD1), lymphocyte activation gene 3 (LAG3), hepatitis a virus cellular receptor 2 (HAVCR2), cytotoxic t-lymphocyte associated protein 4 (CTLA4), and B7-H3 (CD276) (Figure 6F). While individuals with elevated expression of immune checkpoint genes could benefit more from immune checkpoint inhibitors, rapid tumor recurrence may occur due to the presence of other immune checkpoint genes. Consequently, a combination of immune checkpoint inhibitors has been implemented in clinical treatment.

Association between risk subgroups and drug efficacy

To assess how patients with varying risk scores respond to common chemotherapy drugs, “oncoPredict” package was used to analyze IC50 values for LGG patients in the TCGA cohort, specifically concerning temozolomide and carmustine. Patients with higher risk scores exhibited lower IC50 values for both temozolomide (Figure 7A) and carmustine (Figure 7B), suggesting a potential higher sensitivity to alkylating agents. Additionally, the study identified kinase inhibitor (axitinib), cyclin-dependent protein kinase (CDK) inhibitor (ribociclib), and poly-ADP ribose polymerase (PARP) inhibitor (olaparib) as potential candidate drugs for treating high-risk score patients (Figure 7C-7E). Meanwhile, targeting the PI3K pathway (AZD2014), the MAPK pathway (ERK2440), and Heat shock protein 90 (luminespib) were proposed as effective strategies for low-risk score patients based on the results (Figure 7F-7H).

Figure 7 Drugs sensitivity analysis in the high and low-risk group. Predicted sensitivity scores of TMZ, carmustine, axitinib, ribociclib and olaparib, which were candidate drugs for the high-risk group (A-E). Predicted sensitivity scores of AZD2014, ERK2440 and luminespib, which were candidate drugs for the low-risk group (F-H). ****, P<0.0001. TMZ, temozolomide.

Validation of prognostic genes

We examined the expression of four prognostic genes using the HPA database, focusing on immunohistochemistry images of protein expression in LGG samples compared to normal cerebral cortex tissue (Figure 8). Among these genes, CPT2, GPD1, and HSD17B10 proteins were found to be highly expressed in LGGs, while UROS was low or not detected in LGGs.

Figure 8 CPT2, GPD1, HSD17B10 and UROS protein expression in cerebral cortex samples and LGG samples from HPA (https://www.proteinatlas.org/). The links to the normal and tumor tissues of each protein are provided for CPT2 (https://www.proteinatlas.org/ENSG00000157184-CPT2/tissue/cerebral+cortex#img; https://www.proteinatlas.org/ENSG00000157184-CPT2/cancer/glioma#img), GPD1 (https://www.proteinatlas.org/ENSG00000167588-GPD1/tissue/cerebral+cortex#img; https://www.proteinatlas.org/ENSG00000167588-GPD1/cancer/glioma#img), HSD17B10 (https://www.proteinatlas.org/ENSG00000072506-HSD17B10/tissue/cerebral+cortex#img; https://www.proteinatlas.org/ENSG00000072506-HSD17B10/cancer/glioma#img), and UROS (https://www.proteinatlas.org/ENSG00000188690-UROS/tissue/cerebral+cortex#img; https://www.proteinatlas.org/ENSG00000188690-UROS/cancer/glioma#img), respectively (Scale bar, 200 µm; immunohistochemical staining). HPA, Human Protein Atlas; LGG, lower-grade glioma.

Discussion

LGGs represent a heterogeneous group of neoplasms with the potential for malignant transformation (20). At present, the treatment methods in LGGs encompass surgery, radiotherapy, and chemotherapy (21). Additionally, there is growing interest in immunotherapy and targeted therapy for gliomas. Despite these advancements, the optimization of treatment strategies remains challenging in clinical practice (22). Postoperative adjuvant regimens—primarily involving radiotherapy and chemotherapy—are generally tailored based on clinicopathological risk stratification. Nevertheless, OS rates remain suboptimal, and there is a clear lack of reliable guidance for implementing immunotherapy or targeted therapy approaches.

Emerging research has underscored the significance of metabolic reprogramming, with a particular focus on FAs within neoplastic cells and the tumor microenvironment (TME), in the evolution of cancer, response to treatment, recurrence, and metastasis (23,24). Cui et al. have proposed that aberrant FAM promotes gastric cancer proliferation (25). Moreover, perturbations in FAM within tumor cells have been linked to therapeutic resistance (26). Signatures related to glucose and amino acid metabolism have enhanced our understanding of the molecular pathways in gliomas, offering new insights for glioma treatment strategies (27,28). Previous research has highlighted the significant correlation between FAMRGs and the malignancy, prognosis, and immunophenotype of gliomas (18). Notably, FAM in tumors encompasses anabolism, catabolism, connections to ferroptosis, and the generation of signaling molecules (29). Our study primarily focuses on the prognostic significance of FAMRGs, the underlying molecular pathways, and their interactions with the TME and the drug responsiveness in LGGs. These findings not only deliver novel biological insights that extend beyond current clinicopathological indicators but also provide a critical framework for clinicians to develop individualized treatment plans.

In this study, gene expression and survival data from LGG patients in the TCGA database were analyzed using bioinformatics methods. Ten DEGs with prognostic value were identified and further narrowed down to seven using LASSO regression for subsequent multivariate Cox regression analysis. Subsequently, four genes, including CPT2, UROS, GPD1, and HSD17B10, were screened to develop a prognostic model.

CPT2, a type of acyltransferase linked to FA oxidation (FAO), has been identified as upregulated in GBM and recurrent cases with unfavorable prognoses (30). Radioresistant GBM cell lines have been observed to rewire the metabolism by increasing FAO rate, including CPT2 enzymatic activity (31). UROS plays a role in the heme biosynthesis pathway and is associated with congenital erythropoietic porphyria. Research by Nawaz et al. discovered that miR-4484 upregulation reduces the colony-forming and migratory abilities of glioma cells, correlating positively with UROS expression (32). Aligning with this finding, our study observed lower expression levels of UROS in the high-risk group. GPD1, an enzyme dependent on NAD⁺/NADH that bridges carbohydrate and lipid metabolism (33), is upregulated during early tumor development stages, particularly in GBM (34). Our study revealed higher expression in the high-risk group, which suggests a potential for this group to progress to higher grade gliomas. HSD17B10, an enzyme located in the mitochondria that converts 17-estradiol into estrone and is involved in FAM (35), has been linked to poor outcomes in osteosarcoma patients in a previous study (36), consistent with our study findings.

Moreover, our multivariate Cox regression analyses based on the TCGA and CGGA databases both confirmed the independent predictive significance of the risk score for LGGs. To enhance clinical applicability, we constructed a nomogram model designed to forecast 1-, 3-, 5-, and 8-year survival probabilities. Our FAM-related model exhibited comparable prognostic efficacy when compared to other prognostic models for LGGs developed in previous studies (37,38). Therefore, the FAM-related signature emerged as a reliable prognostic marker that could potentially assist in clinical decision-making.

GSVA revealed an enrichment of immune response-associated pathways in the high-risk group, indicating a significant interplay between FAM and the tumor immune microenvironment (TIME). CIBERSORT analysis indicated elevated infiltration of regulatory T cells, resting CD4⁺ memory T cells, and M2 macrophages in the high-risk group, which are known to be positively correlated with immunosuppression and malignant progression (39-41). In contrast, the low-risk group exhibited higher levels of B_cells_naive, T_cells_CD4_naive, and NK_cells_resting, which is consistent with the correlation analysis. Immune cells have the capacity to modulate the microenvironment through the secretion of various cytokines. They can either impede tumor advancement and establish an unfavorable microenvironment for tumor growth or facilitate tumor progression and create a conducive microenvironment for tumor development (41). Previous research has indicated that FAM influences the proliferation and functionality of immune cells, each exhibiting distinct FAM patterns (42). For instance, FAM has been implicated in promoting M2-like tumor-associated macrophages polarization, contributing to an immunosuppressive microenvironment (43). Similarly, our results also indicated an enrichment of immunosuppressive cells in the high-risk group.

As is well recognized, immune checkpoint inhibitors are becoming a primary treatment modality by restructuring the immunosuppressive TME (44). However, immunotherapies, including immune checkpoint inhibitors for gliomas, should be administered exclusively within the context of well-designed clinical trials (45). The expression of immune checkpoints is crucial for immune evasion and determines the effectiveness of immune checkpoint inhibitor therapies. Ding et al. highlighted the specific role of HAVCR2 in the immunoregulation of glioma (46). Our investigation revealed a positive association between the high-risk group and elevated expression of immune checkpoint-related genes, suggesting that immune checkpoint inhibitors as a novel therapeutic approach may yield better outcomes in this group compared to the low-risk group.

Furthermore, we identified 196 medications with significantly differing IC50 values between the risk groups. This analysis was performed using the “oncoPredict” R package, a robust and widely recognized tool for predicting drug sensitivity based on genomic data. Our risk assessment indicated that temozolomide and carmustine were more appropriate for the high-risk group, in line with the National Comprehensive Cancer Network’s (NCCN) recommendations (11). For every risk group, the targeted medications that might have some therapeutic effects for LGGs were also examined. Therefore, based on our findings, we propose that, in order to treat LGGs effectively on an individual basis, it is essential to consider the patient’s gene profile, including FAM and other relevant indicators.

However, this study is not without limitations. Firstly, we created and validated a prognostic model using public databases, but we did not verify it using LGGs in vitro or in vivo. Secondly, owing to data constraints, immune character analysis and drug sensitivity were not confirmed in the testing set.

Conclusions

In brief, we have developed a FAM-related model for prognostic prediction and immune phenotyping of LGGs. It performed well in both TCGA and CGGA databases and provided insights into the potential effectiveness of immune and targeted therapies for personalized treatment approaches.

Acknowledgments

None.

Footnote

Reporting Checklist: The authors have completed the TRIPOD reporting checklist. Available at https://tcr.amegroups.com/article/view/10.21037/tcr-2025-1320/rc

Peer Review File: Available at https://tcr.amegroups.com/article/view/10.21037/tcr-2025-1320/prf

Funding: The study was supported by Beijing-Tianjin-Hebei Basic Research Cooperation Special Projects (No. 23JCZXJC00090/J230003).

Conflicts of Interest: All authors have completed the ICMJE uniform disclosure form (available at https://tcr.amegroups.com/article/view/10.21037/tcr-2025-1320/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. Bray F, Laversanne M, Sung H, et al. Global cancer statistics 2022: GLOBOCAN estimates of incidence and mortality worldwide for 36 cancers in 185 countries. CA Cancer J Clin 2024;74:229-63. [Crossref] [PubMed]
2. Ostrom QT, Patil N, Cioffi G, et al. CBTRUS Statistical Report: Primary Brain and Other Central Nervous System Tumors Diagnosed in the United States in 2013-2017. Neuro Oncol 2020;22:iv1-iv96. [Crossref] [PubMed]
3. Stupp R, Hegi ME, Mason WP, et al. Effects of radiotherapy with concomitant and adjuvant temozolomide versus radiotherapy alone on survival in glioblastoma in a randomised phase III study: 5-year analysis of the EORTC-NCIC trial. Lancet Oncol 2009;10:459-66. [Crossref] [PubMed]
4. Louis DN, Perry A, Reifenberger G, et al. The 2016 World Health Organization Classification of Tumors of the Central Nervous System: a summary. Acta Neuropathol 2016;131:803-20. [Crossref] [PubMed]
5. Ostrom QT, Bauchet L, Davis FG, et al. The epidemiology of glioma in adults: a "state of the science" review. Neuro Oncol 2014;16:896-913. [Crossref] [PubMed]
6. Xu E, Patterson J, Angione A, et al. Low-Grade Glioma Clinical Trials in the United States: A Systematic Review. Life (Basel) 2024;14:1133. [Crossref] [PubMed]
7. Jakola AS, Myrmel KS, Kloster R, et al. Comparison of a strategy favoring early surgical resection vs a strategy favoring watchful waiting in low-grade gliomas. JAMA 2012;308:1881-8. [Crossref] [PubMed]
8. Claus EB, Horlacher A, Hsu L, et al. Survival rates in patients with low-grade glioma after intraoperative magnetic resonance image guidance. Cancer 2005;103:1227-33. [Crossref] [PubMed]
9. Schiff D. Low-grade Gliomas. Continuum (Minneap Minn) 2017;23:1564-79. [Crossref] [PubMed]
10. Etxaniz O, Carrato C, de Aguirre I, et al. IDH mutation status trumps the Pignatti risk score as a prognostic marker in low-grade gliomas. J Neurooncol 2017;135:273-84. [Crossref] [PubMed]
11. Horbinski C, Nabors LB, Portnow J, et al. NCCN Guidelines® Insights: Central Nervous System Cancers, Version 2.2022. J Natl Compr Canc Netw 2023;21:12-20. [Crossref] [PubMed]
12. Koundouros N, Poulogiannis G. Reprogramming of fatty acid metabolism in cancer. Br J Cancer 2020;122:4-22. [Crossref] [PubMed]
13. Kao YC, Ho PC, Tu YK, et al. Lipids and Alzheimer's Disease. Int J Mol Sci 2020;21:1505. [Crossref] [PubMed]
14. Xicoy H, Wieringa B, Martens GJM. The Role of Lipids in Parkinson's Disease. Cells 2019;8:27. [Crossref] [PubMed]
15. Tang Y, Tian W, Xie J, et al. Prognosis and Dissection of Immunosuppressive Microenvironment in Breast Cancer Based on Fatty Acid Metabolism-Related Signature. Front Immunol 2022;13:843515. [Crossref] [PubMed]
16. Fu Y, Wang B, Fu P, et al. Delineation of fatty acid metabolism in gastric cancer: Therapeutic implications. World J Clin Cases 2023;11:4800-13. [Crossref] [PubMed]
17. Miska J, Chandel NS. Targeting fatty acid metabolism in glioblastoma. J Clin Invest 2023;133:e163448. [Crossref] [PubMed]
18. Qi Y, Chen D, Lu Q, et al. Bioinformatic Profiling Identifies a Fatty Acid Metabolism-Related Gene Risk Signature for Malignancy, Prognosis, and Immune Phenotype of Glioma. Dis Markers 2019;2019:3917040. [Crossref] [PubMed]
19. Li L, Leng W, Chen J, et al. Identification of a copper metabolism-related gene signature for predicting prognosis and immune response in glioma. Cancer Med 2023;12:10123-37. [Crossref] [PubMed]
20. Bauchet L, Ostrom QT. Epidemiology and Molecular Epidemiology. Neurosurg Clin N Am 2019;30:1-16. [Crossref] [PubMed]
21. Cancer Genome Atlas Research Network. Comprehensive, Integrative Genomic Analysis of Diffuse Lower-Grade Gliomas. N Engl J Med 2015;372:2481-98. [Crossref] [PubMed]
22. Mohile NA, Messersmith H, Gatson NT, et al. Therapy for Diffuse Astrocytic and Oligodendroglial Tumors in Adults: ASCO-SNO Guideline. J Clin Oncol 2022;40:403-26. [Crossref] [PubMed]
23. Nenkov M, Ma Y, Gaßler N, et al. Metabolic Reprogramming of Colorectal Cancer Cells and the Microenvironment: Implication for Therapy. Int J Mol Sci 2021;22:6262. [Crossref] [PubMed]
24. Petiti J, Arpinati L, Menga A, et al. The influence of fatty acid metabolism on T cell function in lung cancer. FEBS J 2025;292:3596-615. [Crossref] [PubMed]
25. Cui MY, Yi X, Zhu DX, et al. The Role of Lipid Metabolism in Gastric Cancer. Front Oncol 2022;12:916661. [Crossref] [PubMed]
26. Chuang HY, Lee YP, Lin WC, et al. Fatty Acid Inhibition Sensitizes Androgen-Dependent and -Independent Prostate Cancer to Radiotherapy via FASN/NF-κB Pathway. Sci Rep 2019;9:13284. [Crossref] [PubMed]
27. Liu YQ, Chai RC, Wang YZ, et al. Amino acid metabolism-related gene expression-based risk signature can better predict overall survival for glioma. Cancer Sci 2019;110:321-33. [Crossref] [PubMed]
28. Zhao S, Cai J, Li J, et al. Bioinformatic Profiling Identifies a Glucose-Related Risk Signature for the Malignancy of Glioma and the Survival of Patients. Mol Neurobiol 2017;54:8203-10. [Crossref] [PubMed]
29. Nagarajan SR, Butler LM, Hoy AJ. The diversity and breadth of cancer cell fatty acid metabolism. Cancer Metab 2021;9:2. [Crossref] [PubMed]
30. Zeng HL, Hu L, Chen X, et al. DIA-MS Based Proteomics Combined with RNA-Seq Data to Unveil the Mitochondrial Dysfunction in Human Glioblastoma. Molecules 2023;28:1595. [Crossref] [PubMed]
31. Jiang N, Xie B, Xiao W, et al. Fatty acid oxidation fuels glioblastoma radioresistance with CD47-mediated immune evasion. Nat Commun 2022;13:1511. [Crossref] [PubMed]
32. Nawaz Z, Patil V, Thinagararjan S, et al. Impact of somatic copy number alterations on the glioblastoma miRNome: miR-4484 is a genomically deleted tumour suppressor. Mol Oncol 2017;11:927-44. [Crossref] [PubMed]
33. Liu R, Feng Y, Deng Y, et al. A HIF1α-GPD1 feedforward loop inhibits the progression of renal clear cell carcinoma via mitochondrial function and lipid metabolism. J Exp Clin Cancer Res 2021;40:188. [Crossref] [PubMed]
34. Rusu P, Shao C, Neuerburg A, et al. GPD1 Specifically Marks Dormant Glioma Stem Cells with a Distinct Metabolic Profile. Cell Stem Cell 2019;25:241-257.e8. [Crossref] [PubMed]
35. Hiltunen JK, Kastaniotis AJ, Autio KJ, et al. 17B-hydroxysteroid dehydrogenases as acyl thioester metabolizing enzymes. Mol Cell Endocrinol 2019;489:107-18. [Crossref] [PubMed]
36. Salas S, Jézéquel P, Campion L, et al. Molecular characterization of the response to chemotherapy in conventional osteosarcomas: predictive value of HSD17B10 and IFITM2. Int J Cancer 2009;125:851-60. [Crossref] [PubMed]
37. Liu B, Liu J, Liu K, et al. A prognostic signature of five pseudogenes for predicting lower-grade gliomas. Biomed Pharmacother 2019;117:109116. [Crossref] [PubMed]
38. Chen J, Li Y, Han X, et al. An autophagic gene-based signature to predict the survival of patients with low-grade gliomas. Cancer Med 2021;10:1848-59. [Crossref] [PubMed]
39. Blitz SE, Kappel AD, Gessler FA, et al. Tumor-Associated Macrophages/Microglia in Glioblastoma Oncolytic Virotherapy: A Double-Edged Sword. Int J Mol Sci 2022;23:1808. [Crossref] [PubMed]
40. Sun Y, Liu L, Fu Y, et al. Metabolic reprogramming involves in transition of activated/resting CD4(+) memory T cells and prognosis of gastric cancer. Front Immunol 2023;14:1275461. [Crossref] [PubMed]
41. Bahrami A, Fereidouni M, Pirro M, et al. Modulation of regulatory T cells by natural products in cancer. Cancer Lett 2019;459:72-85. [Crossref] [PubMed]
42. Lochner M, Berod L, Sparwasser T. Fatty acid metabolism in the regulation of T cell function. Trends Immunol 2015;36:81-91. [Crossref] [PubMed]
43. Sun M, Yue Y, Wang X, et al. ALKBH5-mediated upregulation of CPT1A promotes macrophage fatty acid metabolism and M2 macrophage polarization, facilitating malignant progression of colorectal cancer. Exp Cell Res 2024;437:113994. [Crossref] [PubMed]
44. Wilky BA. Immune checkpoint inhibitors: The linchpins of modern immunotherapy. Immunol Rev 2019;290:6-23. [Crossref] [PubMed]
45. Sahm K, Weiss T. Immunotherapy against gliomas. Nervenarzt 2024;95:111-6. [Crossref] [PubMed]
46. Ding M, Li YA, Lu Z, et al. Identification of Potential Immune Checkpoint Inhibitor Targets in Gliomas via Bioinformatic Analyses. Biomed Res Int 2022;2022:1734847. [Crossref] [PubMed]