Cell death pathway regulation by fatty acid metabolism-related genes in neuroblastoma: a multi-omics analysis identifying CHD5 as a novel biomarker

Abstract

Background

Neuroblastoma in children is commonly found as an extracranial solid tumor with poor prognosis in high-risk cases impeding successful treatment. While dysregulated cell death mechanisms and metabolic reprogramming are hallmarks of cancer progression, the interplay between fatty acid metabolism and cell death pathway regulation in neuroblastoma remains incompletely understood.

Methods

Identifying molecular subtypes influenced by fatty acid metabolism were built by consensus clustering analysis. Independent prognostic genes were identified through random survival forest analysis, acquiring a novel risk signature. Risk signatures were validated internally and externally, and their independent prognostic value, immune landscape, and drug susceptibility were explored. The study systematically analyzed correlations between signature genes and seven major cell death pathways (apoptosis, pyroptosis, ferroptosis, autophagy, necroptosis, cuproptosis, and disulfidptosis), encompassing over 1,200 genes to comprehensively explore the intricate relationships between these molecular signatures and diverse cell death mechanisms. Gene Set Enrichment Analysis (GSEA) was performed to assess pathway-level associations. Utilizing a single-cell dataset of neuroblastoma samples, cells were categorized and labeled based on UMAP analysis. Feature map visualization was employed to display the expression level and allocation of specific genes across various cell populations. Validation of CHD5 expression in NB cells and tissues was confirmed through Western blotting and immunohistochemical staining.

Results

The study identified 42 fatty acid metabolism key enzyme genes whose expression was significantly different within high-risk and non-high-risk neuroblastoma patients, by which acquiring two distinct prognostic clusters associated with fatty acid metabolism. A machine learning approach was used to select 4 hub genes (CHD5, TP63, XKR4, and CTAG1A) for the establishment of a fatty acid metabolism prognostic risk model. Cell death pathway analysis revealed that TP63 exhibited the strongest correlations across multiple death pathways, particularly with necroptosis (r = 0.684, p = 2.80e-23) and pyroptosis (r = 0.647, p = 3.12e-20), while XKR4 showed moderate correlations with autophagy (r = 0.398, p = 2.09e-07) and CHD5 displayed selective associations. High risk score and low risk score groups displayed notable variations in the immune microenvironment, characterized by reduced immune cell infiltration in the high group leading to immune escape, and conversely, heightened responsiveness of the low group to immune checkpoint blockade therapy. Single-cell dataset analysis highlighted significant expression of CHD5 in specific cell populations, suggesting its potential as a marker gene for neuroblastoma. Immunohistochemical staining revealed varying levels of CHD5 expression across different clinical stages of neuroblastoma, with decreased deposition observed as staging advances. Functionally, CHD5 expression was found to inhibit proliferation, migration, and invasion of neuroblastoma cells.

Conclusion

The developed fatty acid metabolism prognostic risk model underscores the significance of fatty acids in neuroblastoma prognosis and immune landscape, thereby facilitating the optimization of chemotherapy and immunotherapy strategies for this disease. The comprehensive analysis of cell death pathways revealed distinct regulatory mechanisms of signature genes, particularly highlighting TP63's central role in coordinating multiple cell death processes. CHD5, as an identified gene inhibiting the proliferation, invasion and metastasis of neuroblastoma cells, serves as a novel tumor biomarker.

Introduction

Neuroblastoma, a prevalent solid tumor in children outside the brain, typically arises during the development of the sympathetic nervous system [1]. It accounts for approximately 15% of cancer-related deaths in this age group, often manifesting around 23 months of age [2]. While patients with non-high-risk neuroblastoma have relatively satisfactory prognoses, with 5-year survival rates of around 97.9 ± 0.5% and 95.8 ± 0.8% respectively, those with high-risk neuroblastoma face significantly lower survival rates at approximately 62.5 ± 1.3%, despite undergoing diverse treatment strategies [3]. Neuroblastoma exhibits remarkable heterogeneity in its clinical presentation and molecular characteristics, with key genetic alterations including MYCN amplification, found in approximately 20% of cases and strongly associated with poor prognosis [4]. The tumors can range from spontaneously regressing lesions to highly aggressive metastatic disease, with most primary tumors occurring in the adrenal medulla (40%) or sympathetic ganglia [5]. Hence, exploring the underlying factors contributing to the discrepancy in outcomes between high-risk and non-high-risk neuroblastoma patients is crucial for further research.

During lipid metabolism, nutrients are converted into metabolic intermediates that are used for multiple functions including the synthesis of cell membranes, storing energy, and generating signaling molecules [6]. Cancer cells exhibit a distinct metabolic phenotype characterized by alterations in lipid metabolism, and neuroblastoma demonstrates particularly notable abnormalities in these pathways. In neuroblastoma, MYCN amplification drives increased fatty acid uptake and biosynthesis through upregulation of fatty acid transport protein 2 (FATP2), while genes involved in unsaturated fatty acid synthesis, such as FASN and ELOVL6, are specifically upregulated in high-risk cases [7]. Interrupting the supply of lipids to cancer cells greatly affects their bioenergetics, intracellular signal transmission, and membrane production [8]. Previous research has shown that most tumors display aberrant lipid metabolism, with neuroblastoma exhibiting distinct alterations in glycerolipid metabolism and steroid hormone biosynthesis, particularly in MYCN-amplified cases [9]. Activated neutrophils with pathologic characteristics, also known as polymorphonuclear myeloid-derived suppressor cells, are crucial in modulating the immune response to cancer [10]. Selective pharmacological inhibition of FATP2 can effectively eliminate polymorphonuclear myeloid-derived suppressor cells' activities and significantly delay tumor progression in mice [11, 12].

Regulated cell death represents a crucial area of research in tumor biology, with currently recognized types including apoptosis, autophagy, pyroptosis, ferroptosis, necroptosis, cuproptosis, and disulfidptosis [13]. These cell death modalities exhibit intricate and profound interactions with fatty acid metabolism. Ferroptosis, for instance, critically depends on lipid peroxidation processes, where the unsaturation degree and oxidative state of fatty acids directly influence the progression of cell death [14]; during autophagy, the remodeling of fatty acid metabolism can modulate mitochondrial function and cellular stress responses[15]; in apoptosis, alterations in fatty acid metabolism can affect cell survival by regulating mitochondrial membrane permeability and cellular energy metabolism [16]. Moreover, pyroptosis [17] and necroptosis [18] are also closely associated with lipid metabolic abnormalities, with the fatty acid metabolic state capable of influencing these cell death modes through regulating inflammatory factor release and mitochondrial function. Despite the growing understanding of cell death mechanisms in various cancer types, the specific interactions between lipid metabolism and regulated cell death in neuroblastoma remain largely unexplored, presenting a significant knowledge gap that urgently requires comprehensive investigation.

The primary objective of this study was to unravel the complex interplay between fatty acid metabolism and neuroblastoma progression, with a particular focus on exploring the mechanisms of regulated cell death. The research aims to comprehensively investigate the expression and functional roles of fatty acid metabolism key enzyme genes, identify potential molecular markers, and develop a novel prognostic model that could provide deeper insights into neuroblastoma pathogenesis. To achieve these goals, a multi-dimensional research approach was employed, including consensus clustering analysis, univariate Cox regression, single-cell sequencing, and functional experiments. An initial detailed analysis focused on gene expression differences related to fatty acid metabolism between high-risk and non-high-risk neuroblastoma patients, successfully categorizing these genes into distinct subtypes. Through comprehensive bioinformatics analyses and experimental validation, CHD5 was identified as a key gene with significant implications for neuroblastoma progression and cell death regulation. By integrating computational predictions with experimental validation, the study aimed to provide novel insights into the molecular mechanisms underlying neuroblastoma and potentially develop more precise diagnostic and therapeutic strategies.

Materials and methods

Samples collection

From May 2013 to June 2023, a total of 40 neuroblastoma tumor tissue samples were collected at the Cancer Institute and Hospital of Tianjin Medical University, each with detailed clinical and prognostic data. This study follows the guiding principles of the Declaration of Helsinki of the World Medical Association and has been approved by the Cancer Research Institute of Tianjin Medical University and the Hospital Research Ethics Committee (approval number E20210027). The tumor tissue samples were collected during surgery and immediately snap-frozen in liquid nitrogen, then stored at −80 °C until further processing. For inclusion in the study, samples had to meet the following criteria: (1) confirmed pathological diagnosis of neuroblastoma; (2) tumor content ≥ 60% as assessed by pathological examination; (3) no prior chemotherapy or radiation therapy before sample collection; (4) complete clinical follow-up data available. Samples were excluded if they showed significant necrosis, had inadequate RNA quality (RIN < 7), or insufficient tissue quantity for analysis. All tissue samples were processed according to standard operating procedures, including controlled thawing and tissue homogenization using a mechanical homogenizer in appropriate lysis buffer for subsequent molecular analyses.

Data preparation and processing

The clinical data and mRNA expression profile of the GSE49710 dataset were obtained from the Gene Expression Omnibus (GEO) repository [19]. The raw microarray data from GSE49710 were preprocessed using the robust multi-array average (RMA) method for background correction, followed by log2 transformation. Low-quality probes and those with missing values were filtered out. In addition, we collected gene expression and clinical data from the UCSC Xena database, which is well known for the Therapeutically Applicable Research to Generate Effective Treatments (TARGET). The E-MTAB-8248 dataset was downloaded from the Array Express database. Using the "Search" function of MSigDB database, we searched "fatty acid metabolism" as a keyword, and obtained a gene set of "FATTY ACID METABOLISM", which contains 103 genes.

To process multiple datasets and remove batch effects, quantile normalization was first performed to ensure comparable distribution of expression values across GSE49710, TARGET-NBL, and E-MTAB-8248 datasets. Subsequently, the ComBat algorithm from the 'sva' R package was employed to eliminate batch effects while preserving biological variation. ComBat was specifically chosen because it has been widely validated for handling batch effects in gene expression data and is particularly effective for cancer datasets. For quality control, principal component analysis (PCA) was conducted before and after batch correction to verify the effectiveness of the harmonization process. Separate training and validation sets were maintained during this process to prevent information leakage and ensure robust validation of the findings.

Clustering analysis

One hundred and three fatty acid metabolism key enzyme genes were extracted from MSigDB. Considering the significant differences between high-risk and non-high-risk groups in the Children's Oncology Group (COG) neuroblastoma risk rating, 42 fatty acid metabolism-related enzyme genes were identified. The Consensus Cluster Plus R package was used to cluster differentially expressed genes associated with fatty acid disorder, using Euclidean distance as the basis. To ensure a secure and dependable subgroup categorization, the K-means clustering algorithm was applied, with a maximum of 5 clusters allowed. The analysis resulted in the identification of two subtypes, for which differential gene expression was examined (with screening criteria set at log2FC > 1 and P < 0.05).

Construction and verification of prognostic risk model

Based on the expression levels of 42 fatty acid metabolism key enzyme genes, samples from the GSE49710 neuroblastoma dataset were classified into groups to identify clusters with comparable gene expression patterns. A risk model was then constructed using features from each subtype. Genes affecting overall survival (OS) were identified through univariate Cox regression analysis. Next, the random forest supervised classification algorithm was used to determine significant signals, leading to the creation of a scoring tool known as the m7Sig risk score, which calculated the sum of each gene's expression value multiplied by its coefficient in the Cox regression model.

To develop the optimal risk model, Kaplan–Meier (KM) analysis was utilized, and -log10(p-value) for all models was compared. Finally, the genes comprising the risk profile were screened. The GSE49710 dataset served as the training set, while E-MTAB-8248 and TARGET-NBL datasets functioned as test sets. The cutoff value for patient risk groups was determined by calculating the median risk score from the training set. Subsequently, patients in both the training and test sets were stratified into high- and low-risk groups accordingly.

The “heatmap” R package helped perform the allocation of patient risk scores and their corresponding survival status map. To assess significant survival differences within high- and low-risk patients, the “Survival” R package was used to construct Kaplan–Meier survival curves. Univariate Cox proportional hazard regression analysis was conducted, incorporating both risk scores and clinical factors. To assess whether these factors had an independent impact on patient prognosis, a multivariate Cox proportional hazard regression analysis was conducted.

Furthermore, each dataset was stratified based on clinical characteristics, allowing the plotting of Kaplan–Meier survival curves for each subgroup to observe prognostic disparities within risk groups. Lastly, the “time ROC” R package was used to construct time-dependent ROC curves, which allowed evaluation of model accuracy in predicting patient survival across various periods.

Enrichment analysis

The biological roles of selected genes were examined using the 'clusterProfiler' R package (Version 4.8.0). This involved conducting enrichment analysis for the Kyoto Encyclopedia of Genes and Genomes (KEGG) and Gene Ontology (GO). Significant findings were identified with a significance threshold of adjusted p-value < 0.05.

Assessment of immune cell infiltration

The default parameters of the IOBR R package's CIBERSORT algorithm were used to compute immune cell scores in 22 tumor microenvironments. The TCGA-STAD data was employed to determine the infiltration proportion of immune cells based on gene expression profiles.

Drug sensitivity

The R package “pRRophetic” and gene expression data were used to predict sensitivity for over a hundred drugs in the GDSC database. The efficacy of drug treatments for patients with neuroblastoma was then assessed using these predicted IC50 values.

Single-cell dataset analysis

The single-cell dataset GSE137804 of UVM was downloaded from the GEO database. Data quality control was carried out. Samples were integrated through SCT correction. By setting the "DIMS" parameter to 20, the t-SNE method was used to reduce data dimensionality, and the "KNN" method was used with a resolution of 1.0 to conduct cell clustering. Subsequently, cells were annotated by various cell surface markers. Finally, the percentage of selected genes in each cell was obtained by importing them through the “PercentageFeatureSet” function.

Cell death pathway correlation analysis

The relationships between the four-gene signature (CHD5, TP63, XKR4, and CTAG1A) and various cell death pathways were investigated through correlation analysis. Gene sets associated with seven types of cell death were collected from published literature and molecular signature databases, encompassing apoptosis (580 genes), pyroptosis (52 genes), ferroptosis (88 genes), autophagy (367 genes), necroptosis (101 genes), cuproptosis (14 genes), and disulfidptosis (23 genes). Correlation coefficients were calculated between our signature genes and the genes involved in each death pathway. Hierarchical clustering was performed to visualize the correlation patterns. Gene Set Enrichment Analysis (GSEA) scoring was performed for each death pathway, and correlations with the signature genes were analyzed to understand the broader pathway-level associations.

Western blotting

The cells were disrupted using RIPA lysis buffer (Solarbio) to obtain protein samples. The concentrations of protein in supernatants were determined by BCA. Protein samples were separated on 8% SDS gels and subsequently transferred to PVDF membranes. Followed by incubation with antibodies targeting CHD5 (dilution 1:1000, ab300437, abcam) (overnight, 4 °C), the membranes were blocked with 5% skimmed milk (1 h, room temperature). The membranes were incubated with secondary antibody, washed three times with TBS-T, and then the band images were visualized using GelView 6000Plus system(Biolight Biotechnology in Guangzhou, China).

Immunohistochemistry (IHC)

The histopathological sections were prepared by cutting them into 4 μm thick slices and then subjecting them to an overnight baking process at 60 °C. Paraffin was then removed, and the sections were dehydrated. Antigen retrieval was performed, and endogenous peroxidase activity was blocked. The slices were incubated with antibodies targeting CHD5 (1:1000 dilution; ab300437; Abcam) in a humid container at 4 °C for 24 h. This was followed by incubation with matching secondary antibodies for 1 h at room temperature. Subsequently, diaminobenzidine (DAB) was added for five minutes until a brown reaction product appeared. Digital images of the stained sections were captured using Leica light microscopes. To evaluate the IHC sections, ImageJ software (Version 1.53 s) was used to measure the area and integrated optical density (IOD). The average optical density (AOD), calculated as AOD = IOD/Area, served as an indicator of CHD5 staining intensity or relative expression level. To ensure unbiased analysis, signal density in tissue regions from five randomly selected fields was quantified without knowledge of sample identity or group assignment. High expression was defined as AOD ≥ 0.3, while low expression corresponded to AOD < 0.3.

Cell lines and cultures

The human SK-N-BE (2) neuroblastoma cell line was obtained from ATCC. The cells were cultured and maintained in a medium consisting of 90% MEM/F12, 10% fetal bovine serum, 1% Glutamine/Gluta-MAX (2 mM), 1% NEAA (1X), 1% Sodium Pyruvate (0.11 g/L, 1 mM), and 1% penicillin/streptomycin. Cells were incubated in a humidified incubator at 37 °C with 5% CO₂.

Cell transfection

We obtained the CHD5-targeting small interfering RNA and a negative control siRNA from General Biol. The transfection of neuroblastoma cells with siRNA was performed according to the manufacturer's guidelines using Lipofectamine 3000 transfection reagent from Invitrogen. After 24 h, transfected cells were harvested for subsequent experiments. Western blot was performed to verify the transfection efficiency.

Cell counting kit 8

Cell proliferation was assessed using the CCK-8 assay. Neuroblastoma cells were cultured in 96-well plates at a consistent temperature of 37 °C with 5% CO₂. The CCK-8 solution (10 μL/well) was introduced into the wells at specified intervals (0, 24, 48, and 72 h) and then left to incubate for 2 h at 37 °C. Using a microplate reader, three readings of absorbance at 450 nm were taken for each sample.

Scratch test

A new six-well plate should be taken out, and a ruler should be used to draw straight lines across each well. Three evenly spaced lines should be drawn for each well. Approximately 300,000 to 600,000 cells should be added into each well, gently shaken to distribute them evenly, and incubated overnight until confluent growth is achieved. On the next day, two perpendicular scratches should be created on the back of the six-well plate using a 10 μL pipette tip, along the ruler on both sides of each well. The vertical alignment of the pipette tip must be maintained, and quick single strokes should be performed without repeating any lines. Cells should be treated with 1 × PBS solution three times to remove any detached or clumped cells. In the control group, serum-free medium should be added, while in the experimental group, serum-free medium containing a specific concentration of drugs (e.g., 5 μM, 10 μM) should be added. The plate should be incubated at 37 ℃ with 5% CO₂. At designated time intervals (such as 0 h, 24 h, and 48 h), samples should be collected and photographs should be captured. Both scratch width and area should be measured and recorded using ImageJ software.

Statistical analysis

The R program (version 4.3.1) was used for statistical analysis. Kaplan–Meier method was used to create survival curves, and the log-rank test was used to compare group differences. To independently evaluate the prognostic significance of risk scores, a Cox regression model combined with other clinical features was applied for both univariate and multivariate analyses. Time-dependent ROC curve analysis was performed to evaluate the predictive value of prognostic characteristics. ROC analysis was used to assess sensitivity and specificity in predicting prognosis, with the area under the curve (AUC) considered as an indicator of prognostic judgment.

Results

Comparative analysis of expression patterns and functional roles of fatty acid metabolism key enzyme genes in neuroblastoma tissues with high-risk and non-high-risk levels

The cohort GSE49710 provided a total of 498 neuroblastoma tumor tissue samples, consisting of 176 high-risk and 322 non-high-risk cases. Differential expression analysis was conducted using the criteria of |log2FC|>0.585 and a significance level of P<0.05, revealing a set of 6494 genes that exhibited differential expression between highrisk and non-high-risk neuroblastoma tumors. Among these genes, there were 490 up-regulated and 6004 down-regulated genes. A subset of 42 differentially- expressed fatty acid metabolism keyenzyme genes was identified by intersecting with the geneset extracted from MSigDB. To visualize the results, we generated a volcano map (Fig. 1A) and a thermogram specifically highlighting the differentially expressed fatty acid metabolism key enzyme genes (Figure B). Protein interaction analysis revealed strong connectivity among several key genes including ACAA1, ACACB, CPT2, FABP3, LPL. Additionally, the key enzyme genes were classified into four distinct groups by calculating the correlation among their expression patterns (Figures C, D). A significant enrichment of functional GO terms was observed for enzymes involved in oxidoreductase activity, specifically targeting the CH-CH group of donors. Acyltransferase activity was found to transfer groups other than amino-acyl groups. These enzymes play crucial roles in processes such as fatty acid elongation, glycerophospholipid metabolism and PPAR signaling pathway (Figs. 1E–H). Analysis of clinical features revealed notable variations in the expression levels of key enzyme genes across different age groups, INSS stages, disease progression, and MYCN status (Fig. 1B).

Fig. 1.

Expression disturbance of fatty acid-related genes in neuroblastomas. A Volcanogram of fatty acid related genes differentially expressed in high-risk and non-high-risk neuroblastomas. B Heat map of differentially expressed fatty acid-related genes. C Based on the differentially expressed fatty acid related genes, the PPI network was constructed by using STRING database. D Expression correlation analysis of genes related to differential fatty acids. E, F, G GO functional enrichment analysis of genes related to differential fatty acids. H KEGG functional enrichment analysis of genes related to differential fatty acids

Identification of molecular subtypes using genes associated with the metabolism of fatty acids

Subtyping can aid in identifying distinct tumor states, facilitating the implementation of personalized treatment approaches. The GSE49710 neuroblastoma dataset was utilized to perform consistency clustering according to the levels of 42 key enzyme genes. This allowed for the identification of sample groups exhibiting similar expressions. Analysis of the cumulative distribution function and delta area plot revealed a smooth change in the CDF curve when k = 2. Consequently, the samples were categorized into two clusters as A and B (Fig. 2A, C). Notably, patients belonging to different fatty acid metabolism subtypes exhibited significant differences in prognosis, with cluster B having a worse survival time compared to cluster A patients (Fig. 2D). By employing SSGSEA to calculate scores for pathways like DNA replication and cell cycle within KEGG gene sets, notable variations were observed between these scores among different molecular subtypes related to fatty acid metabolism (Fig. 2E). Evaluation of immune cell infiltration through CIBERSORT analysis demonstrated marked distinctions in effector memory CD4 and CD8 T cells, CD56 + dim and CD56 + bright natural killer cells, plasmacytoid dendritic cells across subtypes (Fig. 2F). ESTIMATE algorithms were utilized to calculate tumor purity values, imm une and stromal scores (Fig. 2G), providing deeper insights into the immune cell infiltration characteristics within each molecular subtype.

Fig. 2.

Molecular subtype recognition of fatty acid related genes. A, B GSE49710 neuroblastoma samples were analyzed by consistent cluster analysis based on 42 fatty acid related genes. C The consistency matrix heat map when the number of clusters is 2. D Survival curve of patients with fatty acid subtypes. E The scores of related pathways in KEGG gene set in MSigDB database were calculated by SSGSEA. F Using CIBERSORT to calculate the heat map of immune cell infiltration. (G) ESTIMATE was used to calculate immune score, matrix score and tumor purity. ns means p > 0.05, *p <  = 0.05, **p <  = 0.01 and ***p <  = 0.001

Construction of a prognostic signature associated with the metabolism of fatty acids

Given the distinct clinical characteristics and variations in immune microenvironment landscape of the two subtypes, a risk model was developed based on each subtype's signature. Following a univariable Cox regression analysis to identify genes significantly associated with overall survival, 10 genes were screened using a random forest supervised classification algorithm (Fig. 3A, B). After comparing all risk models using Kaplan–Meier (KM) analysis, a four-gene risk signature comprising CHD5, TP63, XKR4, and CTAG1A was identified (Fig. 3C).

Fig. 3.

Construction of FARS and validation. A Volcano plot illustrating the FARS subtype-based genes by Univariable Cox Regression analysis. B Random survival forest analysis screening 10 genes. C After Kaplan–Meier analysis of various combinations, the top 20 signatures are ordered by the p-value. The four-gene signature was established, for it had a relatively great − log10(p value). D FARS risk score analysis of patients in GSE49710 cohort. E Kaplan–Meier OS analysis of neuroblastoma patients in low-risk and high-risk groups. F The time-dependent ROC curves analysis for FARS in GSE49710 cohort. G, H Univariate and multivariate Cox analysis were used to determine whether Score was an independent prognostic factor for OS. (I) TARGET-NBL dataset was used to analyze the relationship between survival status/risk score, mRNA expression heat map of 4 genes and survival time (days)/risk score. J Kaplan–Meier OS analysis of neuroblastoma patients in low-risk and high-risk groups based on TARGET-NBL data set K Time-dependent ROC curve of TARGET-NBL dataset OS. AUC was evaluated at 2 year, 3 years and 5 years, respectively. P value was calculated using the log-rank test. P < 0.001. L, M The TARGET-NBL dataset uses univariate and multivariate Cox analysis to determine whether Score is an independent prognostic factor for OS. N The relationship between survival status/risk score, mRNA expression heat map of 4 genes and survival time (days)/risk score was analyzed by E-MATB-8248 dataset. O Kaplan–Meier OS analysis of neuroblastoma patients in low-risk and high-risk groups based on E-MATB-8248 data set. P Time-dependent ROC curve of E-MATB-8248 dataset OS. AUC was evaluated at 2 year, 3 years and 5 years, respectively. P value was calculated using the log-rank test. P < 0.001. Q, R The E-MATB-8248 dataset uses univariate and multivariate Cox analysis to determine whether Score is an independent prognostic factor for OS

Assessing the predictive effectiveness of the model by analyzing both training and externally validated datasets

Patients were categorized in two sets based on their individual scores, which were calculated using a specific formula. The Kaplan–Meier curve analysis revealed a significantly lower survival probability for high score patients compared to the other patients (Fig. 3D, E). To assess the prognostic accuracy of the model for predicting survival rates, a time-dependent ROC analysis was conducted. As indicated by the area under the ROC curve, the model demonstrated favorable predictive performance with an AUC of 0.74 at 2 years, 0.70 at 3 years, and 0.67 at 5 years (Fig. 3F). Univarate and multivariate Cox regression analyses were conducted to access whether the score fuctioned as an independent prognostic indicator for overall survival. The univariate analysis revealed a notable association between Score obtained from the GSE49710 dataset and overall survival. Following adjustments for other potential influencing factors, the multivariate analysis validated that score continued to be a significant predictor of overall survival (Fig. 3G, H).

To ensure the model's stability, the same algorithm was utilized to calculate each sample's score in both TARGETNBL and E-MATB-8248 datasets. According to median score, neuroblastoma patients were categorized in two groups. Similar to our findings from GSE49710, patients with higher scores had a poor chance of survival compared to the others (as shown in Fig. 3I, J, N, O). Furthermore, our prognostic model demonstrated that AUC was 0.66 for 2-year, 3-year survival rate and 0.70 for 5-year survival rate in TARGET-NBL dataset (Fig. 3K), while it was found to be at AUC of 0.69 for a two years period, an AUC of 0.72 over three years and an AUC of 0.71 over five years in E-MATB-8248 dataset (Fig. 3P). In order to validate these results further univariate as well as multivariate cox analyses were conducted. Univariate cox analysis revealed that the score is evidently correlated with overall survival (Fig. 3M, R) and adjusting for other potential influencing variables using multivariate cox analysis, the score continues to be an independent prognostic factor (Fig. 3L, Q). Model is associated with the tumor’s clinical attributes.

Figure 4 demonstrates the relationship between risk scoring system, immune infiltration, and therapeutic response in neuroblastoma. Our analysis revealed that higher risk scores were significantly associated with adverse clinical characteristics, including INSS stage 4, COG high-risk group, age ≥ 1.5 years, and MYCN amplification. Notably, patients in cluster B exhibited higher scores and poorer prognosis compared to cluster A (Fig. 4A). Utilizing the CIBERSORT algorithm to assess immune cell infiltration, we discovered significant correlations were discovered between several immune cell types, including cytotoxic lymphocytes, NK cells, myeloid dendritic cells, and the risk score. Notably, patients in the high score group displayed limited immune cell infiltration overall (Fig. 4B). Examining gene expression of Immunomodulators within different risk score subgroups revealed significantly lower levels across various types of immunomodulators in the high-score group (Fig. 4C). Analysis of immune checkpoint expression between risk groups revealed distinct patterns across 20 different checkpoint molecules (Fig. 4D). Among these checkpoints, only CD86, PVR, and VTCN1 showed no significant differences between groups (marked as ‘ns’) (Fig. 4D). The remaining 17 immune checkpoint genes demonstrated significantly higher expression in the low-risk group compared to the high-risk group: ADORA2A, BTLA, CD200, CD200R1, CD274, CD276, CD38, CD80, CEACAM1, CTLA4, HAVCR2, KIR3DL1, LAGS3, LAIR1, LGALS9, and NTFE, and PDCD1 (p < 0.05) (Fig. 4D). The consistently lower checkpoint expression in the high-risk group suggests a potential mechanism of immune evasion, supporting the 'cold tumor' phenotype in high-score neuroblastoma and potentially explaining the limited efficacy of immune checkpoint blockade therapy in these patients. SSGSEA correlation analysis revealed a strong association between risk scores and chemokine receptor scarcity, suggesting that reduced chemokine signaling might facilitate neuroblastoma proliferation and metastasis (Fig. 4E). Drug sensitivity analysis using the pRRophetic package revealed differential therapeutic responses across ten compounds between high and low-risk groups. The analysis identified two distinct response patterns: one group of drugs showed higher sensitivity in the low-risk group, including Sepantronium bromide (p < 2.2e-16), Pevonedistat (p = 7.7e-11), AZD4547 (p < 2.2e-16), Dactinomycin (p = 2.4e-06), and Bortezomib (p = 4.9e-07), while another group demonstrated increased sensitivity in the high-risk group, comprising Paclitaxel (p < 2.2e-16), PD0325901 (p < 2.2e-16), Gemcitabine (p = 4e-14), Vinblastine (p = 1.9e-14), and Eg5_9814 (p = 1.4e-14)(Fig. 4F).

Fig. 4.

Relationship between risk scoring system and immune infiltration and immunotherapy. A Analysis of correlation between risk score and clinical characteristics of patients with neuroblastoma. B Heatmap of tumor-related infiltrating immune cells based on MCPCOUNTER, CIBERSORT, CIBERSORT-ABS, EPIC, XCELL, and QUANTISEQ algorithms in the two subtypes of risk score group and different clinical characteristics of patients with neuroblastoma. C Heatmap of different types of immunomodulator in the two subtypes of risk score group. D The difference of gene expression in immune checkpoint between high and low score groups. E Pearson analysis of SSGSEA score and fatty acid risk score. F Correlation analysis between fatty acid risk score and chemotherapy resistance. *p <  = 0.05, ***p <  = 0.001, ****p <  = 0.0001, ns means p > 0.05

Single-cell dataset analysis

Utilizing neuroblastoma samples from the GEO database (GSE137804), a single-cell dataset was employed to group and annotate cells based on UMAP analysis. A total of 160,910 cells from 16 patients were categorized into 37 clusters, ultimately identified as a few major cell types including Schwann cells, T cells, myeloid cells, fibroblasts and neuroendocrine cell(NE) cells (Fig. 5A). Subsequently, the expression pattern and distribution of four model genes (CHD5, TP63, XKR4, and CTAG1A) within the cell population were emerged by feature plot visualization. The findings revealed significant expression of CHD5 in the cell population while XKR4, TP63, and CTAG1A were not present in this cluster. This indicates that CHD5 can be used as a potential biomarker for neuroblastoma identification (Fig. 5B).

Fig. 5.

Single cell sequencing analysis of GSE137804. A UMAP clustering and cell group annotation based on NB samples from single-cell data sets. B Display the expression and distribution of model genes in the cell population by Feature Plot visualization

Integration analysis results of signature genes and cell death mechanisms in neuroblastoma

Through correlation analysis of signature genes with different cell death pathways (Fig. 6A–G), TP63 consistently demonstrated the strongest correlations across multiple death pathways, showing notably positive associations with apoptosis, pyroptosis, ferroptosis, autophagy and necroptosis genes, while XKR4 exhibited moderate correlations, CHD5 showed selective associations, and CTAG1A displayed minimal correlations with most pathway genes. The hierarchical clustering revealed distinct patterns of co-regulation between the signature genes and specific cell death pathway components, suggesting pathway-specific regulatory mechanisms.

Fig. 6.

Correlation analysis of signature genes with cell death pathways. A Correlation heatmap showing relationships between signature genes and apoptosis-related genes. B Correlation heatmap showing relationships between signature genes and pyroptosis-related genes. C Correlation heatmap showing relationships between signature genes and ferroptosis-related genes. D Correlation heatmap showing relationships between signature genes and autophagy-related genes. E Correlation heatmap showing relationships between signature genes and necroptosis-related genes. F Correlation heatmap showing relationships between signature genes and cuproptosis-related genes. G Correlation heatmap showing relationships between signature genes and disulfidptosis-related genes. H GSEA score correlation heatmap between signature genes (CHD5, TP63, XKR4, and CTAG1A) and seven cell death pathways. Color scale represents correlation coefficient from -0.5 (blue) to 1 (dark red). Hierarchical clustering dendrograms show the relationship patterns. Asterisks indicate statistical significance: *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001

GSEA score correlation analysis comprehensively revealed the relationships between the four signature genes and cell death pathways (Fig. 6H). TP63 demonstrated the most significant and extensive correlation patterns: showing the strongest positive correlation with the necroptosis pathway (r = 0.684, p = 2.80e-23), followed by significant positive correlations with pyroptosis (r = 0.647, p = 3.12e-20), autophagy (r = 0.600, p = 6.74e-17), apoptosis (r = 0.553, p = 3.96e-14), and ferroptosis pathways (r = 0.404, p = 1.29e-07), while exhibiting a weak negative correlation with the cuproptosis pathway (r = −0.269, p = 0.0006). XKR4 primarily displayed moderate positive correlations with autophagy (r = 0.398, p = 2.09e-07), necroptosis (r = 0.302, p = 0.0001), and pyroptosis pathways (r = 0.288, p = 0.0002). CHD5 showed significant but relatively weak positive correlation only with the autophagy pathway (r = 0.223, p = 0.005), while showing no significant correlations with other pathways. CTAG1A exhibited no statistically significant correlations with any cell death pathways, with all absolute correlation coefficients less than 0.13 and p-values greater than 0.05. These findings not only reveal the central role of TP63 in regulating multiple cell death pathways but also suggest that different signature genes might influence neuroblastoma progression through specific regulation of distinct cell death mechanisms.

Biological significance of CHD5 in neuroblastoma

A significant association was observed between reduced CHD5 expression and shorter survival time in neuroblastoma patients (Fig. 7A). Increased CHD5 expression was associated with higher immune score, stromal score and estimate score, indicating a weakened immune response as the TME score shows (Fig. 7B). Examination of immune infiltration showed a correlation between CHD5 expression and 17 out of 28 types of immune cells (Fig. 7C). Clinicopathological characteristics, including MYCN status, INSS stage, age, and COG risk group, were investigated in patients with different CHD5 expression levels. Low levels of CHD5 expression were strongly linked to worse outcomes, suggesting a more aggressive cancer phenotype as indicated by these characteristics (Fig. 7D, E). Additionally, the expression level of CHD5was explored for its potential relationship with immune checkpoints to some extent (Fig. 7F). To validate the present findings regarding CHD5 expression in neuroblastoma, immunohistochemical staining demonstrated varying degrees of CHD5 deposition across different clinical stages. Notably, as staging increased, there was less deposition of CHD5 (Fig. 7G). We transfected SK-N-BE (2) cells with siRNA targeting CHD5 in order to assess the significance of CHD5 in neuroblastoma development. Western blot analysis confirmed successful downregulation of CHD5 after transfection for subsequent experiments (Figs. 7H). As shown in Fig. 7J, quantification revealed significant differences in CHD5 protein levels across experimental groups. Compared to IMR32 cells, CHD5 expression was significantly decreased in both SK-N-BE (2) control cells and si-NC cells (*p < 0.05). The si-CHD5 transfected SK-N-BE(2) cells displayed an even more pronounced reduction in CHD5 expression (***p < 0.001 compared to IMR32; ***p < 0.001 compared to control; **p < 0.01 compared to si-NC). No significant difference was observed between control and si-NC groups. Results of CCK-8 assay showed that down-regulating the expression level of CHD5 could significantly promote the growth of SK-N-BE (2) cells (Fig. 7I), which was consistent with our scratch test results demonstrating an increased number of cell clones compared to the control group (Fig. 7K). The present findings suggest that inhibiting the progression of neuroblastoma is possible through modulation of CHD5 expression.

Fig. 7.

Biological significance of CHD5 in neuroblastoma. A The relationship between the high and low expression of CHD5 in neuroblastoma and prognosis B The relationship between the high and low expression of CHD5 and the score of TME C The correlation between the high and low expression of CHD5 and immune cell infiltration. D, E Analysis of the correlation between the high and low expression of CHD5 and the clinicopathological features of patients F The relationship between the high and low expression of CHD5 and immune checkpoints G Expression of CHD5 in different clinical stages of neuroblastoma. H The western blot analysis was performed to validate the downregulation of CHD5 after transfection with siRNA in SK-N-BE(2) cells. I The CCK-8 assay was performed to measure the proliferation capacity of SK-N-BE(2) cells. J Quantitative analysis of Western blot band intensities. (K) Experimental observation of cell scratch (10x)

Biological function of CHD5 in various types of cancer

A comprehensive assessment across multiple cancer types revealed a significant association within the expression of CHD5 and overall survival time in UCEC, LGG, and PAAD patients (Fig. 8A). Furthermore, there was a notable correlation between CHD5 expression and disease-free interval (DFI) in PRAD, UCEC, and STAD cases (Fig. 8B). The study also identified a link between disease-specific survival (DSS) and UCEC as well as LGG patients (Fig. 8C), along with a significant relationship between progression-free interval (PFI) and UCEC, MESO, LGG, and PAAD individuals (Fig. 8D). The present investigation further explored the connection among CHD5, TMB, and MSI. Remarkably strong associations were observed in UCEC, BRCA COAD STAD, and PRAD tumors. These results reminded that CHD5 could be used as an essential marker for immunotherapy detection targeting these specific malignancies (Fig. 8E, F). Lastly, the close association of CHD5 with immune cell infiltration levels across various tumors indicates its involvement in regulating tumor immune responses within the microenvironment (Fig. 8G).

Fig. 8.

Biological significance of CHD5 in pan-cancer. A Relationship between CHD5 expression level and OS B Relationship between CHD5 expression level and disease free interval (DFI) C Relationship between CHD5 expression level and Disease specific survival (DSS) D Relationship between CHD5 expression level and Progression free interval (PFI) E Correlation analysis between CHD5 and Tumor mutation burden (TMB) F Analysis of the correlation between CHD5 and Microsatellite instability (MSI) G The relationship between CHD5 and immune cell infiltration. *p <  = 0.05, **p <  = 0.01, ***p <  = 0.001

Discussion

As the most prevalent extracranial solid malignant tumor in childhood, neuroblastoma is considered to originate in the developing sympathetic nervous system. Over half of neuroblastomas occur in the adrenal gland, with other common sites including the neck, chest, and pelvic cavity. The clinical signs and symptoms vary significantly and are closely associated with the location of the primary tumor, metastasis, and paraneoplastic syndrome. More than 90% of neuroblastoma cases affect children under ten years old, with a median diagnosis age at 18 months [20]. Despite its incidence being only 15 per million cases, neuroblastoma accounts for 15% of all childhood cancer deaths [21]. The Children's Oncology Group (COG) conducts risk assessment on neuroblastoma patients based on potential prognostic factors such as INSS stage, age at diagnosis, differentiation level, MYCN amplification, ALK mutation and chromosome 11 loss [22]. With advancements in research regarding molecular biological mechanisms underlying neuroblastoma occurrence, the five-year OS rate for patients in the low- and medium-risk groups has exceeded 90% after conventional treatment [23]. However, in spite of combined treatments like surgery, maximum tolerated dose chemotherapy, radiation therapy, and combined immunotherapy, the overall survival rate for patients classified as high risk remains below 40%. Furthermore, patients belonging to this high-risk category are still prone to relapse and metastasis even after receiving comprehensive treatment. They also exhibit resistance to chemotherapy, resulting in gradual deterioration during chemotherapy progression. This limits follow-up treatment options considerably. Therefore, it is crucial to identify drug targets that can enhance patient sensitivity.

Cellular metabolism reprogramming is a characteristic feature of cancer. Cancer cells exhibit dysregulated glucose, lipids, and glutamine metabolism to support cell growth, maintain redox balance, and adapt to nutrient and oxygen deprivation [24–28]. Lipid metabolism plays a crucial role in providing energy for cancer cells as they are essential components of cellular membranes and store precursors for biologically active lipid mediators [29–31]. Previous studies have demonstrated that metabolic reprogramming contributes to the development and aggressiveness of neuroblastoma [23]. The analysis revealed two distinct fatty acid metabolism-related subtypes with divergent expression patterns and prognostic outcomes in neuroblastoma. These findings highlight the significance of fatty acid metabolism in assessing the prognosis of neuroblastoma. In order to accurately predict individual prognosis based on differences in fatty acid metabolism, we used random survival forest analysis to search for independent prognostic genes among the differentially expressed key enzyme genes. Subsequently, we developed a four-gene risk model. The performance of this risk model was evaluated using GSE49710, TARGET, and E-MTAB-8248 datasets which demonstrated excellent discrimination ability and accuracy. Furthermore, our risk model exhibited strong associations with previously established prognostic markers while also serving as an independent prognostic factor for neuroblastoma. Additionally, we constructed a nomogram that included INSS stage, risk grade, MYCN amplification, and other clinical features widely recognized as prognostic indicators of neuroblastoma in order to further validate the current findings. In conclusion, the current study presents a novel risk model based on fatty acid metabolism in neuroblastoma which enables reliable and personalized prognosis prediction.

Neuroblastoma patients eventually develop resistance to drugs, despite 60% of them being responsive to chemotherapy [32]. As a result, the survival rate within five years is less than 10%. Chemotherapy resistance is the difficulty of NB treatment in high-risk groups, and also the main reason for the final clinical treatment failure. Therefore, it is particularly important to find new therapeutic drug targets and improve the chemotherapy sensitivity of NB patients. Previous research has demonstrated that a few cancers rely on β-oxidation to maintain their energy supply [33] [34–36]. Researches of breast cancer have indicated that some signal pathways play a role in supporting cancer stem cell characteristics and chemotherapy resistance by promoting β-oxidation [37]. Animal experiments have validated this finding, showing that blocking signal pathway with medication can significantly decrease the number of cancer stem cells and enhance the chemotherapy efficacy [38]. Current investigation focuses on examining the relationship between a risk score related to fatty acid metabolism and chemotherapy resistance in neuroblastoma cells. Significant variations in drug sensitivity were observed between high and low-risk groups. Specifically, pevonedistat and AZD4547 exhibited superior therapeutic effects among individuals with low scores. Consequently, targeting fatty acid metabolism may serve as a novel approach for reversing drug resistance in neuroblastoma cells.

There is a risk signature related to fatty acid metabolism that consists of four genes. CHD5, a central gene, displays strong correlations with tumor characteristics such as size, histological grade, TNM stage, and overall survival. This suggests its potential as a biomarker for predicting tumor prognosis. Positioned at the 1p36 locus, CHD5 acts as a tumor suppressor gene frequently absent or silenced in adult cancers with poor outcomes, though its role in neuroblastoma remains unclear [39]. Analysis of the GSE49710 dataset uncovered a significant association between decreased CHD5 expression in neuroblastoma patients and poor prognosis. Reduced CHD5 levels were detected in neuroblastoma patients with MYCN amplification, high-risk/high-stage disease, and advanced tumors. Immunohistochemical analyses on patient tumor samples confirmed a meaningful link between CHD5 expression and patient outcomes, consistent with dataset findings. CHD5 protein analysis in neuroblastoma cell lines revealed higher levels in IMR-32 and SK-N-BE (2) cell lines. Using siRNA techniques to specifically reduce CHD5 expression in SK-N-BE (2) cells demonstrated that CHD5 inhibition increased neuroblastoma cell proliferation, invasion, and migration abilities.

The present findings demonstrate CHD5's significant role in neuroblastoma progression, but the mechanistic relationship between CHD5 and fatty acid metabolism warrants deeper examination. Recent evidence suggests that CHD5, as a chromatin remodeler, primarily influences cellular processes through epigenetic regulation. CHD5 is involved in neural differentiation and tumor suppression by altering the epigenome to regulate gene expression [40]. This is particularly relevant in neuroblastoma, which is characterized by significant epigenetic changes including DNA methylation and histone modifications that are crucial for transcriptional regulation [41]. The relationship between CHD5 and fatty acid metabolism likely involves complex interactions with other key regulators, particularly MYCN. MYCN amplification, a key driver in neuroblastoma, leads to altered fatty acid metabolism through epigenetic mechanisms, including increased fatty acid uptake and biosynthesis critical for tumor survival [7]. This is evidenced by the upregulation of genes involved in unsaturated fatty acid synthesis, such as FASN and ELOVL2, in high-risk neuroblastoma [42, 43]. The interaction between MYCN and PRC1 in repressing docosahexaenoic acid synthesis further highlights the complex interplay between oncogenic signals and metabolic pathways [43]. While direct regulation of fatty acid metabolism by CHD5 remains to be established, its role as a chromatin remodeler suggests it may influence metabolic pathways indirectly by modulating chromatin accessibility for transcription factors and other regulatory proteins. The epigenetic landscape in neuroblastoma, particularly shaped by factors like MYCN, provides a framework for understanding how CHD5 might affect metabolism-related gene expression through its chromatin remodeling activity [40]. Future studies employing ChIP-seq analysis and metabolomic profiling would help elucidate the specific mechanisms by which CHD5 influences fatty acid metabolism in neuroblastoma cells.

In our correlation analysis of signature genes with cell death pathways, TP63 exhibited the most significant and extensive association pattern. In-depth molecular mechanism research revealed that TP63 plays a multi-faceted regulatory role in neuroblastoma cell death. First, TP63 can transcriptionally regulate apoptosis-related genes such as BAX, PUMA, and NOXA, directly promoting the activation of the mitochondrial peripheral pathway. Second, TP63 can induce the expression of death receptor-related proteins like TRAIL and Fas, activating the extrinsic apoptotic pathway. In necroptosis, TP63 promotes changes in cell membrane permeability and inflammatory factor release by regulating the RIPK1/RIPK3/MLKL axis. Notably, TP63 can also indirectly influence cell death progression by modulating cell cycle checkpoint proteins like p21. The mechanism of XKR4 is equally challenging. Existing research indicates that XKR4 participates in phosphatidylserine flip and cell membrane remodeling, which are critical steps in cell death signal transduction. During autophagy, XKR4 may affect autophagosome formation and fusion by regulating key proteins such as Beclin-1, ATG5, and LC3. In necroptosis, XKR4 may be involved in calcium ion channel regulation, promoting intracellular calcium overload and mitochondrial dysfunction. Research has also found that XKR4 is closely associated with the expression of inflammatory factors like TNF-α and IL-1β, suggesting its potential role in regulating inflammatory cell death. As an important epigenetic regulatory factor, CHD5's cell death regulation mechanism is more complex. Existing research confirms that CHD5 can influence cell death through multiple mechanisms. First, CHD5 acts as a tumor suppressor by regulating apoptosis in neuroblastoma cells, with its expression associated with the upregulation of pro-apoptotic genes and downregulation of anti-apoptotic pathways [39, 44]. Specifically, CHD5 influences apoptosis by repressing genes involved in cell cycle progression, such as WEE1, a G2/M checkpoint gene, which is crucial for preventing uncontrolled cell division [44]. The loss of CHD5 expression, often due to epigenetic silencing or deletion, leads to reduced apoptosis and dysregulation of apoptosis-related pathways, contributing to tumor survival and progression [45, 46]. CHD5 can directly interact with the p53 pathway, promoting p53's transcriptional activity and thereby regulating apoptosis-related gene expression. Second, CHD5 participates in DNA damage repair processes, affecting cellular responses to oxidative stress and genomic instability by regulating key repair proteins like BRCA1 and RAD51. In autophagy regulation, CHD5 may influence autophagy initiation and progression by modulating the mTOR signaling pathway and AMPK activation. Notably, CHD5 may also indirectly regulate the expression of cell death-related genes through epigenetic modifications, such as histone deacetylation and chromatin remodeling. Although CTAG1A did not show significant correlation in cell death pathways, it may play an important role in tumor immunity and antigen presentation. Research indicates that CTAG1A can activate specific T cell subsets, promoting the recognition and elimination of tumor-associated antigens. By activating cytotoxic T cells and natural killer cells (NK cells), CTAG1A indirectly promotes tumor cell death.

The current study demonstrated significant associations between fatty acid metabolism-based risk stratification and tumor immune microenvironment in neuroblastoma. The high-risk group exhibited markedly reduced expression of immune checkpoint molecules, with 17 out of 20 checkpoint genes showing significantly lower expression compared to the low-risk group. This altered fatty acid metabolism profoundly impacts the tumor immune microenvironment through multiple mechanisms. During tumor development, fatty acid metabolism in the tumor microenvironment (TME) involves complex processes of synthesis, oxidation, and uptake of fatty acids, which are crucial for tumor cell survival and proliferation under stress conditions such as hypoxia and nutrient deficiency [47]. The metabolic competition between tumor and immune cells for fatty acids leads to immune cell dysfunction, as fatty acids provide essential energy and synthetic materials for immune cell activation and differentiation [48]. This competition creates an immunosuppressive microenvironment by altering cytokine and chemokine secretion, contributing to a “cold tumor” phenotype that limits immune cell infiltration [49]. Furthermore, increased fatty acid oxidation in the TME suppresses the activation of effector T-cells while promoting regulatory T-cells and myeloid-derived suppressor cells, reinforcing the immunosuppressive state [50]. The therapeutic implications of these metabolic alterations are reflected in our drug sensitivity analysis, which revealed distinct response patterns between risk groups. In the low-risk group, the increased sensitivity to several targeted agents was observed: Sepantronium bromide (p < 2.2e-16), which targets survivin in tumors with lower survivin expression[51]; Pevonedistat (p = 7.7e-11), an NEDD8-activating enzyme inhibitor effective in tumors with less active protein degradation pathways [51]; AZD4547 (p < 2.2e-16), a FGFR inhibitor particularly effective in less aggressive cases with FGFR alterations[51] (Wang et al. [53]); Dactinomycin (p = 2.4e-06), which shows better efficacy in tumors with lower proliferative indices [52]; and Bortezomib (p = 4.9e-07), a proteasome inhibitor more effective in tumors with less complex proteasome activity[53]. Conversely, high-risk tumors showed increased sensitivity to agents targeting rapid cell division and active signaling pathways, including Paclitaxel (p < 2.2e-16) and Vinblastine (p = 1.9e-14), which target microtubule dynamics in rapidly dividing cells [54]; PD0325901 (p < 2.2e-16), targeting the frequently upregulated MAPK pathway [51]; Gemcitabine (p = 4e-14), effective against high DNA replication stress[52]; and Eg5_9814 (p = 1.4e-14), targeting high mitotic activity[51]. Given these complex interactions between fatty acid metabolism, immune response, and drug sensitivity, combination approaches targeting multiple pathways might be necessary for high-risk tumors. Recent evidence suggests that combining immunotherapy with interventions targeting lipid metabolism could potentially increase the efficacy of treatments like adoptive cell therapy and cancer vaccines [55]. These insights highlight the potential value of fatty acid metabolism-based risk stratification in guiding personalized treatment strategies for neuroblastoma patients, particularly in designing combination approaches that address metabolic reprogramming, immune evasion mechanisms, and specific drug sensitivities.

The current discoveries have the potential to enhance our understanding of the mechanisms involved in neuroblastoma's metabolism of fatty acids, leading to improved methods for diagnosing and treating this disease. However, several limitations of our study should be noted. First, the heterogeneity of tumor samples and varying clinical characteristics may affect the generalizability of the present findings. Second, while our predictive model showed promising results, external validation in larger, more diverse patient cohorts is needed to fully establish its clinical utility. Third, the complex interaction between fatty acid metabolism and cell death pathways may be influenced by other molecular mechanisms not covered in our current study. While our study has shown significant prognostic value for CHD5, further validation through large-scale clinical research is necessary to determine its role in enhancing the existing risk stratification system. Additionally, targeted therapy against CHD5 poses a challenge due to its essential role in maintaining normal biological functions in healthy cells. Therefore, it is crucial to identify an appropriate therapeutic window for intervention. Future research should validate our findings in larger, multicenter cohorts to enhance the robustness of our conclusions. The potential clinical management implications of CHD5 in neuroblastoma need to be thoroughly investigated to translate our findings into therapeutic benefits. Further exploration of the detailed molecular mechanisms underlying the interaction between fatty acid metabolism and cell death pathways will deepen our understanding of neuroblastoma progression. Moreover, developing more selective therapeutic strategies that can effectively target CHD5 in tumor cells while sparing normal tissues remains an important goal for improving treatment outcomes.

Conclusions

This comprehensive study systematically explores the intricate relationships between fatty acid metabolism, cell death mechanisms, and neuroblastoma progression. The novel prognostic model reveals the critical role of signature genes in regulating multiple cell death pathways and provides insights into the complex interplay between metabolic reprogramming and cellular fate in neuroblastoma. By identifying CHD5 as a key hub gene with significant implications for cell death regulation and tumor progression, the research uncovers potential molecular targets that could revolutionize diagnostic and therapeutic strategies for this challenging childhood cancer. The study presents a promising framework for personalized treatment approaches that integrate metabolic profiling, cell death mechanisms, and immune landscape characterization

Author contributions

YC and QZ designed the study. JL, YJ (Yubin, Jia), YJ (Yan, Jin), JY, YL, and BZ analyzed and interpreted data. YC drafted the manuscript, and major revised by BZ and QZ. All authors contributed to the article and approved the submitted version.

Funding

This work was supported by grants from the National Key Research and Development Program of China (2018YFC1313000, 2018YFC1313001) and Tianjin Key Medical Discipline (Specialty) Construction Project (TJYXZDXK-009A).

Data availability

Data is provided within the manuscript or supplementary information files.

Declarations

Consent for publication

Informed Consent was obtained from all the participants involved in the study.

Competing interests

The authors declare no competing interests.