Neuroblastoma (NB) is an embryonal tumor with extensive molecular heterogeneity in its biological and clinical presentation (Matthay et al., 2016). Although some NBs resolve spontaneously without treatment, approximately 50% of patients are classified as high risk on diagnosis, and only 50% survive for 5 years (Shohet et al., 2021). Accounting for 7%–10% of all childhood cancers, NB contributes 15% of cancer-related deaths (Irwin and Park, 2015). The pathogenesis of NB is reportedly associated with a number of genetic aberrations in genes regulating the cell cycle, cell proliferation, and programmed cell death (Salemi et al., 2022). Despite the emergence of many new treatment modalities for NB, such as radioimmunotherapy and molecular targeting therapies (Feng et al., 2023; Zafar et al., 2021), the estimated 5-year survival rates for patients with high-risk NB are approximately 50% (Shohet et al., 2021). According to the Children’s Oncology Group guidelines, NB can be divided into four International Neuroblastoma Staging System (INSS) stages (Brodeur et al., 1993). INSS is a postsurgical staging system that uses tumor location in relation to the midline structures, lymph node status, and the extent of previous surgical resection to define the tumor status. Bone marrow assessment and radiographic studies are used to detect metastases. Stage 4 or 4S determination depends on other clinical details (Naranjo et al., 2018).

Owing to the key role of INSS staging in assessing disease severity and prognosis, researchers continue to explore prognostic biomarkers associated with the INSS stage. MYCN gene amplification is more observed in neuroblastomas with higher INSS stages (Stages 3 and 4) (Hansen et al., 2017). Significantly higher TERT expression has been detected in NB samples with INSS Stage 4 than in samples with earlier stages (Akter and Kamijo, 2021). Differences in the 1p/11q (Raitio et al., 2021) and serum inflammatory factors (Zheng et al., 2020) were also identified by the stratification based on INSS staging. However, these studies either did not use transcriptomics and bioinformatics to screen genes at a hierarchical level, as in Secomandi et al., who compared only one gene, the lysosomal protease cathepsin D (Secomandi et al., 2022), or did not use clinically derived specimens for expression validation (Fan et al., 2020). Therefore, identifying specific markers for diagnosis and treatment is of vital importance for improving the prognosis of children with advanced INSS stage of NB.

Over the last few years, high-throughput sequencing and gene microarray technologies have been applied in many medical fields to facilitate the classification of tumors based on histological and clinical data, cancer-related genes, and biological pathways (Satam et al., 2023). Moreover, emerging bioinformatics based on gene expression has recently become an effective method for systematically screening tumor-related genes and exploring the related molecular mechanisms (Hernandez-Lemus et al., 2019). The early detection and prognosis of many diseases depend on biomarkers that can mark changes in the structure or function of systems, organs, tissues, cells, and subcells (Shawraba et al., 2021). Biomarkers can be used to classify NB into different groups, facilitating NB treatment and prognosis (Trigg et al., 2019). Several studies have discussed the prognostic role of biomarkers (Galardi et al., 2018; Glembocki and Somers, 2024; Liu et al., 2022); however, the clinical use of these biomarkers requires clinical evaluation. In this study, we screened patients with NB using the Therapeutically Applicable Research to Generate Effective Treatments (TARGET) and Gene Expression Omnibus (GEO) databases for potential biomarkers of survival prognosis. A novel approach for the prognostic diagnosis of NB was proposed. We identify markers of advanced-stage NB, explore their role in prognosis, and highlight the value of their implementation in a clinical setting. To effectively manage advanced-stage NB, the detection and analysis of prognostic biomarkers can play a vital role in refined risk assessment and the development of targeted therapeutic regimens.

NB has an insidious onset without specific symptoms (Kembhavi et al., 2015), making early-stage diagnosis challenging. Approximately 50% of children with NB show metastases at initial diagnosis (Tolbert and Matthay, 2018), which contributes to a poor prognosis, especially in patients at an advanced INSS stage (Pizzo et al., 2015). The INSS has been an effective predictive tool for the prognosis of NB over the past decade in China (Brodeur et al., 1993). Patients with advanced INSS stages are resistant to current high-intensity therapy and typically have a poor prognosis compared to children in the early stages (Sokol and Desai, 2019; Maris et al., 2007). Researchers have speculated that this may be due to tumors that are more difficult to treat or that have spread to areas that are inaccessible for surgery; however, clinical evidence is lacking. By reviewing the clinical data of 72 patients with NB, Qi et al. found that advanced lung cancer inflammation index and INSS stage were independent prognostic factors for patients with NB (Qi et al., 2022). Keane et al. evaluated the expression of DLG and LIN7 gene families in different INSS stages and found that the disrupted DLG isoform was abundant in the advanced stage of NB. This could control cell polarity and signaling, thereby affecting cancer cell viability (Keane et al., 2021). Further mechanistic studies investigating poor prognosis in children with advanced-stage NB are needed in the future.

In this study, we analyzed NB RNA-seq data from the TARGET and GEO databases to investigate the relationship between DEGs and NB prognosis. During screening, four genes (CDKN3, PCLAF, PRKACB, and CNR1) were identified as being significantly associated with advanced NB, and a risk model based on these four genes showed prognostic value for patient survival. These genes exert biological functions in the cell cycle, immunoinflammatory processes, and the p53 signaling pathway, and expression assays validated that the mRNA and protein levels of CDKN3, PCLAF, PRKACB, and CNR1 in NB tissues correlate with NB staging.

As a member of the dual-specific protein phosphatase family and kinase-associated phosphatases, the cyclin-dependent kinase inhibitor 3 (CDKN3) gene product inhibits the G (1)/S transition by dephosphorylating cyclin-dependent kinases (Ding et al., 2020). In our report, we observed that the level of CDKN3 was elevated in the advanced-stage NB. Several recent studies have reported that CDKN3 plays a regulatory role in the survival and proliferation of cancer cells. Yang et al. found that CDKN3 is a marker of poor prognosis in NB (Yang et al., 2017). Vernaza et al. found that knocking down CDKN3 reduced the expression of the cell proliferation markers Ki67 and proliferating cell nuclear antigen (PCNA), a binding partner of PCLAF, and decreased the colony formation of NB cells (Vernaza et al., 2024). The in vitro functional assays have been conducted in liver hepatocellular carcinoma (LIHC). Q-PCR and Western blotting were used to detect the expression of CDKN3 in LIHC and its adjacent tissues as well as human liver cancer cell lines. It was found that, compared with the adjacent samples, the mRNA expression of the CDKN3 gene was significantly upregulated in 63.9% (23/36) of liver cancer tissues. After silencing the expression of CDKN3, the growth and colony formation ability of cancer cells were significantly inhibited (Li et al., 2023). Moreover, silencing the expression of CDKN3 could induce G0/G1 phase arrest in liver cancer cells, indicating that CDKN3 promotes the proliferation of liver cancer cells by affecting the cell cycle distribution (Lin et al., 2013). Knockdown of CDKN3 led to the downregulation of p53 and p21 protein levels, while the AKT serine/threonine kinase 1 was upregulated (Dai et al., 2016). Therefore, the elevated expression of CDKN3 may reduce the survival of tumor cells and change the sensitivity to therapeutic agents through the AKT/P53/P21 signaling pathway (Dai et al., 2016). These observations facilitate translating genomic and transcriptomic features into molecular mechanisms of high-risk NB.

Proliferating cell nuclear antigen-clamp associated factor (PCLAF) is involved in DNA repair through its binding partner, proliferating cell nuclear antigen (PCNA) (Liu et al., 2021). PCLAF has been reported to play various essential roles in cell proliferation, apoptosis, and cell cycle regulation (Liu et al., 2021). Among the differentially expressed downregulated genes in NB, Liu et al. found that the cell cycle-related pathway genes were most significantly altered in association with the decreased level of PCLAF (Liu et al., 2022a). The patients with elevated levels of PCLAF mRNA had the worst OS and EFS, which was consistent with our results. Importantly, PCLAF knockdown could limit the proliferation of NB cells in vitro and in vivo (Liu et al., 2022a). In eight types of tumors, including cholangiocarcinoma, cervical cancer, and glioblastoma multiforme, the expression levels of PCLAF are significantly higher than those in normal tissues (Liu et al., 2022b). Meanwhile, the DNA methylation level of PCLAF is decreased, indicating an association between hypomethylation and high expression (Liu et al., 2022b). The survival analysis conducted using the GEPIA2 tool showed that the high expression of the PCLAF gene is correlated with the overall survival and disease-free survival of patients, suggesting that it can serve as a potential prognostic biomarker. At the same time, this pan-cancer study also found that the expression of PCLAF is positively correlated with activated CD4⁺ T cells and T helper 2 (Th2) cells. This implies that PCLAF may play a specific role in the process of tumor immune cell infiltration and affect the functions and distribution of immune cells in the microenvironment (Liu et al., 2022b). These results, together with ours, reveal the mechanism through which PCLAF facilitates cell cycle progression, suggesting this gene may be a therapeutic target in NB.

Protein kinase cAMP-dependent catalytic subunit beta (PRKACB) is a member of the serine/threonine protein kinase family and a key effector of cAMP/PKA-induced signal transduction (Soberg et al., 2017). It is involved in numerous cellular processes, including cell proliferation, apoptosis, gene transcription, metabolism, and differentiation (Ahmed et al., 2022; Huang et al., 2024). Shao et al. found significantly lower PRKACB expression in an N-myc proto-oncogene (MYCN)-amplified group than in an MYCN-non-amplified group (Shao et al., 2021). In the present study, PRKACB downregulation was observed in children with poor OS rates. Whole genome-scale integrated analyses of exon arrays demonstrated that PRKACB is a novel cancer-related variant transcript in gastric cancers (Furuta et al., 2012). The expression of PRKACB was downregulated in both colorectal cancer and non-small cell lung cancer (NSCLC) (Yao et al., 2020; Chen et al., 2013). In a study aiming to evaluate the expression pattern of PRKACB in NSCLC, Chen et al. found that PRKACB is downregulated in NSCLC tissues, and the upregulation of this gene decreased the number of proliferative, colony-forming cells and invasive cells, indicating that an upregulated PRKACB level may be an effective way to prevent the progression of NSCLC (Chen et al., 2013). In a large-scale study in 2023 involving a nutrient database and cellular and animal assays, trans-vaccenic acid (TVA) was found to lead to an increased PRKACB expression in immunodeficient nude mice, T cell receptor-α knockout mice, and mice with a CD8⁺ T cell depletion. When the gene was upregulated, the cAMP-PKA-CREB axis was activated, which not only enhanced the function of CD8⁺ T cells but also reversed T cell exhaustion (Fan et al., 2023). This multi-faceted approach provides a new potential direction; that is, the anti-tumor effect could be enhanced by regulating the intake of related nutrients based on their gene regulation mechanisms.

Cannabinoid receptor-1 (CNR1) is a classical receptor of the endocannabinoid system that primarily exists in the central and peripheral nervous systems (Tao et al., 2020). Depending on its binding to its corresponding ligands, CNR1 activates intracellular signaling pathways related to survival and is a potential biomarker of poor prognosis in cancer (Zou and Kumar 2018). In recent years, cannabinoids (CB) have received increasing attention due to their effects on cancer growth. CB2 agonists (JWH-015; JWH-133) have been proved to reduce the size of breast cancer tumors (Blanton et al., 2022). In triple-negative breast cancer (TNBC), CNR1 regulates fatty acid metabolism by modulating the PI3K-AKT/MAPK signaling pathways, thereby affecting the sensitivity of cells toward ferroptosis (Li et al., 2022). Moreover, the combined use of CNR1 antagonists and ferroptosis inducers has demonstrated good anti-tumor effects in TNBC (Li et al., 2022). In studies on the association between liver metabolism and tumors, the activation of CNR1 can specifically induce the expression of the CREBH gene through the phosphorylation of the JNK signaling pathway and the binding of c-Jun to the AP-1 in the promoter of the CREBH and then regulating hepatic gluconeogenesis (Chanda et al., 2011). In neuroblastoma, this signaling pathway may also be involved in regulating the proliferation, metabolism, and other processes of tumor cells. Mu et al. compared 89 DEGs between MYCN amplification and non-amplification groups and suggested that CNR1 plays a prognostic role in NB, although they did not perform any verification experiments (Schmitt et al., 1986). Decock et al. applied two independent genome-wide methylation screening methods to eight NB cell lines and found that CNR1 methylation was associated with classical risk factors, including stage and MYCN status, but not with OS, EFS, or both (Decock et al., 2012). Although the bioinformatic analysis revealed an association between the NB stage and CNR1, remarkably different CNR1 expression levels were not observed in the clinical samples in our study. This may have been owing to tumor heterogeneity or the small sample size. Thus, further analyses are required to clarify the role of CNR1 as an independent prognostic biomarker.

MYCN gene amplification (defined as a copy number greater than 10-fold) is an internationally recognized molecular event in high-risk NB, and this alteration is present in approximately 20%–30% of cases (Qiu and Matthay, 2022). Clinical data indicate that the 5-year survival rate of patients with MYCN gene amplification is significantly lower than that of patients without it (Schwalbe et al., 2017). We used the predictive model constructed from the four DEGs to divide patients into high-risk and low-risk groups. We found that the positive amplification rate of the MYCN gene in the high-risk group was significantly higher than that in the low-risk one. This confirmed that this risk model was significantly correlated with the MYCN molecular type.

The chromosomal ploidy of NB is classified into near-diploid (Diploid) and near-triploid (Triploid). Near-diploid tumors are often accompanied by deletions or gains of chromosomal segments and are associated with MYCN amplification and a high risk of recurrence (Irwin et al., 2021). Near-triploid tumors are more common in infants, mostly low-risk babies, and may resolve spontaneously. Studies have shown that near-diploid tumors have higher genomic instability and are prone to accumulating MYCN amplification, leading to treatment resistance and recurrence (Benard et al., 2008). In this study, through performing Cox, ROC, and DCA analysis, we found that ploidy has independent prognostic value and can provide a more reliable basis for clinical decision-making in NB. Currently, the International Neuroblastoma Risk Group (INRG) system has integrated factors such as MYCN amplification, patient age, tumor stage, and ploidy to guide NB treatment. Intervention strategies targeting the MYCN gene include the following three aspects: (i) directly interfering with MYCN transcription, for instance, using peroxisome proliferator-activated receptor γ (PPAR-γ) antagonists to induce cell differentiation by inhibiting MYCN expression (Nakao-Ise et al., 2023); (ii) metabolic intervention: aldehyde dehydrogenase 18 family member A1 (ALDH18A1) inhibitors can block the MYCN-metabolic feedback loop and have shown significant efficacy in animal models (Guo et al., 2020); (iii) epigenetic regulation: histone deacetylase inhibitors or DNA methyltransferase inhibitors can reverse the epigenetic silencing driven by MYCN (Eppet et al., 2023). Due to the synergistic effect between ploidy and MYCN amplification (Park et al., 2012), multi-omics integration analysis, as well as studies combining MYCN inhibitors with immune checkpoint inhibitors (ICIs), have shown therapeutic potential in long-term follow-up (van Tilburg et al., 2020). In the future, multi-dimensional molecular stratification and targeted intervention will be applied to improve the survival outcomes of high-risk patients.

Immunotherapy has been proved to be an effective treatment for solid tumors in adults (Anderson et al., 2022). However, evaluations of tumor mutational burden have shown that neuroblastoma ranks first among human cancers with the fewest mutations, and it also has the lowest infiltration of immune cells (Anderson et al., 2022). Therefore, due to the “cold” state of NB in terms of immunity and mutation, its immunotherapy seems to have reached a bottleneck (Grobner et al., 2018). Fortunately, the inherent immune evasion mechanism of neuroblastoma is similar to that of adult tumors (Wienke et al., 2021). Anderson et al. mentioned that the immunotherapy of NB should not only focus on increasing tumor antigenicity; addressing the problem of the tumor immunosuppressive microenvironment is also crucial (Anderson et al., 2022).

In the present work, after evaluating the infiltration of immune cells in high- and low-risk patients using the established model, we observed that the infiltration of immune cells in these two groups was different: the numbers of macrophage M0, plasma cells, follicular helper T cells, and Tregs in the high-risk cohort were significantly higher than in the low-risk group. Research by Zhang et al. showed that the infiltration degree of M0 macrophages in hepatocellular carcinoma tissues was remarkably higher than that in normal liver tissues (Zhang et al., 2022), which was consistent with our results. Together with our findings, these results indicate that macrophage infiltration is related to poor prognosis, and immunotherapy may be effective because macrophages can enhance phagocytosis of tumors in high-risk NB. M2 macrophages, which are polarized from M0 macrophages, can release cytokines and chemokines such as transforming growth factor-β (TGF-β), vascular endothelial growth factor (VEGF), and indoleamine-2,3-dioxygenase (IDO). These inhibitory factors substances could suppress the anti-tumor immune response (Iolascon et al., 2000). Meanwhile, these substances will recruit plasma cells and Tregs into the tumor microenvironment, where they directly kill effector T cells or inhibit their functions and facilitate tumor cells escaping immune surveillance (Anderson et al., 2022). Kang et al. reported that the proportion of follicular helper T cells was higher in the high-risk cohorts than in the low-risk NB group (Kang et al., 2021). Mei et al. used single-cell RNA-seq to conduct a comparative analysis of patients with and without metastasis and observed macrophage enrichment, T-cell exhaustion, and increased numbers of Tregs (Mei et al., 2024). Methods need to be found to inhibit the functions or reduce the number of regulatory T cells to break the immunosuppressive state. However, relevant research is still under exploration (Anderson et al., 2022; Zhu et al., 2023).

Cancer immunotherapy has achieved remarkable success due to the application of immune checkpoint inhibitors (ICI) in the treatment of adult tumors and pediatric hematological malignancies (Davis et al., 2017). However, in contrast, most children with solid tumors, including high-risk neuroblastoma, have hardly benefited from ICI (Anderson et al., 2022). Using the model we established, the patients were divided into high- and low-risk groups. We found that the upregulated lymphocyte activation gene 3 (LAG3) was significantly upregulated in the high-risk cohort. It has been observed that the LAG3 gene exerts an inhibitory function on antigen-activated T cells (Camisaschi et al., 2010). In addition to LAG3, two other important checkpoint molecules expressed by tumor cells, namely, the PD-L1 ligand (CD274) and its paralog PDCD1LG2, were significantly decreased in the high-risk groups. By suppressing the expression of PD-L1, tumor cells weaken the signals transmitted to immune cells, thus facilitating tumor escape (Mezzadra et al., 2017). Tang et al. studied the mechanism of PD-1 blockade therapy and found that macrophages are important effector cells in anti-high-risk NB immunity (Tang et al., 2022). PD-1 blockade can lead to the exhaustion of macrophages, a decrease in their phagocytic potency, a reduction in tumor growth, and an increase in survival in mouse models of NB in a macrophage-dependent manner (Gordon et al., 2017). Recent studies have focused on optimizing existing GD2 antibodies, developing novel bispecific antibodies or chimeric antigen receptor T cell (CAR-T) therapies, and exploring the combined application of ICI (NCT05437315, NCT02914405, etc.). Although the efficacy of a single ICI is limited, combined regimens (such as anti-GD2 + anti-PD-1) and novel immunomodulatory strategies are showing potential in clinical application (Kennedy et al., 2023). Immunotherapy combined with chemotherapy has been added to the standard treatment regimen recommended for relapsed or refractory neuroblastoma (Anderson et al., 2022). It will be necessary to verify the safety and long-term efficacy of these therapies through clinical trials. The findings related to immune cell infiltration and checkpoints in this study are expected to serve as targets for drug development, aiming to restore the functions of immune cells that have been damaged during chemotherapy.

This study had several limitations. First, the main sources of clinical information, the online databases, are limited and incomplete; some biological factors, such as individual therapeutic regimens, cannot be obtained. Second, limited by the current research conditions, the verification of the biological functions was only performed in vivo. It is recommended that in further studies, the cancer-promoting mechanisms of these genes should be preferentially verified in vitro.

Our study proposes an INSS stage-related risk model that combines genetic and clinical characteristics to predict the prognosis of patients with NB. Four prognostic genes (CNR1, PRKACB, CDKN3, and PCLAF) were identified, and a prognostic model based on these genes was developed that provides novel insights into the prognostic evaluation of neuroblastoma. In clinical applications, the diversity of biomarkers and their implication on clinical features can facilitate the formation of targeted therapeutic strategies and the development of individualized immunotherapy regimens.