Introduction

Colorectal cancer (CRC) is the most common gastrointestinal tumor, the third most common malignant tumor and the third leading cause of cancer-related death worldwide, which has attracted extensive public health attention1,2. In 2020, there will be more than 19 million new cases and 900 000 deaths globally, accounting for 10% of all new cancers and 9.4% of cancer deaths, respectively1,2. Numerous studies have shown that the 5-year overall survival (OS) rate of CRC is about 60%, and the prognosis of metastatic CRC is even worse, with an overall survival rate of only 14%3. Over the past few decades, the incidence and mortality of many cancers have decreased due to the implementation of screening programs and the development of treatment technologies4. However, the opposite status was observed for early-onset CRC (EOCRC). EOCRC is typically defined as CRC diagnosed in patients younger than 50 years of age. In 2010, EOCRC accounted for 4.8% of colon cancers and 9.5% of rectal cancers5. Recently evidences show that the incidence of CRC in people over 50 years old is decreasing by 3.1% per year, but the incidence of EOCRC is steadily increasing by 2% per year5,6,7,8. By 2030, they are projected to account for 10.9% of all colon cancers and 22.9% of all rectal cancers9. The incidence of EOCRC is on the rise, and its malignant degree is also more serious. EOCRC is more aggressive, more malignant and has a later clinical stage than late-onse CRC (LOCRC)5,6,7,8,9,10,11. Therefore, further research on prognostic biomarkers of EOCRC is of great significance for prognosis determination and individualized treatment.

In the field of cancer biomarker screening, many researchers have conducted many different studies from different perspectives, and many important results have also been achieved12,13,14,15,16,17,18. The immune microenvironment of cancer is still multilayered, complex, and not fully understood19. But anyway, increasing evidences suggest that the immune system is not only involved in the occurrence and development of several cancers, but also closely related to the prognosis and treatment of these cancers20. Immunotherapy targeting the immune system has significantly improved the effectiveness of cancer treatment over the past few decades21. For example, investigators can use immune checkpoint inhibitors, such as anti-CTLA4, anti-PD1, and anti-PDL1 to modulate the immune system, thereby achieving the suppression effects21, including CRC22,23,24.

In the present study, we used comprehensive analysis approaches to perform biomarkers screening studies on the data from The Cancer Genome Atlas Program (TCGA) and Gene Expression Mapping (GEO) database, used other independent samples to perform validation study on the feasibility of candidate biomarkers, and also used the clinical EOCRC samples to carry out validation study on the expression of candidate biomarkers and their related genes. In addition, we also explore the relationship between risk model and the treatment sensitivity of immunotherapy and chemotherapy drugs to provide evidence for the treatment and prognosis of EOCRC.

Material and method

Flow of analysis

The flow chart in Supplementary Figure S1 depicted the data analysis process.

Public data processing

After excluding the samples over 50 years old, 90 samples (8 control samples and 82 EOCRC samples) were obtained from the TGCA data, 79 EOCRC samples were obtained from the GEO39582 data, 21 EOCRC samples were obtained from the GEO17536 data, and 9 EOCRC samples were obtained from the GEO17537 data. Their clinical information was also obtained from the corresponding databases. The Limma algorithm in R (3.6.1) was used to screen differentially expressed genes (DEGs) with |Log2FC|≥0.3 and p-value < 0.05 as the criteria25,26.

Colorectal tissue samples were obtained from patients who underwent surgery at the Second People’s Hospital from March 2022 to August 2024. Patients were required to be younger than 50 years of age. The tumor tissue and adjacent tissue were determined by two clinicians according to the results of pathological examination. All patients provided written informed consent. All procedures involving human participants in this study were in accordance with the Declaration of Helsinki. This study was approved by the Ethics Committee of Hunan University of Medicine (2021(H031781)).

Immune status analysis

The Estimate algorithm in R (3.6.1) was used to evaluate the tumor microenvironment (TME) using default parameters, including Estimate Score, Immune Score, Stromal Score and tumor purity.

The GSVA algorithm in R (3.6.1) was used to score the immune cells and factors with default parameters.

ImmuCellAI (https://guolab.wchscu.cn/ImmuCellAI//#! /) could predict the abundance of 24 immune cell types in samples by default parameters.

Enrichment analysis

DAVID 6.8 (https://davidbioinformatics.nih.gov/tools.jsp) was used to carry out Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment analysis using the default parameters27,28,29.

GSEA were used to perform GO and KEGG enrichment analysis on the expression values of R (3.6.1).

Univariate and multivariate cox regression analysis

According to the median value of gene expression, EOCRC patients were divided into high expression group and low expression group. Under default parameters, univariate and multivariate Cox regression analysis were used to screen candidate biomarkers associated with OS in R (3.6.1). Kaplan-Meier method was used to describe survival status. A nomogram was used to visualize the correlation between different signatures and OS.

Construction of risk model

The selected candidate markers were used to construct a risk model using the following methods17,18. Risk score=\(\:{\sum\:}_{i=1}^{n}\left(Coei*Exp\right)\), where n represents the gene number, Coei represents the coefficient value, and Exp represents the gene expression level.

Genomic heterogeneity analysis

Mutant-allele tumor heterogeneity (MATH), Microsatellite Instability (MSI), and Tumor mutational burden (TMB) scores for individual EOCRC patients were obtained from the Sangerbox database (http://vip.sangerbox.com/login.html).

Immunotherapy and chemotherapy analysis

The Immune Dysfunction and Exclusion (TIDE) (http://tide.dfci.harvard.edu/login/) database was used to evaluate the immunotherapy effect using the candidate DEGs with the default parameters. Oncoprediction package was used to evaluate the chemotherapy effect using the candidate DEGs with the default parameters.

Quantitative real-time PCR

Tissue mRNA was extracted using the RNA Easy Fast Kit (TIANGEN, China) according to the manufacturer’s instructions. Two µg of total RNA was reverse transcribed using Thermo Scientific RevertAid™Kit (Thermo Scientific, USA). Quantitative real-time PCR (qRT-PCR) was performed using the Light Cycler®480 Instrument II (Roche). GAPDH-F: 5-GGAGCGAGATCCCTCCAAAAT-3. GAPDH-R: 5-GGCTGTTGTCATACTTCTCATGG-3. NOS2-F: 5-AGGGACAAGCCTACCCCTC-3, NOS2-R: 5-CTCATCTCCCGTCAGTTGGT-3. CXCL1-F: TTGCCTCAATCCTGCATCCC. CXCL1-R: GTTGGATTTGTCACTGTTCAGCAT. CXCL2-F: TTGTCTCAACCCCGCATCG. CXCL2-R: CAGTTGGATTTGCCATTTTTCAG.

Correlation analysis

The Pearson method in Sanger box was used for correlation analysis and their criteria were |R|≥0.3 and p-value < 0.05.

Statistical analysis

A repeated measure ANOVA followed by an unpaired two-tailed student’s t test was used as indicated. All results are expressed as the mean ± SEM.

Results

Screening of immune-related candidate biomarkers

The ESTIMATE algorithm was used to evaluate the EOCRC tumor microenvironment from TCGA. The results showed that the ESTIMATE score, immune score, and stromal score of EOCRC were significantly lower than those of normal, but the tumor purity was significantly higher (Supplementary Figure S2a). The GSVA algorithm was used to evaluate the scores of EOCRC immune cells and molecules from TCGA. The results showed that the angiogenic activity and stemness score were significantly increased (Supplementary Figure S2b), while the mesenchymal EMT was significantly decreased in EOCRC. There was no significantly difference for tumorigenic cytokines between normal and EOCRC patients (Supplementary Figure S2c). The score status of immune cells is shown in 2d. Compared with the normal group, 22 out of 28 immune cells in EOCRC patients had significantly higher scores, 1 immune cell (CD56bright natural killer cell) had significantly lower scores, and 5 (Activated CD4 T cell, Activated dendritic cell, CD56dim natural killer cell, Immature dendritic cell, and Type 17 T helper cell) had no significant difference in scores. The infiltration status was evaluated by ImmuCellAI. The infiltration score was significantly decreased in EOCRC patients. The results showed the infiltrations of 15 immune cells were significantly decreased while 3 immune cells (Bcell, Monocyte, and NK) were significantly increased (Supplementary Figure S2d).

Limma algorithm was used to screen the differentially expressed genes (DEGs). The result showed that there were 437 DEGs between EOCRC patients and normal in TGCA database, including 272 up-regulated and 165 down-regulated DEGs (Supplementary Figure S3a). These DEGs were overlapped with 1793 immune-related genes (IRGs)30. Among the 437DEGs, 53 were immune-related genes (Supplementary Figure S3b). Similar analyses were performed on data from GSE39582. The result showed that there were 2508 DEGs between EOCRC patients and normal in GSE39582database, including 1591 up-regulated and 917 down-regulated DEGs (Supplementary Figure S3c). Among the 2508 DEGs, 153 were immune-related genes (Supplementary Figure S3d). Of which, only 3 IRDEGs (NOS2, NF2F1, and VEGFC) were consistently expressed in TCGA and GSE39582 (Supplementary Figure S3e). The expressions of those 3 IRDEGs were displayed in Supplementary Figure S3f and Figure S3g.

Evaluation of immune related prognostic risk model

Univariate and multivariate Cox regression analysis were performed for those three DEGs (NOS2, NF2F1, and VEGFC). In the TCGA group, univariate (Fig. 1a) and multivariate (Fig. 1b) Cox regression analysis showed that NOS2 was significantly associated with the survival of EOCRC patients. EOCRC patients with high NOS2 expression show better survival (Fig. 1c). We observed the same pattern for NOS2 in the GEO39582 group (Fig. 1d and f). In addition, we also analyzed the relationship between NOS2 and different clinical features, and found that NOS2 expression was not significantly different among different clinical features (Fig. 1g and n).

Fig. 1
figure 1

Screening of candidate biomarker for EOCRC. a-b, Forest map of univariate (a) and multivariate (b) Cox regression analysis in TCGA group. c, K-M curve of NOS2 for OS in TCGA group. d-e, Forest map of univariate (d) and multivariate (e) Cox regression analysis in GSE39582 group. f, K-M curve of NOS2 for OS in GSE39582 group. g-j, Expression level of NOS2 among different clinical features in TCGA group. k-n, Expression level of NOS2 among different clinical features in GSE39582 group. * p < 0.05. ** p < 0.01. *** p < 0.001. NS, No Significance.

Full size image

Therefore, NOS2 was used as a candidate marker to construct the risk model. Referring to previous studies17,18, Youden index was used to distinguish high-risk and low-risk EOCRC patients. We first conducted a model construction and evaluation study in TCGA group. The risk scores and survival times of EOCRC patients with different risk status are shown in Fig. 2a and b. EOCRC patients with high-risk score had significantly lower NOS2 expression (Fig. 2c) and significantly shorter survival time (Fig. 2d). To understand the prognostic accuracy of the risk model, Receiver operating characteristic (ROC) analysis was performed. The Area under curve (AUC) value of the risk model was 0.710, which was little higher than that of the pathologic Stage (Fig. 2e). Similar analyses were carried out in the GSE39582 group. The similar results were shown in Fig. 2f and j.

Fig. 2
figure 2

Construction and validation of prognostic risk model. a-b, The distribution (a) and survival status (b) of EOCRC patients in TCGA group. c, Expression of NOS2 for high- and low-risk EOCRC patients in TCGA group. d, K-M curve of EOCRC patients with different risk score in TCGA group. e, Comparative analysis of ROC curves for EOCRC in TCGA group. f-g, The distribution (f) and survival status (g) of EOCRC patients in GSE39582 group. h, Expression of NOS2 for high- and low-risk EOCRC patients in GSE39582 group. i, K-M curve of EOCRC patients with different risk score in GSE39582 group. j, Comparative analysis of ROC curves for EOCRC in GSE39582 group. k-l, The distribution (k) and survival status (l) of EOCRC patients in GSE17536 + GSE17537 group. m, Expression of NOS2 for high- and low-risk EOCRC patients in GSE17536 + GSE17537 group. n, K-M curve of EOCRC patients with different risk score in GSE17536 + GSE17537 group. o, Comparative analysis of ROC curves for EOCRC in GSE17536 + GSE17537 group.

Full size image

To further verify the feasibility of NOS2 as a candidate biomarker, we conducted similar studies in GSE17536 and GSE17537 group (Fig. 2k and o). EOCRC patients with a high-risk score did have significantly lower NOS2 expression (Fig. 2m) and significantly shorter survival time expression (Fig. 2n).

We further drew nomograms based on the above findings (Fig. 3a and c, and 3e). Time-dependent ROC curve showed the risk model could well predict the outcome of EOCRC (Fig. 3b and d, and 3f). The survival probabilities of EOCRC patients at 1, 3, 5, and 10 years predicted by the risk model were basically consistent with their actual survival rates (Supplementary Figure S4a-S4i).

Fig. 3
figure 3

Nomogram analysis of risk model in EOCRC. a, The Nomograms were used to predict 3-, 5-, and 10-year survival rates of EOCRC patients in TCGA group. b, Time dependent ROC curve of risk model in TCGA group. c, The Nomograms were used to predict 3-, 5-, and 10-year survival rates of EOCRC patients in GSE39582 group. d, Time dependent ROC curve of risk model in GSE39582 group. e, The Nomograms were used to predict 3-, 5-, and 10-year survival rates of EOCRC patients in GSE17536 + GSE17537 group. f, Time dependent ROC curve of risk model in GSE17536 + GSE17537 group.

Full size image

Enrichment analysis for EOCRC with different risk score

To understand the enrichment status between high-risk and low-risk patients, we re-performed the differential expression analysis. There were 321 DEGs (136 up-regulated DEGs and 185 down-regulated DEGs) and 3265 DEGs (2075 up-regulated DEGs and 1190 down-regulated DEGs) between EOCRC patients with high and low risk scores (Supplementary Figure S5a-S5b). Nine pathways were significantly enriched in the TCGA group (Supplementary Figure S5c). Sixty-eight signaling pathways were significantly enriched in GSE39582 group. The top 10 signaling pathways were shown in Supplementary Figure S5d. A total of six overlapping signaling pathways were significantly enriched based on TCGA (Supplementary Figure S5e) and GEO (Supplementary Figure S5f) databases, including Cytoskeleton in muscle cells, ECM-receptor interaction, Focal adhesion, IL-17 signaling pathway, Legionellosis, and Regulation of actin cytoskeleton.

The ssGSEA algorithm was also used to analyze the 6 signaling pathway (Supplementary Figure S5g-S5h). Only Cytoskeleton in muscle cells and IL-17 signaling pathway still showed significantly difference between high risk EOCRC and low risk EOCRC patients. The score of Cytoskeleton in muscle cells and IL-17 signaling pathway were significantly higher in low risk EOCRC patients (Supplementary Figure S5g-S5h).

Genes involved in these two signaling pathways are shown in Supplementary Figure S6i and 6j. For TCGA group, five genes were significantly enriched in IL-17 signaling pathway while nine genes were significantly enriched in Cytoskeleton in muscle cells. For GEO39582 group, five genes were significantly enriched in IL-17 signaling pathway, while 37 genes were significantly enriched in Cytoskeleton in muscle cells (Supplementary Figure S5i-S5j).

Heterogeneity analysis for EOCRC with different risk score

Mutant-allele tumor heterogeneity (MATH), microsatellite instability (MSI), and tumor mutational burden (TMB) were closely related to tumor treatment, including CRC31,32,33. Previous studies have found that MATH was associated with the risk of colon cancer metastasis31. MSI can be used as a unique clinical and pathological feature of CRC, and is closely related to immunotherapy32. TMB was closely related to immunotherapy of CRC33. The results from TCGA database showed that MATH, MSI and TMB were not significantly different between living EOCRC and dead EOCRC patients (Supplementary Figure S6a, S6e, and S6i). Similarly, they did not differ significantly between high-risk and low-risk EOCRC patients (Supplementary Figure S6b, S6f, and S6j). Pearson correlation analysis showed MATH, MSI, and TMB were not significantly associated with risk, with their r values less than 0.3 and p values more than 0.05 (Supplementary Figure S6c, S6g, and S6k). The K-M curve showed that there was no significant difference in survival between EOCRC patients with high MATH and those with low MATH, and the same situation was also observed in MSI and TMB (Supplementary Figure S6d, S6h, and S6l). In addition, MATH, MSI, and TMB were not significantly different among patients with different clinical phenotypes (Supplementary Figure S7a-S7l).

Immunotherapy and chemotherapy analysis for EOCRC with different risk score

To determine whether NOS2 can be used to guide immunotherapy, we first used TIDE to analyze the immunotherapy sensitivity of EOCRC patients in TCGA and GSE39582. The results from TCGA showed that the TIDE, Exclusion and MDSC scores of EOCRC patients with response were significantly lower (Fig. 4a). The results from GEO39582 showed that the TIDE, Dysfunction, Exclusion and CAF scores of EOCRC patients with response were significantly lower, but the MSI.Expr.Sig scores were significantly higher (Fig. 4b).

Fig. 4
figure 4

Sensitivity analysis of Immunotherapy. a, Levels of immunotherapy related signatures for EOCRC patients with different response in TCGA. a, Levels of immunotherapy related signatures for EOCRC patients with different response in GSE39582. c, Levels of immunotherapy related signatures for EOCRC patients with different risk score in TCGA group. d, Levels of immunotherapy related signatures for EOCRC patients with different risk score in GSE39582 group. * p < 0.05. ** p < 0.01. *** p < 0.001. NS, No Significance.

Full size image

When we categorized the EOCRC patients into high and low risk groups, we found that the scores of TIDE, Exclusion, CAF and TAM.M2 of EOCRC patients with low-risk were significantly decreased, while the scores of IFENG, MSI.Expr.Sig, Merck18 and Dysfunction of EOCRC patients were significantly increased in TCGA group (Fig. 4c). Pearson correlation analysis showed that the risk score was positively correlated with CAF, Exclusion, TAM.M2 and TIDE (Supplementary Figure S8a, S8c, S8g, and S8h), and negatively correlated with Dysfunction, IFENG, Merck18 and TAM.M2 in TCGA group (Supplementary Figure S8b, S8d, S8e, and S8f). In the GEO39582 group, the scores of TIDE, Exclusion and TAM.M2 of EOCRC patients with low-risk were significantly decreased, while the scores of IFENG, MSI.Expr.Sig, Merck18 and MDSC were significantly increased (Fig. 4d). Pearson correlation analysis showed that the risk score was positively correlated with IFENG, Merck, MSI-EXPR.Sig and TIDE (Supplementary Figure S8i, S8j, S8n, and S8o), and negatively correlated with Dysfunction, IFENG and Merck18 (Supplementary Figure S8k, S8l, and S8m).

In addition, we also analyzed the relationship of NOS2 risk model to chemotherapy drug sensitivity. The oncoPredict algorithm was used to analyze the sensitivity of EOCRC patients using DEGs between high- and low-risk EOCRC. Among the 35 drugs, 14 and 25 drugs in TCGA and GSE39582 group showed significant differences in sensitivity between high- and low-risk EOCRC patients, respectively (Fig. 5a and b). Pearson correlation analysis found that in TCGA group, 11 out of 14 drugs had a positive correlation with the risk score, and in GEO39582 group, 15 out of 25 drugs and 4 out of 25 drugs had a significant positive and negative correlation with the risk score, respectively (Fig. 5c and d).

Fig. 5
figure 5

Sensitivity analysis of chemotherapy drugs. a-b, Heatmap of the sensitivity of different drugs in TCGA (a) and GSE39582 (b) group. c-d, Correlation of the sensitivity of different drugs with risk score in TCGA (c) and GSE39582 (d) group.

Full size image

Analysis of potential mechanisms and expression validation

NOS2 was candidate biomarker for EOCRC. To understand the possible function of NOS2 in EOCRC progression, we firstly used the TCGA and GSE39582 group for correlation analysis. Using TCGA data, we screened 441 genes that were significantly correlated with NOS2. Using GEO39582 data, we screened 1430 genes that were significantly associated with NOS2 expression. In both TCGA and GEO39582 groups, there were 33 genes with consistent correlations. Functional enrichment analysis of GO and KEGG was performed using DAVID 6.8, and the results showed that 22 BP, 6 CC, 6 MF, and 12 KEGG were significantly enriched (Supplementary Figure S9a-S9d). KEGG results showed that multiple immune-related signaling pathways were enriched to, including hsa04668: TNF signaling pathway, hsa04657: IL-17 signaling pathway, hsa04062: Chemokine signaling pathway, and hsa04060: Cytokine-cytokine receptor interaction, suggested that NOS2 may affect the process of EOCRC by influencing the immune system, and may be used as a prognostic biomarker for EOCRC. The genes highly associated with these immune related pathways were CXCL1 and CXCL2 (Supplementary Figure S9e). CXCL1 and CXCL2 were positively correlated with NOS2 (Fig. 6a and d). QPCR results showed that the expressions of NOS2, CXCL1, and CXCL2 in cancer tissues was significantly lower than that in paracancerous tissues (Fig. 6e and h).

Fig. 6
figure 6

Validation analysis of NOS2 and its related genes. a-b, Correlation of NOS2 with CXCL1 (a) and CXCL2 (b) in TCGA group. c-d, Correlation of NOS2 with CXCL1 (c) and CXCL2 (d) in GSE39582 group. e, Clinical information for EOCRC in the clinical validation group. f-h, Expression of NOS2 (f), CXCL1 (g), and CXCL2 (h) in EOCRC patients. * p < 0.05. ** p < 0.01. *** p < 0.001.

Full size image

Discussions

CRC is the third most common cancer worldwide, and the disease mainly affects the elderly, with the majority of cases occurring in people over 50 years old34. Recently studies have shown an increasing trend in the incidence and mortality of EOCRC35. It is of great significance to find suitable biomarkers for EOCRC. The comprehensive analysis of present study found that the immune-related gene NOS2 was significantly decreased in dead EOCRC patients and was independently associated with poor prognosis of patients.

Nitric oxide (NO) is an active gas molecule that is widely expressed in mammals. NO plays an important function in vivo as the first massager. The key enzymes of NO synthesis are divided into three subtypes: neuronal NOS (nNOS), endothelial NOS (eNOS) and inducible NOS (iNOS, NOS2). Under normal conditions, the content of NO in the body is low. High concentrations of NO can cause cancer by producing nitrosamine carcinogens, damaging DNA, inhibiting DNA repair system, and also inhibiting cancer metastasis36. Previous studies have found that NOS2 has a pro-tumor role37. Many studies have shown that NOS2 is expressed in many cancers. NOS2 is highly expressed in breast cancer, liver cancer, prostate cancer, pancreatic cancer, and gastric cancer38,39,40,41,42,43. Tumor cells induce tumor cell proliferation, invasion, and angiogenesis via NOS-derived NO44,45. The amount and activity of NOS in tumor tissue were positively correlated with the malignant degree of tumor46. In a previous study of 111 tumors, there was a significant correlation between NOS2 positivity in stromal and tumor cells and local microvessel density and apoptotic index47. High concentrations of nitric oxide (NO) mediate tumor transformation in oncogenes and chemically induced tumorigenesis models; Chronic inflammation and continuous exposure to moderate-high concentrations of NO produced by NOS2 are thought to promote neoplastic transformation44. Interestingly, NOS2 also acts as a tumor suppressor37. NOS2 expression was negative in oral soft tissue sarcoma48. Low levels of NOS2 generate tumors in the pancreas and cause liver metastasis, whereas high levels of NOS2 do not cause metastasis49. Inhibition of NOS2 expression can promote distal liver metastasis in mice49. In CRC, previous studies have also found that inhibition of NOS2 promotes cancer50. Du et al. found that NOS2 knockdown induced morphological changes in CRC cells in vitro and in vivo, and promoted the invasion and migration of CRC cells51. In addition, they found that NOS2 expression was reduced in 90% of liver metastatic CRCs compared with only 6% of primary CRCs51. Overexpression of NOS2 inhibits tumorigenesis in mice with thyroid cancer52. Overexpression of NOS2 increased cisplatin-induced cell death in DU145 and PC338 cells53. Also in CRC, a previous study found that high expression of NOS2 could promote cancer progression. Overexpression of NOS2 increases the sensitivity of colon cancer HT-29 cells to radiotherapy54. In xenograft models of human colon and ovarian cancer, cells overexpressing NOS2 increased apoptosis in xenograft tumors36.

NOS2 is rarely expressed in normal tissues, and its expression occurs due to hypoxia, endotoxin, cytotoxin, tumor tissue, or inflammatory cells involved in pathological processes55. Tumor growth depends not only on the cancer cells themselves, but also on the tumor microenvironment. Tumor immune microenvironment (TIME) refers to the surrounding microenvironment in which tumor cells exist, including immune cells, fibroblasts, bone marrow-derived inflammatory cells, various signaling molecules and extracellular matrix. TIME is a complex system, which has both pro-tumor and anti-tumor effects56. Several studies in this study showed that NOS2 was closely related to immunity and cancers44,45. Bioinformatics analysis and clinical sample validation study found that NOS2-related genes CXCL1 and CXCL2 were significantly decreased in EOCRC patients. CXC chemokines have been identified as potential prognostic markers and potential therapeutic targets for a variety of cancers, including CRC57,58. CXCL1 and CXCL2 belong to CXC chemokines and have been found to be used as indicators of tumor characteristics and survival in CRC patients58. Interestingly, previous studies found that CXCL1 and CXCL3 were significantly upregulated in CRC59,60. In this study, we found that CXCL1 expression was reduced in patients with EOCRC, which is inconsistent with previous studies. However, the study of Lina Hu et al. showed that CXCL1 expression was related to age61. Whether the significantly reduced expression of CXCL1 in EOCRC is also age-related remains to be further investigated. Secondary, we focus on EOCRC in the present study. Although the proportion of EOCRC in CRC is on the rise, the proportion of EOCRC in CRC is estimated to be only about 10.9% of all colon cancers and 22.9% of all rectal cancers by 20239. Whether CXCL1 expression in EOCRC is masked by this proportional state remains to be further investigated. Finally, previous studies indicated that silencing CXCL1 can inhibit the proliferation, migration and invasion of CRC cells, and promote cell apoptosis59. Overexpression of CXCL3 promotes the activation of ERK signaling pathway, thereby promoting tumor development60. However, whether the decreased expression of CXCL1 is related to the occurrence and development of EOCRC remains to be further studied. Therefore, it is necessary to collect more EOCRC patients and study the expression of NOS2 and CXCL1 factors. CXCL3, a member of the CXC chemokine subfamily, acts as a potent chemokine for neutrophils, coordinating recruitment and migration of white blood cells while triggering inflammatory responses62. Recent studies have revealed a key role of CXCL3 in carcinogenesis62. In the tumor microenvironment, CXCL3 derived from tumor and stromal cells intricately regulates cell behavior through autocrine and paracrine interactions, primarily through interactions with its receptor CXCR262. Activation of signaling cascades such as ERK/MAPK, AKT, and JAK2/STAT3 highlights the propensity of CXCL3 to favor the tumorigenic process62. However, CXCL3 has a dual function and can exert anti-tumor effects under certain conditions. In summary, the expression of CXCL1 and CXCL2 in EOCRC is inconsistent with the previously reported expression in CRC, which is highly needed to be confirmed by further studies.

The role of NOS2 in oxidative stress has been widely recognized. Reactive oxygen species and active nitrogen species activate signaling pathways, such as NF-kB63. There is a positive feedback loop between NOS2/NO and NF-kB63. Previous studies have also demonstrated that activated NF-κB pathways can activate the expression of genes, such as IL-6, IL-8, CXCL1, CXCL3, CXCL5, CCL20, CD5463, suggesting that NOS2/NO has a positive regulation for CXCL1 and CXCL3. In this study, we found that CXCL1 and CXCL2 expression were significantly reduced in EOCRC. The expression pattern of CXCL1 and CXCL2 was similar to that of NOS2 in CRC and EOCRC. NOS2, CXCL1, and CXCL2 were up-regulated in CRC but down-regulated in EOCRC59,60. The expression of NOS2, CXCL1 and CXCL2 is different in different stages of CRC, and we speculate that their effects may be different in different stages of cancer, but their specific roles need to be further studied. Combining the previous studies and the present results, we speculate that NOS2 may serve as a prognostic indicator of EOCRC and that NOS2 may affect the progression of EOCRC by affecting the immune system. However, further studies are needed to confirm the specific role of NOS2 in EOCRC.

Conclusion

The aim of the present study was to investigate the expression and prognostic significance of NOS2 in EOCRC. The NOS2 prognostic model was a good independent prognostic biomarker for EOCRC, which provides a new perspective for predicting the prognosis of EOCRC. EOCRC patients with different risk score have different sensitivities to immunotherapy and various chemotherapy drugs, which can provide guidance for clinical treatment of EOCRC patients. In addition, in vivo validation analysis showed that NOS2 and its related genes CXCL1 and CXCL2 were significantly decreased in EOCRC patients, suggesting that these three IRDEGs may be involved in the occurrence and development of EOCRC, but their role and mechanism in the development of EOCRC need to be further explored.