Hypoxia is a feature of tumor physiology, specifically that of mechanisms associated with the acquisition of some malignant attributes, such as metastasis, invasion7–9, and drug resistance10. In these processes, hypoxia-related genes act on the corresponding pathways or on the regulating immune cells. Hypoxia-inducible factor (HIF) is activated during hypoxia; some immunosuppressive factors, such as vascular endothelial growth factor, are HIF target genes, which affect both angiogenesis and immunosuppression11. Hypoxia also results in upregulated EGFR expression, which promotes ligand-independent epidermal growth factor receptor signaling12,13. This process increases the rate of tumor glycolysis, resulting in metabolic competition14. In our study, we screened for hypoxia-related genes in the gut and found that the core genes (ALDOB, GPC1, ALDOC, and SLC2A3) were closely related to patient prognosis. The rate of aldolase-B and fructose-bisphosphate B-driven fructose metabolism is significantly increased in patients with colon cancer and liver metastasis, and in those with colorectal villous polyps15,16. GPC1 has also been shown to be overexpressed in various malignancies17,18. A recent study has reported GPC1 enrichment in tumor-derived exosomes17. In melanoma, NME1 has been shown to inhibit metastasis by activating ALDOC transcription19. Solute carrier family 2, member 3 can increase glucose uptake in anoxic cells and, thus, increase the rate of glycolysis. These findings suggest that hypoxia-related gene expression levels are closely related to tumor development and metabolism20; therefore, they were included in our hypoxia-related gene model.
With a deepening understanding of the mechanisms of hypoxia, its influence on tumor prognosis is also increasingly being understood21,22. To further investigate the relationship between the expression of hypoxia-related genes and patient prognosis, we evaluated patient prognosis by taking the product sum of the expression levels and the coefficients of ALDOB, GPC1, ALDOC, and SLC2A3 in TCGA and GEO databases as the risk score. The prognosis of patients in the high- and low-risk groups differed significantly; the survival rate of patients in the low-risk group was significantly higher than that of their counterparts. However, the ROC curve analysis, which aimed to evaluate the accuracy of the survival estimates, showed that the curves obtained from the GEO database corresponded poorly to the observed prognosis. Thus, we also analyzed the influence of other factors, including age, sex, TNM staging, and our proposed risk score, on patient prognosis. We concluded that our proposed risk score showed a good correspondence to patient prognosis. However, the results of the multivariate analysis in the GEO database were not statistically significant. This finding suggests that the risk score alone cannot be used as an independent prognostic factor. The samples in the GEO database were all colorectal adenocarcinoma samples that had been obtained in France. Because the tumor type and region are very specific, these samples may not accurately reflect the relationship between hypoxia and colon cancer prognosis. Moreover, the prognosis of colon cancer is related to disease stage and clinical, histological, genetic, and molecular factors, among others. These factors should be considered in future studies.
We next applied GSEA to identify pathways enriched in the high- and low-risk groups in the two databases. We found that most of the enriched pathways were related to inflammation, immune response, and apoptosis. Hypoxia and cell death in tumor tissue produce large amounts of cell debris and trigger the release of inflammatory factors, which can attract macrophages and monocytes, and can induce macrophage polarization. After polarization, macrophages secrete inflammatory factors23. These findings suggest a close relationship between hypoxia, inflammation, and the immune response. In addition, some apoptosis-related genes, such as p53, which is a tumor suppressor gene, are closely related to tumor apoptosis. Nilay et al.24 found that mutations in p53 in gastric and esophageal cancer cells can induce hypoxia signaling. This finding was confirmed in a study involving nude mice24.
The results of a large number of previous studies and those of the present study have shown that hypoxia can recruit immune cells in the tumor microenvironment. Hypoxia-induced tumor-derived cytokines, such as IL-10 and transforming growth factor-beta, can induce tumor-associated macrophages to differentiate into M2 macrophages with immunosuppressive effects25. When monocytes are stimulated by inflammatory factors, such as interferon-gamma and lipopolysaccharide, they activate M1 macrophages, which can secrete inflammatory factors, such as IL-6 and tumor necrosis factor-alpha, and can phagocytize invasive pathogens and tumor cells23,26. Hypoxia can also lead to immune escape through the role of immune cells5; for example, hypoxia can reduce T-cell activity27. Hypoxia is closely related to immune cell function. In our study, we observed significant differences in the levels of activated CD4 memory T cells and M0 macrophages between the high- and low-risk groups in TCGA database. In the GEO database, the levels of infiltration of activated CD4 memory T cells and M0 macrophages, those of activated and resting mast cells, and those of neutrophils differed significantly between the high- and low-risk groups. Both databases showed significant differences in the infiltration rates of activated CD4 memory T cells and M0 macrophages between the high- and low-risk groups. When we screened the genes that regulated these two kinds of cell function, we found different levels of CCL2/4/5, CSF1, and CX3CL1 expression between the high- and low-risk groups. Further analyses confirmed the significant correlation between the expression level of each of these genes and the risk of hypoxia.
Three of these five genes encode chemokines, which play a chemotactic role in immune cells, such as natural killer cells and monocytes, which are closely related to tumor development. CCL2 and CCL5 play important roles in prostate cancer metastasis and drug resistance28,29, while CCL4 is associated with the clinical characteristics of breast cancer30. Previous studies have shown that CX3CL1 can recruit natural killer T and dendritic cells to participate in antitumor immunity31,32. CX3CL1 is expressed in various tumors33–35; however, its antitumor effects remain controversial. Specifically, Tsang33 reported that CX3CL1 expression was positively correlated with the rate of lymphocyte infiltration and negatively correlated with tumor survival. Colony-stimulating factor 1 is expressed in various tumor cells36. Its secretion can induce the differentiation of M2-type macrophages and recruit immune cells into the tumor37. The five genes identified in the present study are closely related to immune cells.