This study utilized bioinformatics methods to screen out genes related to lysosomes in colorectal cancer and constructed a prognostic model.

The colorectal cancer data sets were downloaded from TCGA and GEO databases, respectively. The R package was used to screen out the differentially expressed LRGs related to CRC survival, and univariate Cox regression and LASSO regression were used to screen the related prognostic genes to construct a prognostic risk model. Univariate Cox regression analysis was used to construct a clinical prognostic model, and independent prognostic and survival analysis was performed. The performance of the model was verified by K-M survival curve and ROC curve analysis. Finally, the relationship between ATP6V1G2, DLG4, LZTS1 gene and protein expression and pathological features and prognosis of colorectal cancer was further verified by in vitro experiments.

Seven related prognostic genes were screened out, including ATP6V1G2, LZTS1, CLU, PDGFRA, CLVS2, DLG4 and RAMP1, and a prognostic model of colorectal cancer was constructed based on these genes. K-M survival curve showed that the prognosis of low-risk group was significantly better than that of high-risk group (χ²=15.14, P=0.0001), which was verified in GEO database. The ROC curve showed that the AUC values of 5-year and overall survival time were greater than 0.65 (0.67, 0.66), indicating that the model had certain predictive ability. GO enrichment analysis showed that LRGs were mainly involved in autophagy, endocytosis, macroautophagy, cell degradation and other processes, thereby affecting the occurrence, development and transformation of colorectal cancer. KEGG enrichment analysis showed that Rap1 signaling pathway, endocytosis and salivary secretion were mainly enriched. The results of in vitro experiments showed that ATP6V1G2 and DLG4 were highly expressed (ATP6V1G2: t=−6.847, P＜0.05; DLG4: t=−6.324, P＜0.05) and LZTS1 was lowly expressed (t=5.568, P＜0.05) in colorectal cancer. The high expression of ATP6V1G2 and DLG4 and low expression of LZTS1 were significantly correlated with the depth of tumor invasion (ATP6V1G2: χ²=5.333, P=0.021; DLG4: χ²=6.522, P=0.011; LZTS1: χ²=6.095, P=0.014), lymph node metastasis (ATP6V1G2: χ²=7.065, P=0.008; DLG4: χ²=5.265, P=0.022; LZTS1: χ²=5.224, P=0.022) and patient age (ATP6V1G2: χ²=4.844, P=0.028; DLG4: χ²=6.332, P=0.012; LZTS1: χ²=4.969, P=0.026). The higher the expression levels of ATP6V1G2 and DLG4, and the lower the expression level of LZTS1, the worse the prognosis of the patients will be. ROC curve showed that ATP6V1G2, DLG4 and LZTS1 alone and in combination had predictive value for the diagnosis and prognosis of colorectal cancer (ATP6V1G2: AUC=0.886, P＜0.001; DLG4: AUC=0.781, P=0.003; LZTS1: AUC: 0.667, P=0.029; Combination of the three: AUC=0.894, P＜0.001).

This study successfully constructed a prognostic model of colorectal cancer based on 7 lysosome-related genes, which can better predict the prognosis of patients. The high expression of ATP6V1G2 and DLG4 and low expression of LZTS1 can be used as potential molecular markers for poor prognosis of colorectal cancer.