Expression of GPX4 in different cancers
TIMER, GEPIA, GENT2 and TNMPlot online databases were used to analyze the expression level of GPX4 in different types of human cancer and normal tissues. First, we analyzed TCGA RNA sequence data using TIMER database to evaluate GPX4 gene expression in various cancers. The results showed that GPX4 was differentially expressed in tumor tissues and normal tissues. Compared with adjacent normal tissues, the expression of GPX4 was significantly up-regulated in BLCA, COAD, ESCA, HNSC, KICH, KIRC, KIRP, LIHC, LUAD, PRAD, READ, STAD, THCA and UCEC. However, GPX4 expression was significantly down regulated in BRCA (Fig. 1a). Next, we used GEPIA2 database, which integrates gene expression data from TCGA and GTEx to determine the expression level of GPX4 in 48 human tumors and 337 normal tissues. According to this result, compared with adjacent normal tissues, the expression of GPX4 in DLBC, PAAD, STAD, THYM and UCEC was significantly increased, however, the expression of GPX4 was significantly decreased in LAML and TGCT (Fig. 1b). In GENT2 database, based on U133_ Plus_ 2.0 The expression profile of microarray platform (GPL570) found that compared with adjacent normal tissues, the expression of GPX4 gene in the following cancer types was significantly up-regulated: Adrenal, Esophah, Liver, Lung, Ovary, Pancreas, Prostate, Rectum, Renal, Sky, Stomach, Thyroid and Uterus were significantly down regulated in AML, Breath and Testis (Fig. 1c). We further evaluated the expression level of GPX4 in human cancer using TNMplot database (Fig. 1d), and found that the expression level of GPX4 sent significant changes in most cancer tissues. In these databases, the expression of GPX4 in many cancers has changed significantly, and most of them are significantly up-regulated. Because only GEPIA database has DLBCL data organized normally, DLBCL data can only be obtained in GEPIA.
Table 1 GPX4 prognostic risk model data were obtained by Lasso Cox regional analysis on GSE10846 and GSE181063
|
GSE10846
|
GSE181063
|
Genes
|
HR (univariable)
|
HR (univaariable)
|
GPX4
|
0.06 (0.00-1.00, p=.050)
|
0.01 (0.00-0.09, p<.001)
|
CDCA7
|
1.21 (0.24-6.08, p=.814)
|
9.59 (3.01-30.53, p<.001)
|
P> 0.05 is considered not significant
Expression Analysis of GPX4 in DLBCL
We further analyzed the gene expression level of GPX4 in DLBCL using the online database GEPIA. According to this result, compared with normal DLBCL tissues, GPX4 is up-regulated in cancer tissues. In the TCGA dataset boxplot, we found that there was an obvious stratification between DLBCL tissue and normal tissue samples, indicating that GPX4 might be a potential diagnostic marker of DLBCL (Fig. 2a). Therefore, we verified the difference of GPX4 expression in GSE56315. Through R4.2.1 software, we found that GPX4 is a significantly different up-regulated gene (Fig. 2b), and analyzed the ROC curve. GSE56315 also showed that there is a good distinction between the expression of GPX4 in cancer samples and normal tissue samples, and AUC=0.998. Therefore, GPX4 is a potentially good diagnostic marker of DLBCL (Fig. 2c). In addition, we found that GPX4 has significantly different expression in TP53 mutant and TP53 wild type of DLBCL (Fig. 2d). TP53 mutation is a common factor of DLBCL mutation, representing the poor progress of DLBCL. The increased expression of GPX4 may be related to alleviating this progress trend.
Prognostic analysis of GPX4 in DLBCL
In order to evaluate the prognostic role of GPX4 in DLBCL, we used R4.2.1 to conduct Lasso Cox regional analysis on GSE181063 and GSE10846 to obtain GPX4 prognostic risk model data (Table 1), and conducted univariate analysis. The results showed that GPX4 had a significant impact on the prognosis of DLBCL, and then KM curve also showed that there was a significant difference between the survival curves of patients with high and low expression of GPX4, which was verified in TCGA data set (P<0.05) (Fig. 3a-c).
Transcription Factor Prediction and Co expression Network Analysis of GPX4 in DLBCL
In order to obtain knowledge about the biological functions of GPX4 in DLBCL, first we use, first we use the Cistrome DB database( http://cistrome.org/db/#/ )For transcription factor prediction, chr19:1103993:1106778 was selected to explore that the top 20 transcription factors in the score rankings affected the expression of GPX4 (Fig. 4a). Then, GeneMANIA was used to build a PPI network of 21 genes centered on GPX4 (Fig. 4b). GO function enrichment and KEGG pathway analysis were carried out for these 21 genes. Significantly rich GO terms include Glutathione peroxidase activity, Leukotriene metabolic process, Reactive oxygen species metabolic process, while the significantly rich KEGG pathway is Arachidonic acid metabolism, which has been reported to be closely related to the immune regulation of lymphoma in recent years (Fig. 4c). These results suggest that GPX4 may be an important regulator of DLBCL progression by participating in important metabolic pathways related to lymphoma and immune cell infiltration.
Correlation Analysis of GPX4 and DLBCL Key Genes
We analyzed the differential genes (Log2>4 and P<0.01) that were up-regulated in DLBCL ranking by TCGA data, and then verified them by GSE56315 (Fig. 5a). A total of 124 differential genes have been verified and enriched through Metascape to study potential biological pathways (Fig. 5b). Utilize STRING database( https://string-db.org/ )The network relationship diagram of 124 differential genes in DLBCL was obtained, and the key functional gene modules were obtained through the MCODE plug-in in the Cytoscape software. We took the module with the highest total score (Score>14), and a total of 16 key genes were obtained (Figure 5c). Spearman correlation analysis was used to analyze the correlation between GPX4 and the six key MCODE genes of DLBCL (Fig. 5d). The results showed that all these genes had a negative correlation with GPX4, indicating that the upregulation of the key genes of this module promoted the occurrence and development of DLBCL, and there was a significant negative correlation between GPX4 and CDCA7. Then we conducted a prognostic analysis of CDCA7 and found that CDCA7 was a risk factor (Fig. 5e-f and Table 1), CDCA7 gene encodes 371 amino acid proteins, which are abnormally expressed in various tumor tissues. The correlation results showed that the high expression of GPX4 in DLBCL inhibited the high expression of CDCA7, thus protecting the occurrence and development of DLBCL and the prognosis of patients. At present, there is no report on the role of CDCA7 in DLBCL. CDCA7 has a significant prognostic difference in 1311 DLBCL patients, but there is no significant difference in 181 clinical samples from patients treated with CHOP and 233 clinical samples from patients treated with rituximab CHOP. This may be related to the sample size and treatment methods, but it is undeniable that CDCA7 may be a potential key molecule in the pathogenesis and prognosis of DLBCL.
Correlation between GPX4 and immune cell infiltration
Tumor microenvironment has been proved to play an important role in tumorigenesis. Since there is a significant negative correlation between CDCA7 and GPX4, we used TIMER to determine whether the expression of GPX4, CDCA7 and their GPX4 copy number changes in DLBCL is related to immune cell infiltration. We found that GPX4 is significantly negatively related to immune B cell infiltration (Figure 6a), and CDCA7 is significantly positively related to immune B cell infiltration (Figure 6b); The change of the copy number of GPX4 was significantly correlated with the infiltration of immune B cells and immune macrophages (Fig. 6c), and the change of the copy number of CDCA7 was significantly correlated with the infiltration of immune macrophages (Fig. 6d). Therefore, the immune interaction between GPX4 and CDCA7 at the expression level is mainly related to the infiltration of immune B cells, while the immune interaction at the copy number variation level is mainly related to immune macrophages.