CXCL9 screening and expression
The intersection of the TCGA-UCEC and GEO datasets (GSE17025 and GSE62678) yielded 198 up-regulated differential genes (Figure 2G). Cox univariate analysis combined with clinical information from EC patients yielded 25 overlapping genes associated with OS (Figure 2H), DSS and PFI, while multivariate analysis yielded only CXCL9 associated with OS, DSS and PFI in EC patients (Figure 2I). The flow chart was shown in Figure 1. In the TCGA pan-cancer cohort, CXCL9 was highly expressed in 21 of the 26 tumors compared to normal tissue (Figure 3A). Meanwhile, in the TCGA-UCEC datasets, CXCL9 showed high expression in EC tissues (Figure 3B). Using the UALCAN dataset, we also analyzed the correlation between CXCL9 expression in tumor samples and its clinical and pathological indicators, including age, race, stage and histological type (Figure 3C-F). The results showed that CXCL9 expression was only correlated with the expression of stage. Expression was higher in the low stage than in the high stage.
Survival analysis
Prognostic data for EC patients were downloaded from the TCGA database. Univariate COX regression analysis showed that CXCL9, age, histological type, stage and pathological grade were predictors of OS (P <0.05); CXCL9, histological type, stage and pathological grade were predictors of DSS (P <0.05); CXCL9, histological type, stage and surgery were predictors of PFI (P<0.05) (Table1). These predictors were included in Cox multivariate analysis, which showed that CXCL9, age, and stage were associated with OS (P<0.05) (Figure 4A); CXCL9, stage and pathological grade were associated with DSS (P<0.05) (Figure 4B); and CXCL9, stage and surgery were associated with PFI (P<0.05) (Figure 4C). Then, Kaplan-Meier analysis was used to investigate the relationship between CXCL9 expression and the prognosis of EC patients. The results showed that high CXCL9 expression was associated with prolonged survival (Figure 4D-F).
Nomogram development and validation
Cox regression multivariate results were used to construct nomograms of OS, DSS and PFI for 1-,3-,5-and8-year (Figure5). The C-indexes for nomograms were 0.775(95% CI 0726-0.824), 0.84 (95% CI 0.799-0.881) and 0.72 (95% CI 0.667-0.773), respectively. The ROC curves showed that nomograms had better value in predicting the 5-year survival of patients compared to the individual variables. The AUC values of the area under the ROC curves of OS, DSS and PFI nomograms in 5 years were 0.805, 0.845 and 0.736, respectively (Figure 6A-C). The results showed that the model predicted good discriminatory power. The calibration curve results showed that the predicted standard curve and the actual survival probability curve were close to the 45 degrees diagonal, demonstrating the high applicability of the prediction model. (Figure 6D-F). In DCA curve analysis, it was obvious that the prediction model showed better results (Figure 6G-I).
Protein interactions
CXCL9-related genes were linked using GeneMANIA and STRNG, containing a total of 21 genes and 1060 links (Figure 7A-B). The TIMER website was then used for correlation analysis of CXCL9-related genes. The results showed that 15 genes (CCL2, CCL5, CCL11, CCL13, CCL18, CCL19, CCL21, CCL26, CXCL10, CXCL11, CXCL12, CXCR3, PF4, STAT1 and XCL1) were associated with CXCL9 (P <0.05) and 5 genes (CCL28, CXCL2, CXCL14, CXCL17 and PPBP) were not associated with CXCL9 (P >0.05) (Figure 7C-F).
Somatic cell mutations
A total of 610 EC specimens were selected from the cBioPortal database (Figure 8A), of which 13 specimens had CXCL9 mutations, with a mutation rate of 2.1%. There were 9 cases of missense mutation, 3 cases of truncating mutation and 1 case of deep deletion (Figure 8B). The results indicate that the CXCL9 gene was predominantly missense mutation in EC, but its mutation rate is not high. 13 patients with clinical information showed a mean age of 56 and a predominantly G3 pathological classification, with only one death. In addition, genes significantly associated with CXCL9 mutations in EC were obtained from muTarget software (P<0.01), including 43 down-regulated genes and 40 up-regulated genes. We selected the top five up-regulated genes, which included GPX4 (FC = 1.57, P = 7.82e-04), LONP1 (FC = 1.58, P = 8.47e-04), NDUFA6 (FC = 1.51, P = 7.48e-04), SLC18B1 (FC = 2.08, P = 5.66e-65) and ZNF117 (FC = 2.33, P = 7.90e-64) (Figure 8 C-G). A complete list of all CXCL9 mutation-associated genes can be found in Supplementary Table 1.
Immunoassay
The immune cell infiltration score of CXCL9 was assessed in 9555 tumor samples from 39 tumor types using the CIBERSORT algorithm(Figure 9A).The expression data of CXCL9 and 150 marker genes of the five immune pathways (chemokine (41), receptor (18), MHC (21), Immunoinhibitor (24), Immunostimulator (46)) were further extracted in each sample, and the pearson correlation between CXCL9 and the marker genes of the five immune pathways was calculated(Figure 9B);In addition, the expression data of CXCL9 and 60 marker genes of two types of immune checkpoint pathways (Inhibitory (24), Stimulatory (36)) were extracted in each sample, and the pearson correlation between CXCL9 and immune checkpoint marker genes was calculated(Figure 9C).We also assessed the correlation between CXCL9 expression and 22 immune cells in endometrial cancer tissues by the CIBERSORT algorithm, showing that CD8 T cells, CD4 T cells, CD4 memory T cells, T cell follicular helper cells, NK cells, M1 macrophages, and dendritic cells had significantly higher immune function in the high expression group (Figure 9D).Finally the correlation of CXCL9 with six immune cells and the correlation of six immune cells with the prognosis of EC patients were analyzed using the TIMER method, which showed that B cells and CD+8 T cells were correlated with the prognosis of EC patients (Table 2, Figure 9 E-F).TISIDB was used to explore the correlation of CXCL9 with the expression of human immune cells, immunomodulators and other chemokines in human cancers (Supplementary Figure 1). These results suggest that CXCL9 plays an important role in the immune microenvironment.
Gene enrichment analysis
GO and KEGG enrichment analysis showed that CXCL9 is enriched for functions such as immune response regulation signaling pathways, the external side of the plasma membrane, immune receptor activity, cell adhesion and other pathways, extracellular structures and extracellular matrix organization (Figure 10A-B).GSEA analysis showed that CXCL9 was associated with functions such as angiogenesis and hypoxia, in relation to chemokine and T cell receptor signaling pathways (Figure 10C-D).In addition, the JAK-STAT signaling pathway, P13K-Akt signaling pathway, MAPK signaling pathway and PPAR signaling pathway were also identified as signature signaling pathways with high enrichment (Figure 10E-H).