3.1 Clinicopathologic Significance of the Estimate Scores
To reveal the relationship between the component of immune or stromal cells and the clinical outcomes, survival analysis was performed for ImmuneScore, StromalScore, and ESTIMATEScore, respectively. As presented in Figure 1A, the percentage of immune components was positively correlated with the overall survival rate despite that StromalScore and ESTIMATEScore made no difference (Figure 1B, C). These results manifested that the immune composition in TME was a more reliable indicator of the prognosis for EC patients. We further estimate the correlation between the three scores and clinicopathologic parameters. The results demonstrated that tumors of EC patients over 65 years old had lower StromalScore and ESTIMATEScore than younger patients (P < 0.01 and <0.05, respectively) even though ImmuneScore did not differ in terms of age as shown in Supplementary Figure 3A, B, C. As for the pathological grade, it was worth noting that our data exhibited consistently decreasing trends in the three scores as tumor progressed even though not all comparisons among groups with different grades were statistically significant (Supplementary Figure 3D, E, F). These results implied that advanced age (> 65) and higher grade of EC, mainly grade Ⅲ, tended to have less immune and stromal component, possibly signified poor prognosis.
3.2 DEGs Shared by ImmuneScore and StromalScore and Functional Enrichment
In order to explore the variation of gene profile in the wake of the alteration of the immune or stromal component in TME, we compared the expression level of genes between the high- and low-score groups. Among them, 716 DEGs (552 up-regulated and 164 down-regulated) were determined regarding ImmuneScore, 731 DEGs were screened out from StromalScore, which mainly consisted of highly expressed genes. As shown in the heatmaps, the top 50 DEGs were displayed based on ImmuneScore and StromalScore, and the gene expression pattern showed an apparent difference between groups (Figure 1D, E). After the intersection, 386 genes were shared in common by ImmuneScore and StromalScore, with 366 up-regulated and the rest down-regulated (Figure 1F, G). Then, functional enrichment analysis was carried out based on these overlapped genes. Our results indicated that the identified genes were chiefly enriched in the regulation of lymphocyte activation, differentiation, and proliferation in the biological process (BP). About the cellular component (CC) of GO analysis, these genes encoding proteins were the main components of the external side of the plasma membrane, immunological synapse, granule membrane, and protein complex involved in cell adhesion. As for molecular function (MF), the DEGs prevailingly enrolled in immune or cytokine receptor activity (Figure 1H). KEGG analysis revealed that the chemokine signaling pathway and cytokine-cytokine receptor interaction were most relevant to these DEGs (Figure 1I). It thus appeared that the overall function of these DEGs primarily focused on immune-related activities, which essentially suggested that the involvement of immune modulation was a notable feature of TME in EC.
3.3 Intersection Analysis of PPI Network and Cox Regression
To elucidate the underlying mechanism, we explored the PPI network constructed by the STRING database, and thus, the interaction between genes was identified. On this basis, the MCODE plugin of Cytoscape software was applied to seek the hub gene cluster. Finally, 56 genes were identified with 794 edges, displayed in Figure 2A, and the bar plot in Figure 2B showed the top 40 genes, ranked by the number of adjacent nodes. Additionally, to identify prognosis-related genes among 386 DEGs in EC, univariate Cox regression analysis was performed, and the results manifested that the expression level of 16 genes was highly correlated with the survival outcome as shown in the forest plot (Figure 2C). Subsequently, two genes, TNFRSF4 and S1PR4, were screened out after intersection analysis based on the above PPI network and univariate Cox regression analysis (Figure 2D).
3.4 Prognostic Significance and Indicative Role in TME Remodeling of TNFRSF4 in EC
According to previous studies on EC, we found a gene, namely TNFRSF4, barely reported on EC but was a high-profile target on other cancers. TNFRSF4 was first discovered on the surface of activated CD4+ T cells, which played a vital role in immune regulation in multiple cancers as a crucial immune checkpoint. In the following analysis, our data indicated that TNFRSF4 showed higher expression levels in tumor tissues than normal tissues either in unpaired or paired samples (P < 0.001 and P = 0.028, respectively), as shown in Figure 3A, B. To confirm the expression of this gene, six microarrays were eventually involved in the current study from the GEO database, as displayed in Supplementary Table 3. In total, 255 tumor samples and 126 normal tissues were included. Regarding the expression trends of TNFRSF4, a comprehensive meta-analysis was carried out to quantify its expression according to these datasets accurately. The calculation results were presented in Figure 4A, for which the fixed-effects model was applied given the existed homogeneity (I2 = 19%, τ2 = 0.0277, P = 0.29). Results also indicated that the expression of TNFRSF4 was dramatically augmented in the EC group (SMD = 0.34, 95% CI [0.11; 0.57], Z = 2.87, P = 0.0041). The publication bias was also assessed, as displayed in Figure 4B. The funnel plot indicated no bias of the publications (z = 0.94, P = 0.348). Later, a sensitivity test was generated, and no items were determined to exert possible effects on the results (Figure 4C). Taken together, it was reasonable to determine the up-regulated expression trend of TNFRSF4 in EC tissues.
Interestingly, the expression level decreased as aged when compared with the younger group (Figure 3C). Similarly, tumors of grade Ⅱ or Ⅲ exhibited a consistently less amount of TNFRSF4 when compared with tumors of grade Ⅰ (Figure 3D). More importantly, high expression of TNFRSF4 was significantly associated with a better prognosis (Figure 3E). Then, GSEA was implemented for the high and low-expression groups of TNFRSF4. As shown in Figure 3F, the genes in the TNFRSF4 high-expression group were chiefly enriched in immune-related activities, such as IFN-γ response, IL-2/STAT5 signaling, and IL-6/JAK/STAT3 signaling. In terms of the TNFRSF4 low-expression group, the genes were mainly involved in the G2M checkpoint, mitotic spindle, mTORC1 signaling, and protein secretion (Figure 3G). For the C7 collection defined by MSigDB, the immunologic gene sets and multiple immune functional gene sets were enrolled in the high and low TNFRSF4 expression groups (Figure 3H, I). These results suggested that TNFRSF4 might be a potential indicator for the status of TME.
3.5 Correlation of TNFRSF4 with the Abundance of TICs
Given the above findings, the component of TICs of each sample was further estimated using the CIBERSORT algorithm to detect the pertinence relation of TNFRSF4 expression with the immune microenvironment, and 22 kinds of immune cell profiles in EC samples were identified, as shown in the bar plot (Figure 5A). It was observed that immune cells in EC were mainly composed of T cells and macrophagocytes. Besides, the correlation among the immune cells was also displayed. The results showed that the proportion of CD8+ T cells was negatively correlated with the presence of CD4+ memory resting T cells and M0 macrophagocytes (correlation coefficient = -0.55 or -0.63, respectively). Conversely, the content of CD8+ T cells was positively related to the scale of CD4+ memory-activated T cells, as seen in Figure 5B.
To verify the primary immune cells affected by TNFRSF4, the difference and correlation analysis were carried out, and results demonstrated that a total of 5 kinds of TICs differed between the high and low-TNFRSF4 expression groups, including CD8+ T cells, regulatory T cells, resting dendritic cells, eosinophils, and neutrophils (Figure 6A). Furthermore, seven kinds of immune cells were correlated with the expression of TNFRSF4, as observed in Figure 6B. In addition, four types of immune cells, including CD8+ T cells, regulatory T cells, eosinophils, and neutrophils, were identified to be vitally interconnected with TNFRSF4 (Figure 6C). T-cell-mediated immune modulation in the immune microenvironment partly depends on the subsets of T cells. In this part, we analyzed the potential relationship between TNFRSF4 and the subsets of T cells. Results showed that TNFRSF4 was positively correlated with subsets of T cells, such as Th1 (R = 0.35), Th17 (R = 0.15) and Treg (R = 0.6), but negative correlated with Th2 (R = -0.12) (Supplementary Figure 4A, B). These results hinted that the expression of TNFRSF4 had much to do with the immune activity of TME in EC.
3.6 Identify the Expression of TNFRSF4 and Its Correlation with Immune-related Genes.
From the TCGA mRNA expression profiles, we found that the expression of TNFRSF4 was positively correlated with CD4 (R = 0.51), CD8A (R = 0.47), and FOXP3 (R =0.59) (Figure 7A, B, C). In IHC staining, representative images of TNFRSF4, CD4, CD8, and FOXP3 were exhibited in Figure 7D, TNFRSF4, CD4, and CD8 proteins were over-expressed in EC tissues (mainly on tumor immune infiltrating cells) compared with paired adjacent normal tissues or even unpaired normal tissues (Figure 7E, F). However, we did not detect any evident expression of FOXP3 at the protein level. Besides, there was a significant statistical correlation of TNFRSF4 with CD4 and CD8, except for clinicopathologic parameters (such as age and histological grade) (Supplementary Table 4). To assess the diagnostic value of TNFRSF4 in EC, the ROC diagnosis model was performed. Surprisingly, we uncovered that TNFRSF4 had a higher diagnostic significance in either the TCGA dataset (the AUC value = 0.715) (Figure 7G) or clinical specimens (the AUC value = 0.777) (Figure 7H), even compared with the diagnostic capability of CD4 or CD8.
Based on the results from IHC, we further carried out m-IHC to comprehensively investigate the relationship among TNFRSF4, CD4, CD8, and FOXP3 in the TME. As shown in Figure 7I, TNFRSF4, CD4, and CD8 mainly expressed on the tumor stroma and displayed a co-localization pattern observed in both the separate and merged images. Similarly, no reliable fluorescence staining with FOXP3 was observed. Altogether, these data implied the potential role of TNFRSF4 in immune microenvironment remodeling and diagnostic performance for EC.