3.1. Analysis RILPL2 expression in EC patients
By analyzing the data from TCGA, the expression of RILPL2 in the paracancerous samples was remarkably higher than that in the tumor specimens (P < 0.001, Fig. 1A). Based on GSE17025 dataset, we observed that RILPL2 was also decreased in EC samples compared with paracancerous samples (P < 0.001, Fig. 1B). In addition, we further validated RILPL2 expression using our recruited cohort, and the result showed that RILPL2 in tumor samples was obviously lower compared with normal control (P = 0.013, Fig. 1C). Moreover, the expression level of RILPL2 protein in human endometrium and EC specimens were compared calculating the data obtained from the HPA database, and the result showed that RILPL2 protein was loss in EC samples (Fig. 1D). To further validate the protein level of RILPL2, the CPTAC analysis in UALCAN was applied. Not surprisingly, the result validated the downregulated RILPL2 in EC samples agian (P < 0.001, Fig. 1E).
3.2. Association between RILPL2 expression and clinicopathologic features
A correlation analysis was performed between RILPL2 expression and corresponding clinical characteristics. As presented in Fig. 2, decreased expression of RILPL2 remarkably related to the multiple factors, including age (P = 0.033, Fig. 2A), grade (P < 0.001, Fig. 2B), the tumor histological type (P < 0.001, Fig. 2C), and clinical stage (P < 0.001, Fig. 2D). Moreover, as exhibited in Table 1, logistic regression analysis utilizing the median of RILPL2 expression as a classification of the dependent variable indicated that reduced RILPL2 expression was significantly correlated with high grade (Grade 2 vs. Grade 1, P = 0.005; Grade 3 vs. Grade 1, P < 0.001; Grade 4 vs. Grade 1, P < 0.001), histological type (Mix vs. Endometrial, P = 0.042; Serous vs. Endometrial, P < 0.001) and pathological stage (Stage Ⅲ vs. Stage Ⅰ, P < 0.001; Stage Ⅳ vs. Stage Ⅰ, P < 0.001).
Table 1
Logistic regression analysis between RILPL2 expression and clinical features.
Clinical characteristics | Odds ratio | P value |
Age | ≤ 60 vs. >60 | 0.99 (0.98–1.01) | 0.483 |
Grade | 2 vs. 1 | 0.42 (0.23–0.77) | 0.005 |
| 3 vs. 1 | 0.17 (0.10–0.28) | < 0.001 |
| 4 vs. 1 | 0.06 (0.01–0.26) | < 0.001 |
Histology | endometrial vs. mix | 0.40 (0.16–0.95) | 0.042 |
| endometrial vs. serous | 0.40 (0.10–0.28) | < 0.001 |
Stage | II vs. I | 0.57 (0.31–1.03) | 0.064 |
| III vs. I | 0.34 (0.22–0.53) | < 0.001 |
| IV vs. I | 0.22 (0.09–0.51) | < 0.001 |
3.3. Prognostic value of RILPL2 in EC patients and Cox regression analysis
Based on the fact of RILPL2 downregulation in EC, we further focused on its prognostic role in patients with EC. Kaplan-Meier analysis of the TCGA dataset demonstrated that the patients with high RILPL2 expression led to significantly better OS (P < 0.001) DFS (P < 0.001), and DSS (P < 0.001) in EC patients compared to the low RILPL2 expression group (Fig. 3A-C). Univariate Cox analysis revealed that RILPL2 (HR = 0.578, 95%CI: 0.463–0.721, P < 0.001) were low-risk factor, while age, stage, histological type, and grade were high-risk factors (Fig. 3D). Furthermore, multivariate Cox analysis showed that RILPL2 (HR = 0.747, 95%CI: 0.574–0.972, P = 0.030) was independently related to OS (Fig. 3E), which implied that RILPL2 could be an independent prognostic predictor for EC.
3.4. The relevance of RILPL2 expression with immune infiltration
Based on the median expression value of RILPL2, 552 EC specimens obtained from the TCGA database were classified into high and low expression cohorts (H-RILPL2 and L-RILPL2 groups). It indicated that the L-RILPL2 group had a higher immune score and stromal score than the H-RILPL2 group via ESTIMATE analysis, while the tumor purity score was inferior (Fig. 4A).
The TIMER analysis exhibited that RILPL2 expression had remarkablely positive correlations with B cells, CD4 + T cells, macrophages, and DCs in EC (Fig. 4B). Moreover, we evaluated the correlations between RILPL2 expression and immune markers of different TIICs subtypes in EC tissues using TIMER database. The analysis exhibited that RILPL2 expression were correlated with the expression of marker genes of CD8 + T cells, T cells (general), B cells, monocyte, TAMs, M2 macrophage, neutrophils, NK cells, DCs, Th1 cells, Th2 cells, Th17 cells, Treg cells, and T cell exhaustion to varying degrees. Whereas, It revealed unrelated relationship between the expression of RILPL2 and the expression of gene markers for M1 macrophage and Tfh cell in EC samples (Table 2).
Table 2
Correlation analysis between RILPL2 and related gene markers of immune cells.
Description | Gene markers | R value | P value |
CD8 + T cell | CD8A | 0.141 | 0.001 |
| CD8B | 0.105 | 0.014 |
T cell (general) | CD3D | 0.211 | < 0.001 |
| CD3E | 0.198 | < 0.001 |
| CD2 | 0.193 | < 0.001 |
B cell | CD19 | 0.246 | < 0.001 |
| CD79A | 0.157 | < 0.001 |
Monocyte | CD86 | 0.247 | < 0.001 |
| CSF1R | 0.319 | < 0.001 |
TAM | CCL2 | 0.135 | 0.002 |
| CD68 | 0.157 | < 0.001 |
| IL10 | -0.111 | 0.010 |
M1 macrophage | NOS2 | -0.021 | 0.631 |
| IRF5 | 0.025 | 0.562 |
| PTGS2 | 0.072 | 0.092 |
M2 macrophage | CD163 | 0.128 | 0.003 |
| VSIG4 | 0.213 | < 0.001 |
| MS4A4A | 0.215 | < 0.001 |
Neutrophils | CEACAM8 | 0.005 | 0.915 |
| ITGAM | 0.294 | < 0.001 |
| CCR7 | 0.182 | < 0.001 |
Natural killer cell | KIR2DL1 | 0.071 | 0.096 |
| KIR2DL3 | 0.066 | 0.125 |
| KIR2DL4 | 0.196 | < 0.001 |
| KIR3DL1 | 0.022 | 0.615 |
| KIR3DL2 | 0.035 | 0.419 |
| KIR3DL3 | 0.079 | 0.066 |
| KIR2DS4 | 0.028 | 0.517 |
Dendritic cell | HLA-DPB1 | 0.325 | < 0.001 |
| HLA-DQB1 | 0.218 | < 0.001 |
| HLA-DRA | 0.333 | < 0.001 |
| HLA-DPA1 | 0.301 | < 0.001 |
| CD1C | 0.419 | < 0.001 |
| NRP1 | 0.288 | < 0.001 |
| ITGAX | 0.262 | < 0.001 |
Th1 cell | TBX21 | 0.155 | < 0.001 |
| STAT4 | 0.150 | < 0.001 |
| STAT1 | -0.053 | 0.214 |
| IFNG | 0.026 | 0.540 |
| TNF | 0.028 | 0.516 |
Th2 cell | GATA3 | -0.032 | 0.452 |
| STAT6 | 0.208 | < 0.001 |
| STAT5A | 0.271 | < 0.001 |
| IL13 | -0.020 | 0.641 |
Tfh cell | BCL6 | 0.024 | 0.582 |
| IL21 | 0.012 | 0.773 |
Th17 cell | STAT3 | 0.200 | < 0.001 |
| IL17A | 0.013 | 0.758 |
Treg cell | FOXP3 | 0.119 | 0.005 |
| CCR8 | 0.086 | 0.046 |
| STAT5B | 0.141 | 0.001 |
| TGFB1 | 0.032 | 0.453 |
T cell exhaustion | PDCD1 | -0.010 | 0.808 |
| CTLA4 | 0.169 | < 0.001 |
| LAG3 | -0.041 | 0.334 |
| HAVCR2 | 0.244 | < 0.001 |
| GZMB | -0.032 | 0.454 |
To distinguish the variance of the distribution of 22 TIICs between the two groups, CIBERSORT algorithm was applied to analyze EC cases from the TCGA database. The violin plot manifested the ratio differentiation of 22 TICs between EC tumor specimens with L- or H- RILPL2 expression. T cells CD4 memory activated, T cells follicular helper, T cells regulatory, Dentritic cells resting, and Dentritic cells activated were primary immune cells having a significant relationship with RILPL2 expression (Fig. 4C). The proportion of all these 5 immune cells was increased in H-RILPL2 group compared with L-RILPL2 group.
3.5. Methylation and genetic alterations of RILPL2 in EC
The DNA methylation level of RILPL2 was obtained from the TCGA database, and the differentially expressed methylation levels of RILPL2 between EC and paracancerous specimens were analyzed. The methylation level of RILPL2 in the normal samples was notably lower than that in the tumor samples (P < 0.001, Fig. 5A). Besides, the total methylation level of RILPL2 gene and methylation level of each site all showed significantly negative correlation with its expression (P < 0.001, Fig. 5B, Fig. S1), which indicated that high methylation level was a significant cause for RILPL2 low expression in EC. Besides, low amplification and mutantion rates of RILPL2 were found in EC patients (Fig. 5C). Thus, genetic alterations might not be crucial for the dysregulation of RILPL2 in EC.
3.6. Analysis of the potential mechanisms of RILPL2
In order to build a weighted co-expression network and identify modules and genes related to RILPL2, 2,573 DEGs between EC and paracancerous tissues from the the TCGA database were submitted to WGCNA. After a series of adjustments for WGCNA parameters, the DEGs were clarified into 9 modules by average linkage hierarchical clustering (Fig. S2A-B, Fig. 6A-B). Among these modules, the turquoise module hinted the highest negative correlation with RILPL2 expression (Cor=-0.45, P < 0.001) (Fig. 6C), which might indicated the RILPL2-involved potential oncogenic mechanisms. Subsequently, forty-two genes in the turquoise module were reserved as key genes (GS > 0.2 and MM > 0.8) (Fig. 6D).
To further understand the potential oncogenic mechanisms asssociated to these hub genes, we next conduct GO and KEGG analysis to perform. “chromosome segregation”, “chromosome region”, and “DNA-dependent ATPase activity” were the significantly important GO terms for cellular components, biological processes and molecular functions, respectively (Fig. 7A). “cell cycle” was the most significant pathway in the KEGG pathway analysis (Fig. 7B). Besides, we established a PPI network with these hub genes and found that a large numbers of cell cycle related genes play key roles (Fig. 7C).