A construction and comprehensive analysis of ceRNA networks and infiltrating immune cells in papillary renal cell carcinoma

Abstract

Background: As the second most common malignancy in adults, papillary renal cell carcinoma (PRCC) has shown an increasing trend in both incidence and mortality. Effective treatment for advanced metastatic PRCC is still lacking. In this study, we aimed to establish competitive endogenous RNA (ceRNA) networks related to PRCC tumorigenesis, and analyze the specific role of differentially expressed ceRNA components and infiltrating immune cells in tumorigenesis.

Methods: CeRNA networks were established to identify the key ceRNAs related to PRCC tumorigenesis based on the 318 samples from The Cancer Genome Atlas database (TCGA), including 285 PRCC and 33 normal control samples. The R package, "CIBERSORT," was used to evaluate the infiltration of 22 types of immune cells. Then we identified the significant ceRNAs and immune cells, based on which two nomograms were obtained for predicting the prognosis in PRCC patients. Finally, we investigated the co-expression of PRCC-specific immune cells and core ceRNAs via Pearson correlation test.

Results: COL1A1, H19, ITPKB, LDLR, TCF4, and WNK3 were identified as hub genes in ceRNA networks. Four prognostic-related tumor-infiltrating immune cells, including T cells CD4 memory resting, Macrophages M1, and Macrophages M2 were revealed. Pearson correlation test indicated that Macrophage M1 was negatively related with COL1A1 (p < 0.01) and LDLR (p < 0.01), while Macrophage M2 was positively related with COL1A1 (p < 0.01), TCF4 (p < 0.01), and H19 (p = 0.032). Two nomograms were conducted with favorable accuracies (area under curve of 1-year survival: 0.935 and 0.877; 3-year survival: 0.849 and 0.841; and 5-year survival: 0.818 and 0.775, respectively).

Conclusion: The study constructed two nomograms suited for PRCC prognosis predicting. Moreover, we concluded that H19-miR-29c-3p-COL1A1 axis might promote the polarization of M2 macrophages and inhibit M1 macrophage activation through Wnt signaling pathway, collaborating to promote PRCC tumorigenesis and lead to poor overall survival of PRCC patients.

Keywords: PRCC; ceRNA network; immune cells; nomogram.

FIGURE 1

The analysis steps of this study

FIGURE 2

The differentially expressed genes in PRCC. The heatmap (A) and the volcano plot (B) of all differentially expressed mRNAs; the heatmap (C) and the volcano plot (D) of all differentially expressed miRNAs; the heatmap (E) and the volcano plot (F) of all differentially expressed lncRNAs. PRCC, papillary renal cell carcinoma; mRNAs, messenger RNAs; miRNAs, microRNAs; lncRNAs, long non‐coding RNAs

FIGURE 3

(A) Gene ontology function enrichment circle and (B) Pathway enrichment cycle of differentially expressed mRNAs

FIGURE 4

(A) The experimental validated ceRNAs networks. (B) LDLR, (C) COL1A1, (D) has‐miR‐130b‐3p, and (E) PDGFRB had significantly prognostic values

FIGURE 5

(A, B) Through lasso regression analysis, six molecules (COL1A1, H19, ITPKB, LDLR, TCF4, and WNK3) were integrated into the (C) Cox proportional hazards model. (D) Nomogram based on the Cox hazard model for predicting the OS of PRCC patients. (E) RNAs in the panel (C) exhibited survival capability. (F) Calibration curve and (G) ROC curve for suggesting the prediction ability and accuracy of nomogram. The area under curve of 1‐year, 3‐year, and 5‐year survival was 0.935, 0.849, and 0.818, respectively. PRCC, papillary renal cell carcinoma; ROC, receiver operating characteristic

FIGURE 6

(A) Bar plot showing immune cell types and relative percent in PRCC tissues. (B) Heatmap of tumor‐infiltrating cells in PRCC tissues and control group tissues. On the top, the blue annotation represents the normal control group, and the red annotation represents PRCC patients. (C) Violin plot by Wilcoxon rank‐sum test for comparison of cells’ proportion between the normal and PRCC tissues. It revealed that the proportions of the B cells naive (p < 0.001), plasma cells (p = 0.006), resting memory CD4T cells (p < 0.001), M1 macrophages (p < 0.001), and resting dendritic cells (p < 0.001) in PRCC were reduced, and memory B cells (p < 0.001), CD8T cells (p = 0.038), follicular helper T cells (p = 0.014), M0 Macrophages (p < 0.001), M2 Macrophages (p < 0.001), and activated dendritic cells (p < 0.001) were increased in PRCC tissues. (D) Correlation analysis of different tumor‐infiltrating cells in PRCC samples. PRCC, papillary renal cell carcinoma

FIGURE 7

Relevance of 22 immune cells’ types with clinic and prognosis of PRCC. (A‐E) The proportion of M2 macrophage differed between M0 and M1 (p = 0.044), N0 and N1 (p < 0.0001), alive and dead (p = 0.0001), T1 and T3 (p < 0.0001), and stage I and stage III (p < 0.0001). (F‐I) The proportion of M1 macrophages differed between N1 and N0 (p = 0.035), T3 and T2 (p = 0.0004), dead and alive (p = 0.0009), and stage III and stage I (p = 0.0004). (J‐N) Kaplan–Meier analysis indicated that resting memory CD4T cells (p = 0.019) and M1 macrophages (p < 0.001) were associated with favorable survival status while activated memory CD4T cells (p < 0.001), Tfh cells (p = 0.011), and M2 macrophages (p < 0.001) were linked with poor survival status. PRCC, papillary renal cell carcinoma

FIGURE 8

Based on (A, B) lasso regression, four immune cells, resting memory CD4T cells, M1 macrophages, M2 macrophages, and activated dendritic cells were included in (C) the multivariate Cox risk regression model. (D) Kaplan–Meier analysis based on the Cox regression model exhibited prognostic capability (p < 0.001). (E) Nomogram based on the Cox regression model for outcome prediction of PRCC patients. (F) ROC and (G) calibration curves suggested excellent accuracy and discrimination of the nomogram. The area under curve of 1‐year, 3‐year, and 5‐year survival is 0.877, 0.841, and 0.775. (H) A heatmap visualized the infiltration of four tumor‐related immune cells in the low‐risk score and high‐risk score groups. PRCC, papillary renal cell carcinoma; ROC, receiver operating characteristic

FIGURE 9

Co‐expression analysis of immune cells and key genes by Pearson's correlation coefficient. (A) The co‐expression heatmap of immune cells and key genes. (B) M1 macrophages were negatively associated with (B) COL1A1 (R = −0.27, p < 0.001) and (C) LDLR (R = −0.29, p < 0.001). M2 macrophages were positively associated with (D) COL1A1 (R = 0.32, p < 0.001), (E) TCF4 (R = 0.27, p < 0.001), and (F) H19 (R = 0.13, p = 0.032). (G) Resting memory CD4T cells and H19 were negatively correlated (R = −0.25, p < 0.001)