The incidence of kidney cancer among which renal cell carcinoma accounts for 90% worldwide occupies a portion of 2.4%, accounting to 338,000 new cases and 144,000 deaths each year in total(1). ccRCC is a complex tumor with different clinical and pathological features, genetic variation, DNA methylation profiles, and RNA and proteomic signatures, which are closely related to the prognosis of ccRCC patients(27). Nevertheless, TNM staging system, the most used risk assessment system for ccRCC patients, failed to take these genomic variation of ccRCC in to consideration which makes it not perfect for accurately predicting prognosis of ccRCC patients(28).
The novel hypothesis of gene expression regulation has been confirmed that each RNA combining with the same MREs could interact with or compete each other, which could help to further identify the mechanism of different kind of diseases especially cancer. The disturbance of the equipoise of ceRNA network was of vital importance for tumorigenesis. In gallbladder cancer, the lncRNA PVT1 which was positively related to malignancies and worse overall survival time was up-regulated in gallbladder cells. PVT1 and HK2 act as a ceRNA of miR-143 which could regulate aerobic glucose metabolism in GBC cells, promoting cell proliferation and metastasis (29). PTAR acts as a ceRNA of miR-101 which promotes tumorigenicity and metastasis in vivo in Ovarian cancer(30). LncRNA DANCR functions as a ceRNA in osteosarcoma which could promote cell proliferation and metastasis(31). And MT1JP, a down-regulated lncRNA in gastric cancer, was associated with malignant tumor phenotypes and survival of gastric cancer. MT1JP which severs as a ceRNA regulating FBXW7 expression could influence the progression of gastric cancer(32). Thus, ceRNA network containing crucial biomarkers was of vital importance in tumorigenesis.
Importantly, lncRNA-miRNA-mRNA dys-regulated ceRNA network displayed vital role in predicting disease prognosis. For example, in pancreatic cancer, 11 lncRNAs have been found and validated to function good in predicting prognosis(33). In the study of Soft tissue sarcoma (STS), seven genes (LPP-AS2, MUC1, GAB2, hsa-let-7i-5p, hsa-let-7f-5p, hsa-miR-101-3p and hsa-miR-1226-3p) in a recurrent STS-specific ceRNA network associated with recurrence and survival was identified based on the TCGA database containing 259 primary sarcomas and 3 local recurrence samples(34). In Hepatocellular carcinoma (HCC), MCM3AP-AS1 functioned as a ceRNA of miR-194-5p was a novel oncogenic lncRNA, which indicated poor clinical outcomes in patients with HCC, MCM3AP-AS1 could be a potential prognostic biomarker and therapeutic target for HCC(35). In glioblastoma multiforme, lung cancer, ovarian cancer and prostate cancer, based on the networks, the target mRNAs are normally up-regulated by the sponge lncRNAs after being released from miRNA control, and only a fraction of sponge lncRNA-mRNA regulatory relationships and hub lncRNAs are shared by the four cancers. Moreover, most sponge lncRNA-mRNA regulatory relationships show a rewired mode between different cancers, suggesting that different cancers had varied ceRNA networks(36). Although there are many studies on ceRNA networks in numerous cancers; however, few of them are on ccRCC. In addition, the sample sizes have not been large enough and most of them only focus on only lncRNAs.
In this study, the trait correlated and differential expression mRNAs and lncRNAs in ccRCC were identified by WGCNA and DESeq2 including data from TCGA and GTEx, we combined the differential expression genes with trait correlated genes and obtained ccRCC-specific genes including 2191 mRNAs and 1377 lncRNAs, followed by a construction of ceRNA network including 363 lncRNAs, 3 miRNAs and 7 mRNAs by MiRcode, StarBase, miRTarBase and TargetScan databases. To promote explore the relationships with prognosis of these 370 genes (mRNA and lncRNA in the ceRNA network), a gene signature with eight genes, namely MPP5, WT1-AS, AC114316.1, AC103719.1, AL162377.1, HS1BP3-IT1, LINC02657, AC015909.1, was constructed by univariate cox proportional hazard regression, lasso and multivariate cox proportional hazard regression analysis. we also ensued the discriminations and accuracy of the gene signature with C-index and time-dependent ROC curve which all suggesting that the eight genes in the model could act as biomarkers based on the patients’ prognosis.
Among the eight genes in the gene signature, the exclusive mRNA MPP5 which is associated with the membrane-associated guanylate kinase family helping the construction of cell polarity had been validated to be related to maintenance of cell polarity, invasion, cell division in prostate cancer(37), meanwhile, disruption of apical protein MPP5 which could negatively regulate YAP/TAZ abundance and activity might promote enrichment of oncogenic YAP and TAZ in HCC(38). MPP5 whose loss is a hallmark of cancer is crucial for tissue organization corresponding to the down-regulated expression in ccRCC. Long noncoding RNA WT1-AS which functioned as a potential tumor suppressor is related to poor survival in cervical squamous cell carcinoma(39)(40)(41) and triple-negative breast cancer (TNBC)(42). For lncRNA HS1BP3-IT1, it may be a prognosis biomarker for cholangiocarcinoma(43), laryngeal cancer(44) respectively. Therefore, our prediction of the ceRNA network had great confirmation of previous studies, meanwhile, by conducting Kaplan-Meier survival curves analysis for our gene signature, we confirmed that the high-risk group displayed a worse overall survival compared with low-risk group and Time-dependent ROC curve of our gene signature displayed a good performance in all group.
Nomograms are widely used as prognostic tools in oncology and medicine. By including various prognosis-associated variables and generating survival probability nomogram can help clinicians make a better treatment decision.(45). In present study, by including gene signature based on the ceRNA network and other related clinical characteristics into a stepwise cox model, a concise nomogram for prognostic prediction of ccRCC patients based on the eight-gene signature, ages, N stage, and AJCC stage was established, of course, an accurate estimation of the nomogram is validated by C-index and calibration curve with bootstrap methods which suggested its perfect discriminations and calibrations.
This study identified a eight-gene signature on the basis of dys-regulated ceRNA network, which will help us better understand dys-regulated ceRNA network mediated ccRCC and suggest therapeutics to ccRCC. Consequently, a nomogram with a good accuracy (C-index 0.81; 95% CI 0.70 ~ 0.89) based on genes in the dys-regulated ceRNA network in ccRCC is developed to predict prognosis of ccRCC patients compared with Jiang’s nomogram (C-index 0.79;95% CI 0.75–0.82)(46). Although few novel prognosis-prediction systems based on gene expression have been developed for ccRCC patients, each previous study mainly focused on lncRNA molecular where we don’t know whether there exist a regulation function in ccRCC, such as Jiang’s (46) and Zhang’s (47) nomogram, it’s necessary to develop a comprehensive analysis to prognostic prediction on the basis of ceRNA network.
However, there are still some limitations in our study. First, although the ceRNA network was constructed based on TCGA and GTEX cohort, the development of the prognostic nomogram only using a cohort got from TCGA, and there were not a external validated dataset. Second, we ignored some other potential clinical characteristics. Although these limitations, the nomogram based on a comprehensive dys-regulated ceRNA network analysis in ccRCC can promote clinicians to make a better and accurate treatment decision.
In conclusion, a dys-regulated ceRNA network based on ccRCC-specific genes was constructed followed by a development of nomogram predicting 1-, 3- and 5-year OS of ccRCC patient. The nomogram included both some clinical characteristics and ccRCC-specific gene signatures, so it will help clinicians make a better and accurate treatment decision more.