KIRP reallenge due to high histologic heterogeneity, poor prognosis, and limited treatment options. KIRP is the second most prevalent phenotype of RCC, however, the carcinogenesis mechanism of KIRP is not fully understood. Much of the previous research on KIRP genes has focused on some known cancer-related genes of KIRC. There is a lack of epigenetic biomarkers, and most of the prognosis of KIRP biomarker research are focused on the mRNA, lncRNA, and miRNA. For instance, Lan et al. identified seven lncRNAs that could predict the prognosis of KIRP; Luo et al. identified hsa-mir-3199-2 and hsa-mir-1293 as novel prognostic biomarkers for KIRP; and Gao et al. found that five mRNAs (CCNB2, IGF2BP3, KIF18A, PTTG1, and BUB1) can predict KIRP patient survival. Although some prognostic markers of KIRP have been found in previous studies, the results are not consistent and there is no analysis at multiple omics levels. Therefore, reliable molecular signatures are needed to predict the survival of KIRP patients.
In our study, we identified 47 methylated gene sites from nine differentially methylated genes (DMGs), with opposite differential expression patterns. We used various statistical approaches to identify a four methylated sites signature from the 47 methylated sites. The signature that we selected can separate the KIRP patients into high‑risk and low‑risk groups with significantly different survival times in the training and test datasets, indicating it has a powerful prediction ability. The independence of the selected DNA methylation genes signature in predicting OS in the entire dataset was identified using multivariable Cox regression analysis, which confirmed that the risk score of DNA methylation sites signature was independent of OS.
The ROC curve showed that the AUC is 0.791 in the training group and 0.742 in the test group. Considering that larger AUC usually indicates better prediction power, this result further demonstrated that the DNA methylation signature in our study is a high accuracy novel prognostic marker and has important clinical value. In addition, our DNA methylation sites signature did not depend on other clinical features. Moreover, we validated the methylation gene sites signature and methylation sites in the TCGA and GEO cohorts and demonstrated their ability to predict overall survival of KIRP patients. We also established a methylation gene sites nomogram including methylation gene sites signature and clinical-related risk factors (e.g. stage and age) to predict OS. So, our study results help in understanding the development KIRP and for developing tailored therapy, and ultimately may contribute to an increase in survival rates of KIRP patients.
In addition, we analyzed the function of the selected DNA methylation genes. The four methylation sites cg04448376, cg24387542, cg08548498, and cg14621323, were located in UTY, LGALS9B, SLPI, and PFN3, respectively. SLPI is a gene encoding secretory leukocyte protease inhibitor, a 11.7 KDa serine protease inhibitor, and is one of the members of the whey acidic protein four-disulfide core family[37, 38]. SLPI can reduce the activities of trypsin, neutrophil elastase, chymotrypsin and cathepsin G. Therefore, SLPI may be a potential tumor marker to predict the prognosis. Previous studies have shown that SLPI is related to tumor metastasis. In some high-risk, aggressive or metastatic tumors, such as pancreatic, uterine cervix, papillary thyroid, and ovarian cancers, SLPI was often found to be highly expressed[40–41]. However, in bladder tumors, nasopharyngeal carcinoma and some breast cancers, SLPI has a low expression. SLPI has high expression in gastric cancer cells with serosa invasion, and SLPI overexpression in gastric cancer cell lines can improve the cell migration and invasion rate. These observations are entirely consistent with the previous observation that the expression of SLPI in tumors is often associated with poor prognosis. As far as we know, the present study is the first report of SLPI methylation sites as a prognostic biomarker in human KIRP. SLPI is hypomethylated and overexpressed in this report just like the previous study demonstrated, so we believe that SPLI is a favorable biomarkers for prognosis.
UTY is located on the Y chromosome and can encode a demethylase and was reported to be an epigenetic-related gene. UTY is essential in the development of teratoma through regulation of epigenetic changes[42–43]. In urothelial bladder cancer (UBC), 22.8% (8/35) of patients were found to have a reduced UTY copy number and cell proliferation was found to increase in a UTY knockout. UTY also plays an important role in some regulatory pathways, such as the NF-B and p53 pathways[44, 45]. In this study, we showed that methylation of UTY was significantly associated with KIRP survival, and that UTY can act as a survival-related methylation biomarker for KIRP.
For LGALS9B and PFN3, there is very little known about their regulatory mechanisms. The LGALS9B gene was initially thought to represent a pseudogene of galectin 9, however, the association between it and tumors is unclear. This gene is one of two similar loci on chromosome 17p similar to galectin 9 and is now thought to be a protein-encoding gene. We have found that its functions are primarily associated with protein binding. Thus, we suspect that it is similar in function to galectin 9. Galectin-9 was reported to be related to metastasis and immunosuppression. In terms of its role in cancer, galectin-9 can affect different aspects in the process of tumor growth and aggressiveness, such as metastasis and immunomodulation[46–47]. In breast carcinoma, liver cancer and cervical tumors, LGALS9 expression affects disease prognosis[48–51]. PFN3, one of the isoforms of Profilin, is an actin binding protein. Previous studies show that PFN3 is expressed in the brain, testis, and kidney. Genetic variation of PFN3 is significantly related to nephrolithiasis of Japanese individuals. Although the functions of LGALS9B and PFN3 are not very clear, they are significantly associated with KIRP survival. Our study indicates that UTY, LGALS9B, SLPI, and PFN3 have important roles in KIRP.
The limitations of this study need to be recognized. First, the samples of our study are entirely retrospective and inherent biases may influence the results. Hence, we may have lost signatures that are potentially correlated with KIRP survival. Secondly, we have not further searched the mechanism of action of these DNA methylation genes in KIRP. Finally, although we identified the selected DNA methylation sites as a powerful prediction signature, applying it in a clinical setting will require more research. Despite these drawbacks, the significant and consistent correlation between our four methylated site signature and overall survival in two independent datasets indicated that it is a potentially powerful prognostic marker for KIRP.