Identification of a Novel Tumor Microenvironment Prognostic Signature for Bladder Urothelial Carcinoma

Abstract

BackgroundThe tumor microenvironment (TME) regulates the proliferation and metastasis of solid tumors and the effectiveness of immunotherapy against them. We investigated the prognostic role of TME-related genes based on transcriptomic data of bladder urothelial carcinoma (BLCA) and formulated a prediction model of TME-related signatures.MethodsMolecular subtypes were identified using the non-negative matrix factorization (NMF) algorithm based on TME-related genes from the TCGA database. TME-related genes with prognostic significance were screened with univariate Cox regression analysis and lasso regression. Nomogram was developed based on risk genes. Receiver operating characteristic (ROC) curve and decision curve analysis (DCA) were used for inner and outer validation of the model. Risk scores (RS) of patients were calculated and divided into high-risk group (HRG) and low-risk group (LRG) to compare the differences in clinical characteristics and PD-L1 treatment responsiveness between HRG and LRG.ResultsWe identified two molecular subtypes (C1 and C2) according to the NMF algorithm. There were significant differences in overall survival (OS) (p<0.05), progression-free survival (PFS) (p<0.05), and immune cell infiltration between the two subtypes. A total of eight TME-associated genes (CABP4, ZNF432, BLOC1S3, CXCL11, ANO9, OAS1, FBN2, CEMIP) with independent prognostic significance were screened to build prognostic risk models. Age (p<0.001), grade (p<0.001), and RS (p<0.001) were independent predictors of survival in BLCA patients. The developed RS nomogram was able to predict the prognosis of BLCA patients at 1, 3, and 5 years more potentially than the models of other investigators according to ROC and DCA. RS showed significantly higher values (p = 0.047) in patients with stable disease (SD)/progressive disease (PD) compared to patients with complete response (CR)/partial response (PR).ConclusionsWe successfully clustered and constructed predictive models for TME-associated genes and helped guide immunotherapy strategies.