Risk Classification of Bladder Cancer by Gene Expression and Molecular Subtype

Our study used NanoString technology, a high-throughput platform measuring gene expression at the mRNA level to identify a set of genes predictive of clinical outcomes in bladder cancer patients. Twenty-seven differentially expressed genes were correlated with clinicopathological variables including molecular subtypes (luminal, basal, null/double-negative), histological subtypes (conventional urothelial carcinoma or carcinoma with variant histology), clinical subtype (NMIBC and MIBC), tumor stage category (Ta, T1 and T2-4), tumor grade, PD-L1 expression (high vs. low expression), and clinical risk categories (low, intermediate, high, and very high). Then, two risk models integrating the molecular subtypes and the level of expression of TP53, CCND1 and MKI67 were developed. These models provided a score ranging from 0 (best prognosis) to 7 (worst prognosis) that could be used to predict patient' outcome and guide treatment decisions in bladder cancer.Abstract: This study evaluated a panel including the molecular taxonomy subtype and the expression of 27 genes as a diagnostic tool to stratify bladder cancer patients at risk of aggressive behavior, using a well-characterized series of non-muscle invasive bladder cancer (NMIBC) as well as muscle-invasive bladder cancer (MIBC). The study was conducted using the novel NanoString nCounter gene expression analysis. This technology allowed us to identify the molecular subtype and to analyze the gene expression of 27 bladder-cancer-related genes selected through a recent literature search. The differential gene expression was correlated with clinicopathological variables, such as the molecular subtypes (luminal, basal, null/double negative), histological subtype (conventional urothelial carcinoma, or carcinoma with variant histology), clinical subtype (NMIBC and MIBC), tumor stage category (Ta, T1, and T2-4), tumor grade, PD-L1 expression (high vs. low expression), and clinical risk categories (low, intermediate, high and very high). The multivariate analysis of the 19 genes significant for cancer-specific survival in our cohort study series identified TP53 (p = 0.0001), CCND1 (p = 0.0001), MKI67 (p < 0.0001), and molecular subtype (p = 0.005) as independent predictors. A scoring system based on the molecular subtype and the gene expression signature of TP53, CCND1, or MKI67 was used for risk assessment. A score ranging from 0 (best prognosis) to 7 (worst prognosis) was obtained and used to stratify our patients into two (low [score 0-2] vs. high [score 3-7], model A) or three (low [score 0-2] vs. intermediate [score 3-4] vs. high [score 5-7], model B) risk categories with different survival characteristics. Mean cancer-specific survival was longer (122 + 2.7 months) in low-risk than intermediate-risk (79.4 + 9.4 months) or high-risk (6.2 + 0.9 months) categories (p < 0.0001; model A); and was longer (122 + 2.7 months) in low-risk than high-risk (58 + 8.3 months) (p < 0.0001; model B). In conclusion, the molecular risk assessment model, as reported here, might be used better to select the appropriate management for patients with bladder cancer.

Automated summary

This study used NanoString technology to identify a set of genes predictive of clinical outcomes in bladder cancer patients. Two risk models were developed based on the molecular subtypes and expression levels of TP53, CCND1, and MKI67, which could be used to predict patient outcomes and guide treatment decisions.