Introduction: Molecular subtyping of bladder cancer has revealed luminal tumors generally have a more favorable prognosis. However, some aggressive forms of variant histology, including micropapillary, are often classified luminal. In previous work, we found long non-coding RNA (lncRNA) expression profiles could identify a subgroup of luminal bladder tumors with less aggressive biology and better outcomes. In the present study, we aimed to investigate whether lncRNA expression profiles could identify high-grade T1 micropapillary bladder cancer with differential outcome.
Methods: LncRNAs were quantified from RNA-seq data from a HG T1 bladder cancer cohort that was enriched for primary micropapillary cases (15/84). Unsupervised consensus clustering of variant lncRNAs identified a three-cluster solution, which was characterized using a panel of micropapillary-associated biomarkers, molecular subtypes, gene signatures and survival analysis. A single-sample genomic signature was trained using lasso-penalized logistic regression to classify micropapillary-like gene-expression, as characterized by lncRNA clustering. The genomic classifier (GC) was tested on luminal tumors derived from the TCGA cohort (N=202). Primary endpoints were overall, progression-free and high-grade recurrence-free survival.
Results: Primary micropapillary HGT1 showed decreased FGFR3, SHH and p53 pathway activity relative to tumors with conventional urothelial carcinoma. Many bladder cancer-associated lncRNAs were downregulated in micropapillary tumors, including UCA1, LINC00152, and MALAT1. Unsupervised consensus clustering resulted in a lncRNA cluster (LC1) with worse prognosis that was enriched for primary micropapillary histology and the Luminal Unstable (LumU) molecular subtype. Interestingly, LC1 appeared to better identify aggressive HG T1 disease, compared to stratifying outcomes using primary histologic characteristics. A signature trained to identify LC1 cases showed good performance in the testing cohort, identifying seven cases with significantly worse survival (p < 0.001).
Conclusions: Using the lncRNA transcriptome we identified a subgroup of aggressive HGT1 bladder cancer that was enriched with micropapillary histology. These data suggest that lncRNAs can facilitate the identification of aggressive micropapillary-like tumors, potentially improving patient management.