Questions

Urothelial cancer (UC) is a challenging disease with a wide tumor-biological spectrum. Most molecular classifications cover only muscle-invasive bladder cancer and are transcriptome-based, relating only indirectly to the therapeutically relevant and more stable protein level. Also, the advent of new targeted therapies requires quantitative data about the tumor specificity of their target proteins. We turned to the proteome to address these questions.

Methods

We performed deep proteomic profiling of a comprehensive cohort by optimized tandem mass tag-labelled liquid chromatography-coupled tandem mass spectrometry. Data acquisition was validated internally with immunoblotting and externally by bioinformatic reclassification within existing, filtered transcriptomic data. After bioinformatics, the top cluster defining proteins were quantified immunohistochemically under real-world conditions. Protein profiles were individualized and separately evaluated with drug repurposing libraries. Cell viability assays were performed for a panel of twelve UC cell lines to validate these predictions in vitro.

Results

We analyzed 434 samples with 242 tumors and 192 paired normal mucosae, covering all stages of UC from pTa to >pT2. 9542 proteins were quantified and revealed five distinct proteomic subtypes. These were validated internally and externally, showing relevant survival stratification also in the TCGA dataset. The proteomic subtypes were independent from pathological groups with relevant stratification of progression- free and overall survival (low vs. high-risk: median 103 vs. 27 months). Tumor specificity of all proteins was highly heterogeneous across stages and subtypes. As an example, the therapeutic target NECTIN4 was generally overexpressed only in non-muscle-invasive UC. Drug repurposing revealed several new candidate drugs, each specific to different proteomic subtypes. In vitro data showed increased sensitivity by subtype in line with four out of seven representative predictions.

Conclusion

Proteomic subtypes add independent prognostic information and carry predictive value for several newly identified adjuvant drug candidates. The tumor specificity of biomarkers and drug targets is highly dependent on stage and subtype and calls for quantitative patient-specific testing.