Abstract 2190: Integrative genomic and transcriptomic characterization of metastatic urothelial carcinoma

Metastatic urothelial carcinoma (mUC) is a lethal cancer with limited therapeutic options available. To identify novel targets for therapy, large-scale sequencing efforts are needed. The Cancer Genome Atlas (TCGA) initiative substantially improved our knowledge on the genomic and transcriptomic characteristics of primary UC (Robertson et al. Cell 2017), however, the molecular landscape of mUC still remains largely unexplored. We performed whole genome sequencing (WGS) on 116 biopsies of UC metastases and matched mRNA sequencing (RNAseq) for 90 samples. We applied genomic alteration analysis on the WGS data and consensus clustering on the RNAseq data. We compared the molecular alterations of mUC with the data of the TCGA cohort and found that most affected driver genes were similar between primary UC and mUC. However, we identified CNTNAP5, RARG and MGP as exclusively mutated driver genes in the metastatic setting, and TP53 alterations were more prevalent in mUC than in primary UC. APOBEC induced mutation signatures in coding and non-coding regions were identified in 91% of the mUC samples and correlated with ploidy, mutational burden and copy number alterations. Based on the etiology of observed mutational signatures, five genomic subtypes were identified in mUC and validated in the TCGA cohort. APOBEC mutagenesis was the main mutational process in genomic subtype 1, which was the most common subtype (67%, N = 78). Moreover, APOBEC deamination mutations were found to be increased over time in paired biopsies from eight patients, indicating that APOBEC enzyme activity is ongoing in the metastatic setting. Unsupervised clustering and comparing phenotypic markers between the groups revealed five clinically-relevant transcriptomic subtypes: luminal-a (20%), luminal-b (20%), stroma-rich (24%), basal/squamous (23%) and a non-specified phenotype (12%). The basal/squamous and stroma-rich subtypes showed highly similar molecular characteristics as the TCGA subtypes, but with enrichment of the stroma-rich and reduction of the basal/squamous subtype. Based on genomic alterations, potential therapeutic targets were identified in 111/116 mUC patients, of which FGFR3 alterations (18%) and fusions (6%), and CDKN2A (44%), ERBB2 (20%), and TSC1 (14%) alterations were most common. By integrating the genomic and transcriptomic data, we propose potential novel therapeutic options per transcriptomic subtype, like addition of a TGF-β inhibitor to immune checkpoint inhibition fo