Background: Expression-based molecular subtypes thought to be intrinsic in bladder cancer have been widely reported, carrying important potential clinical treatment implications. Histologically, bladder cancers are also heterogeneous diseases, with a large portion of urothelial carcinomas exhibiting divergent differentiation. Previous subtyping efforts have been carried out using predominantly fresh frozen tissue samples, potentially obscuring this known differentiation heterogeneity. Methods: Here we performed targeted multiplexed, amplicon-based DNA and RNA sequencing on 100 formalin-fixed paraffin-embedded (FFPE) bladder cancer samples (including 12 paired urothelial / squamous lesions). High-confidence somatic point mutations, short insertions/deletions (indels), and copy number alterations were detected using the DNA component of the Oncomine Comprehensive Assay (OCP). Targeted RNA sequencing was carried out using a custom Ampliseq panel comprised of 8 housekeeping genes and 103 target genes assessing major transcriptional programs as identified from publically available data. Results: By DNA analysis, we observe frequent TP53 (35%) and activating hotspot PIK3CA (23%) somatic mutations across the cohort, as well as targetable high-level (log-2 copy number ratio > = 1.5) focal amplifications of ERBB2 (3%) or EGFR (3%) in a subset of samples. We report a novel approach for detecting sub-gene copy-number alterations, and confirm several detectable multi-exon losses using whole transcriptome RNA sequencing. Pairing targeted RNA expression analysis with DNA-based alterations, we show high level expression of EGFR and ERBB2 in focally-amplified samples. Most importantly, we show that despite identical prioritized somatic genomic alterations, we observe divergent expression-based profiles in 3 of 12 (25%) paired urothelial and squamous samples. Conclusions: Taken together, these results highlight the importance of molecular heterogeneity in bladder cancer and suggest important considerations for using existing expression-based clustering approaches to guide clinical treatment decisions.