Background: Large-scale efforts have sought to define the genomic landscape of urothelial carcinomas of the bladder with the goal of identifying novel therapeutic targets. Many bladder cancers harbor mutations in genes that regulate chromatin state, such as KDM6A, ARID1A and KMT2D. There remains uncertainty as to the timing at which alterations in chromatin modifying genes (CMG) arise during the evolution of urothelial cancer. Methods: We leveraged a prospective genomic profiling initiative to characterize driver alterations in urothelial cancer. To define the timing at which mutations arose during disease pathogenesis, we performed whole exome (WES) or targeted sequencing analysis of matched pairs of primary and metastatic tumors. Results: CMG mutations were identified in 76% of UC samples in the prospective MSK-IMPACT tumor sequencing cohort (N=1057); most frequent in KDM6A (31%), KMT2D (28%) and ARID1A (27%). Comparison of primary low-grade (N=65), high-grade (N=742) and metastatic (N=250) samples revealed KDM6A, FGFR3 and KMT2C were more frequently mutated in low-grade tumors. WES and targeted sequencing of 63 matched pairs demonstrate a high degree of concordance among likely oncogenic mutations. Notably, 100% (18/18) of pairs were concordant for mutations in KDM6A suggesting that, when present KDM6A mutations arise early during tumor development. Mutations in ARID1A were present only in the metastatic samples of a subset of tumors (29%) suggesting that ARID1A mutations arise later in disease pathogenesis. Several additional oncogenic drivers were found in only the metastatic sample (FGFR3, TSC1, PIK3CA) of select tumor pairs. Conclusions: Primary and metastatic pairs were largely concordant for known and likely oncogenic mutations. Evolutionary analysis of sequencing data suggested that KDM6A mutations arise early in tumor development whereas ARID1A mutations were enriched and present only in the metastatic samples of a subset of patients. Despite the high frequency of FGFR3 mutations in low grade tumors, discordance of FGFR3 mutational status was observed in small fraction of patients suggesting that profiling of metastatic tumors or cell free DNA would be preferable to identify patients for FGFR inhibitor therapy.