Tracking minimal residual disease with urine tumor DNA in muscle-invasive bladder cancer after neoadjuvant chemotherapy.

e16514 Background: Standard-of-care for muscle-invasive bladder cancer (MIBC) consists of neoadjuvant chemotherapy (NAC) followed by radical cystectomy. The inability to noninvasively assess minimal residual disease (MRD) after NAC limits our ability to offer bladder-sparing treatment. We perform urine tumor DNA (utDNA) analysis to identify pathologic complete response (pCR) at the time of cystectomy in patients receiving NAC. Methods: We applied CAPP-Seq to urine cell-free DNA samples acquired on the day of radical cystectomy from 19 MIBC patients treated with NAC. utDNA variant-calling was performed without prior tumor mutational knowledge using a panel of 49 consensus driver genes mutated in MIBC. The utDNA level for each patient was represented by the duplex-supported non-silent driver mutation with the highest variant allele fraction (vAF) after removing germline variants. We also serially tracked utDNA variants in two patients before, during, and after NAC. Results: Comparing patients with residual disease detected in their cystectomy specimen ( n = 10) to those who achieved a pCR ( n = 9), median utDNA levels were 2.4% vs. 0%, respectively ( P = 0.006). Using an optimal utDNA threshold to define MRD detection, positive utDNA MRD was highly correlated with the absence of pCR ( P = 0.003). Analysis of two patients’ serial urine samples revealed utDNA dynamics that were consistent with treatment responses in real-time. In one patient who ultimately achieved a pCR, four non-silent driver mutations were detectable pre-NAC, including ERCC2 N238S (7.8% vAF) associated with increased chemosensitivity. One week after starting NAC, ERCC2 N238S increased by 1.6-fold in urine, as did PIK3CA E726K which increased by 8.4-fold. Four weeks post-NAC, however, all mutations previously detected in this patient’s urine became undetectable, consistent with the patient’s pCR and long-term disease-free survival. Conversely, another patient harbored two non-silent driver mutations in PLEKHS1 (1.9% vAF) and KMT2D (4.9% vAF) pre-NAC. One week after starting NAC, both mutations decreased dramatically by 8.0- and 4.3-fold, respectively. By three weeks post-NAC, however, these mutations progressively increased by 5.2-fold on average, which correlated with a lack of pCR as well as post-treatment disease progression. Two newly detected non-silent driver mutations in ARID1A and ERBB2 also emerged on NAC and persisted following completion of chemotherapy , likely reflecting the development of treatment resistance. Conclusions: utDNA MRD after NAC but before radical cystectomy for MIBC correlated significantly with pathologic response, which could help personalize patient selection for bladder-sparing treatments in the future. Serial monitoring of utDNA variants during NAC can reveal dynamic mutational changes that reflect real-time treatment responses as well as ultimate disease-free survival.