Glutathione dynamics is a potential diagnostic and therapeutic trait determining the neoadjuvant chemotherapy response to bladder cancer

Radical cystectomy with preoperative cisplatin-based neoadjuvant chemotherapy (NAC) is a current standard care for muscle-invasive bladder cancers (MIBCs). However, the response rate remains relatively low, and current clinico-pathologic and molecular classifications are inadequate to prospectively stratify MBIC patients regarding the NAC response. Here, we showed that dysregulation in glutathione (GSH) pathway is fundamental for the NAC resistance of MBICs. Comprehensive analysis of transcriptome profiles from four independent cohorts revealed that GSH metabolism and immune response genes were significantly enriched in NAC resistant and sensitive MIBC tumors, respectively. A machine learning based tumor/stroma classifier was applied for high-throughput digitalized immunohistochemistry analysis, identifying GSH dynamic proteins including glutaminase-1 (GLS1) were associated with the NAC resistance and unfavorable clinical characteristics. Consistently, GSH dynamics was activated in cisplatin resistant human MIBC cells, stimulating their proliferation, stemness features, and invasion. Combination treatment of GSH dynamics modulators and cisplatin significantly suppressed tumor growth in an orthotopic xenograft animal model. Our results demonstrate the prognostic and therapeutic values of GSH dynamic mediators in determining the clinical features and response to NAC of patients with MIBCs. Overall design: This ASAN cohort included 105 MIBC (cT2-4aN0M0) patients who underwent TUR and NAC at least 3 cycles were retrospectively collected. The ASAN cohort was classified into two groups showing marked contrast in NAC response: NAC response (R) patients with the absence of residual invasive tumor and NAC nonresponse (NR) patients with large residual MIBC (4.9 ± 2.5 cm) in the radical/partial cystectomy specimens. To investigate a tumor specific molecular nature determining NAC response, we performed laser capture microdissection (LCMD) of tumors from the formalin-fixed and paraffin-embedded (FFPE) TURBT specimens of 39 patients (N=18 in NR group and N=21 in R group) from ASAN cohort and then compared their genome-wide transcriptome profiles.