Background: The value of PD-L1 to predict durable responses to immune checkpoint inhibition (ICI) in metastatic urothelial carcinoma (mUC) is inconsistent. We hypothesize that the use of archived primary tumor material (PRIM) for PD-L1 testing in clinical trials not properly reflecting the metastatic disease status (MET) contributes to this clinical issue. Objective: To analyze the predictive and prognostic value of PD-L1, spatial immunephenotypes and MHC-I determined in patient-matched PRIM/MET. Methods: PD-L1 (Ventana IC-Score, combined positivity score), spatial immunephenotypes (midi-plex digital spatial immuneprofiling) and MHC-I were examined in 154 mUC patients with at least one available pretreatment MET (138 patient-matched PRIM/MET pairs). 119 patients received first-line platinum-based chemotherapy, and 50 patients received second-line immunecheckpoint inhibition. PD-L1 expression, spatial immunephenotypes and MHC-I status of (patient-matched PRIM and) pretreatment MET were correlated to chemotherapy and ICI response and outcomes. Results: Discordance rates in patient-matched PRIM/MET amounted 25/30%, 36% and 49% for PD-L1 (CPS10/IC5%), immunephenotypes and MHC-I (loss versus preserved), respectively. Correlations with chemotherapy and ICI responses were observed for immunephenotypes and MHC-I status determined in MET (not for PD-L1 alone), but not in PRIM. In case of ICI, patients with cytotoxic tumor immune microenvironments (TIME) showed durable responses with disease control rates of 90% and a hazard ratio for disease progression/death of 0.05 (95%-CI:0.01-0.65) versus patients with immunedepleted MET (DCR 29%). MET MHC-I status added incremental value to predict durable ICI responses: Combination of MHC-I based (auto-)antigen expression of tumor cells with spatial immunepehnotypes in pre-treatment MET improved predictive and prognostic impact for response and outcome prediction of mUC patients undergoing first-line platinum-based chemotherapy and second-line immunecheckpoint inhibition. Limitations include the partly retrospective design and the lack of MET multisampling on individual patient level. Conclusions: The TIME is subject to substantial dynamics during metastatic evolution. MET immunephenotypes and MHC-I statuses show promising potential to predict chemotherapy and durable ICI responses, while the PRIM TIME does not. Thus, future clinical trials should rather rely on pre-treatment MET-biopsies reflecting the current immunological disease state than on PRIM.