Abstract Purpose: Merkel cell carcinoma (MCC) is an aggressive neuroendocrine skin cancer, which can be effectively controlled by immunotherapy with PD-1/PD-L1 checkpoint inhibitors. However, a significant proportion of patients are characterized by primary therapy resistance. Predictive biomarkers for response to immunotherapy are lacking. Experimental Design: We applied Bayesian inference analyses on 41 patients with MCC testing various clinical and biomolecular characteristics to predict treatment response. Further, we performed a comprehensive analysis of tumor tissue–based immunologic parameters including multiplexed immunofluorescence for T-cell activation and differentiation markers, expression of immune-related genes and T-cell receptor (TCR) repertoire analyses in 18 patients, seven objective responders, and 11 nonresponders. Results: Bayesian inference analyses demonstrated that among currently discussed biomarkers only unimpaired overall performance status and absence of immunosuppression were associated with response to therapy. However, in responders, a predominance of central memory T cells and expression of genes associated with lymphocyte attraction and activation was evident. In addition, TCR repertoire usage of tumor-infiltrating lymphocytes (TILs) demonstrated low T-cell clonality, but high TCR diversity in responding patients. In nonresponders, terminally differentiated effector T cells with a constrained TCR repertoire prevailed. Sequential analyses of tumor tissue obtained during immunotherapy revealed a more pronounced and diverse clonal expansion of TILs in responders indicating an impaired proliferative capacity among TILs of nonresponders upon checkpoint blockade. Conclusions: Our explorative study identified new tumor tissue–based molecular characteristics associated with response to anti–PD-1/PD-L1 therapy in MCC. These observations warrant further investigations in larger patient cohorts to confirm their potential value as predictive markers.