Antibody blockade of programmed death-1 (PD-1) or its ligand, PD-L1, has led to unprecedented therapeutic responses in certain tumor-bearing individuals, but PD-L1 expression's prognostic value in stratifying cancer patients for such treatment remains unclear. Reports conflict on the significance of correlations between PD-L1 on tumor cells and positive clinical outcomes to PD-1/PD-L1 blockade. We investigated this issue using genomically related, clonal subsets from the same methylcholanthrene-induced sarcoma: a highly immunogenic subset that is spontaneously eliminated in vivo by adaptive immunity and a less immunogenic subset that forms tumors in immunocompetent mice, but is sensitive to PD-1/PD-L1 blockade therapy. Using CRISPR/Cas9-induced loss-of-function approaches and overexpression gain-of-function techniques, we confirmed that PD-L1 on tumor cells is key to promoting tumor escape. In addition, the capacity of PD-L1 to suppress antitumor responses was inversely proportional to tumor cell antigenicity. PD-L1 expression on host cells, particularly tumorassociated macrophages (TAM), was also important for tumor immune escape. We demonstrated that induction of PD-L1 on tumor cells was IFNg-dependent and transient, but PD-L1 induction on TAMs was of greater magnitude, only partially IFNg dependent, and was stable over time. Thus, PD-L1 expression on either tumor cells or host immune cells could lead to tumor escape fromimmune control, indicating that total PD-L1 expression in the immediate tumor microenvironment may represent a more accurate biomarker for predicting response to PD-1/PD-L1 blockade therapy, compared with monitoring PD-L1 expression on tumor cells alone.