Release of immunoreactive negative regulatory factors such as immune checkpoint limits antitumor responses. PD-L1 as a significant immunosuppressive factor has been involved in resistance to therapies such as chemotherapy and target therapy in various cancers. Via interacting with PD-1, PD-L1 can regulate other factors or lead to immune evasion of cancer cells. Besides, immune checkpoint blockade targeting PD-1/PD-L1 has promising therapeutic efficacy in the different tumors, but a significant percentage of patients cannot benefit from this therapy due to primary and acquired resistance during treatment. In this review, we described the utility of PD-L1 expression levels for predicting poor prognosis in some tumors and present evidence for a role of PD-L1 in resistance to therapies through PD-1/PD-L1 pathway and other correlating signaling pathways. Afterwards, we elaborate the key mechanisms underlying resistance to PD-1/PD-L1 blockade in cancer immunotherapy. Furthermore, promising combination of therapeutic strategies for patients resistant to PD-1/PD-L1 blockade therapy or other therapies associated with PD-L1 expression was also summarized.

Background. Ovarian cancer (OC) is a malignancy exhibiting high mortality in female tumors. Glycosylation is a posttranslational modification of proteins but research has failed to demonstrate a systematic link between glycosylation‐related signatures and tumor environment of OC. Purpose. This study is aimed at developing a novel model with glycosylation‐related messenger RNAs (GRmRNAs) to predict the prognosis and immune function in OC patients. Methods. The transcriptional profiles and clinical phenotypes of OC patients were collected from the Gene Expression Omnibus and The Cancer Genome Atlas databases. A weighted gene coexpression network analysis and machine learning were performed to find the optimal survival‐related GRmRNAs. Least absolute shrinkage and selection operator regression (LASSO) and Cox regression were carried out to calculate the coefficients of each GRmRNA and compute the risk score of each patient as well as develop a prognostic model. A nomogram model was constructed, and several algorithms were used to investigate the relationship between risk subtypes and immune‐infiltrating levels. Results. A total of four signatures (ALG8, DCTN4, DCTN6, and UBB) were determined to calculate the risk scores, classifying patients into the high‐and low‐risk groups. High‐risk patients exhibited significantly poorer survival outcomes, and the established nomogram model had a promising prediction for OC patients’ prognosis. Tumor purity and tumor mutation burden were negatively correlated with risk scores. In addition, risk scores held statistical associations with pathway signatures such as Wnt, Hippo, and reactive oxygen species, and nonsynonymous mutation counts. Conclusion. The currently established risk scores based on GRmRNAs can accurately predict the prognosis, the immune microenvironment, and the immunotherapeutic efficacy of OC patients.