In silico discovery of pathogenic PD-L1 nsSNVs with altered glycosylation and immunotherapy binding

Immune checkpoint inhibitors (ICIs), particularly anti-PD-L1 monoclonal antibodies, block extracellular interactions between programmed death ligand-1 (PD-L1) and programmed cell death protein-1 (PD-1) to enhance antitumour immunity. Here, we present a streamlined workflow integrating public databases, bioinformatics tools, and in silico molecular dynamics (MD) simulations and in vitro experimental assays to identify consequential pathogenic nonsynonymous PD-L1 variants (PD-L1SNVs). Four variants predicted as pathogenic across six bioinformatic tools (SIFT, PolyPhen-2, PANTHER, SNP-PhD, SNP&GO, and Pmut) destabilized the anti-PD-L1 ICI atezolizumab-drug binding epitope in MD studies, correlating with reduced drug affinity. Live-cell assays directly linked PD-L1SNVs to intracellular trafficking defects and aberrant glycosylation, even distal from N-glycosylation sites. Allostery modeling (AlloSigMA) linked these disruptions to long-range structural perturbations. Our work establishes bioinformatics-driven predictions of functionally impactful pathogenic PD-L1SNVs and highlight how glycosylation/trafficking defects may serve as key mechanisms compromising therapeutic efficacy. These insights in predicting and validating pathogenic PD-L1SNVs has significant implications for guiding the personalized selection of effective PD-L1-targeted therapies in clinical settings. One Sentence Summary: Uncovering the implication of non-synonymous single nucleotide variants (nsSNVs) on the structure and function of Programmed Death Ligand-1.