Identification of Kukoamine B from Lycii Radicis Cortex as a PD-L1 inhibitor against lung cancer

Lycii Radicis Cortex (LRC), a medicinal and edible resource, has shown anti-lung cancer activity by disrupting the PD-1/PD-L1 interaction and reactivating T cells, but its active constituent remains unclear. Herein, we trained five machine learning models to predict the IC₅₀ values for PD-L1 inhibitors. Among them, the random forest regression algorithm showed superior performance and was used to prioritize LRC-derived candidates for experimental validation. Homogeneous time-resolved fluorescence (HTRF) and surface plasmon resonance (SPR) assays confirmed that KB binds to PD-L1 with a dissociation constant (Kd) of 18 μM and inhibits the PD-1/PD-L1 interaction (IC₅₀ = 1.687 μM). Molecular docking, molecular dynamics simulations, and co-immunoprecipitation analyses revealed that KB forms stable interactions with key PD-L1 residues. In co-culture assays, KB markedly reduces PD-L1-induced T cell apoptosis and enhances T cell cytotoxicity against lung cancer cells. In a lung carcinoma mouse model, KB significantly inhibited tumor growth with a favorable safety profile. Single-cell RNA sequencing and flow cytometry further indicated that KB remodels the tumor immune microenvironment and potentiates CD4 + and CD8 + T cells effector functions. These findings identify KB as an LRC-derived PD-L1 inhibitor and support AI-assisted prioritization of dietary immunomodulators from natural products.