Structure-guided discovery of potent and selective DGKalpha inhibitors for targeted cancer therapy

Introduction: Cancer remains a major cause of morbidity and mortality in North America. Diacylglycerol kinases (DGKs) are key regulators of lipid-mediated signaling, converting diacylglycerol (DAG) to phosphatidic acid (PA), two bioactive second messengers that orchestrate diverse cellular pathways. The DGK families comprise ten isoforms classified into five subtypes. Among these, DGKalpha-a type I isoform-plays critical roles in immune regulation, neuronal signaling, and membrane remodeling. In the context of cancer, DGKalpha enhances tumor cell proliferation and survival while suppressing cytotoxic T-cell and natural killer (NK) cell activity, thereby facilitating immune evasion. DGKalpha also promotes PD-L1 expression, further reinforcing tumor immune escape. Dysregulated DGKalpha signaling similarly contributes to immune dysfunction. To address these pathological roles, we aim to develop highly selective DGKalpha inhibitors by exploiting a distinct allosteric pocket to achieve precise and isoform-specific modulation of the enzyme. Methods: We generated high-confidence structural models of DGK isoforms including (alpha, beta) using state-of-the-art AI prediction platforms, complemented by molecular dynamics (MD) simulations and binding free-energy methodologies. Structural models of DGKalpha and DGKbeta were constructed in the presence of biologically relevant cofactors (ATP, Ca²⁺, Zn²⁺, Mg²⁺) to capture catalytically competent conformations. Each model underwent extensive validation through long-timescale MD simulations, and dominant conformational ensembles were extracted via clustering analyses. A generative-design pipeline was subsequently deployed to create and prioritize novel small-molecule inhibitors based on predicted potency, isoform selectivity, and physicochemical suitability. Top-ranked candidates were further interrogated through molecular docking, MD refinement, and free-energy calculations to characterize their binding poses. Results: The DGKalpha structural models demonstrated strong concordance with available in vitro data, supporting their suitability for downstream computational analyses. Molecular docking identified a previously uncharacterized allosteric pocket capable of accommodating the generative-AI-derived compounds. MM-PBSA analyses of MD trajectories provided binding free-energy estimates and revealed key residues that mediate ligand engagement and stabilize the inhibitor-protein complexes. Conclusion: S532, L556, H606, Y558, and F559 mediate ligand recognition, enabling elucidation of the mechanistic basis that underlies the interactions of two lead chemotypes. Several designed molecules exhibit marked isoform selectivity, displaying substantially stronger predicted affinity for DGKalpha than for DGKbeta. Together, these insights provide a structural framework for the rational development of selective DGKalpha inhibitors.