Identification of highly specific bacterial L-asparaginase through genome-wide computational screening and experimental validation

Bacterial L-asparaginases are key therapeutics for acute lymphoblastic leukemia (ALL), but their clinical efficacy is limited by off-target L-glutaminase co-activity, which contributes to severe side effects. This study aimed to identify highly specific bacterial L-asparaginases with minimal L-glutaminase co-activity.

We screened ~ 85,000 prokaryotic genomes using a workflow integrating AlphaFold2 structure prediction, molecular docking, binding-energy estimation, and molecular dynamics simulations. Six top-ranked enzymes were selected for experimental validation.

RhAII protein from Acidobacteriota sp. showed high specificity toward L-asparagine (L-Asn), with a Michaelis constant (K_M) of 12 ± 2 µM and turnover number (k_cat) of 34 ± 2 s⁻¹, and negligible L-glutaminase activity. Molecular dynamics (MD) simulations revealed a relationship between enzyme activity and the conformational dynamics of the flexible gating element (FGE), characterized by residue-level root mean square fluctuation (RMSF), and L-Asn positional fluctuations, evaluated by ligand root mean square deviation (RMSD) relative to the protein. High L-asparaginase activity was associated with optimal FGE dynamics and low L-Asn RMSD, consistent with stable substrate positioning within the active site. In contrast, enzymes with altered FGE dynamics relative to the reference EcAII enzyme from Escherichia coli, such as PoAII from Planctomycetota sp., showed poor L-asparaginase performance (K_M 5797 ± 1143 µM; k_cat 1.2 ± 0.1 s⁻¹).

Genome- and structure-guided screening identified RhAII as a promising antileukemic candidate for further validation in biological models. A key limitation of the computational workflow was the lack of correlation between predicted substrate binding energies and catalytic efficiency, highlighting the importance of active-site dynamics in enzyme kinetics.