Positional scanning and computational modeling reveal determinants of legumain transpeptidase activity

Legumain is a lysosomal cysteine protease that plays a central role in antigen processing for MHC class II presentation. Beyond its canonical proteolytic activity, legumain can also function as a peptide ligase or transpeptidase. This non-canonical activity becomes particularly relevant under pathophysiological conditions such as cancer and neurodegeneration, where legumain translocates to near-neutral pH environments, including the cytosol, nucleus, and extracellular space, that favor ligation over hydrolysis. Here, we combine in vitro positional scanning with in silico substrate profiling to elucidate the molecular determinants governing human legumain-mediated peptide cyclization. We identify glycine residues at the P1' and P1" positions and basic residues at P2'/P2″ as key drivers of efficient cyclization. Guided by these insights, we designed an optimized substrate exhibiting substantially enhanced cyclization efficiency. Computational analysis not only recapitulated the experimental observations but also predicted a covalent inhibition mechanism involving a P1' cysteine, revealed a k_cat-tuning switch embedded within the substrate, and highlighted its potential for developing high-performance fluorogenic substrates. Together, these findings establish the structural basis for the rational design of activity-selective legumain probes and inhibitors, with potential applications in therapeutic development and diagnosis.