Mucin-type O-glycosylation is a ubiquitous post-translational modification that imparts proteins with unique properties crucial for cellular signaling, immune response, and numerous diseases including cancer. Among enzymes involved in O-glycan processing, the GH101 family, which encompasses the endo-α-N-acetylgalactosaminidase from Bifidobacterium longum (EngBF), are the only known enzymes capable of processing O-glycan core structures. As such, these enzymes represent a valuable resource for the investigation and manipulation of O-glycans, which is currently hindered by the limited availability of effective techniques and tools.

To expand the enzymatic toolbox for studying extended O-glycans, we employed a rational approach guided by sequence and structural information to engineer EngBF. We constructed catalytically promiscuous variants of EngBF that exhibit activity towards extended sialyl-core 1 substrates, enhancing the enzyme’s versatility and broadening its potential applications in the study of complex O-glycans. By using mutational studies and computational methods, we investigated the molecular basis of catalysis, providing insights into the active site of EngBF and guiding future efforts to enhance its activity and specificity. However, the inherent structural complexity in GH101 enzymes, which is largely attributed to their large size, poses a challenge to more exhaustive enzyme engineering using directed evolution. To overcome this, we utilized a deep learning-based methodology to create a minimal enzymatic unit. The approach resulted in the creation of thermostable scaffolds containing the EngBF active site with a compact three-dimensional structure that is considerably smaller than the wildtype GH101 enzyme. Overall, our work represents significant progress in developing a versatile platform for investigating complex O-glycans. By expanding the enzymatic toolbox and optimizing the size of the enzyme, we have created a valuable starting point for future enzyme engineering efforts.