Molecular design of non-leloir furanose-transferring enzymes from an alpha-L-arabinofuranosidase: A rationale for the engineering of evolved transglycosylases

The vast biodiversity of glycoside hydrolases (GHs) constitutes a reservoir of readily available carbohydrate-acting enzymes that employ simple substrates and hold the potential to perform highly stereopecific and regioselective glycosynthetic reactions. However, most GHs preferentially hydrolyze glycosidic bonds and are thus characterized by a hydrolysis/transglycosylation partition in favor of hydrolysis. Unfortunately, current knowledge is insufficient to rationally modify this partition, specifically mutating key molecular determinants to tip the balance toward transglycosylation. In this study, in the absence of precise knowledge concerning the hydrolysis/transglycosylation partition in a hydrolytic GH51 alpha-L-arabinofuranosidase, we describe how an iterative protein engineering approach has been used to create the first "non-Leloir" transarabinofuranosylases. In the first step, random mutagenesis yielded a point mutation (R69H) at a position that is highly conserved in clan GH-A. Characterization of R69H revealed that this enzyme displays high transglycosylation activity but severely reduced (61-fold) activity on pNP-alpha-L-arabinofuranoside. Upon recombination of R69H with other point mutations selected using semirational or in silico approaches, transfer rates close to 100% and transarabinofuranosylation yields of the main (1 2)-linked oligosaccharide product of 80% (vs 11% for the wild-type) were obtained. Combining data presented here with knowledge drawn from the literature, we suggest that the creation of non-Leloir transglycosylases necessarily involves the destabilization of the highly developed transition states that characterize the predominantly hydrolytic exo-acting GHs; this is an efficient way to prevent ubiquitous water molecules from performing the deglycosylation step.