The exploration of chemo-enzymatic routes to complex carbohydrates has been hampered by the lack of appropriate enzymatic tools having the substrate specificity for new reactions. Here, we used a computer-aided design framework to guide the construction of a small, diversity-controlled library of amino acid sequences of an α-transglucosylase, the sugar binding subsites of which were re-engineered to enable the challenging 1,2-cis-glucosylation of a partially protected β-linked disaccharide allyl (2-deoxy-2-trichloroacetamido-β-D-glucopyranosyl)-(1→2)-α-L-rhamnopyranoside, a potential intermediate in the synthesis of Shigella flexneri cell-surface oligosaccharides. The target disaccharide is not recognized by the parental wild-type enzyme and exhibits a molecular structure very distinct from that of the natural α-(1→4)-linked acceptor. A profound reshaping of the binding pocket had thus to be performed. Following the selection of 23 amino acid positions from the first shell, mutations were sampled using RosettaDesign leading to a subset of 1515 designed sequences, which were further analyzed by determining the amino acid variability among the designed sequences and their conservation in evolutionary-related enzymes. A combinatorial library of 2.7 × 10 4 variants was finally designed, constructed, and screened. One mutant showing the desired and totally new specificity was successfully identified from this first round of screening. Impressively, this mutant contained seven substitutions in the first shell of the active site leading to a drastic reshaping of the catalytic pocket without significantly perturbing the original specificity for sucrose donor substrate. This work illustrates how computer-aided approaches can undoubtedly offer novel opportunities to design tailored carbohydrate-active enzymes of interest for glycochemistry or synthetic glycobiology.

Neutral drift (also called purifying selection) is an attractive approach to generate polymorphic variant libraries for enzyme engineering. Here, we have applied this strategy to modify the substrate specificity of a transglucosylase. Our model enzyme, the amylosucrase from Neisseria polysaccharea, is a glucosylation biocatalyst of prime interest because it uses the widespread substrate sucrose as a glucosyl donor and shows broad acceptor promiscuity. A library of 440 functional amylosucrase variants was generated after four rounds of neutral drift at a low mutation rate. The functional variations present in this library were investigated by assaying the ability of these variants to use an alternative glucosyl donor (pnitrophenyl-α-D-glucopyranoside, pNP-Glc) and to glucosylate a range of acceptors (including methyl-α-L-rhamnopyranoside, which is not naturally recognized by the parental enzyme). The impact of these mutations on the thermal stability of the variants was also assessed. Large variations of acceptor promiscuity were observed, ranging from the complete loss of detectable activity to a 2-fold increase relative to the parental enzyme. Variants showing increased catalytic efficiency toward the alternative pNP-Glc donor were also identified. Specifically, one variant combining four unprecedented amino acid changes was 25-fold more efficient at utilizing pNP-Glc than the parental enzyme and acquired glucosylation activity toward methyl-α-L-rhamnopyranoside. Enzymes with improved thermal stability were also identified. Overall, our work demonstrates that neutral drift is an effective and powerful strategy to engineer transglycosylases with enhanced or even acquired substrate specificities from small-sized functional libraries compatible with accurate low-throughput multi-parameter analyses.