Glycosylation is one of the most common reactions in the biosphere, yet a particularly challenging one for conventional synthetic chemistry. Indeed, the need to control both regio- and stereo-selectivity leads to a succession of reactions, including protecting group manipulations and bond activations, resulting in low chemical yields, poor atom economy, and large amounts of waste. Conversely, enzymatic glycosylation occurs in a single reaction with unprotected sugars and acceptors and lends perfect control over stereoselectivity (Nidetzky et al., 2018). Provided with the appropriate enzyme, full control over regioselectivity, as well as quantitative chemical yields are also feasible. In Nature, glycosylation is primarily catalyzed by glycosyltransferases, enzymes that transfer a saccharide from an activated sugar donor to an acceptor molecule. These enzymes are organized in >100 distinct glycosyltransferase families in the CAZy database (Coutinho et al., 2003; Lombard et al., 2014), with all enzymes within a family sharing phylogeny, structural fold, and generally mechanism. The β-glycosylation of natural products is mainly achieved by enzymes from glycosyltransferase family 1 (GT1) (Louveau and Osbourn, 2019). These GT1s are inverting enzymes using α-nucleotide sugars as donors, most commonly UDP-sugars, and are thus also termed UGTs, for UDP-dependent glycosyltransferases (Ross et al., 2001). They catalyze the formation of O-, N-, S- or C-glycosidic bonds. O-glycosylations are the most common reactions and are usually promoted by a His-Asp catalytic dyad sharing a proton abstracted from the acceptor (Scheme 1) (Brazier-Hicks et al., 2007; Teze et al., 2021). The N- and S- mechanisms are slightly different (Teze et al., 2021), and the C-glycosylation mechanism is related but yet to be firmly established (Gutmann and Nidetzky, 2013; Putkaradze et al., 2021). GT1 enzymes are relatively promiscuous, being able to act on a variety of natural products (Offen et al., 2006; Chen et al., 2015; Zhang et al., 2022), and most GT1s are active against polyphenols (Yang et al., 2018).

GT1s have received considerable interest as tools for biotechnological glucosylation (Nidetzky et al., 2018; Vasudevan and Lee, 2020). Indeed, the possibility to use sucrose synthase for forming UDP-Glc from UDP and sucrose, and using lysates from the enzyme’s production as UDP providers, makes β-glucosylation an economically feasible process (Wang et al., 2012; Schmölzer et al., 2016; Liu and Nidetzky, 2021). However, their stability—a crucial industrial property—has only been scarcely characterized (Fujiwara et al., 2009; Gao et al., 2020). In a few recent cases (Petermeier et al., 2021; Bidart et al., 2022), we observed instability, seemingly not intrinsic but dependent on experimental conditions, particularly enzyme and acceptors concentrations. Indeed, a nonlinear behavior was observed upon enzyme dilution, particularly at acceptor substrate concentrations in the millimolar range (Petermeier et al., 2021; Bidart et al., 2022). In order to investigate how widespread this peculiar behavior is within GT1-catalyzed reactions, we analyzed the effect of 16 reaction conditions on end-point reaction yields from 18 distinct GT1 enzymes, each against three different polyphenol acceptors.

The 18 GT1 enzymes have 24–40% pairwise identity after multiple sequence alignment via Clustal Omega (Sievers and Higgins, 2014). Nine of these enzymes have been previously described in the literature: PtUGT1 (Teze et al., 2021), ZmUGT708A6 (Ferreyra et al., 2013), ZmUGT706F8 (Bidart et al., 2022), the GT1s from Arabidopsis thaliana (AtUGT72E2, At71C1, At71D1) (Yang et al., 2018), RhGT1 (Wang et al., 2013), Gm88E3 (Liu and Nidetzky, 2021), and MtUGT78G1 (Modolo et al., 2007). Among the nine GT1 enzymes that were not previously described, five already had designated names (Zm71B1, Os88C1, Lc72B10, Fi88A10, and Fe88J1), and the remaining four are named according to the UGT naming convention (Mackenzie et al., 2005) preceded by two letters referring to genus and species (e.g., ZmUGT88C10). UniProt accession numbers and melting temperatures (T _m) of the 18 GT1 enzymes are provided in Supplementary Table S1.

These 18 GT1 enzymes are described here for their activity against three acceptors (Figure 1), representing different classes of polyphenols of biotechnological interest, that is, coumarins (scopoletin), stilbenes (trans-resveratrol), and flavones (apigenin). Interestingly, we found that each of the 18 enzymes is active against each of the chosen acceptors, and in most cases (44/54), analytical yields of glycosylation >50% are reached. Greater than 90% of yields are obtained for at least one condition in about half of the enzyme-acceptor pairs (Figures 2, 3).

While most of the curves display the classical dependency on enzyme concentration of a reaction catalyzed by enzymes with low total turnover numbers—that is, a linear or sublinear increase in the product as a function of enzyme concentration—half of the enzyme: substrate pairs (27/54) display dilution-induced inactivation behavior with a superlinear dependency on enzyme concentration (Figures 2, 3). At low enzyme concentrations (e.g., 10 mg/L, circa 150 nM), the no-to-little reaction is observed, yet doubling the enzyme concentration far more than double the observed yields. The full dataset of yields as a function of enzyme and acceptor concentrations is available in the supplementary material (Supplementary Table S2).

Importantly, this behavior is also related to acceptor concentration, being more prevalent at 400 μM than at 50 μM. It is particularly pronounced with apigenin, for example, Fe88J1, PtUGT1, ZmUGT88C10, AtUGT72E2, At71C1, RhGT1, and ZmUGT708A6 (Figures 2, 3). It is also observed with resveratrol (e.g., ZmUGT708A6 or ZmUGT706F8) and scopoletin (e.g., ZmUGT88C10 or AtUGT72E2). Interestingly, while glycosylation of apigenin and resveratrol regularly (19/36) reaches full conversion of the acceptors, the glucosylation of scopoletin results in an equilibrium (Scheme 1), with a maximum yield depending on the acceptor concentration. At the highest acceptor concentration, nearing donor and acceptor equimolarity (500 and 400 μM, respectively), the maximal yields observed are around 50% (Figures 2, 3). This allows for the observation of hydrolysis in 5/18 GT1 enzymes in our dataset, being particularly pronounced for SlUGT72B68. Indeed, while the formation of scopoletin-glucoside from UDP-Glc and the formation of UDP-Glc from scopoletin are in equilibrium, the hydrolysis of UDP-Glc by the enzyme is irreversible (Scheme 1). There seems to be a weak relationship between the intrinsic stability of the enzyme, represented by its melting temperature (T _m), as the three enzymes seemingly unaffected by the conditions were the relatively stable ZmUGT72G3 (T _m = 56.4 ± 0.1⁰C/T _m = 329.6 ± 0.1 K), ZmUGT72G4 (T _m = 61.2 ± 0.4⁰C/T _m = 334.4 ± 0.4 K), and At71D1 (ND). Conversely, ZmUGT708A6 (T _m = 37.4 ± 0.1⁰C/T _m = 310.6 ± 0.1 K) and ZmUGT88C10 (T _m = 42 ± 0.3⁰C/T _m = 315.2 ± 0.3 K) were most affected by conditions. Considering ∼0.008 kJ/mol/residue (Rees and Robertson, 2001), and an average length of GT1 enzymes of c. 500 residues, a ΔT _m of 1 K roughly equates to stabilization of 1 kcal/mol, thus between the most and least stable enzymes in our dataset a difference as large as 25 kcal/mol is observed. Enzyme-substrate interactions are generally thought to be stabilizing, which is the rationale behind the use of differential scanning fluorimetry as a basis for identifying enzyme-substrate pairs (Niesen et al., 2007). We assessed whether polyphenol acceptors modified the T _m of our proteins, and did not observe a significant change in either direction (Supplementary Figure S2). ZmUGT708A6, which displays chemostability issues in presence of all three acceptors, would even appear to present slightly higher T _m in presence of resveratrol and apigenin (Supplementary Figure S1).

In this article, we demonstrate the widespread yet not widely reported phenomena of dilution-induced inactivation and low chemostability towards their own acceptors of GT1 enzymes. These effects are important and can introduce biases in both the kinetic study and discovery efforts for GT1 enzymes. The latter is of particular importance since one of the major obstacles to a wider biotechnological application of glycosyltransferases is the characterization of their acceptor scope. While one might be enticed to assess acceptors at high concentrations to detect catalysts with low affinity (high K _S), or at low enzyme concentrations to be cost-efficient, our results demonstrate that this would result in a significant number of false negatives. While we report the effect, we do not offer a mechanistic explanation. Protein destabilization by small molecules generally occurs at much higher concentrations than the effects reported here at submillimolar concentrations (Singh et al., 2011). Molecular crowding, occasionally invoked to rationalize dilution-induced inactivation, occurs at much higher concentrations (Miklos et al., 2011; Wang et al., 2012; Cohen and Pielak, 2017). Conversely, the enzyme’s adsorption onto equipment (vessel, glassware, tips, etc.) is a concern for trace concentrations or up to the nanomolar range, several orders of magnitude lower than our data and therefore not likely to account for our observations. Furthermore, GT1 enzymes are monomeric, clearly demonstrated by size exclusion chromatography and several crystallographic structures (Wetterhorn et al., 2016). Furthermore, we performed a native PAGE analysis of Fi88A10 at different concentrations, up to 2.5 μg/μl (>300-fold higher concentration than the highest concentration used in the enzymatic assays). Even at high concentration, Fi88A10 appeared monomeric, ruling out dilution-induced oligomerization disruption as an explanation (Supplementary Figure S2). The effects seemed to affect enzymes presenting either high (e.g., At71C1 on apigenin) or low (e.g., ZmUGT708A6 on apigenin) specificity. To assess whether initial rates were affected, we determined these rates for three enzymes (Fi88A10, At71C1, and ZmUGT708A6) at defined conditions (20 or 80 mg/L enzymes, 50 or 400 μM apigenin). We observed a strong effect on the chemo instability, as no rates/reaction could be observed within 10 min for Fi88A10 and ZmUGT708A6 at 400 μM acceptor concentration, particularly compared to the reasonable rates (8–60 per min) observed at 50 μM apigenin (Supplementary Table S3). Conversely, the initial rates of At71C1 in presence of 50 μM apigenin could not be determined, as >30% of the acceptor was converted within 0.5 min even at 20 mg/L enzymes.

Here, the synergistic effect with the chemostability at moderately high acceptor concentrations, together with the fact that each enzyme presents various behaviors depending on the acceptor, indicates that specific phenomena related to GT1 enzymes are behind our observations. Conceivably, their relatively large, solvent-exposed hydrophobic acceptor site (Brazier-Hicks et al., 2007; Teze et al., 2021) could be involved. Appreciation of these effects is important both for mechanistic and kinetic studies of GT1 enzymes, their biotechnological use (scale-up to high acceptor concentrations), as well as for discovery efforts.