Family 1 Glycosyltransferase UGT706F8 from Zea mays Selectively Catalyzes the Synthesis of Silibinin 7-O-β-d-Glucoside

Abstract

Regioselective glycosylation is a chemical challenge, leading to multistep syntheses with protecting group manipulations, ultimately resulting in poor atom economy and compromised sustainability. Enzymes allow eco-friendly and regioselective bond formation with fully deprotected substrates in a single reaction. For the selective glucosylation of silibinin, a pharmaceutical challenged with low solubility, enzyme engineering has previously been employed, but the resulting yields and k_cat were limited, prohibiting the application of the engineered catalyst. Here, we identified a naturally regioselective silibinin glucosyltransferase, UGT706F8, a family 1 glycosyltransferase from Zea mays. It selectively and efficiently (k_cat = 2.1 ± 0.1 s^–1; K_M = 36.9 ± 5.2 μM; TTN = 768 ± 22) catalyzes the quantitative synthesis of silibinin 7-O-β-d-glucoside. We solved the crystal structure of UGT706F8 and investigated the molecular determinants of regioselective silibinin glucosylation. UGT706F8 was the only regioselective enzyme among 18 glycosyltransferases found to be active on silibinin. We found the temperature optimum of UGT706F8 to be 34 °C and the pH optimum to be 7–8. Our results indicate that UGT706F8 is an efficient silibinin glycosyltransferase that enables biocatalytic production of silbinin 7-O-β-d-glucoside.

Short abstract

We demonstrate enzymatic quantitative synthesis of silibinin-7-O-β-d-glucoside, a potential pharmaceutical, without environmentally challenging synthetic chemistry.

Introduction

Silibinin is a natural product of the milk thistle, Silybum marianum, and is a mixture of two stereoisomers of a chiral polyphenol, silybin A and silybin B. It is used as a hepatoprotectant, to treat mushroom intoxication, and as a veterinary medicine.¹ Its water solubility is low (430 mg/L), and its bioavailability has been improved by different strategies, such as esterification by succinate^2,3 or glycosylation.¹ Silibinin has five O-glycosylation sites (3, 5, 7, 20, and 23). Position 23 has been chemically glycosylated with glucose, galactose, lactose and maltose. All four glycosides had increased water solubility (4–30 times) and an enhanced hepatoprotective effect as assayed in rat hepatocytes.⁴ Their therapeutic potential was later substantiated in other cell cultures.⁵ Silibinin is an interesting target for enzymatic glycosylation, given that enzymes—in addition to requiring milder reaction conditions—can catalyze regioselective glycosylation on molecules presenting multiple unprotected glycosylation sites of similar reactivities in a single reaction. Silibinin has been glycosylated on position 7 with opium poppy cell cultures⁶ and with an engineered GH3 enzyme, although in the latter case the glycosylation site and efficiency were not reported.⁷ In Nature, this type of reaction is predominantly carried out by enzymes of the UDP-dependent glycosyltransferase family (UGTs),⁸ phylogenetically belonging to family 1 glycosyltransferases within the CAZy database.⁹ These are Leloir glycosyltransferases displaying the GT-B fold,¹⁰ generally presenting a His-Asp catalytic dyad and a large hydrophobic substrate-binding pocket,^8,11 although family 1 glycosyltransferases without the Asp do exist (e.g., UGT75 and UGT84 subfamilies).¹² They use UDP-activated sugars as donors and are capable of glycosylating a vast array of natural product acceptors, many of which have industrial relevance.¹³ Positions 7 and 3 in silybin A were recently shown to be glycosylated by Siraitia grosvenorii UGT74AC2.¹⁴ This enzyme produces a mixture of silibinin 3-O-β-d-glucoside, 7-O-β-d-glucoside, and 3,7-O-β-d-diglucoside. By point mutation, the authors engineered regiospecific variants. Unfortunately, these enzyme variants are very slow, with k_cat values of 0.20–0.47 min^–1 and chemical yields of 39%–57%, making it unsuitable for application as a biocatalyst.

Experimental Section

Materials

Buffers, standard reagents, and silibinin (≥98% mixture of silybin A and B) were purchased from Sigma-Aldrich.

Expression and Purification of UGT706F8

The full-length histidine-tagged DNA sequence of UGT708F8 was synthesized and cloned into a pET-28a(+) vector by Genscript (USA). The protein was expressed in a 1 L culture of One Shot BL21 Star (DE3) E. coli cells (ThermoFisher Scientific, USA). The cleared lysate was purified by nickel affinity chromatography, and greater than 90% pure UGT706F8 was stored in 25 mM 4-(2-hydroxyethyl)-1-piperazineethanesulfonic acid (HEPES), pH 7, 50 mM NaCl, and 1 mM dithiothreitol (DTT). Preparations for crystallography were further treated with the tobacco etch virus (TEV) protease to remove the hexahistidine tag before crystallization and concentrated to greater than 10 mg/mL.

Crystallographic Structure Determination and Analysis

UGT706F8 was cocrystallized with 5 mM UDP (Sigma-Aldrich) in 20% PEG 6000, 100 mM Tris, pH 8.1, and 200 mM NaCl. Data were collected at the P14 beamline at the PETRA III synchrotron (DESY, Hamburg, Germany). Data were processed to 1.6 Å resolution with XDSAPP¹⁵ using XDS¹⁶ in P2₁, and the structure was determined with molecular replacement using PHASER¹⁷ from the Phenix software package.¹⁸ As a search model, we used the structure of the homologous enzyme PtUGT1 from Polygonum tinctorium (5NLM), which has 38.3% sequence identity to UGT706F8. Manual model rebuilding and refinement was done iteratively with Coot(19) and phenix.refine,²⁰ and the final model had R/R_free of 16.7/18.8. Processing and refinement statistics can be found in Table S2. The structure is deposited in the Protein Data Bank with ID 7Q3S.

Kinetics Studies

Here, 0.5 mM UDP-Glc was used as the sugar donor and 0.1 mM silibinin as the acceptor, except for the Michaelis–Menten analysis, where the silibinin concentration was varied between 0 and 0.25 mM. All reactions were performed without stirring. Reactions were performed at 293 K in a 50 mM phosphate-citrate buffer, pH 8, except for the pH characterization, which was conducted in phosphate-citrate, HEPES, and lysine buffers. Reactions were prepared from a 10 mM stock solution of silibinin in DMSO using 15–120 nM enzyme (15 nM for Michaelis–Menten analysis and temperature optimum, 30 nM for the time course reaction, and 120 nM for the pH optimum conditions), and 5 μL was analyzed undiluted via reverse phase HPLC. Reactions for Michaelis–Menten analysis and optimal temperature determination were quenched by thermal denaturation for 30 s at 95 °C, while time-course reactions and pH optimum determination reactions were performed in HPLC vials, and the ongoing reaction was analyzed every 6 min. The analyses were performed on an Ultimate 3000 series apparatus (Dionex) equipped with a kinetex 2.6 μm C18 100 Å 100 mm × 4.6 mm analytical column (Phenomenex) maintained at 40 °C, and the obtained chromatograms at 280 nm were processed via Chromeleon 7.2.9 (Dionex). Milli-Q water and acetonitrile containing 0.1% formic acid were used as mobile phases A and B, respectively, according to the multistep gradient displayed in Figure S1. K_M and k_cat parameters were determined using the function MM.2 of the drc package in R 3.6.2 (Rstudio version 1.2.1335, RStudio, Inc.). These parameters were determined using the assumption that the reaction rate was constant for the first 20 min and that a single point measurement at 20 min reflected initial rates. Here, 80%–95% of the acceptor substrate remains after the reaction, which supports this assumption. The linearity observed using time points at 0, 2, 4, 6, 8, and 10 min for initial rate determinations at pH 6.8–8.3 seem to support this assumption (Figure S13). However, this approach may result in an underestimation of k_cat. To study the potential amount of silibinin glucoside produced by unit of catalyst (UGT706F8), reactions were made in units of 500 μL, using 100 mM buffer phosphate-citrate, pH 8, 0.5 mM UDP-Glc, 0.1 mM or 0.3 mM silibinin, and eight different enzyme concentrations (12.6, 9.4, 6.3, 3.14, 1.57, 0.785, 0.39, and 0.20 mg/L).

Enzymatic Synthesis of Silibinin 7-O-β-d-Glucoside by UGT706F8 for Structure Determination by NMR

Silibinin was dissolved in deuterated dimethyl sulfoxide (DMSO-d6) to a concentration of 10 mM. We chose deuterated DMSO both in case we would need to monitor the reaction directly by NMR at a later stage and in case some DMSO would be extracted when we analyzed the end point reaction by NMR. The enzymatic reaction was carried out at 30 °C and 600 rpm in 40 mL of HEPES buffer, pH 7, containing 27.5 μg/mL of UGT706F8, 0.5 mM silibinin, 200 mM sucrose, 1 mM UDP, and 20 μg/mL of sucrose synthase. After 16 h, the reaction was quenched with 1:1 (v:v) ethyl acetate, and the organic phase was extracted with 60 mL ethyl acetate. It was washed once with water and dried on MgSO₄, followed by removal of the ethyl acetate using a rotary evaporator. The formation of 7-O-β-d-glucoside and its purity (>99%) was assessed with HPLC.

Silibinin 7-O-β-d-Glucoside Structure Determination

The silibinin glucoside sample was dissolved in DMSO-d6 and transferred to an NMR tube. The NMR data were acquired on an Avance III (799.75 and 201.11 MHz for ¹H and ¹³C, respectively) equipped with a 5 mm TCI ¹H/(¹³C, ¹⁵N) CryoProbe or a Bruker Avance IIIHD (400.13 and 100.61 MHz for ¹H and ¹³C, respectively) equipped with a 5 mm CryoProbe Prodigy. ¹H and ¹³C spectra are provided in Figures S2 and S3, respectively. The following 2D experiments were performed: ¹H–¹³C heteronuclear single quantum coherence (HSQC, Figure S4), ¹H–¹³C heteronuclear multiple bonds correlation (HMBC, Figure S5), rotating frame Overhauser effect spectroscopy (ROESY, Figure S6), and heteronuclear single quantum coherence total correlated spectroscopy (HSQC-TOCSY, Figure S7). All experiments were carried out using standard Bruker pulse sequences. The data were processed using Bruker Topspin 4.0.6.

Results and Discussion

To discover an enzyme for regioselective silibinin glucoside production, we screened 18 in-house UGTs from phylogenetic group E²¹ (Table S3). Only one, UGT706F8, exclusively produced one product with high chemical yields greater than 99% analytical yields according to both ¹H NMR and reverse phase chromatography.

Structure Determination of Enzymatic Product

The signal assignment is reported in Table S1 and is in agreement with the published assignment of 7-O-β-d-glucosylated silibinin²² in CD₃OD, except that we observed and assigned the signals corresponding to exchangeable protons. Two products can be observed by NMR, corresponding to the glucosylation of both silibinin A and B at position 7. The ratio between the signals arising from the different products was found to be approximately 48:52 by NMR (Scheme 1 and Figures S2–S9) and 51.5:48.5 (silibinin A:silbinin B) by HPLC (Figure S10, considering an identical absorption coefficient for both isomers). Comparing the ¹H NMR spectrum of the glucosylated silibinin sample with that of silibinin, one of the three phenol signals has disappeared, as shown in Figure S6. The missing signal corresponds to the phenol at position 7 of silibinin, as it is the only phenol not capable of forming an intramolecular hydrogen bond, resulting in signal broadening due to fast exchange. This was verified by HMBC (Figure S5), where both of the remaining phenol protons have strong correlations to carbon in their respective regions of the molecule, as shown in Table S1 (e.g., the phenolic proton of position 5 has strong correlations with carbons 6, 9, 10, and 7). Interestingly, various receptors and enzymes,²³ including glycosyltransferases,²⁴ interact differently with silybin A and silybin B, while UGT706F8 seems equally proficient at glycosylating O7 of both diastereomers (Figure S10). It might be because O7 is far away from the asymmetric carbons 7′ and 8′ (>10 Å) compared to the other positions reported to be glycosylated in literature (3, 9′, or 4′′). Further, given the shallow active site of UGT706F8, the E ring of enzyme-bound silibinin is likely outside the active site of UGT706F8, and its configuration might therefore be irrelevant.

Scheme 1. Structure of Silibinin A 7-O-β-d-Glucoside (Top) and Silibinin B 7-O-β-d-Glucoside (Bottom), Formed by UGT706F8 from Silibinin.

Presented according to the systematic numbering of silibinin.²⁴

Structure Determination of UGT706F8

To understand the structural determinants for the regioselectivity of UGT706F8, we solved its crystal structure in complex with UDP (PDB ID: 7Q3S) (Figure 1 and Table S2). UGT706F8 crystallized in the P2₁ space group with one molecule in the asymmetric unit. The crystals diffracted to 1.6 Å, and the structure was refined to R/R_free of 16.7/18.8. The structure displays the GT-B fold with two Rossmann domains¹⁰ common for UGTs (Figure 1A). The active site consists of a UDP-Glc binding site (Figure 1B), a hydrophobic acceptor binding pocket (Figure 1C), and the conserved catalytic histidine (His19). The catalytic residue is usually activated by an aspartate, and the proton abstracted from the acceptor is shared by these two residues.^11,25 However, in UGT706F8, a glycine incapable of activating the catalytic residue is found in this position (Gly119, Figure 1B and C), and no other carboxylate in the vicinity could perform this role. Interestingly, UGT706F8 contains a tryptophan in the active site (Trp206, Figure 1C) in the vicinity of a residue that was shown to promote regioselective silibinin glycosylation by UGT74AC2 when mutated to a tryptophan (Leu200 in UGT74AC2).¹⁴ Similarly to Leu200Trp in UGT74AC2, the bulky Trp206 might restrict silibinin binding modes and thereby contribute to regioselectivity.

Figure 1.

(A) Overall fold of UGT706F8 shown in cartoon representation. (B) Zoom on donor site with His19 and Gly119 highlighted. (C) Zoom on acceptor binding pocket, with His19, Gly119, and Trp206 highlighted.

Biochemical Characterization of UGT706F8

To enable the application of UGT706F8 in a biocatalytic process toward silibinin 7-O-β-d-glucoside, we screened a range of reaction conditions. UGT706F8 has a temperature optimum of 34 °C (Figure 2A and B). A sharp decrease of obtained yields is observed at temperatures above 34 °C, while the effect on initial rates is milder, likely due to the denaturation of UGT706F8 at temperatures greater than 34 °C. Since the pH profile (Figure 2C and D), displaying a maximum from 7.3 to 8 followed by a sharp decrease at higher pH, combined with the pK_a of silibinin 7-OH (experimentally reported as 7.7 or 7.95)²⁴ does not support a mechanism without enzyme-assisted deprotonation, it is assumed that His19 abstracts a proton without the assistance of another residue. Despite lacking one of the canonical catalytic residues, UGT706F8’s catalytic properties are within the typical range for UGTs²⁶ (k_cat = 2.1 ± 0.1 s^–1 and K_M = 36.9 ± 5.2 μM (Figure 2E). The k_cat is 600-fold higher than that of the Leu200Trp engineered variant of UGT74AC2,¹⁴ the only other enzyme known to catalyze the regioselective synthesis of silibinin 7-O-β-d-glucoside. Furthermore, UGT706F8 catalyzes this reaction with quantitative yields (Figures 2C and F and 3). The total turnover number (TTN) in the phosphate-citrate buffer at pH 8 at 23 °C is 768 ± 22 (Figure 3). The nonlinear relationship between product and enzyme concentration for the lower part of the curve (Figure 3) is likely due to low stability of the protein in highly diluted solutions (<2 μg/mL), a phenomenon we recently reported for a related group E UGT from Polygonum tinctorium.²⁷ Compared to the amounts of silibinin used for medication³ (e.g., 20 mg/kg/day for treating mushroom poisoning), our current titers (Figures 2 and 3 and Figure S11) should prove useful and enable the study and application of silibinin 7-O-β-d-glucoside.

Figure 2.

Biochemical characterization of UGT706F8. (A) Temperature profile corresponding to analytical glycosylation yields of 100 μM silibinin after 20 min in the presence of 15 nM enzyme in a 50 mM phosphate-citrate buffer, pH8. (B) Temperature profile corresponding to relative initial rates against 100 μM silibinin in the presence of 15 nM enzyme in a 50 mM phosphate-citrate buffer, pH 8.3. (C) pH profile corresponding to analytical glycosylation yields of 100 μM silibinin after 40 min at 293 K in the presence of 120 nM enzyme. (D) pH profile corresponding to initial rates against 100 μM silibinin after at 293 K in the presence of 15 nM enzyme. (E) Michaelis–Menten curve for the synthesis of silibinin 7-O-β-d-glucoside in the presence of 15 nM enzyme at 293 K, 50 mM phosphate-citrate buffer, pH8. Analytical yields after 20 min reaction time are used as a proxy for initial rates. (F) Time course of silibinin 7-O-β-d-glucoside formation in the presence of 30 nM enzyme at 293 K in 50 mM phosphate-citrate buffer, pH8. Solid lines are a visual guide between the measured data points, except in panel E. Chromatograms corresponding to panels C and F have been added as Figure S12 and S13, respectively, and plots for initial rate determinations corresponding to panel D have been added as Figure S14.

Figure 3.

Silibinin 7-O-β-d-glucoside titers measured after 24 h reaction time, using 0.1 mM silibinin as substrate at different enzyme concentrations. Reactions were performed at 293 K in 100 mM phosphate-citrate buffer, pH 8, in the presence of 0.5 mM UDP-Glc, at a 0.5 mL volume in HPLC vials with no stirring.

Conclusion

UGT706F8 was identified as a regioselective silibinin glycosylating enzyme, capable of producing silbinin 7-O-β-d-glucoside with quantitative yields using reasonable enzyme loads. The crystal structure revealed a peculiar catalytic machinery lacking the aspartate that is usually part of the His-Asp catalytic dyad of this clade of plant GT1 enzymes. While enzyme engineering can be employed to obtain selective variants from promiscuous GT1 enzymes, this study demonstrates the discovery of a naturally regioselective and potent biocatalyst, paving the way for the sustainable production of a silibinin derivative with pharmaceutical potential. It hints at the need for larger discovery efforts of natural GT1 enzymes to obtain biocatalysts able to produce glucosides of choice with the activity levels required for production.

Glossary

Abbreviations

GT1

family 1 glycosyltransferase

UGTs

UDP-dependent glycosyltransferase family

HSQC

heteronuclear single quantum coherence

HMBC

heteronuclear multiple bonds correlation

ROESY

rotating frame Overhauser effect spectroscopy

HSQC-TOCSY

heteronuclear single quantum coherence total correlated spectroscopy

NMR

nuclear magnetic resonance

DMSO

dimethyl sulfoxide

DTT

dithiothreitol

Supporting Information Available

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acssuschemeng.1c07593.

• Plasmid sequence; experimental procedures; ¹H NMR, ¹³C NMR, HSQC, HMBC, ROESY, HSQC-TOCSY spectra of the β-d-glucosylated silibinin; ¹H spectrum of the phenolic region; zoom of select ¹³C NMR signals; chromatogram of silibinin A and B and its glucosylation by UGT706F8; silibinin 7-O-β-d-glucoside titers; chromatograms corresponding to Figure 2B; chromatograms corresponding to Figure 2D; plots for initial rate determinations of UGT706F8 in phosphate-citrate buffers, pH 6.8–8.3; Fo-Fc difference electron density map showing the residual active site density; HPLC chromatogram of the reaction mixture used for the NMR structural determination of the product; NMR assignment of the β-d-glucosylated silibinin; data collection and refinement statistic data of UGT706F8; screen of GT1s for regioselective silibinin glucosylation, PDB ID: 7Q3S (PDF)

Author Contributions

The manuscript was written through contributions of all authors. All authors have given approval to the final version of the manuscript.

The Novo Nordisk Foundation Grants NNF18OC0034744, NNF10CC1016517, and NNF20CC0035580, the Danish National Research Foundation (Grant DNRF124), and Grant 7129-00003B from the Danish Agency for Science, Technology and Innovation through the instrument center DanScatt.

The authors declare no competing financial interest.