Reaction intensification for biocatalytic production of polyphenolic natural product di‐C‐β‐glucosides

Polyphenolic aglycones featuring two sugars individually attached via C‐glycosidic linkage (di‐C‐glycosides) represent a rare class of plant natural products with unique physicochemical properties and biological activities. Natural scarcity of such di‐C‐glycosides limits their use‐inspired exploration as pharmaceutical ingredients. Here, we show a biocatalytic process technology for reaction‐intensified production of the di‐C‐β‐glucosides of two representative phenol substrates, phloretin (a natural flavonoid) and phenyl‐trihydroxyacetophenone (a phenolic synthon for synthesis), from sucrose. The synthesis proceeds via an iterative two‐fold C‐glycosylation of the respective aglycone, supplied as inclusion complex with 2‐hydroxypropyl β‐cyclodextrin for enhanced water solubility of up to 50 mmol/L, catalyzed by a kumquat di‐C‐glycosyltransferase (di‐CGT), and it uses UDP‐Glc provided in situ from sucrose by a soybean sucrose synthase, with catalytic amounts (≤3 mol%) of UDP added. Time course analysis reveals the second C‐glycosylation as rate‐limiting (0.4–0.5 mmol/L/min) for the di‐C‐glucoside production. With internal supply from sucrose keeping the UDP‐Glc at a constant steady‐state concentration (≥50% of the UDP added) during the reaction, the di‐C‐glycosylation is driven to completion (≥95% yield). Contrary to the mono‐C‐glucoside intermediate which is stable, the di‐C‐glucoside requires the addition of reducing agent (10 mmol/L 2‐mercaptoethanol) to prevent its decomposition during the synthesis. Both di‐C‐glucosides are isolated from the reaction mixtures in excellent purity (≥95%), and their expected structures are confirmed by NMR. Collectively, this study demonstrates efficient glycosyltransferase cascade reaction for flexible use in natural product di‐C‐β‐glucoside synthesis from expedient substrates.


| INTRODUCTION
Polyphenols (e.g., flavonoids, anthrones) are a large class of natural products widely distributed in plants (phytochemicals) (Abbas et al., 2016;Durazzo et al., 2019;Rasouli et al., 2017). They are generally considered important in a healthy human diet and contribute to the efficacy of traditional medicines (Luo et al., 2021;Mithul Aravind et al., 2021;Shakoor et al., 2021;C. Sun, Zhao, et al., 2020). Polyphenol biosynthesis often involves the late-stage attachment of one or more sugar residues to the aglycone core structure Thuan & Sohng, 2013). A distinct natural motif of polyphenol glycosylation is that of the di-C-glycoside. This involves two sugars individually linked to the same benzene ring of the polyphenol via a C-glycosidic bond (Kitamura et al., 2018; Y. Q. Zhang et al., 2022). The type of glycosylation most often used is that of C-β-D-glucosyl ( Figure 1) (C. F. Liu, 2022;Y. Q. Zhang et al., 2022).
The di-C-glycosides of polyphenols have attracted attention for the unique chemical properties and bioactivities they can offer (Xiao et al., 2016;Zeng et al., 2013). Generally, the C-glycosidic linkage is resistant to hydrolysis (Putkaradze et al., 2021;Tegl & Nidetzky, 2020;Y. Q. Zhang et al., 2022). The C-glycosyl substituents further activate the aromatic ring of the polyphenol to readily donate an electron towards antioxidant function (Kanamori et al., 2018;Mannem et al., 2020;Marrelli et al., 2014;Materska, 2014;Wen et al., 2017;Xiao et al., 2014;Xie et al., 2020). Contrariwise, an O-glycosyl substituent can destroy the antioxidant properties of the polyphenol, depending on the glycosylation site used (Xiao, 2015). The polyphenol bioavailability linked to its solubility in water is strongly altered through the appendage of the di-C-β-glucoside motif (Sato et al., 2020). Although various polyphenol di-C-β-glucosides have been reported from different plant species (Oualid & M. S. Silva, 2012), their natural abundance is not sufficient for large-scale isolation from the immediate source (Shie et al., 2010). Low availability often limits even the systematic early-stage assessment of their biological activity as basis for the development of future uses (Kaur & Kaur, 2014).
Synthetic routes for the facile installment of the di-C-β-glucoside motif on different polyphenol core structures are therefore highly desirable. Here we show a biocatalytic process technology for an efficient and flexible di-C-β-glucoside production.
The biosynthesis of di-C-β-glucosides involves a special subclass of sugar nucleotide-dependent (Leloir) glycosyltransferases (GTs) . The enzymes, here referred to as CGTs, use uridine-5′diphosphate (UDP)-glucose for site-selective C-β-glycosylation of the polyphenol acceptor substrate (Figure 1) (Chong et al., 2022;Dai et al., 2021;Gao et al., 2022;Putkaradze et al., 2021;Tegl & Nidetzky, 2020;Y. Q. Zhang et al., 2022). The iterative two-fold F I G U R E 1 Reaction scheme for the synthesis of a polyphenol-di-C-β-glucoside by C-glycosyltransferase FcCGT, coupled with UDP-Glc production and regeneration by a sucrose synthase GmSuSy. PTHAP, 2-phenyl-2′,4′,6′-trihydroxyacetophenone; UDP, uridine-5′-diphosphate; UDP-Glc, UDP-glucose. glycosylation leading to the di-C-β-glucosides may arise from the coordinated action of separate CGTs (Feng et al., 2021;Y. Sun, Chen, et al., 2020;, but as recent works have shown, a single dual-specific di-CGT can also be sufficient to form the fully di-glycosylated product. Within a rapidly growing number (≥67) of plant CGTs isolated and characterized (Bao et al., 2022;Y. Q. Zhang et al., 2022), only a few enzymes manage the difficult task of iterative glycosylation to release the di-C-glucoside product: FcCGT, Fortunella crassifolia and CuCGT, Citrus unshiu (Ito et al., 2017); MiCGTb, Mangifera indica (Chen, Fan, et al., 2018); GgCGT, Glycyrrhiza glabra (M. Zhang et al., 2020); DcaCGT, Dendrobium catenatum (Ren et al., 2020); AbCGT, Aloe barbadensis ; AaCGT, Anemarrhena asphodeloides (Huang et al., 2022), and AdCGT, Angelica decursiva (Z. Wang et al., 2022). Compared to CGTs performing only a single C-glycosylation, a relatively open and spacious acceptor substrate-binding pocket appears to be requirement of the protein structure for di-CGT activity (Chen, Fan, et al., 2018;. The known di-CGTs were analyzed for substrate scope in diversity-oriented small-scale syntheses done as part of their characterization (Chen, Fan, et al., 2018;Ito et al., 2017;. While the enzymes hold promise as selective catalysts of the di-C-β-glucoside formation, it remained to be shown that di-CGTs are applicable to synthetic glycosylation reactions that fulfill the requirement of being amenable to a biocatalytic production at larger scale in principle. In particular, glycosylation from substrate more expedient than UDP-Glc and reaction intensification to obtain the di-C-β-glucoside as single product in high yield and concentration were important tasks for development in the current study. Indeed, earlier attempts  to synthesize di-C-βglucosides in CGT-expressing E. coli cells gave the target product in relatively low yield (≤2.8%, based on the limiting acceptor concentration) and rather insufficient amount (≤0.04 g/L). Here, we show the FcCGT for an efficient synthesis of the di-C-β-glucoside of phloretin, a dihydrochalcone of the flavonoid class and representative polyphenol of plant origin (Behzad et al., 2017). To become truly efficient, the enzymatic di-glycosylation required intensification (≥100-fold) along three lines of reaction engineering: (1) cascade reaction development for an overall glycosylation that proceeds fully to the di-glycosylated product and uses an expedient substrate, here sucrose ( Figure 1); (2) aglycone solubility enhancement through means compatible with enzyme activity and stability, here encapsulation in 2-hydroxypropyl-β-cyclodextrin (HPCD; H. Schmölzer et al., 2018); and (3) di-C-β-glucoside product stabilization during the reaction, as discovered in the current study. A biocatalytic process technology for the selective production of the phloretin-3′,5′-di-C-β-glucoside in high reaction yield (≥95%), concentration (∼40 mM; 24 g/L), and good recovery rate (≥80%) as a high-purity product (≥98%) is shown. Replication potential of the technology is demonstrated for production of the 3′,5′-di-C-β-glucoside of 2-phenyl-2′,4′,6′-trihydroxyacetophenone (PTHAP; Figure 1) which is equally efficient as the phloretin di-glycosylation by the criterions used.

| Encapsulation of acceptor substrates by HPCD
HPCD inclusion complexes of phloretin were prepared according to a protocol from literature (Schmölzer et al., 2018), with the molar ratio of 1.25:1 for HPCD to phloretin. The PTHAP inclusion complexation was performed with a modified protocol from that described in the literature (Schmölzer et al., 2018), using the molar ratio of 2:1 for HPCD to PTHAP. The actual concentrations of encapsulated acceptor substrates were determined based on calibration curves (Supporting Information: Figure S1). Full details of the inclusion complex formation and concentration measurements are given in the Supporting Information under "Encapsulation of acceptor substrates by HPCD".

| Enzyme production
The codon-optimized synthetic gene of FcCGT (Gene UGT708G1; Supporting Information: Figure S2) in a pET-28a expression vector was purchased from GenScript (Germany) and transformed into E. coli

| Enzyme activity assays
The specific activity of FcCGT towards free and HPCD-encapsulated phloretin/PTHAP was determined in an enzymatic assay containing 1.0 mM acceptor substrate and 5.0 mM UDP-Glc. The initial rates for mono-and di-C-glycosylation were calculated from the corresponding time courses (Supporting Information: Figures S4 and S5). The activity of GmSuSy on UDP (2.0 mM) was determined from a time course (Supporting Information: Figure S6) of the reaction with 500 mM sucrose. Full details of the activity assays are provided in the Supporting Information under "Enzyme activity assays". based on literature protocols (Lemmerer et al., 2016;Schmölzer et al., 2018). Aliquots (80 µL) of the thawed reaction mixture were additionally subjected to liquid-liquid extraction with organic solvents (80 µL; 1-butanol, 1-octanol, diethyl ether, ethyl acetate, dichloromethane). The extraction was repeated three times, organicand aqueous phases collected and analyzed on TLC (Supporting Information: Figure S19A). Up-scaled extraction with 1-butanol: Reaction mixture (0.7 mL) was extracted with 1-butanol (3 × 3.0 mL), the organic phases combined and solvent removed under reduced pressure on a Laborota 4000 rotary evaporator (Heidolph) at 40°C.
The solid material was dissolved in acetone (2.0 mL) followed by addition of deionized water (2.0 mL) and the samples analyzed by TLC (Supporting Information: Figure S19B). For the final product isolation via TLC, the remaining reaction mixture (1.20 mL) was loaded onto silica plates (20 cm × 20 cm, layer thickness 2 mm; Merck). The target product was isolated from the TLC plate (Supporting Information: Figure S20) as described in the Supporting Information under "Di-C-β-glucoside product isolation". The purity and identity of phloretin-3′,5′-di-C-β-glucoside were analyzed on HPLC (Supporting Information: Figure S21) and nuclear magnetic resonance (NMR) (Supporting Information: Figure S22).

| Biocatalytic reaction system
The C-glycosyltransferase FcCGT was used for its reported activity of di-β-C-glycosylation of both phloretin and PTHAP (Ito et al., 2017).
Among the various CGTs reported in the literature, FcCGT appeared to be a promising candidate enzyme to be used in a study of reaction intensification. The FcCGT was isolated from Escherichia coli expression culture at 3.3 mg protein/L (Supporting Information: Figure S3). The purified FcCGT (see the Supporting Information Methods for the procedure used) had a specific activity (mono-C-glycosylation of phloretin from UDP-Glc; Supporting Information: Figure S4A) of 3.02 U/mg (±1%; N = 2). The specific activity for the further C-glycosylation of nothofagin (Supporting Information: Figure S4A) was ∼7.4-fold lower at 0.41 U/mg (±10%; N = 2). Earlier work on FcCGT described the enzyme as more active towards nothofagin (3.57 U/mg) than phloretin (2.04 U/mg) (Ito et al., 2017).
In other di-glycosylating O-or C-glycosyltransferases, however, the second glycosylation was slower considerably than the first (Chen, Fan, et al., 2018;. Using PTHAP as the acceptor, FcCGT had a specific activity of 3.25 U/mg (±6%; N = 2) and 0.48 U/mg (±7%; N = 2) for the first and the second C-glycosylation, respectively (Supporting Information: Figure S5A). Standard enzyme assays performed with the acceptor-HPCD complex instead of the free acceptor gave specific activities of 1.82 U/ mg (±8%; N = 2) and 0.09 U/mg (±10%; N = 2) for the first and second Cglycosylation on the phloretin (Supporting Information: Figure S4B).
Compared to the free phloretin, the inclusion complex showed lower apparent activity in particular in the second C-glycosylation (4.6-fold decrease). Interestingly, therefore, the FcCGT exhibited even higher apparent activity towards PTHAP-HPCD than free PTHAP, with specific activities of 3.92 U/mg (±3%; N = 2) and 0.65 U/mg (±3%; N = 2) determined for the first and second C-glycosylation, respectively (Supporting Information: Figure S5B). Phloretin and PTHAP may differ in various ways regarding their inclusion complexes with HPCD in the nonglycosylated as well as the mono-C-glycosylated form (e.g., molecular orientation and dynamic equilibrium of the host-guest molecule interaction), as generally suggested by literature on complexation by cyclodextrins (Cesari et al., 2020;Fang & Bhandari, 2010;Gratieri et al., 2020;Kellici et al., 2016;Zheng et al., 2005) While interesting, an exploration into the molecular origins of the observed effects was not within the scope of the current study. With both acceptors, however, feasibility of inclusion complexation for FcCGT substrate solubility enhancement was demonstrated in view of the intended di-C-glycoside synthesis.

| Product stability during the enzymatic reaction
Using initially a low concentration of the acceptor (1.0 mM) in free form or as the HPCD complex, we monitored the time course of C-glycoside formation in the FcCGT reaction, now aiming at full conversion into the doubly C-glycosylated product. Unexpectedly, we discovered the di-C-β-glucoside to decompose rapidly in the reaction, irrespective of whether phloretin or PTHAP was used ( Figure 2a). The corresponding mono-C-β-glucoside appeared to be stable. Inclusion complexation of the acceptor substrate did not enhance the stability of the di-C-glycoside formed. Analyzing the phloretin-3′,5′-di-C-β-glucoside in more detail, we showed that after extended reaction (≥8 h) nearly all of the compound was degraded. pH variation in the range 6.0-8.0 was not effective to achieve stabilization of the di-C-glycoside (Figure 2b), nor was exchange of the buffer salt, Tris/K 2 HPO 4 (Figure 2c). Considering the decomposition to possibly arise from oxidation events, we examined the effect of reducing agent and antioxidant. The addition of tris(2carboxyethyl)phosphine (TCEP, 2.0 mM) and L-ascorbic acid (1.0 mM) was not sufficient to prevent the di-C-glucoside decomposition ( Figure 2d). Finally, 2-mercaptoethanol (12.5 mM) proved highly efficient in preventing the decomposition, and phloretin-3′,5′-di-C-β-glucoside was found intact even after 24 h of reaction ( Figure 2e). Applying the same conditions to the glycosylation of PTHAP, we show the corresponding 3′,5′-di-C-β-glucoside is also stabilized effectively (Figure 2f). Overall, therefore, these results demonstrate that enzymatic synthesis of the di-C-β-glucosides of phloretin and PTHAP necessitates stabilization of the product against oxidative decomposition. Low stability is ascribed to the inductive effect of the two C-β-glucosyl moieties in the context of a highly substituted, and thus activated, aromatic system. Specific applications of the di-C-glycosides that harness their reactivity toward use as pharmacological agents (e.g., antioxidants) or photosensitizers can thus be envisaged (Kanamori et al., 2018;Mannem et al., 2020;Marrelli et al., 2014;Materska, 2014;Wen et al., 2017;Xiao et al., 2014;Xie et al., 2020).

| Enzyme cascade reaction for di-C-glycosylation of phloretin
For synthesis, we combined the di-C-glycosylation of phloretin (HPCD complex) by FcCGT with in situ production of UDP-Glc from sucrose ( Figure 1). The sucrose synthase from soybean (Glycine max; GmSuSy) was used (Supporting Information: Figure S3B). Except for 2-mercaptoethanol added (10 mM), the basic reaction conditions ( Figure 3) were taken from our earlier work on enzyme cascade glycosylation using GmSuSy (H. Schmölzer et al., 2018). This included UDP used at 0.5 mM and sucrose present in large excess (500 mM). As shown in Figure 3a, phloretin was fully converted into the 3′,5′-di-C-β-glucoside. Consistent with the results of the activity assays described earlier, the mono-C-β-glucoside (nothofagin) formation rate was much faster (∼8-fold) than the di-Cglycoside formation rate (Table 1,  The intensification was aligned to the process tasks of complete conversion of the acceptor substrate into the di-C-glycoside product and minimized decomposition of the target product during the reaction. The selected tasks implied a particular focus on fast reaction for selective transformation, while economy of enzyme use, despite its undeniable importance for a comprehensive optimization, had lower priority. A summary of the results is given in Table 1. Compared to the 10 mM reaction, rate acceleration by increases in enzyme loading and temperature (30°C→45°C) was important at higher phloretin concentrations (e.g., 30 mM; FcCGT stability appeared to be a main factor limiting the conversion obtained. Change in the UDP concentration (0.5-3.0 mM) was ineffective (Table 1, entry 6), suggesting that 0.5-1.0 mM were sufficient. Increase in the GmSuSy loading to change the mass ratio with FcCGT from the usually used 2:1 to 4:3 (Table 1, entry 8) did not affect the phloretin conversion. In all 50 mM reactions, the phloretin was fully converted to nothofagin, but the conversion into the 3′,5′di-C-β-glucoside did not proceed to completion (Supporting Information: Figures S9-S14). The initial rates of release of the mono-and the di-C-glycoside were largely unaffected by the change in reaction conditions used (Table 1, entries 4-9). The C-glycosylation of the intermediary nothofagin evidently was too slow in the 50 mM reactions to exclude the effect of time-dependent loss of FcCGT activity. Effect noted by Welner and co-workers that family GT1 natural product GTs can show low stability toward their acceptor substrate (Teze et al., 2022), could be relevant here. Evidence that the UDP-Glc/UDP ratio was shifted to a higher value (e.g., 1.0→1.8; Table 1 Table 1, entry 10), giving a space-time yield of ∼10 mM/h for the di-C-glycoside product release. The UDP-Glc/UDP ratio was balanced at ∼1.2 ( Figure 4c and Table 1, entry 10).

Significant (~60%) decomposition of UDP and UDP-Glc to uridine
and UMP were observed, indicating the high enzyme concentration used to play a role in the donor substrate degradation (Figure 4d).

| Enzyme cascade reaction for di-C-glycosylation of PTHAP
Using conditions established for the phloretin reaction, we examined enzyme cascade transformation for synthesis of the PTHAP-3′,5′-di-C-β-glucoside. Table 1 (entries 11 and 12) shows the results. The conversion of PTHAP (30 mM, HPCD complex) was considerably faster (~3-fold) in both glycosylation steps than the comparable conversion of phloretin (Table 1; entries 11 and 3). It proceeded to a quantitative yield of the di-C-β-glucoside product (Supporting Information: Figure S15A). The UDP-Glc/UDP ratio in the PTHAP reaction (Supporting Information: Figure S15B and Table 1, entry 11) was notably higher than in the phloretin reaction (Supporting Information: Figure S8 and Table 1, entry 3), indicating a relatively higher rate of UDP-Glc release by the GmSuSy when PTHAP was used. We increased the PTHAP concentration to 50 mM while supplying the same amount of enzyme (Table 1, entry 12) and show full conversion in 24 h (~85% in 8 h; Figure 5a,b). Considering the requirement of full conversion of the acceptor substrate to fulfill the set process tasks, the reaction at 30 mM PTHAP gave a space-time yield (~10 mM/h) substantially higher (5-fold) than that of the reaction at 50 mM PTHAP (~2 mM/h). It seems that the high PTHAP concentration resulted in an inhibition of FcCGT in the second C-glycosylation step of the overall transformation (Table 1, entries 11 and 12). Full-fledged optimization of the space-time yield of the di-C-glycoside product was beyond the scope of the current study. However, the evidence shown suggests the existence of an optimum substrate concentration for the space-time yield higher than 10 mM and lower than 50 mM.
In the reaction with 50 mM PTHAP, the UDP-Glc donor substrate of FcCGT exhibited interesting dynamics, as shown in Figure 5c. At reaction start, there was an almost instantaneous release of UDP-Glc, corresponding to~80% of the total UDP present.
The UDP-Glc concentration then dropped rapidly to~50% of the total UDP, reflecting the usage of the UDP-Glc in the fast mono-Cglycosylation of the PTHAP. As the second C-glycosylation was slower than the first, the UDP-Glc consumption rate decreased and the UDP-Glc concentration increased again to~50% of the total UDP supplied. Evidence that the initial UDP-Glc concentration was not The initial rates were determined from the linear part of the time courses by multiplying the slope of the linear regression (%/min) with the substrate concentration (mM) giving the initial rate in mM/min. dryness and re-dissolved in acetone with the aim of increasing the purity of the desired product. Phloretin-di-C-glucoside was not found in the acetone extracts, yet it could be identified (together with sucrose, fructose, and HPCD) on TLC when the crystals from 1-butanol extraction were further dissolved in water (Supporting Information: Figure S19B).
Attempts to obtain the PTHAP-3′,5′-di-C-β-glucoside by analogous procedure of preparative TLC were unsuccessful due to complete decomposition of the product while re-extracting it from the silica gel. Using column chromatography on a silica C18 resin, the di-C-glycoside was eluted together with the HPCD.
Re-chromatography of the concentrated product fractions on the same C18 column did not yield the di-C-glycoside in sufficient purity (data not shown). Isolation of the PTHAP-3′,5′-di-C-β-glucoside was eventually accomplished by a combination of C18 and silica 60 columns. The silica 60 column operated with the TLC eluent as the mobile phase succeeded in separating the PTHAP-3′,5′-di-C-βglucoside from the other compounds, but it also led to a silica contamination. The C18 column operated with water was used to wash out the dissolved silica. The target compound was eluted with acetonitrile. The explorative purification procedure here used gave the authentic PTHAP-3′,5′-di-C-β-glucoside (NMR data in Supporting Information: Figure S25) in high purity (≥95%, HPLC data in Supporting Information: Figure S24), thus fulfilling the basic requirements for follow-up applications. The product yield (4.9 mg;~20%) is considered preliminary at this point, due to the number of purification steps applied. A streamlined procedure would involve two-column chromatographic steps only, silica 60 first, C18 afterwards.
Lastly, we performed a preliminary investigation on the spontaneous decomposition of the di-C-β-glucoside products. The isolated phloretin-3′,5′-di-C-β-glucoside was analyzed in incubations at pH 8.0 (2.0 mM; 1.0 mL in water). Extensive set of HPLC-UV/MS data suggests degradation into multiple species apparently bearing a multihydroxylated aromatic core structure (Supporting Information: Figures S26 and S27). Degradation seems to involve loss of the two glucose residues from the aromatic core in ways that must await further study for clarification.