Structural insights into the catalytic mechanism of the phenylethanoid glycoside rhamnosyltransferase UGT79G15 from Rehmannia glutinosa

Abstract

Phenylethanoid glycosides (PhGs) are a group of important natural products widely distributed in medicinal plants and known for their remarkable pharmacological properties. Uridine diphosphate (UDP) glycosyltransferase 79G15 (UGT79G15) from Rehmannia glutinosa catalyzes the conversion of osmanthuside A to osmanthuside B, a key intermediate in the PhG biosynthetic pathway, through the formation of a (1→3) glycosidic bond. In this study, we present the crystal structures of UGT79G15 in its apo form, UDP-bound form, and, notably, its ternary complex containing UDP and a mimic acceptor, forsythiaside A, within its active site. Structural and comparative analyses revealed that UGT79G15 possesses a distinctive funnel-shaped acceptor-binding pocket with a small auxiliary cavity capable of accommodating the 4′-hydroxycinnamoyl group of PhGs, explaining the enzyme’s regiospecificity toward the 3′-OH of the acceptor. Additional structural examination and site-directed mutagenesis identified key residues that recognize and stabilize UDP-rhamnose and the sugar acceptor. Among the variants generated, I204W exhibits enhanced catalytic efficiency for osmanthuside A conversion, reaching up to 2.2-fold higher activity than the wild type. This study provides mechanistic insight into the donor specificity and acceptor regioselectivity of PhG 1,3-rhamnosyltransferase and expands the structural understanding of plant UGTs.

UGT79G15, a 1,3-rhamnosyltransferase from Rehmannia glutinosa, catalyzes the formation of osmanthuside B, an essential intermediate in the phenylethanoid glycoside (PhG) biosynthetic pathway. This study reports the crystal structures of UGT79G15 in its apo form, UDP-bound form, and ternary complex form containing UDP and a mimic acceptor, forsythiaside A, providing new structural insights into its catalytic mechanism and informing the engineering of variants with improved catalytic performance.

Introduction

Phenylethanoid glycosides (PhGs) are a class of hydrophilic natural products widely distributed throughout the plant kingdom. To date, nearly 600 PhGs have been isolated and characterized from medicinal plants such as Cistanche tubulosa, Forsythia suspensa, and Rehmannia glutinosa (Fuji et al., 2024). PhGs have drawn significant attention owing to their diverse pharmacological properties, including antioxidant, antibacterial, antimetastatic, antiviral, and neuroprotective effects (Xue and Yang, 2016). Structurally, PhGs comprise a β-D-glucopyranose core linked to both an aromatic acid unit and a hydroxyphenylethyl unit. The glucose moiety can be modified by additional sugar residues; rhamnose is the most common second sugar, typically attached to the C-3′ or C-6′ position of the core glucose. Caffeic acid, cinnamic acid, and ferulic acid are the predominant aromatic acids found in PhGs and are usually conjugated at the C-4′ position of the glucose core (Xue and Yang, 2016; Fuji et al., 2024). The type of sugar bound to the aglycon moiety affects the overall activity of PhGs (Xue and Yang, 2016).

In recent years, studies have identified key enzymes involved in the pathways upstream of PhG biosynthesis (Zhou et al., 2020; Li et al., 2022); salidroside is regarded as a key intermediate (Torrens-Spence et al., 2018; Zhou et al., 2020). Additional investigations have confirmed the involvement of multiple catalytic enzymes in the p-coumaroylation, glycosylation, and hydroxylation of salidroside, ultimately forming verbascoside or echinacoside (Yang et al., 2021; Harris et al., 2023; Yao et al., 2024; Huang et al., 2025). Among these enzymes, osmanthuside A 1,3-rhamnosyltransferases catalyze the synthesis of a critical intermediate, osmanthuside B, through the formation of a (1→3) glycosidic linkage between the rhamnose moiety and the central glucose of PhG (Yang et al., 2021; Yao et al., 2024; Huang et al., 2025). These enzymes have been identified in a range of PhG-producing plants, implying important role in PhG biosynthesis.

Crystal structures are essential for elucidating enzyme–substrate interactions and catalytic mechanisms, as well as guiding rational enzyme engineering. Since the first determination of the crystal structure of plant-derived uridine diphosphate (UDP) glycosyltransferase 71G1 (UGT71G1), approximately 40 plant glycosyltransferase crystal structures have been resolved (Shao et al., 2005; Wang et al., 2023; Jiang et al., 2025). Despite the large number of identified UGT enzymes, only a few studies have focused on their crystal structures, mainly due to the challenges associated with obtaining high-quality crystals. This difficulty is particularly evident for ternary complexes containing acceptors bound to the enzyme. Among glycosyltransferases with available structures, the apo form or the binary complex form with UDP at the sugar-donor-binding site is most frequently observed (Yao et al., 2022). To date, only a few UGTs have been crystallized in the ternary complex form, including Glycyrrhiza glabra C-GT complexed with UDP/phloretin or UDP/nothofagin (Zhang et al., 2020) and UGT76G1 complexed with UDP/RebB (Liu et al., 2020). These complex crystal structures have provided valuable insight into the catalytic mechanisms of GTs. Recently, the crystal structure of Callicarpa macrophylla GT3, a homolog of UGT79G15, was reported; however, it was not determined in the ternary complex form (Yao et al., 2024), and the detailed acceptor–substrate recognition mechanism of this enzyme remains unclear. The scarcity of such crystal structures has hindered comprehensive understanding of the structure–function relationships and glycosylation mechanisms of PhG UGTs. It is therefore crucial to clarify the three-dimensional structural determinants governing substrate specificity in PhG rhamnosyltransferases.

UGT79G15 from R. glutinosa, a previously reported rhamnosyltransferase, catalyzes regiospecific glycosylation of the 3′-OH of osmanthuside A (Yang et al., 2021). In this study, we determined the crystal structures of UGT79G15 in its apo form, UDP-bound form, and UDP/mimic (forsythiaside A)-bound form. Structural analysis enabled the construction of multiple variants, through which we identified key residues involved in recognizing and stabilizing UDP-rhamnose (UDP-Rha) and the sugar acceptor. Furthermore, we revealed a distinct acceptor-binding pocket and critical residues responsible for 3′-rhamnosylation of the central glucose moiety of PhGs.

Results and discussion

Crystallization and overall structures of UGT79G15 and its complexes

To explore the structural basis of substrate recognition and the catalytic mechanism of UGT79G15, we attempted to obtain a ternary complex of the enzyme. Accordingly, co-crystallization and crystal-soaking experiments were performed using the native substrate osmanthuside A. Unfortunately, the ternary complex could not be obtained because osmanthuside A was unstable in buffer; its cinnamoyl group spontaneously migrated among multiple hydroxyl groups on the central glucose moiety (Yang et al., 2021; Yao et al., 2024). These issues were successfully resolved using a mimic substrate, forsythiaside A, which contains a 6′-rhamnosyl group that prevents spontaneous migration of the caffeoyl group. Forsythiaside A was successfully recognized by the enzyme as a sugar acceptor and converted to poliumoside (Figure 1A; Supplemental Figure 1). We obtained crystal structures of three forms of UGT79G15: apo (UGT79G15_apo), UDP-bound (UGT79G15_UDP), and UDP/forsythiaside A-bound (termed UGT79G15_UDP_FA), with high resolutions of 2.4, 1.8, and 2.12 Å, respectively (Supplemental Table 1).

Figure 1.

Function and crystal structure characteristics of UGT79G15.

(A) Schematic representation of the reaction catalyzed by UGT79G15.

(B) Ribbon representation of the UGT79G15 crystal structure showing forsythiaside A and UDP at the active site with a narrow entrance. The N-terminal domain is shown in green, and the C-terminal domain in blue.

(C) Superposition of the structures of UGT79G15_apo (orange), UGT79G15_UDP (green), and UGT79G15_UDP_FA (lilac).

(D) Acceptor-binding pocket of UGT79G15.

The structure of UGT79G15 was found to contain a GT-B fold, a characteristic feature of all known plant UGTs. Its N-terminal domain (NTD) and C-terminal domain (CTD) form a narrow catalytic cleft that permits the binding of specific sugar donors and acceptors (Wang et al., 2023). In UGT79G15_UDP_FA, the NTD, which contains the substrate-binding pocket, predominantly interacts with forsythiaside A, whereas the CTD primarily associates with UDP (Figure 1B). The NTD consists of seven β-strands (Nβ1–Nβ7) surrounded by nine α-helices (Nα1–Nα9, residues 6–252 and 441–456), whereas the CTD comprises six β-strands and 10 α-helices (Cβ1–Cβ6 and Cα1–Cα10, residues 253–440). The three structures were highly similar, indicating that binding of UDP and forsythiaside A induced no significant conformational changes in the enzyme (Figure 1C).

Catalytic mechanism of UGT79G15

In UGT79G15_UDP_FA, the 3′-OH group of forsythiaside A is positioned at the bottom of the active pocket, where it forms hydrogen bonds with H21, which functions as a key catalytic residue and general base. UGT79G15 catalyzes glycosylation specifically at the 3′-OH group of the glucose moiety, despite the presence of several potential O-glycosylation sites within the PhG. The H21–D118 (2.9 Å) motif, a highly conserved catalytic dyad in plant UGTs (Wang et al., 2023), was located near the glycosylation site of forsythiaside A. To investigate the catalytic mechanism of UGT79G15, we constructed UGT79G15_UDP-Rha_FA/OA models by molecular docking of Rha and osmanthuside A into the UGT79G15_UDP_FA structure (Figures 2A and 2B). Molecular docking analysis showed that forsythiaside A binds in an orientation that places its 3′-OH group near the anomeric C-1′ of the rhamnose moiety of UDP-Rha, explaining the transfer of the sugar to the C-3′ position of the substrate (Figure 2C). However, the distance between the 3′-OH of forsythiaside A and the anomeric C-1′ of rhamnose exceeded 7 Å, implying that forsythiaside A moves closer to the anomeric carbon during the catalytic process. Glycosylation by glycosyltransferases involves multiple conformational states—holo, donor-bound, substrate-bound, product-bound, and product-release—each characterized by substantial movements of proteins and ligands. Distinct binding conformations have been observed in the crystals of UGT76G1 (Liu et al., 2020), and structural studies of UGT74AN2 revealed that the Nα4 helix and loop-c undergo rotation of approximately 7.4° and displacement of ∼2.4 Å (Huang et al., 2022). We speculate that UGT79G15 follows a similar catalytic trajectory. In our docked models based on the crystal structure, the relatively long distance between the 3′-OH group and C-1′ of rhamnose likely represents the initial binding state. The substrate-binding pocket of UGT79G15 is large enough to accommodate conformational shifts of forsythiaside A or osmanthuside A and to permit movement of the catalytic motif toward UDP-Rha until a reactive distance is achieved.

Figure 2.

Analysis of the catalytic mechanism of UGT79G15.

(A) Docking structure of UGT79G15_UDP-Rha_forsythiaside A.

(B) Docking structure of UGT79G15_UDP-Rha_osmanthuside A.

(C) Residues interacting with UDP-Rha and forsythiaside A.

(D) Residues interacting with UDP-Rha and osmanthuside A.

(E) Proposed catalytic mechanism of UGT79G15.

(F) Validation of the catalytic site of UGT79G15 using osmanthuside A or forsythiaside A as substrates and UDP-Rha as the sugar donor. N.D., not detected.

Single-point mutants of the enzyme were constructed by site-directed mutagenesis at residues H21 and D118, then tested using forsythiaside A and osmanthuside A as sugar acceptors. The H21A variant exhibited no detectable glycosylation activity, whereas the H21K variant retained modest activity, likely due to the basic nature of lysine, which carries a positive charge similar to that of histidine. Variants D118H and G20A also showed a substantial reduction in enzymatic activity (Figure 2F). These results align with the proposed S_N2-like nucleophilic substitution reaction mechanism, in which His acts as the catalytic base by deprotonating the hydroxyl oxygen of the acceptor substrate, and Asp serves as a charge stabilizer after proton abstraction (Wang et al., 2023) (Figure 2E). The conserved G20 residue, which lacks a side chain, is essential for catalysis and facilitates the movement of the neighboring His residue (Huang et al., 2022). Notably, the D118H variant (charge reversal) retained over 20% of the relative activity of the wild-type (WT) enzyme (Figure 2E). In the rhamnosyltransferase UGT89C1, H21 functions as the sole catalytic residue because the corresponding Asp residue is absent. It has been suggested that H21 and the sugar acceptor form an acceptor–His dyad, where H21 alone acts as the catalytic residue (Zong et al., 2019). However, the catalytic mechanism in the D118H variant of UGT79G15 remains unclear. The structural context and precise role of H118 in this mutant require structural validation to confirm its positioning and potential participation in a catalytic dyad.

Recognition and binding of UDP-Rha

In the structures of both complex forms, the UDP molecules were closely superimposed and localized within the CTD (Figure 1C). The UDP-binding site contains a highly conserved motif characteristic of the plant secondary product GT family (Wang et al., 2023) (Figure 3). The UDP molecule resides within a hydrophobic channel formed by aromatic residues, including Trp, Tyr, and Phe. The uracil loop of UDP is sandwiched between Q337 and W334, forming a parallel π–π interaction. Additionally, the uracil ring forms a hydrogen bond with V335, positioned 2.9 Å away, through its nitrogen atom. Q337 and E360 each form hydrogen bonds with the ribose ring of UDP at distances of 3.0 and 2.8 Å, respectively.

Figure 3.

UDP molecule and its interaction with UGT79G15.

(A) Comparison of the UDP-binding pockets of UGT79G15_apo (wheat) and UGT79G15_UDP (light cyan).

(B) Comparison of the distances between UDP and surrounding residues in UGT79G15_apo (wheat) and UGT79G15_UDP (light cyan).

(C) Two-dimensional interaction analysis of UDP and its surrounding residues.

By superimposing the UDP-binding pockets of UGT79G15_apo and UGT79G15_UDP, we found that UDP binding induced a slight displacement of several residues (including W334 and E360) in opposite directions, resulting in a reduced binding pocket size (Figure 3A). The distances between UDP and the surrounding residues—S276, W334, V335, G356, S357, and E360 (Figure 3B)—were reduced by approximately 0.2–0.8 Å, particularly for G276, S357, W334, and V335, which are positioned near the diphosphate moiety and the uracil loop of UDP, respectively. These findings indicate a localized conformational adjustment upon ligand binding and suggest a substrate-induced fit mechanism in which nearby residues shift slightly to create a more compact cavity that secures UDP more tightly. Two-dimensional interaction analysis using LIGPLOT (Wallace et al., 1995) (Figure 3C) showed that the diphosphate moiety of UDP forms hydrogen bonds with residues S276, H352, G356, and S357, whereas the uracil loop interacts with V335, situated between the NTD and CTD.

The sugar donor selectivity of plant UGTs is known to depend on hydrogen-bonding interactions between the hydroxyl groups of the sugar and certain amino acid residues within the enzyme’s active site. When UDP-glucose, UDP-xylose, UDP-arabinose, UDP-N-acetylglucosamine, and UDP-glucuronic acid were tested as sugar donors, no product was detected, indicating that UGT79G15 is strictly specific for UDP-Rha (Supplemental Figure 2). In the UGT79G15_UDP-Rha_FA model (Figure 2A), the rhamnosyl group is accommodated in a pocket formed by W16, H21, D118, V139, Q277, and Q377. Among these residues, Q377—part of the conserved plant secondary product GT motif—forms a stable interaction with the rhamnose moiety (2.5 Å) (Figure 2C). Notably, in UGT89C1, the first plant rhamnosyltransferase with a resolved crystal structure, the corresponding residue to Q377 is H357, which was identified as a key residue for UDP-Rha recognition (Zong et al., 2019). Moreover, residue V139 in UGT79G15 occupies the equivalent position of P147 in UGT89C1 and may form a hydrophobic interaction with the methyl group of rhamnose. In plant glucosyltransferases, this position is occupied by a highly conserved threonine residue, critical for UDP-glucose recognition (Supplemental Figure 3) (Offen et al., 2006; Yao et al., 2024). Mutation of V139 to threonine (V139T) did not confer activity toward UDP-glucose but almost abolished catalytic activity for UDP-Rha (Supplemental Figure 1B), suggesting that V139 plays an important role in UDP-Rha recognition by UGT79G15.

In the docking model of UGT79G15_UDP-Rha_OA, residue Q277 interacted not only with UDP-Rha (3.2 Å) but also with the 4-OH group of osmanthuside A (2.9 Å) (Figure 2D), suggesting a critical role in UGT79G15 activity. We speculate that the side-chain amide group (-CONH₂) of Q277 recognizes the sugar receptor through hydrogen bonding and stabilizes the transition-state structure. When Q277 was replaced with Glu, Asp, or Asn (i.e., residues without -NH₂ or with a shorter side chain), the enzymatic activities of the Q277D, Q277E, and Q277N variants were substantially reduced. These results confirm that Q277 is essential for UGT79G15 function. The amide group of Q277 may serve a dual role as both a hydrogen bond donor (N–H) and acceptor (C=O), acting as a key mediator linking the substrates to the enzyme’s active site. Disruption of either interaction weakens binding with the donor or acceptor. Mutation of residue Q377, which surrounds the rhamnose moiety, to histidine or asparagine (Q377H and Q377N) resulted in an almost complete loss of catalytic activity because Q377 forms strong interactions with the rhamnose ring through its longer side chain (Figure 2F). These results highlight the essential roles of residues V139, Q277, and Q377 in binding UDP-Rha and catalyzing the rhamnose transfer reaction.

The unique pocket for acceptor binding and roles of key residues in catalytic activity

The structure of UGT79G15_UDP_FA was determined at a resolution of 2.12 Å. This complex revealed that forsythiaside A binds within a spacious hydrophobic pocket composed of several aromatic and nonpolar residues, including W16, H21, I84, I86, F119, I140, I149, F200, I204, and Q277 (Figure 4). The sugar-acceptor-binding pocket of UGT79G15 forms a unique heterogeneous, funnel-shaped cavity that considerably differs from the acceptor-binding pockets of other plant UGTs utilizing terpenoids or flavonoids as native substrates, such as the V-shaped pocket of UGT89C1 and the U-shaped pocket of UGT74AN3 (Zong et al., 2019; Huang et al., 2023) (Figure 5). At the base of the spacious substrate-binding cavity, a smaller hydrophobic niche formed by residues F119, I140, I149, F200, and I204 was identified (Figure 4). This feature, termed an extended accessory pocket, appears sufficiently large to accommodate the 4′-caffeoyl group of forsythiaside A. In the structure of C. macrophylla GT3 (Yao et al., 2024), a homolog of UGT79G15 from C. macrophylla, a similar accessory cave was found, suggesting that this structural element is conserved among osmanthuside A 1,3-rhamnosyltransferases. To date, such an accessory pocket has not been identified in any plant UGTs outside this subgroup. The residues forming the funnel-shaped pocket substantially differ from those of other glycosyltransferases, especially at positions W16, I149, I204, and Q277 (Supplemental Figure 4). To understand the evolutionary origin of UGT79G15, a phylogenetic analysis was conducted using glycosyltransferases involved in PhG biosynthesis and previously characterized glycoside-specific glycosyltransferases that catalyze sugar-chain elongation in plant natural products, such as flavonoids and terpenoids. The homologs of UGT79G15 formed a distinct branch closely related to flavonoid glycoside-specific glycosyltransferases (Supplemental Figure 4). These results suggest that the funnel-shaped pocket reflects adaptation to PhG substrates.

Figure 4.

Substrate-binding pocket of UGT79G15.

(A) Binding mode of UDP and forsythiaside A within the substrate-binding channel of UGT79G15.

(B) Residues interacting with UDP and forsythiaside A around the binding pocket.

(C) Two-dimensional interaction analysis of forsythiaside A and its surrounding residues.

Figure 5.

Cross-sectional diagrams of acceptor-binding pockets in UGT79G15, UGT89C1, and UGT74AN3.

(A) Funnel-shaped pocket of UGT79G15 with the substrate forsythiaside A bound inside.

(B) V-shaped pocket of UGT89C1 with the substrate quercetin bound inside.

(C) U-shaped pocket of UGT74AN3 with the substrate resibufogenin bound inside.

Our experimental results verified that the hydroxycinnamoyl group attached to the 4′-OH position of PhGs is necessary for the catalytic activity of UGT79G15. When calceolarioside B, which bears a caffeoyl group at the 6′ position, was used as the substrate, no glycosylated product was detected in the enzymatic reaction (Supplemental Figure 1A). This finding suggests that the accessory pocket accommodates only a hydroxycinnamoyl group located at the 4′ position of PhGs. van der Waals interactions between the benzoic ring of forsythiaside A and hydrophobic residues within the sugar-acceptor-binding pocket stabilize the substrate such that its 3′-OH is oriented toward the anomeric carbon of UDP-Rha. Additionally, as shown in the two-dimensional interaction analysis (Figure 4C), a hydrogen bond between the 2′-OH group of forsythiaside A and the amino group of W16 (3.0 Å) further anchors the substrate in the correct orientation within the acceptor-binding pocket.

To investigate the impact of active-site residues lining the acceptor-binding pocket and to identify potential mutations that could improve catalytic performance, structure-guided variants were generated and biochemically evaluated using osmanthuside A and forsythiaside A as sugar acceptors. Alanine-scanning mutagenesis of the aforementioned key residues produced variable effects on conversion rates (Weiss et al., 2000). Most alanine substitutions caused only minor changes in glycosylation activity; variants W16A, F119A, and Y195A showed considerable reductions in glycosylation capabilities. When F119 and Y195 were modified to F119Y and Y195F, respectively, only minimal effects on enzymatic activity were observed. We hypothesize that the aromatic rings of Phe and Tyr at positions 119 and 195 substantially contribute to the hydrophobic nature of the acceptor-binding pocket. In the W16A variant, the hydrogen bond between the substrate and residue 16 was lost, resulting in reduced enzymatic activity.

Mutations of the hydrophobic residues I84, I86, I140, and I149 to other amino acids with comparable side chains (I84V, I84L, I86V, I86L, I140V, I140L, I149V, and I149L) resulted in only minor changes in enzyme activity. We hypothesize that, due to the hydrophobic nature of leucine and valine, the overall size of the substrate pocket and the hydrophobic interactions with the substrate remained largely unchanged. When these residues were substituted with bulkier side chains, the resulting variants—I84F/I84Y, I86F/I86Y, I140F/I140Y, and I149F/I149Y—exhibited lower enzymatic activity (Figure 6; Supplemental Table 2). Kinetic analyses of selected mutants confirmed a decline in catalytic efficiency (k_cat/K_M) (Table 1). These results suggest that introducing bulky hydrophobic residues restricts the space within the substrate-binding pocket and sterically interferes with the interaction between the enzyme and osmanthuside A or forsythiaside A, ultimately decreasing enzyme activity. When the polar amino acids Asp and Glu were introduced, the corresponding variants (I84D/I84E, I86D/I86E, I140D/I140E, and I149D/I149E) only slightly altered enzyme activity. The effects of these acidic substitutions were less pronounced than those of Phe and Tyr, which have large aromatic side chains, likely due to their greater distance from the substrate.

Figure 6.

Catalytic activities of UGT79G15 and its mutants.

(A) Conversion rate of the glycosylated product obtained using osmanthuside A as the substrate and UDP-Rha as the sugar donor.

(B) Conversion rate of the glycosylated product obtained using forsythiaside A as the substrate and UDP-Rha as the sugar donor.

Table 1.

Kinetic parameters of UGT79G15 toward osmanthuside A.

Protein	kcat, s−1	KM, mM	kcat/KM, mM−1·s −1
WT	2.84 ± 0.34	0.59 ± 0.14	4.77 ± 1.8
I204W	2.82 ± 0.17	0.27 ± 0.046	10.52 ± 2.5
I204A	2.89 ± 0.10	0.38 ± 0.031	7.61 ± 0.89
H21K	0.31 ± 0.07	1.33 ± 0.58	0.23 ± 0.05
D118H	0.29 ± 0.09	1.52 ± 0.77	0.21 ± 0.11
I84V	1.94 ± 0.30	0.68 ± 0.26	2.84 ± 1.2
I84F	0.34 ± 0.10	1.44 ± 0.78	0.24 ± 0.13
I86V	1.43 ± 0.27	0.75 ± 0.33	1.92 ± 0.43
I86F	1.29 ± 0.27	0.79 ± 0.37	1.63 ± 0.22
I140V	1.34 ± 0.28	0.78 ± 0.35	1.84 ± 0.46
I140F	0.93 ± 0.19	0.87 ± 0.38	1.07 ± 0.24
I149V	1.21 ± 0.24	0.84 ± 0.11	1.33 ± 0.12
I149F	0.99 ± 0.13	0.80 ± 0.23	1.24 ± 0.28

Among the variants obtained from alanine scanning, only the I204A variant exhibited slightly enhanced catalytic activity toward osmanthuside A glycosylation compared with the WT enzyme. Residue I204 is located in the funnel-shaped accessory pocket, as shown in Figure 4. To further evaluate its role, I204 was replaced with residues of differing physicochemical properties. Substitution with a hydrophobic residue (Leu, Val, Phe, or Met) or a hydrophilic residue (Asp, Asn, Ser, Thr, Glu, His, or Tyr) produced no significant change in the conversion rate for osmanthuside A but reduced activity toward forsythiaside A (Figure 6; Supplemental Table 2). However, mutation of I204 to Trp, which introduces a larger side chain containing a nitrogen atom, increased the conversion rate by approximately 30% compared with the WT enzyme. To determine whether this enhancement was due to altered affinity for the sugar acceptor, K_M values for osmanthuside A were measured for the I204W and I204A variants, as well as the WT enzyme (Table 1). The K_M value of I204A did not significantly differ from that of the WT enzyme, whereas the I204W substitution resulted in a two-fold decrease in the K_M value for osmanthuside A without a notable change in k_cat (Table 1). Consequently, the catalytic efficiency (k_cat/K_M) of I204W was 2.2-fold higher than that of the WT enzyme.

We speculate that the larger side chain of Trp enhances the hydrophobic interactions of the I204W variant with the coumaroyl group of osmanthuside A, thus shortening the distance between the enzyme and its substrate (Supplemental Figure 5). Furthermore, the nitrogen atom of Trp forms hydrogen bonds with osmanthuside A, further increasing the enzyme’s affinity for the substrate. Taken together, these results highlight the critical influence of residue 204 on substrate binding. No significant improvement in activity was observed for the I204W mutant when forsythiaside A was used as the substrate. This phenomenon may be due to the altered acceptor-binding pocket exerting minimal effect on enzyme catalysis, given that the larger size of forsythiaside A likely limits its accommodation within the pocket. Steric hindrance from the C-6′ rhamnosyl group of forsythiaside A might further restrict its binding to the I204W variant.

In this study, forsythiaside A was used as a mimic acceptor in crystallographic analysis instead of the native substrate osmanthuside A. Structurally, the main difference between the two compounds is that forsythiaside A contains a rhamnosyl group attached to the C-6′ position of the core glucose unit. In the crystal structure of UGT79G15_UDP_FA, the rhamnosyl group of forsythiaside A was oriented toward the cleft between the CTD and NTD; no apparent interaction was observed between the rhamnosyl group and the enzyme surface. Molecular docking results indicated that osmanthuside A occupies the same acceptor-binding pocket as forsythiaside A.

In summary, UGT79G15, an osmanthuside A 1,3-rhamnosyltransferase, plays an important role in the biosynthesis of PhGs. In this study, the crystal complex of UGT79G15 bound to UDP and forsythiaside A was determined for the first time, revealing a distinctive funnel-shaped pocket with a small accessory cavity in the acceptor-binding site. This pocket is essential for recognizing the distinctive core structure of PhG. However, the natural substrate osmanthuside A could not be co-crystallized with UGT79G15 due to its instability, suggesting that some binding residues differ from those observed with forsythiaside A. Both forsythiaside A and osmanthuside A fit into the funnel-shaped pocket in a similar configuration; forsythiaside A serves as a structural mimic of osmanthuside A. UGT79G15 catalyzes the rhamnosylation of the C-3′ hydroxyl group of the core glucose unit in both substrates with strict regioselectivity, implying that they share similar substrate recognition and catalytic mechanisms.

Structure-guided mutagenesis combined with enzymatic assays shed light on the substrate recognition and catalytic mechanisms of PhG 1,3-rhamnosyltransferases. Residue I204 affected the affinity of the enzyme for osmanthuside A, and the variant I204W exhibited enhanced catalytic efficiency. These findings enrich the structural understanding of plant UGTs and provide new insights into the catalytic mechanisms of UGTs involved in natural product biosynthesis.

Because the efficiency of native enzymes is often insufficient for practical applications, protein engineering offers a promising approach to improve catalytic performance and generate novel enzymes for specific purposes. This work provides a foundation for designing novel UGT79G15 variants with enhanced catalytic properties for PhG production through synthetic biology strategies.

Methods

Chemicals and reagents

UDP-Rha, osmanthuside B, and poliumoside were purchased from Acmec (Shanghai, China). Osmanthuside A was prepared as described in our previous study (Yang et al., 2021). Unless otherwise stated, all other chemicals and reagents were obtained from Aladdin (Shanghai, China).

Protein expression and purification

To obtain crystals of UGT79G15 for crystallization experiments, WT UGT79G15 was cloned into the pET28a vector (Novagen) to generate pET28a-UGT79G15, which expressed UGT79G15 with a 6×His tag at the N terminus. The plasmid was transformed into Escherichia coli BL21(DE3) cells, and protein expression was induced with isopropyl-β-D-thiogalactopyranoside. Cells were harvested by centrifugation at 5000 × g for 20 min, and all subsequent purification steps were conducted at 4°C. The cell pellet was resuspended in lysis buffer (25 mM Tris–HCl [pH 7.5], 500 mM NaCl, 20 mM imidazole, and 10% v/v glycerol) and disrupted using a French press. Cell debris was removed by centrifugation at 17 000 × g for 1 h. The supernatant was loaded onto a Ni-NTA column (Cytiva) equilibrated with the same buffer, and target proteins were eluted at approximately 100 mM imidazole. The eluted proteins were concentrated to 1 ml and applied to a Superdex 75 16/100 size-exclusion column (210 ml; Cytiva) equilibrated with gel filtration buffer (10 mM Na₂HPO₄, 10 mM KH₂PO₄ [pH 7.5], and 5 mM dithiothreitol [DTT]) at a flow rate of 1 ml/min. Purity at each stage was verified by sodium dodecyl sulfate–polyacrylamide gel electrophoresis (SDS–PAGE) (Supplemental Figure 6). The purified proteins were concentrated to 33 mg/ml in buffer containing 10 mM Na₂HPO₄, 10 mM KH₂PO₄ (pH 7.5), and 10 mM DTT for crystallization screening.

Proteins used for enzymatic assays were purified as follows. Cultured cells were pelleted by centrifugation at 10 000 × g for 20 min, resuspended in lysis buffer (50 mM Tris–HCl [pH 7.5], 500 mM NaCl, 20 mM imidazole, 1 mM β-mercaptoethanol, and 10% v/v glycerol), and lysed via ultrasonication. The lysate was clarified by centrifugation at 30 000 × g for 20 min and loaded onto a Ni-NTA column (3 ml resin; Smart Life Sciences, China). After washing with 30 ml of buffer A1 (50 mM Tris–HCl [pH 7.5], 500 mM NaCl, 20 mM imidazole, and 10% v/v glycerol), the 6×His-tagged protein was eluted with 20 ml of buffer B1 (50 mM Tris–HCl [pH 7.5], 500 mM imidazole, and 500 mM NaCl). The eluate was further purified using a StrepTactin 4FF column (Smart Life Sciences, China) pre-equilibrated with buffer A2 (100 mM Tris–HCl [pH 8.0] and 150 mM NaCl), washed with 100 ml of buffer A2, and eluted with buffer B2 (100 mM Tris–HCl [pH 7.4], 150 mM NaCl, and 2.5 mM desthiobiotin). Protein purity and concentration were assessed by SDS–PAGE and the bicinchoninic acid assay (Biosharp, China), respectively.

Crystallization, data collection, structure determination, and refinement

All crystallization experiments were conducted at 25°C using the sitting-drop vapor diffusion method. In general, 1 μl of 33 mg/ml UGT79G15 was mixed with 1 μl of reservoir solution in 48-well Cryschem plates and equilibrated against 100 μl of the reservoir at 25°C. Crystals of UGT79G15 were obtained using condition No. 88 of the Bioxtal screen kit (Xtal Quest; 0.4 M NaNO₃ and 30% w/v PEG 3350) with the sitting-drop vapor method. Within 2–5 days, the crystals reached dimensions suitable for X-ray diffraction analysis.

X-ray diffraction data were collected at beamlines BL10U2, BL17B, BL18U1, and BL19U1 of the National Facility for Protein Science in Shanghai (NFPS) at the Shanghai Synchrotron Radiation Facility (SSRF). Crystals were mounted in cryoloops and soaked in cryoprotectant solution prior to data collection at 100 K. Diffraction images were processed using HKL2000 (Otwinowski and Minor, 1997), and crystal structures were determined by molecular replacement with the Phaser program (McCoy et al., 2007) from the Phenix suite (Liebschner et al., 2019); the structure of glycosyltransferase from the Oryza sativa Japonica group (PDB: 7ES2) (Zhang et al., 2021) served as the search model. Further refinement was performed using the programs phenix.refine (Afonine et al., 2012) and Coot (Emsley and Cowtan, 2004). Prior to structural refinement, 5% of randomly selected reflections were reserved for the calculation of R_free (Brünger, 1992) as a validation parameter. Data collection and refinement statistics are summarized in Supplemental Table 1. All figures were prepared using PyMOL (http://pymol.sourceforge.net/).

Site-directed mutagenesis

UGT79G15 variants were generated using a Fast Site-Directed Mutagenesis Kit (TIANGEN, Beijing, China) in accordance with the manufacturer’s protocol. The presence of the intended mutations was confirmed by DNA sequencing (Azenta, Suzhou, China). Primer pairs used to introduce specific mutations are listed in Supplemental Table 3.

UGT79G15 enzyme activity and kinetics assays

WT UGT79G15 and all enzyme variants used in the assays were expressed and purified as described above. Enzymatic assays were performed in 100 μl reaction mixtures containing 50 mM Tris–HCl (pH 6.5), 1 mM donor substrate, 0.5 mM acceptor substrate (osmanthuside A or forsythiaside A), and 0.5 μg purified enzyme. The reactions were incubated at 30°C for 10 min, terminated by adding 100 μl ice-cold methanol, and centrifuged at 15 000 × g for 10 min. Supernatants were analyzed via high-performance liquid chromatography (HPLC). HPLC analyses of the reaction mixtures were performed on a SilGreen C18 reverse-phase column (4.6 × 250 mm) at a flow rate of 1.0 ml/min. Glycosylation products were eluted using a linear gradient from 5% methanol and 95% water (both containing 0.1% formic acid) to 80% methanol and 20% water over 30 min, followed by 100% methanol for 10 min. The column temperature was maintained at 40°C. Glucoside products were identified by comparing their retention times and mass spectra with those of authentic standards using HPLC and further confirmed by LC–high-resolution mass spectrometry on a Bruker MicrOTOF-II spectrometer equipped with an electrospray ionization source.

To determine kinetic parameters, reaction mixtures (100 μl) containing 50 mM Tris–HCl (pH 6.0), acceptor substrates (50–500 μM), 1 mM UDP-Rha, and 0.5 μg purified UGT79G15 were incubated at 30°C for 2 min. Reactions were quenched with ice-cold methanol, centrifuged at 15 000 × g for 10 min, and analyzed by HPLC. All assays were performed in triplicate. Initial reaction velocities were fitted to the Michaelis-Menten equation using nonlinear regression in GraphPad Prism 8.

Molecular docking

The automated docking program AutoDock Vina was used to dock UDP-Rha or osmanthuside A into the active site of UGT79G15. The crystal structure of UGT79G15 complexed with UDP and forsythiaside A served as the reference model, with unnecessary ligands removed prior to docking. Default genetic algorithm parameters were applied to control the docking process. All docking calculations were confined to the predicted binding pocket by defining the active site around residue His21. PyMOL was used to visualize and identify the lowest-energy docking conformations. Hydrogen bonds and van der Waals interactions between ligands and the enzyme were analyzed to determine the optimal binding mode.

Data and code availability

All data supporting the conclusions of this study are included in the main text and/or Supplemental Information. Additional information related to this work is available from the corresponding authors upon reasonable request. The coordinates and structure factors have been deposited in the RCSB Protein Data Bank under accession codes PDB: 9UEI, 9UF7, and 9UFI, corresponding to UGT79G15 in the apo form, in complex with UDP, and in complex with UDP plus forsythiaside A, respectively.

Funding

This work was supported by the National Key Research and Development Program (2020YFA0907800), the Innovation Fund of Haihe Laboratory of Synthetic Biology (22HHSWSS00023), the Science and Technology Program of Tianjin (24ZYJDSS00300), the National Natural Science Foundation of China (31970065), the Tianjin Synthetic Biotechnology Innovation Capacity Improvement Project (TSBICIP-PTJJ-007-07), and the Competitive Support Program of the Tianjin Institute of Industrial Biotechnology, Chinese Academy of Sciences (JZ2024-037).

Acknowledgments

We thank the staff of the BL10U2, BL17B, BL18U1, and BL19U1 beamlines at the National Facility for Protein Science in Shanghai (NFPS), Shanghai Synchrotron Radiation Facility (SSRF), for assistance with data collection. We are also grateful to Senior Engineer Li Qian from the Structural Biology Platform of the Tianjin Institute of Industrial Biotechnology for her guidance and support in protein crystallization experiments. The authors declare no competing interests.

Author contributions

T.L., W.L., and X.Y. designed the study. R.M., H.W., Y.W., and Y.Z. performed the experiments and collected the data. R.M., H.W., Y.Z., J.H., Z.L., Y.C., W.L., T.L., and X.Y. analyzed the data. Y.Z., H.W., R.M., W.L., T.L., and X.Y. wrote the manuscript. All authors discussed the results and approved the final version of the manuscript.

Published: September 25, 2025