Efficient whole-cell biotransformation for gastrodin production via glycosyltransferase engineering and process optimization

Abstract

Gastrodin is a naturally occurring phenolic glycoside derived from Gastrodia elata, which is widely used as an edible medicinal plant in functional foods and dietary supplements in East Asia. Owing to its antioxidant, neuroprotective, and metabolic regulatory activities, gastrodin has attracted increasing attention as a valuable plant-derived glycoside for food and nutraceutical applications. However, its sustainable production is hindered by the limited activity of plant-derived uridine diphosphate-dependent glycosyltransferases (UGTs), the cytotoxicity of p-hydroxybenzyl alcohol (pHBA), and insufficient intracellular supply of UDP-glucose (UDPG). In this study, an efficient whole-cell biotransformation strategy was developed by integrating glycosyltransferase engineering with host metabolic optimization to enhance gastrodin biosynthesis. A triple mutant UGT variant, M4 (N94L/N221G/I343V), exhibiting 50.1% higher catalytic activity than the wild-type enzyme, was obtained through experimental screening. The cytotoxic effects of pHBA were systematically evaluated to define an optimal substrate loading range for high efficiency conversion. Furthermore, intracellular UDPG availability was enhanced using a push-pull strategy that combined a UDPG regeneration module with deletion of the competing pgm gene. Under optimized conditions, the engineered strain achieved a gastrodin titer of 79.9 mM (equivalent to 22.9 g/L) in a 1 L fermenter within 12 h, corresponding to a space-time productivity of 1.9 g/L/h. This work provides a sustainable and scalable biomanufacturing approach for the production of plant-derived phenolic glycosides.

Graphical abstract

1. Introduction

Plant-derived phenolic glycosides represent an important class of bioactive compounds widely distributed in edible and medicinal plants, where they contribute to antioxidant capacity [1,2], stress tolerance [3], and nutritional functionality [4]. Many of these compounds are increasingly incorporated into functional foods [5], dietary supplements [6], and nutraceutical formulations [7] due to their favorable safety profiles and diverse physiological benefits. Among them, gastrodin (4-hydroxybenzyl alcohol β-d-glucopyranoside) is a characteristic phenolic glycoside derived from Gastrodia elata, a traditional edible and medicinal plant that has long been consumed as both food and herbal material in East Asia [8,9]. Gastrodin is recognized for its broad biofunctional properties, including antioxidant activity [10], modulation of oxidative stress [11], and protective effects on the nervous and metabolic systems [12]. Compared with its aglycone, p-hydroxybenzyl alcohol (pHBA), glycosylation significantly improves its aqueous solubility, chemical stability, and bioavailability, making gastrodin particularly attractive as a food-compatible functional ingredient [12,13]. Consequently, gastrodin has gained growing interest in the development of functional foods, nutraceuticals, and plant-based health products, in addition to its traditional medicinal use [14,15].

With the expanding demand for gastrodin-enriched products, the development of sustainable and scalable production strategies has become increasingly important. At present, gastrodin is mainly obtained through plant extraction, chemical synthesis, or biological production. Compared to biological routes, chemical synthesis suffers from several drawbacks, including low regioselectivity, formation of multiple byproducts, harsh reaction conditions, and the need for protection/deprotection steps during glycosylation [16,17], resulting in high cost and environmental burden. Extraction from G. elata is also constrained by the slow growth cycle, limited availability of raw materials, and low yield [18], making it difficult to meet industrial-scale demand. Consequently, the biological production of gastrodin has become a major research focus in recent years. Biological production of gastrodin includes microbial fermentation, whole-cell biotransformation, and enzymatic conversion (Table 1). Among them, fermentation can directly synthesize gastrodin from simple carbon sources, however, the long cultivation period, complex metabolic regulation of central metabolism, and the accumulation of structurally related aromatic intermediates [[19], [20], [21]], which together complicate downstream purification and increase production cost. In the synthesis of gastrodin, biocatalytic processes typically use pHBA and UDP-glucose (UDPG) as glycosyl acceptors and donors, respectively, to form the β-d-glycosidic bond. This single-step glycosylation can be readily implemented in the whole-cell biotransformation system, in which product formation is decoupled from cellular growth. Consequently, pHBA and UDPG can be supplied in a modular and independently controlled manner, avoiding the tight coupling between precursor availability and biomass formation that is inherent to microbial fermentation systems. Although a variety of uridine glycosyltransferases (UGTs) have been reported to catalyze this reaction, many exhibit low catalytic efficiency or insufficient tolerance to high concentrations of pHBA, which strongly inhibits cell growth and enzyme activity [20,22,23]. Some engineered E. coli strains expressing specific UGTs achieved moderate titers, but the overall conversion remained limited by substrate toxicity and cofactor imbalance [24]. Enzymatic production using purified glycosyltransferases can improve product selectivity, yet the requirement for expensive UDPG donors and poor enzyme stability restricts industrial application [25]. Taken together, the major bottlenecks containing in the biosynthesis of gastrodin remain the need for highly active and stable UGTs, the cytotoxicity of pHBA, and the cost of glycosyl donors.

Table 1.

Reported biosynthetic strategies and maximum titers of gastrodin.

Production host	Production strategy	Maximum titer (g/L)	References
Saccharomyces cerevisiae	Microbial fermentation	2.10	[26]
Yarrowia lipolytica	Microbial fermentation	13.40	[21]
E. coli	Microbial fermentation	0.55	[20]
E. coli	Whole-cell biotransformation	4.56	[24]
E. coli	Whole-cell biotransformation	22.90	This study

To overcome these limitations, recent advances have increasingly focused on enzyme engineering and cellular metabolic optimization. Several studies have attempted to enhance UGTs activity toward pHBA glycosylation, however, most reported enzymes display modest catalytic efficiencies or poor stability under biotransformation conditions. Improving UGT performance through rational or semi-rational mutagenesis is therefore an important direction for increasing gastrodin productivity. In addition to the low UGT activity, the biosynthesis of gastrodin is inherently constrained by the cytotoxicity of pHBA. Therefore, excessive concentrations of pHBA can markedly reduce gastrodin production, making substrate control and feeding strategies crucial for efficient biotransformation. Meanwhile, maintaining adequate intracellular UDPG availability is essential for sustaining high glycosylation flux. Engineering strategies that strengthen UDPG regeneration or prevent its consumption by competing pathways have demonstrated promising potential for elevating glycosides output.

In this study, we first expressed the AtUGT gene of Arabidopsis thaliana in Escherichia coli BL21(DE3) to construct the whole-cell catalyst. Through random mutagenesis and experimental screening, a triple mutant M4 (N94L/N221G/I343V) was identified as a key variant with markedly enhanced catalytic activity toward pHBA. To further mitigate the cytotoxic effects of the substrate on whole-cell biotransformation for gastrodin production, we systematically evaluated its inhibitory profile and defined the optimal substrate loading range that maintains cellular viability while maximizing conversion efficiency. In addition, intracellular UDPG availability was strengthened by introducing a UDPG regeneration module and knocking out pgm. By integrating these engineering strategies, an optimized whole-cell biocatalyst capable of efficient gastrodin synthesis was obtained. Biotransformation in a 1 L fermenter with batch feeding for 12 h resulted in a final gastrodin titer of 79.9 mM (equivalent to 22.9 g/L) with a productivity of 1.9 g/L/h, demonstrating a high-efficiency and environmentally friendly route for gastrodin production with strong potential for large-scale industrial application.

2. Materials and methods

2.1. Bacterial strains, plasmids, and cultivation conditions

The strains and plasmids employed in this study are listed in Table S1. The UGT gene (Protein ID: NP_192016.1) from A. thaliana was used as the template for heterologous expression. E. coli Top10 was used for routine molecular cloning, while E. coli BL21(DE3) served as the host strain for UGT overexpression, and plasmid pET-28a (+) served as the expression vector for both the wild-type and mutant enzymes. E. coli strains were first pre-cultured in 5 mL of Luria-Bertani (LB) medium at 37 °C and subsequently transferred into Terrific Broth (TB) medium for induction. When cultures reached an optical density (OD₆₀₀) of 0.8, 0.25 mM isopropyl-β-d-thiogalactopyranoside (IPTG) was added and induction was performed at 25 °C for 24 h. All media were supplemented with kanamycin (50 μg/mL) when necessary.

2.2. Construction and screening of random mutation libraries for AtUGT

The random mutant library was generated by error-prone PCR using Rapid Taq Master Mix (Vazyme, Nanjing, China) supplemented with 0.1 mM Mn²⁺ to enhance the mutation rate. The AtUGT gene was used as the template, and PCR amplification was performed in a 50 μL reaction system following the recommendations of manufacturer. The amplified mutant fragments were purified using the AxyPrep™ Gel Extraction Kit (Axygen, Union City, USA) and subsequently assembled into the linearized pET-28a (+) vector by homologous recombination with the ClonExpress Ultra One Step Cloning Kit (Vazyme, Nanjing, China). The recombinant products were transformed into chemically competent E. coli Top10 cells and plated on LB agar containing 50 μg/mL kanamycin. Following DNA sequencing of 20 random variants, the mutation frequency was determined approximately 3–5 substitutions per 1000 bp. After incubation at 37 °C for 12 h, all colonies were pooled and cultured overnight. Plasmids extracted from this mixed culture were used to transform chemically competent E. coli BL21(DE3), which was then subjected to color or fluorescence screening as described below.

The AtUGT mutant screening strategy employed in this study was developed based on a previously reported method with minor modifications [20]. In brief, individual colonies harboring wild-type and mutant variants of AtUGT were inoculated into 96-deep-well microtiter plates containing 500 μL LB medium supplemented with kanamycin (50 μg/mL). The plates were incubated at 37 °C and 900 rpm overnight. Subsequently, 20 μL of each culture was transferred into fresh 96-deep-well plates containing 600 μL LB medium with 50 μg/mL kanamycin and incubated under the same conditions. When the cultures reached an OD₆₀₀ of 0.6–0.8, protein expression was induced by adding IPTG to a final concentration of 0.25 mM, followed by incubation at 25 °C for 24 h. Cells were harvested by centrifugation at 5000 rpm for 10 min at 4 °C, and the resulting pellets were resuspended in 200 μL M9 medium for further experimentation.

Two high-throughput assays were developed using p-nitrophenol (PNP) and 4-methylumbelliferone as activity reporters. Cultures were incubated at 37 °C for 4 h before colorimetric or fluorescence measurements were performed. For the colorimetric assay, 1 g/L PNP was used as the substrate, and glycosylation activity was monitored by measuring the decrease in absorbance at 400 nm, enabling rapid visual discrimination based on color change. For the fluorescent assay, 1 mM 4-methylumbelliferone was used as the substrate, and fluorescence was subsequently measured (λ_ex = 350 nm, λ_em = 460 nm) using a Biotek Cytation 3 imaging plate reader (Biotek Instruments Inc., Winooski, VT, USA). Variants exhibiting significant changes in absorbance or fluorescence relative to the wild-type control were selected as potential hits for further validation.

2.3. Enzyme expression, purification, and activity measurement

Recombinant E. coli BL21(DE3) harboring AtUGT expression plasmids were cultivated in LB medium at 37 °C until the OD₆₀₀ reached 0.6–0.8. Protein expression was induced with 0.25 mM IPTG followed by incubation at 16 °C for 20 h. Cells were harvested by centrifugation (6000 rpm, 10 min, 4 °C) and resuspended in lysis buffer (50 mM Tris-HCl, 50 mM NaCl, 25 mM imidazole, pH 7.0). Cell disruption was performed by sonication, and lysates were clarified by centrifugation (12,000 rpm, 30 min, 4 °C). The supernatant was subjected to His-tag affinity purification using a His-tag Protein Purification Kit (Sangon, China). Protein purity and concentration were assessed by SDS-PAGE and the Bradford method, respectively.

Enzymatic assays were performed in 100 μL reaction mixtures containing 50 mM Tris-HCl buffer (pH 7.5), 5 mM UDPG, 5 mM pHBA, and 100 μg of purified enzyme. Reactions were incubated at 35 °C for 30 min and terminated by addition of 400 μL methanol. Gastrodin formation was quantified by high-performance liquid chromatography (HPLC). One unit (U) of AtUGT activity was defined as the amount of enzyme generating 1 μmol of gastrodin per min under the assay conditions. The optimal reaction temperature was determined by conducting enzyme assays from 30 to 55 °C at 5 °C intervals. The optimal pH was evaluated at 50 °C using the following buffers: 200 mM acetic acid–sodium acetate buffer (pH 4.0–5.0), 200 Na₂HPO₄–NaH₂PO₄ buffer (pH 5.0–7.0), 50 mM Tris–HCl buffer (pH 7.0–9.0), 50 mM glycine–NaOH buffer (pH 9.0–11.0), 50 mM Na₂HPO₄–NaOH buffer (pH 11.0–12.0). Kinetic parameters were measured at pH 10.0 and 50 °C using pHBA concentrations ranging from 25 to 5000 μM. Values of K_m and k_cat were obtained by nonlinear regression to the Michaelis-Menten equation.

2.4. Protein structural modeling and analysis

The reported crystal structure of AtUGT (PDB ID: 2VCE) was retrieved from the RCSB Protein Data Bank [27]. Substrate molecules pHBA and UDPG were docked into the active site using AutoDock Vina, and the lowest-energy pose with an appropriate binding conformation was selected for further analysis. Protein-ligand interactions were visualized and examined using PyMOL. Molecular dynamics (MD) simulations were performed with GROMACS 2021 employing the CHARMM36 force field [28,29].

2.5. Construction of a whole-cell biotransformation system

Cultures for biotransformations were conducted in 250 mL Erlenmeyer flasks containing 50 mL TB medium inoculated at 1% (v/v) from an overnight culture. E. coli BL21(DE3) harboring pUGT18 and pACYCDuet-cscB-Basp-UgpA was grown at 37 °C and 250 rpm until an optical density of ∼0.8 was reached, at which point 0.25 mM IPTG was added. Following 24 h of induction at 25 °C, cells were pelleted, washed twice with the designated biotransformation buffer (3.0 g/L NaH₂PO₄, 12.8 g/L K₂HPO₄, 0.5 g/L NaCl, 1.0 g/L NH₄Cl, 0.24 g/L MgSO₄, 0.015 g/L CaCl₂) and resuspended to a final OD₆₀₀ of 20. Sucrose and pHBA were then added to a final concentration of 10 mM each, and 10 mL of the resulting cell suspension was transferred into 50 mL Erlenmeyer flasks and incubated under 37 °C and 200 rpm for whole-cell catalysis.

2.6. Genome manipulation

For gene knockout, a CRISPR/Cas9-mediated genome editing system was employed [30]. Accordingly, the deletion of pgm can be used as an instance. To construct the editing cassette for pgm deletion, the upstream and downstream homologous DNA regions (300-600 bp each) were amplified from the genome of E. coli BL21(DE3) using primer pairs pgm-L-F/pgm-L-R and pgm-R-F/pgm-R-R, respectively. The two homology arms were subsequently fused with the corresponding external primers to generate the complete donor DNA fragment. The resulting donor fragment, together with the pTargetF plasmid carrying the pgm-specific protospacer sequence, was co-electroporated into BL21(DE3) cells harboring pCas. After electroporation, cells were recovered at 37 °C for 3 h and plated on LB agar supplemented with kanamycin (50 mg/L) and spectinomycin (50 mg/L), yielding pgm knockout candidates. Positive transformants were identified by colony PCR and verified by DNA sequencing. Using the same strategy, deletions of agp and ushA genes were constructed.

2.7. Toxicity assay of pHBA

To evaluate the toxicity of pHBA toward E. coli BL21(DE3), overnight seed cultures were diluted into fresh LB medium to an initial OD₆₀₀ of 0.05 containing pHBA at final concentrations of 0, 10, 20, 30, 40, or 50 mM. Two hundred microliter aliquots of each culture were transferred into 96-well microplates and incubated at 37 °C with continuous shaking. Cell growth was monitored by measuring OD₆₀₀ at regular intervals, and the resulting growth curves were used to assess the inhibitory effects of pHBA on BL21(DE3).

2.8. Analytical methods

The concentrations of fructose, sucrose, pHBA, and gastrodin were quantified by HPLC using an Agilent 1260 Infinity II LC system (Agilent Technologies, Santa Clara, CA, USA). Fructose and sucrose were analyzed with a refractive index detector equipped with an HPX-87H organic acid column (Bio-Rad, Hercules, CA) under the following conditions: 30 mM H₂SO₄ mobile phase, 0.5 mL/min flow rate, and a column temperature of 50 °C. Gastrodin and pHBA were resolved using a ZORBAX Eclipse XDB-C18 column (4.6 × 250 mm, 5 μm; Agilent Technologies, USA) coupled to a UV detector, with an 80:20 (v/v) mobile phase of water containing 0.1% trifluoroacetic acid (solvent A) and methanol (solvent B). The flow rate was maintained at 1.0 mL/min, the column temperature at 30 °C, and detection was performed at 225 nm. The ATP content was determined using an ATP bioluminescent assay kit (Jiancheng Bioengineering Institute, Nanjing, China) based on the luciferin-luciferase reaction.

2.9. Batch-fed whole-cell biotransformation of gastrodin

Batch-feeding whole-cell catalytic reactions were carried out in 1 L bioreactor containing a total working volume of 500 mL. The reaction mixture consisted of 20 mM sucrose and 20 mM pHBA as initial substrates, along with 20 g/L wet cells. The cultivation was carried out at 37 °C with an agitation speed of 800 rpm. To sustain substrate availability during the reaction, solid pHBA was supplemented to maintain its concentration below 20 mM. At designated time points, 0.1 mL aliquots were withdrawn and immediately dilute with 0.9 mL of deionized water. The resulting samples were centrifuged to remove cell debris, and the supernatants were prepared for HPLC analysis as described above. All reactions were performed in triplicate.

3. Results and discussion

3.1. Directed evolution of AtUGT and screening

In a previous study, AtUGT from A. thaliana was identified as the most efficient catalyst for the glucosylation of pHBA among UGTs from four plant sources [22]. Although AtUGT exhibited the highest activity, its native catalytic efficiency in whole-cell biotransformation was still insufficient for high-level gastrodin production. Therefore, an efficient high-throughput screening strategy was urgently needed to facilitate the directed evolution of AtUGT, thereby enabling rapid identification of variants with improved catalytic activity. Inspired by the fluorescence-based screening method established by Bai et al. which employed 4-methylumbelliferone as a surrogate substrate for UGT73B6 engineering, we attempt to find an easily assayable analog suitable for microplate-based high-throughput screening [20]. Building on this concept, p-nitrophenol (PNP) was adopted as a chromogenic surrogate substrate. As its close structural alignment with pHBA, aside from the substitution of the C1 hydroxymethyl group with a nitro group, enables it to serve as an effective substrate analog for assessing variants with enhanced catalytic activity toward pHBA. Moreover, the disappearance of its characteristic yellow coloration upon glycosylation provides a readily quantifiable optical readout, enabling rapid assessment of glucosyltransferase activity in whole-cell assays.

The concept of this high-throughput color-based glycosyltransferase screen strategy was initially validated by catalyzing the glycosylation of PNP using recombinant E. coli overexpressing AtUGT. Whole cells were incubated with PNP, and a marked reduction in absorbance at 400 nm was observed, consistent with the masking of the phenolic hydroxyl group (Fig. 1A). HPLC analysis further confirmed the formation of the corresponding p-nitrophenyl β-d-glucopyranoside, validating that PNP is a reliable surrogate substrate for activity screening (Fig. S1). After that, a mutant library of AtUGT was generated by error-prone PCR, and approximately 100 variants were subjected to initial screening using a PNP-based colorimetric assay. In this assay, PNP was directly incubated with fermented BL21(DE3) cells expressing AtUGT variants, and the decrease in absorbance at 400 nm was quantified after 4 h reaction period as a measure of glucosyltransferase activity. However, all variants exhibited colorimetric changes comparable to those of the wild-type enzyme (data not shown), preventing the distinction of improved candidates. To investigate this unexpected uniformity, the same assay was performed using the parental E. coli BL21(DE3) strain lacking AtUGT expression. Surprisingly, BL21(DE3) displayed intrinsic activity toward PNP, indicating the presence of endogenous enzymes capable of catalyzing PNP glycosylation. This background reactivity ultimately precluded the use of PNP as a reliable screening substrate for AtUGT directed evolution.

Fig. 1.

High-throughput screening strategies for the directed evolution of AtUGT. (A) Colorimetric screening using p-nitrophenol (PNP) as a surrogate substrate. The glycosylation of PNP by AtUGT using UDPG yields p-nitrophenyl β-d-glucopyranoside (PNPG), accompanied by the disappearance of the characteristic yellow color of PNP. Whole-cell assays were performed using recombinant E. coli expressing AtUGT, with increasing concentrations of PNP (0-1 g/L). Representative microplate images and absorbance values at 400 nm are shown. Parental E. coli BL21(DE3) exhibited intrinsic background activity toward PNP, resulting in comparable absorbance decreases and thereby limiting the applicability of PNP for selective screening of AtUGT variants. (B) Fluorescence-based screening using 4-methylumbelliferone. Glycosylation of 4-methylumbelliferone by AtUGT produces 4-methylumbelliferyl β-d-glucoside, leading to fluorescence quenching that enables sensitive activity detection in whole-cell assays. Left panel: relative activities of AtUGT variants containing single amino acid substitutions identified from error-prone PCR screening. Right panel: relative activities of selected double and triple mutants constructed to evaluate synergistic effects. Activities are normalized to the wild-type enzyme (set as 100%). Data are presented as mean ± SD from three independent experiments.

Given that endogenous glycosyltransferase-like activity in BL21(DE3) compromised the specificity of the PNP-based assay, we next implemented a fluorescence-based screening strategy to enable selective identification of improved AtUGT variants (Fig. 1B). An error-prone PCR library was generated, yielding over 2000 colonies for high-throughput evaluation. Whole-cell reactions were conducted using 4-methylumbelliferone as the surrogate substrate, and glucosyltransferase activity was quantified by monitoring fluorescence quenching during bioconversion (Fig. S2). From this screen, eleven variants exhibiting greater fluorescence loss than the wild-type enzyme were identified (Table S2). Sequencing analysis revealed that these variants encompassed eight single substitutions (G44T, S84T, N94L, E96S, P133S, N221G, I343V, and S461A) and three double substitutions (A43P/V419I, N94L/Y272G, and G35N/G107S). Notably, the N94L mutation appeared twice among distinct clones, suggesting a mutation hotspot in a structurally sensitive region. To define the contributions of individual residues, each substitution was introduced independently into AtUGT. Four variants, A43P, N94L, N221G, and I343V displayed significantly enhanced catalytic activity, improving gastrodin formation by 7.9%, 20.1%, 12.5%, and 24.1%, respectively, relative to the wild-type. Structural analysis revealed that these beneficial substitutions are positioned at distal loci (>10 Å), consistent with emerging evidence that remote mutations can modulate enzyme function by tuning conformational dynamics, long-range communication networks, or the stability of catalytically competent states (Fig. S3) [31,32]. Among these sites, I343 lies adjacent to the conserved plant secondary product glycosyltransferase (PSPG) motif, where substitution to valine may reduce steric hindrance due to the smaller side chain, permitting a more optimal catalytic configuration and leading to increased activity. Residues N94 and N221 are positioned in the acceptor-binding domain, and their substitutions likely adjust regional flexibility in a manner that promotes more productive positioning of the phenolic substrate. A43, located at the N-terminal loop, may affect global structural motions that couple to turnover, thereby contributing to the improved activity observed across these variants. Motivated by these mechanistic insights, double- and triple-mutant combinations were constructed to explore potential synergistic effects. The N94L/N221G, N94L/I343V, and N221G/I343V variants increased catalytic activity by 35.8%, 48.7%, and 16.2%, respectively, while the triple mutant N94L/N221G/I343V exhibited a 50.1% improvement over the wild-type enzyme. Although these mutations were not introduced through structure-guided design, their additive effects illustrate the substantial and often underappreciated potential of distal-site mutations to elevate AtUGT catalytic performance, offering a foundation for future rounds of engineering.

3.2. Analysis and characteristics of wild-type and AtUGT (N94L/N221G/I343V)

To comprehensively evaluate the catalytic properties of AtUGT and its engineered variant, the effects of reaction temperature and pH on enzymatic activity were systematically examined using the wild-type enzyme and the triple mutant M4 (N94L/N221G/I343V). As shown in Fig. 2A and B, both enzymes exhibited highly similar activity profiles in response to variations in temperature and pH, retaining substantial catalytic activity over a broad temperature range from 30 to 55 °C and across a wide pH window from 5.0 to 12.0. Notably, the optimal activity for both the wild-type enzyme and the M4 mutant was observed at 50 °C and pH 8.0. The nearly identical temperature and pH dependent trends indicate that the triple mutations do not perturb the intrinsic thermal stability or the ionization requirements of the catalytic machinery. To quantitatively elucidate the impact of the M4 mutant on catalytic efficiency, steady-state kinetic parameters toward pHBA were determined (Fig. 2C and D). Compared with the wild-type enzyme, the M4 mutant exhibited a markedly increased catalytic turnover number (k_cat = 7.84 min⁻¹ vs 4.04 min⁻¹) and a significantly enhanced catalytic efficiency (k_cat/K_m = 3.41 vs 1.59 mM⁻¹ min⁻¹). In parallel, a decreased Michaelis constant was observed for the M4 mutant (K_m = 2.30 vs 2.55 mM), indicating improved substrate affinity. Collectively, these results demonstrate that the superior catalytic performance of the M4 mutant arises from a synergistic enhancement of both substrate binding and catalytic turnover, likely resulting from favorable conformational rearrangements within the substrate-binding pocket induced by the combined mutations.

Fig. 2.

Comparative characterization of the wild-type AtUGT and the triple mutant M4 (N94L/N221G/I343V). Effect of (A) temperature and (B) pH on the catalytic activity of the wild-type enzyme and the M4 mutant. Michaelis-Menten kinetics of the (C) wild-type AtUGT and (D) M4 mutant toward pHBA. Data are presented as mean ± SD from three independent experiments.

Collectively, these results demonstrate that the enhanced catalytic performance of the M4 mutant arises from a synergistic improvement in substrate affinity and turnover efficiency, which are most plausibly attributed to the cumulative effects of distal-site mutations that modulate enzyme dynamics and long-range functional communication rather than direct alterations of the active site.

3.3. MD analysis of the engineered AtUGT (N94L/N221G/I343V)

To elucidate how distal mutations enhance the catalytic performance of the M4 variant, MD simulations were performed for the wild-type enzyme and the M4 mutant (N94L/N221G/I343V) in complex with pHBA and UDPG for 100 ns. Backbone root-mean-square deviation (RMSD) analysis showed that both systems reached equilibrium within the initial ∼40 ns and remained stable thereafter (Fig. 3A and S4). All subsequent analysis were therefore conducted using the equilibrated trajectories. The M4 variant exhibited slightly reduced RMSD fluctuations compared to the wild-type, suggesting improved conformational stability. In line with this observation, the M4 variant displayed a reduced radius of gyration (Rg) and solvent-accessible surface area (SASA) throughout the simulation (Fig. 3B and C), indicative of a more compact and less solvent-exposed global architecture. These results suggest that the distal mutations promote structural compaction rather than local rigidification.

Fig. 3.

Molecular dynamics analysis of the wild-type AtUGT and the triple mutant M4 (N94L/N221G/I343V). Time evolution of (A) backbone RMSD, (B) Rg, (C) SASA, and (D) RMSF profiles of the wild-type enzyme and the M4 variant during 100 ns MD simulations. Number of hydrogen bonds formed between the enzyme and (E) pHBA or (F) UDPG during the simulations. DCCM of the (G) wild-type enzyme and the (H) M4 mutant.

Residue-level dynamics were further examined using root-mean-square fluctuation (RMSF) analysis (Fig. 3D) [33]. While the overall RMSF profiles of the two systems were comparable, several residues exhibited markedly reduced fluctuations in the M4 variant, including S78, E106, and R109 in the N-terminal region, K247 and Q248 at the inter-domain interface, and Y315 and F316 proximal to the active site. These residues are positioned in regions implicated in acceptor access, domain coupling, and substrate stabilization, respectively, indicating selective rigidification of functionally relevant modules. This redistribution of dynamics was accompanied by enhanced enzyme-substrate interactions. Hydrogen-bond analysis revealed that the M4 variant formed more persistent hydrogen bonds with both pHBA and UDPG than the wild-type enzyme (Fig. 3E and F). Detailed hydrogen-bond occupancy analysis (Table S3) further showed that the catalytic residue Glu388 engages the substrate through both carboxylate oxygens (OE1 and OE2) in the M4 variant, whereas only a single interaction is dominant in the wild type. In addition, a weak but reproducible hydrogen bond involving Tyr315 emerged in the M4 variant, consistent with its reduced flexibility.

Dynamic cross-correlation map (DCCM) analysis revealed a pronounced reorganization of collective motions in the M4 variant (Fig. 3G and H). In particular, strengthened correlated and anti-correlated motions were observed between the N-terminal acceptor-access region, the catalytic core, and the UDP-binding domain, indicating enhanced long-range communication between functional modules. These changes suggest that distal mutations reshape the internal dynamic network, facilitating more efficient coupling between substrate binding and catalysis. Collectively, these results indicate that the enhanced catalytic activity of the M4 variant arises from mutation-induced compaction and dynamic reprogramming, which stabilize key functional regions, optimize substrate positioning, and promote a catalytically favorable conformational ensemble.

3.4. Whole-cell biotransformation of sucrose and pHBA into gastrodin

Building on the improved catalytic performance of triple mutant M4 (N94L/N221G/I343V), we next aimed to establish a cost-effective whole-cell platform for gastrodin production that eliminates reliance on expensive exogenous UDPG. To address this bottleneck, a UDPG regeneration system capable of converting inexpensive disaccharide sucrose into the required glycosyl donor was implemented. Inspired by previous work, a dual-modular strain was constructed in which the UDPG regeneration module consisted of sucrose permease (CscB) from E. coli W, sucrose phosphorylase (Basp) from Bifidobacterium adolescentis, and uridylyltransferase (UgpA) from Bifidobacterium bifidum, thereby enabling intracellular generation of UDPG from sucrose [34]. The second modular encoded the engineered AtUGT variant M4 (N94L/N221G/I343V), responsible for catalyzing pHBA glucosylation (Fig. 4A). This modular arrangement ensured efficient donor regeneration while enabling independent control of catalytic enzyme levels for gastrodin biosynthesis.

Fig. 4.

Whole-cell biotransformation of sucrose and pHBA into gastrodin.(A) Schematic illustration of the dual-modular whole-cell system for gastrodin biosynthesis. The UDPG regeneration module comprises CscB, Basp, and UgpA, enabling intracellular UDPG generation from sucrose. The glucosylation module encodes the engineered AtUGT variant M4 (N94L/N221G/I343V) for pHBA glucosylation. (B) Time-course profiles of pHBA consumption and gastrodin formation during whole-cell biotransformation catalyzed by E. coli GAS1. Data are presented as mean ± SD from three independent experiments.

The resulting whole-cell catalyst, designated E. coli GAS1, achieved robust coexpression of both the UDP-glucose module and the engineered AtUGT under induction at 25 °C with 0.25 mM IPTG. For biotransformation, GAS1 cells were harvested and resuspended to an OD₆₀₀ of 20 and incubated at 37 °C with 10 mM sucrose and pHBA. As shown in Fig. 4B, the catalyst produced 7.8 mM (2.2 g/L) gastrodin within 4 h, corresponding to a substrate conversion of 95.5%. These results demonstrate that coupling remote-site-engineered AtUGT variants with an efficient sucrose-driven UDPG regeneration system enables a practical and economically viable whole-cell process for gastrodin biosynthesis.

3.5. Effect of pHBA concentration on whole-cell gastrodin production

After confirming that the engineered strain GAS1 could efficiently convert sucrose and pHBA into gastrodin, whole-cell biotransformation was performed at varying pHBA concentrations to optimize production efficiency. The reactions were conducted using 10, 20, 30, 40, and 50 mM pHBA with a 1:1 molar ratio of pHBA to sucrose to ensure adequate UDPG supply. When 10 mM pHBA was supplied, complete conversion to gastrodin was achieved within 5 h, corresponding to a production rate of 1.46 mM/h (Fig. 5A). Increasing the substrate to 20 mM yielded 11.41 mM gastrodin after 5 h (3.3 g/L), with a conversion rate of 78.65% and an enhanced productivity of 2.19 mM/h (Fig. 5B). However, further elevation of pHBA markedly impaired whole-cell catalysis. At 30, 40, and 50 mM pHBA, only 8.30, 4.78, and 2.88 mM gastrodin were produced (Fig. 5C–E), and the corresponding productivity declined sharply to 1.66, 0.96, and 0.58 mM/h, respectively (Fig. 5F). This pronounced reduction suggested that high pHBA concentrations exert inhibitory effects on the biocatalytic system.

Fig. 5.

Effect of p-hydroxybenzyl alcohol (pHBA) concentration on whole-cell gastrodin production. Time-course profiles of gastrodin formation during whole-cell biotransformation catalyzed by E. coli GAS1 at nominal initial pHBA concentrations of (A) 10 mM, (B) 20 mM, (C) 30 mM, (D) 40 mM, and (E) 50 mM, respectively. The initial pHBA concentrations shown in the figure are analytically determined. Sucrose was supplied at a 1:1 molar ratio to pHBA in all reactions. (F) Productivity of gastrodin at different initial pHBA concentrations. Data are presented as mean ± SD from three independent experiments.

To investigate whether substrate toxicity accounted for this phenomenon, the impact of pHBA on cell viability was evaluated. Similar to observations reported for other aromatic compounds [35,36], pHBA exhibited dose-dependent cytotoxicity (Fig. S5A). In the presence of 50 mM pHBA, cell viability decreased to 55.40%, indicating substantial growth inhibition (Fig. S5B). These results demonstrate that pHBA concentrations above 30 mM are detrimental to whole-cell catalysis, likely due to compromised cellular integrity and metabolic activity. Based on these findings, 20 mM pHBA was selected as the optimal substrate concentration for subsequent experiments.

3.6. Effect of cell density on gastrodin biosynthesis via ATP modulation

To determine whether the increasing catalyst loading could enhance whole-cell conversion of pHBA to gastrodin, we systematically varied the biomass concentration from 10 to 80 OD units. With 20 mM pHBA supplied, gastrodin formation initially increased with biomass, rising from 6.84 mM at 10 OD to 9.14 mM at 20 OD. However, further increases in cell density caused a progressive decline in product titer, yielding only 6.75, 5.54, and 5.12 mM at 40, 60, and 80 OD, respectively (Fig. 6A). This non-linear behavior indicated that catalyst abundance was not the rate-limiting factor and suggested the presence of a metabolic bottleneck. After systematically investigating the whole-cell catalytic metabolic network of gastrodin and finding that it contains two ATP-dependent steps, we subsequently examined whether ATP availability could restrict flux. First, sucrose uptake through the proton coupled transporter requires ATP-driven proton extrusion, imposing one ATP cost per sucrose imported [37,38]. Second, the UDPG cycle is an energy-demanding process. The UgpA catalyzes the formation of UDPG from glucose-1-phosphate (G1P) and UTP, after which AtUGT converts UDPG and pHBA into gastrodin and UDP. The subsequent phosphorylation of UDP to UTP requires one ATP equivalent and closes the UDPG regeneration loop [39]. Taken together, the conversion of sucrose and pHBA into gastrodin theoretically requires two ATP molecules per product formed (Fig. S6A), making it a potential flux-limiting variable.

Fig. 6.

Effect of cell density on whole-cell gastrodin biosynthesis and intracellular ATP availability. Gastrodin titers obtained at different cell densities (OD₆₀₀ = 10–80) during whole-cell biotransformation catalyzed by E. coli GAS1 in (A) flat-bottom flasks and (C) baffled flasks with 20 mM pHBA. Intracellular ATP levels measured at different cell densities under (B) flat-bottom flask and (D) baffled flask conditions. Data are presented as mean ± SD from three independent experiments.

To test whether ATP availability constrained whole-cell catalysis, intracellular ATP levels were measured across all catalyst densities. Consistent with our hypothesis, the ATP levels remained extremely low (<0.1 μM/OD) regardless of cell concentration (Fig. 6B), indicating severe ATP limitation in flat-bottom flasks. To directly test ATP dependence, the respiratory chain was disrupted using lysed cells to abolish ATP regeneration. Under identical reaction conditions, lysed cells produced no detectable gastrodin, confirming that ATP availability rather than enzyme abundance or cofactor pools constrains whole-cell catalytic flux (Fig. S6B).

Because ATP generation in aerobic E. coli is highly oxygen-dependent, we next increased oxygen transfer by replacing flat-bottom flasks with baffled flasks. Under these conditions, gastrodin production became proportional to biomass, and the maximal titer of 9.94 mM was achieved at the highest catalyst loading (Fig. 6C). The shift from a non-linear to a positive monotonic relationship between biomass and gastrodin titer demonstrates that the productivity decline at high cell density in flat-bottom flasks was caused by oxygen-limited ATP regeneration, which restricted both sucrose uptake and UDPG cycling. Furthermore, as shown in Fig. 6D, the intracellular ATP significantly improved in baffled flasks and remained high at 0.08 mM/OD even at OD₆₀₀ of 80. Taken together, these results demonstrate that flux in the whole-cell biotransformation of gastrodin is governed primarily by intracellular ATP availability, which is tightly linked to the oxygen transfer rate.

3.7. Engineering host strains with reduced carbon loss and elevated UDPG pool

In principle, strain GAS1 is stoichiometrically capable of converting sucrose and pHBA into equimolar gastrodin. Meanwhile, the fructose released during sucrose phosphorolysis becomes the sole energy source that supports ATP generation, which is required to drive sucrose uptake and sustain continuous UDPG regeneration for glycosylation. In the previous section, we demonstrated that increasing the whole cell catalyst density can improve the titer of gastrodin in baffled flasks, however, nearly half of the supplied pHBA remains unconverted. It is worth noting that the flat bottom flasks produced more gastrodin than baffled flasks at low cell densities. This unexpected trend indicated there being additional factors, beyond catalyst loading and ATP supply, that constrained the overall performance of the biotransformation system.

To dissect this discrepancy, we quantified residual sucrose and fructose in both flask types. As biomass increased, the concentrations of sucrose and fructose declined gradually with lower residual levels in baffled flasks due to higher respiratory activity (Fig. S7). Nevertheless, the sucrose depletion and gastrodin formation did not exhibit the expected 1:1 stoichiometric relationship. The stoichiometric mismatch between substrate consumption and product formation implied that a substantial fraction of carbon flux was redirected into endogenous metabolism instead of being channeled toward UDPG-dependent glycosylation. Literature reports show that E. coli possesses several endogenous enzymes that naturally consume G1P and UDPG, thereby competing with the heterologous glycosylation pathway (Fig. 7A). These include: (i) phosphoglucomutase (pgm), a central carbohydrate-metabolism enzyme that interconverts glucose-6-phosphate (G6P) and G1P [40]; (ii) glucose-1-phosphatase (agp), whose primary role is to hydrolyze G1P into glucose and inorganic phosphate [41]; and (iii) UDPG hydrolase (ushA), an enzyme that degrades UDPG to UMP and G1P [42]. Together, these reactions deplete the intracellular glycosyl-donor pool and restrict flux toward gastrodin synthesis.

Fig. 7.

Engineering host strains to reduce carbon loss and enhance the intracellular UDP-glucose pool. (A) Schematic illustration of endogenous competing pathways consuming G1P and UDPG in E. coli, including phosphoglucomutase, glucose-1-phosphatase, and UDPG hydrolase. (B) Gastrodin titers obtained during whole-cell biotransformation catalyzed by engineered strains with deletions of pgm, agp, or ushA. Data are presented as mean ± SD from three independent experiments.

To quantify the impact of these competing pathways, we individually deleted pgm, agp, and ushA to generate strains GAS2-4, using E. coli GAS1 as the wild-type control. Whole-cell biotransformation showed that knockout of agp, and ushA had no effect on gastrodin titer. By contrast, knockout of the pgm gene increased gastrodin titers to 13.8 mM, corresponding to improvements of 38.7% relative to wild-type (Fig. 7B). Collectively, these results show that preventing the diversion of G1P into endogenous competitive pathways is essential for maximizing whole-cell glycosylation efficiency. By removing endogenous branching reactions, we rationally rewired central carbon metabolism and increased the intracellular availability of UDPG, which in turn accelerated the conversion of sucrose and pHBA into gastrodin.

3.8. High-yield synthesis of gastrodin by batch-feeding whole-cell catalysis

To establish a high-intensity whole-cell biotransformation system for gastrodin production, the recombinant strain GAS2 was selected due to its maximal UDPG availability and superior catalytic performance in shake-flask screening. After approximately 24 h of cultivation, cells were immediately used for whole-cell catalysis. For fed-batch whole-cell biotransformation, GAS2 was cultivated in a 1 L bioreactor at a 40% inoculum level. Agitation was maintained at 500 rpm and 1000 rpm, respectively, to assess the impact of oxygen transfer during the whole catalytic process. The reaction was initiated by simultaneously adding 20 g/L wet cells, 20 mM sucrose, and 20 mM pHBA. To maintain the concentration of the limiting aromatic acceptor and to prolong catalytic turnover, solid pHBA was intermittently supplied to maintain its concentration below 20 mM throughout the biotransformation process. Specifically, in the high-DO system, pHBA and sucrose were supplemented at 1.5 h intervals due to the higher initial catalytic activity, whereas feeding was performed every 3 h in the low-DO system. Over the entire 12 h reaction, the total cumulative pHBA input reached 90.5 mM and 70.0 mM in the high-DO and low-DO systems, respectively. This feeding strategy ensured sustained precursor availability while preventing inhibitory effect of toxic aromatic substrates. Under an agitation speed of 500 rpm, the biotransformation proceeded smoothly for 12 h, resulting in the production of 51.4 mM gastrodin (equivalent to 14.7 g/L), corresponding to a space-time productivity of 1.2 g/L/h. In contrast, when the agitation speed was increased to 1000 rpm, the production of gastrodin increased markedly. After 12 h of whole-cell catalysis, the gastrodin titer reached 79.9 mM (equivalent to 22.9 g/L), achieving a significantly higher productivity of 1.9 g/L/h (Fig. 8A and B). Taken together, the improved oxygen transfer at higher agitation speed enhanced intracellular ATP availability, thereby facilitating UDPG regeneration and promoting gastrodin biosynthesis in the whole-cell catalytic system.

Fig. 8.

Fed-batch whole-cell biotransformation of gastrodin by the recombinant strain GAS2 under different agitation speeds. Time courses of gastrodin production and residual pHBA during whole-cell catalysis at agitation speeds of (A) 500 rpm and (B) 1000 rpm in a 1 L bioreactor. The reaction was initiated by the simultaneous addition of 20 g/L wet cells, 20 mM sucrose, and 20 mM pHBA. Solid pHBA was intermittently supplied to maintain its concentration below 20 mM throughout the biotransformation. Data are presented as mean ± standard deviation of three independent experiments.

4. Conclusions

In summary, we have established an efficient whole-cell biotransformation platform for gastrodin production through glycosyltransferase engineering combined with engineered cell modification. First, directed evolution of the AtUGT from A. thaliana yielded the triple mutant M4 (N94L/N221G/I343V), which exhibited 50.1% enhanced catalytic activity compared to the wild-type enzyme. Second, systematic evaluation of pHBA cytotoxicity enabled the determination of an appropriate substrate addition amount, ensuring high reaction rates without compromising cell viability. Third, by simultaneously reinforcing UDPG regeneration and eliminating major endogenous consumption routes, we significantly elevated the intracellular UDPG pool, thereby enhancing glycosylation flux. Finally, application of the engineered strain in a 1 L fermenter under a fed-batch mode enabled sustained catalysis, yielding 22.9 g/L gastrodin within 12 h. This result demonstrates that the combined optimization of biocatalyst, metabolic context, and reaction strategy can effectively overcome the long-standing barriers of low UGT activity, substrate cytotoxicity, and insufficient UDPG supply. Overall, this study provides a robust technological foundation for the large-scale and economically viable industrial biosynthesis of gastrodin.

CRediT authorship contribution statement

Xun Wang: Writing – original draft, Funding acquisition. Jiale Zhang: Investigation, Data curation. Tao Li: Visualization, Software, Resources. Fei Wang: Supervision, Funding acquisition. Zhiguo Wang: Project administration. Xun Li: Writing – review & editing, Investigation, Conceptualization.

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Appendix A. Supplementary data

The following is the Supplementary data to this article: