De novo biosynthesis of Asperosaponin VI in Saccharomyces cerevisiae

Abstract

Asperosaponin VI (ASA VI), the primary bioactive triterpenoid saponin marker of Dipsacus asper Wall. (Chinese Pharmacopoeia 2020), possesses significant neuroprotective, anti-inflammatory, and osteogenic activities. However, its low natural abundance limits large-scale production. In this study, we reported the first complete biosynthetic reconstruction of ASA VI in Saccharomyces cerevisiae using a modular synthetic biology strategy. The pathway included in situ UDP-arabinose (UDP-Ara) biosynthesis via heterologous expression of AtUGDH3, GuUXS2, and GuUXE1; triterpenoid scaffold generation through ERG9, ERG1, CqBAS1, and CqCYP716A78 for oleanolic acid (OA) production; and downstream modifications including C-23 hydroxylation by multicopy-expressed CaCYP714E19, stepwise glucosylation at C-28 by CaUGT73AD1 and CaUGT73C8, and C-3 arabinosylation by AsUGT99D1 to yield ASA VI. LC-MS analysis confirmed ASA VI biosynthesis and the accumulation of key intermediates (OA, HED, HED-28-Glc, and HED-28-Glc-Glc). Although production remained at trace levels (395 ng/L), pathway analysis suggested that the downstream glycosylation steps and UDP-Ara supply could be the major rate-limiting factors. This work established a microbial chassis for the sustainable synthesis of ASA VI and related arabinosylated saponins.

Graphical abstract

1. Introduction

Asperosaponin VI (ASA VI) is an oleanane-type triterpenoid saponin and the principal bioactive constituent of Dipsacus asper Wall., as documented in the Chinese Pharmacopoeia (2020) [1]. ASA VI has been reported to exhibit a wide range of pharmacological activities, including neuroprotective, anti-inflammatory, osteogenic, hepatoprotective, and metabolic regulatory effects [[2], [3], [4]]. Recent pharmacological investigations have further demonstrated its therapeutic potential in models of alcohol-induced liver injury [5], diabetic osteoporosis [6], and male reproductive dysfunction [7]. These multifunctional properties highlight ASA VI as a promising candidate for clinical development. However, large-scale production of ASA VI remains severely restricted by its low natural abundance, slow plant growth, and labor-intensive purification from D. asper tissues.

To overcome the limitations of plant-based production, synthetic biology strategies have been increasingly adopted to reconstruct plant biosynthetic pathways in microbial hosts [[8], [9], [10]]. Microbial cell factories, such as Saccharomyces cerevisiae, offer advantages including rapid growth, genetic tractability, and scalable fermentation. Heterologous biosynthesis of complex phytochemicals has enabled the sustainable production of numerous natural products, including β-amyrin, oleanolic acid (OA), glycyrrhetinic acid and ginsenosides [[11], [12], [13], [14], [15]]. These efforts demonstrated the feasibility of engineering microbial platforms for triterpenoid saponin biosynthesis.

In the case of ASA VI, its biosynthesis in planta is believed to proceed from the triterpenoid scaffold OA. Previous studies have significantly enhanced OA production in S. cerevisiae by leveraging flux redistribution and chassis optimization strategies, achieving titers exceeding 4 g/L under fed-batch fermentation [16]. While these strains represent valuable biosynthetic platforms, we did not adopt them in the current study, as OA biosynthesis was not identified as a major bottleneck. Feeding experiments with exogenous OA showed no significant increase in ASA VI titer, suggesting that the key limitations lie downstream, particularly in glycosylation steps and sugar donor biosynthesis. Consequently, our design focused on reconstituting the complete ASA VI biosynthetic pathway, rather than enhancing OA accumulation at this stage. Nevertheless, the availability of high-yielding OA chassis offers a strong foundation for future yield optimization and potential scale-up of complex saponin production.

Although these high-yielding strains were not employed as the starting chassis in this study, they establish the practical feasibility of extending OA-based biosynthetic routes toward complex saponins such as ASA VI.

Nevertheless, the downstream biosynthetic steps from OA to ASA VI remain largely unresolved. Although the C-23 hydroxylation activity of cytochrome P450 enzymes involved in triterpenoid modification, such as MtCYP72A68v2 and CaCYP714E19, has been independently validated in yeast systems [17,18], their coordinated integration into a complete ASA VI biosynthetic pathway has not yet been demonstrated. Furthermore, arabinosylation at the C-3 position requires the UDP-Ara-dependent glycosyltransferase AsUGT99D1 [19], whose substrate specificity and catalytic efficiency in yeast remain insufficiently characterized. Adding to the complexity, S. cerevisiae lacks an endogenous pathway for UDP-Ara biosynthesis, rendering this sugar donor unavailable during typical fermentation processes. While UDP-Ara production has been achieved in Escherichia coli and in vitro enzymatic systems through engineered salvage or recycling pathways [[19], [20], [21]], no prior report has described the successful reconstitution of a complete de novo UDP-Ara biosynthetic module in yeast for application in arabinosylated triterpenoid biosynthesis.

In this study, we reconstructed the complete biosynthetic pathway of ASA VI in S. cerevisiae using a modular approach. The biosynthetic pathway was organized into three functional modules. First, the UDP-Ara module was constructed by expressing AtUGDH3, GuUXS2, and GuUXE1 to enable in situ production of this non-native sugar donor in yeast. Second, the triterpenoid scaffold module was established to produce OA through the co-expression of ERG1, ERG9, CqBAS1, and CqCYP716A78. Third, the downstream modification module utilized CaCYP714E19, whose high catalytic activity was further leveraged by multi-copy expression, to catalyze the C-23 hydroxylation of OA and produce hederagenin (HED). This was followed by a stepwise glycosylation cascade: CaUGT73AD1 and CaUGT73C8 mediated two-step glucosylation at the C-28 carboxyl group of HED, forming HED-28-Glc-Glc, and AsUGT99D1 subsequently catalyzed arabinosylation at the C-3 hydroxyl group to complete ASA VI biosynthesis. LC-MS analysis confirmed the production of ASA VI and key intermediates. This work established the first functional arabinosylated triterpenoid biosynthetic system in yeast, demonstrating the feasibility of UDP-Ara synthesis and its application in saponin production.

2. Materials and methods

2.1. Strains and chemicals

The S. cerevisiae WAT11 strain (Purchased from HonorGene) was selected as the engineering host.

Authentic OA and HED standard (HPLC purity ≥98 %), ASA VI standard (HPLC purity ≥98 %) was procured from Maide Biotech Co., Ltd. (Shanghai, China). Molecular biology reagents were obtained from certified suppliers: Phanta Max Super-Fidelity DNA Polymerase (Vazyme) and FastPure Gel DNA Extraction Kit from Vazyme Biotech Co., Ltd. (Nanjing, China), TIANprep Rapid Mini Plasmid Kit (TIANGEN), Cycle-Pure Kit (Omega), EASYspin Plant RNA Rapid Extraction Kit (Adeli Biotechnology), HiScript III 1st Strand cDNA Synthesis Kit (Vazyme), Gel Extraction Kit (Omega); all other analytical-grade chemicals (≥99 % purity) were purchased from Sangon Biotech Co., Ltd. (Shanghai, China).

2.2. DNA cloning

The complete nucleotide sequences were obtained from the National Center for Biotechnology Information database (https://www.ncbi.nlm.nih.gov) and are listed in Table S1.

Each expression cassette for genome editing with CRISPR-Cas9 consisted of upstream homology arms (∼1000 nt), promoter, desired gene, terminator and downstream homology arms (∼1000 nt). Assembly of the pieces was achieved by fusion PCR. Specifically, primary fragments with overlapping sequences were obtained first by PCR using the primers given in Table S3. PCR products were purified and subjected to another PCR reaction without any primers to generate the full-length fusion gene. The fusion gene was then used as a template of the final PCR using primers. The resulting fusion products were then used for yeast transformation.

Guide (g)RNA plasmids (Table S2) used in the present study were previously constructed [22].

2.3. Integration of genes by the CRISPR-Cas9 system

The general procedure for gene integration using the CRISPR-Cas9 system in S. cerevisiae has been previously described [23]. Strains used for gene integration were constructed and confirmed by Sanger sequencing (Sangon Biotech). Donor DNAs were constructed and purified using a DNA Extraction Mini Kit (Vazyme Biotech Co., Ltd) before transformation. Gene segments and donor DNAs were introduced into yeast cells via the lithium acetate method [24].

2.4. Construction of strains

The various sources of cyclase, oxidase, and glycosyltransferase are summarized in Table 1. All gene fragments were amplified by PCR and purified using either gel extraction or column purification with a DNA Extraction Mini Kit from Vazyme Biotech Co., Ltd. The gene constructs were then synthesized and sequenced by Sanger sequencing (Sangon Biotech). All S. cerevisiae strains listed in Table S4 were constructed using the high-efficiency LiAc/SS carrier DNA/PEG protocol.

Table 1.

Different sources of enzymes.

	gene	function	source
Sugar donor chassis	AtUGDH3	UDP-Glc dehydrogenase	Arabidopsis thaliana
	PsUGE2	UDP-Glc 4-epimerase	Pisum sativum Linn.
	AtUXS2	UDP-GlcA decarboxylase	Arabidopsis thaliana
	GuUXE1	UDP-Xly 4-epimerase	Glycyrrhiza uralensis
	GuUXS2	UDP-GlcA decarboxylase	Glycyrrhiza uralensis
OA chassis	CqbAS1	β-amyrin synthase	Chenopodium quinoa
	CYP716A78	28-site oxidase	Chenopodium quinoa
	ERG1	squalene monooxygenase	Saccharomyces cerevisiae
	ERG9	Squalene Synthase	Saccharomyces cerevisiae
Synthesis of ASA VI	MtCYP72A68v2	23-site oxidase	Medicago sativa L.
	CYP714E19	23-site oxidase	Centella asiatica
	AtCYP71A16	23-site oxidase	Arabidopsis thaliana
	AsUGT99D1	Arabic glycotransferase	Avena sativa L.
	CaUGT73AD1	glucosyl transferase	Centella asiatica
	CaUGT73C8	glucosyl transferase	Centella asiatica
	CaUGT94M2	glucosyl transferase	Centella asiatica

The UDP-Ara biosynthetic module was constructed by integrating the gene encoding AtUGDH3 into the 911b site using the CRISPR-Cas9 system, with strain WAT11 serving as the chassis cell to create the engineered strain WA1 [25]. Similarly, the genes encoding PsUGE2 and AtUXS3 were integrated into the 607b site by the CRISPR-Cas9 system, using strain WA1 as the chassis cell to generate the engineered strain WA2. Finally, the genes encoding GuUXE1 and GuUXS2 were integrated into the 607b site by the CRISPR-Cas9 system, using strain WA1 as the chassis cell to produce the engineered strain WA3.

The OA chassis was constructed by integrating the genes encoding CqbAS1 and CqCYP716A78 into the 607b site using the CRISPR-Cas9 system, with strain WAT11 serving as the chassis cell to create the engineered strain WA6 [25]. Subsequently, the genes encoding ERG9 and ERG1 were integrated into the 911b site by the CRISPR-Cas9 system, using strain WA6 as the chassis cell to generate the engineered strain WA7.

We then integrated the sugar donor and OA-producing genes into the same strain to construct strain WA9, which was used for the subsequent verification of additional genes. In the verified chassis capable of producing OA and sugar donors, we gradually tested the third module gene responsible for catalyzing the synthesis of ASA VI. First, we verified three oxidase genes that catalyze the oxidation of OA at C-23 to form a hydroxyl group, generating HED. These genes included MtCYP72A68v2 from M. truncatula, CaCYP714E19 from C. sinensis, and AtCYP71A16 from A. thaliana. We transferred each of these genes into WA9 to create strains WA10, WA11, and WA12, respectively. To enhance HED production, we co-expressed two genes responsible for HED synthesis from OA in a single strain, resulting in strain WA13 and WA14. Subsequently, we introduced CaUGT73AD1 and either CaUGT73C8 or CaUGT94M2 into strain WA14, constructing strains WA15 and WA16, respectively, to verify the functions of these two genes.

Finally, the functional genes were integrated into different sites of the WA15 strain to construct strain WA17. Yeast transformation was carried out using the lithium acetate method [24]. Details of the strains used in this study are provided in Fig. 1. Primers and genes are listed in Supporting Information (Table S1).

Fig. 1.

Lineage of stains constructed in this study.

2.5. Growth medium and culture conditions

S. cerevisiae strains were grown on yeast extract-peptone-dextrose (YPD) medium (20 g/L glucose, 10 g/L yeast extract, 20 g/L peptone). All media were autoclaved for 20 min at 115 °C before use.

Biocatalysis was performed using recombinant yeast strains as follows. First, inoculate the recombinant yeast strain into 2 mL of YPD medium and incubate overnight at 30 °C. Then, transfer 1 mL of the culture with an OD₆₀₀ = 0.2 into YPG medium and incubate for 12 h at 30 °C. Afterward, add 100 μL of GAL and incubate for an additional 12 h. Then, add 200 μL of GAL and incubate for 48 h. Subsequently, the culture was extracted three times with ethyl acetate and stored for later use. All experiments were performed in biological triplicates for all strains.

2.6. LC-MS analysis

The prepared samples were analyzed using a Agilent Technologies 6410 Triple Quad LC/MS instrument equipped with a Waters Acquity Poroshell 120 EC-C18 column (3 mm × 100 mm). Mobile phase A consisted of water with 0.01 % ammonium acetate, while mobile phase B was acetonitrile. The flow rate was set to 0.3 mL/min, and the column temperature was maintained at 30 °C. The mass spectrometry conditions were as follows: sheath gas flow rate, 30 Arb; auxiliary gas flow rate, 8 Arb; spray voltage, 3.2 kV; and ion transfer tube temperature, 320 °C, with nitrogen as the carrier gas. In Full scan mode, negative ion mode was employed using an electrospray ionization (ESI) source. The mass-to-charge ratio (m/z) acquisition range was 50–1200 m/z. A single-point external standard method was employed for the quantitative analysis of OA, HED and ASA VI.

3. Results

3.1. Design of the combinatorial biosynthetic pathway for ASA VI in S. cerevisiae

Based on recent advances in the synthetic biology of pentacyclic triterpenoids, we designed a combinatorial biosynthetic pathway for de novo production of ASA VI in S. cerevisiae. ASA VI is a bioactive pentacyclic triterpenoid synthesized via the mevalonate (MVA) or methylerythritol phosphate (MEP) pathway [26,27]. In this pathway, 2,3-oxidosqualene is cyclized into β-amyrin by β-amyrin synthase (β-AS), and subsequently modified by a series of tailoring enzymes, including C-23 hydroxylase, C-28 hydroxylase, C-3 arabinosyltransferase, and a two-step C-28 glucosyltransferase that successively adds two glucose units to complete the glycosylation (Fig. 2). As S. cerevisiae lacks the native machinery to supply arabinose donors required for arabinosylation, we introduced a heterologous arabinose donor biosynthetic pathway. Endogenous glucose is first oxidized to UDP-glucuronic acid (UDP-GlcA) by UDP-glucose (UDP-Glc) dehydrogenase (UGDH), then converted to UDP-xylose by UDP-xylose synthase (UXS) and finally epimerized to UDP-Ara by UDP-xylose epimerase (UXE) [19,28]. This modification enables yeast to internally supply both glucose and arabinose sugar donors necessary for complete glycosylation of ASA VI. To implement the ASA VI biosynthetic pathway in yeast, we screened and assembled functional genes from various plant sources. The full pathway was reconstructed by introducing the following genes into S. cerevisiae: ERG1 and ERG9 (to enhance precursor flux), CqBAS1 (β-amyrin synthase) [29], CqCYP716A78 [29], MtCYP72A68v2 [17], and CaCYP714E19 (P450 hydroxylases for position-specific oxidation) [18], AsUGT99D1 (C-3 arabinosyltransferase) [30], CaUGT73AD1 [31], and CaUGT73C8 (glycosyltransferases responsible for stepwise glucose addition at C-28) [32]. Additionally, genes required for UDP-Ara biosynthesis, including AtUGDH3, GuUXS3, and GuUXE1 [19], were co-expressed. All genes were placed under the control of galactose-inducible promoters to enable tunable expression. This combinatorial biosynthetic approach integrated key enzymatic functions from multiple species into a unified yeast platform, establishing a foundational strategy for the microbial synthesis of ASA VI and structurally complex saponins through pathway reconstitution.

Fig. 2.

Proposed combinatorial biosynthesis pathway of ASA VI in S. cerevisiae. Module I: UDP-Ara is synthesized by introducing UGDH, UXS, UGE or UXE. Module II: Overexpression of key enzymes in the MVA pathway and introduction of genes for OA synthesis. Module III: synthesis of ASA VI was achieved by introducing CYP and UGT.

3.2. Engineering of the UDP-Ara biosynthetic pathway in S. cerevisiae

Glucose- and arabinose-derived sugar donors are required for ASA VI biosynthesis, yet S. cerevisiae lacks the native pathway for arabinose donor formation [33]. To address this, we constructed a heterologous three-enzyme module enabling in vivo conversion of UDP-Glc to UDP-Ara via sequential oxidation, decarboxylation, and epimerization steps catalyzed by UGDH, UXS, and UXE, respectively.

Based on an extensive literature survey, candidate genes were selected from several plant species. A. thaliana contributed AtUGDH3 and AtUXS3, G. uralensis provided GuUXS2 and GuUXE1, and PsUGE2 [19], was obtained from P. sativum (Fig. 3A). These genes were modularly assembled to construct two sugar donor pathways, WA2 and WA3, both incorporating AtUGDH3 but differing in the downstream enzymes employed (Fig. 3B). To evaluate the functionality of the arabinose donor synthesis module, the arabinosyltransferase gene AsUGT99D1 from A. sativa [19], which catalyzes the transfer of UDP-Ara to the C-3 hydroxyl group of triterpenoids, was introduced into WA2 and WA3, yielding strains WA4 and WA5, respectively.

Fig. 3.

Construction of UDP-Ara biosynthesis pathway in S. cerevisiae. (A) Schematic diagram of the engineered UDP-Ara biosynthetic pathway. (B) Genotypes of strains WA4 (WAT11AtUGDH3:PsUGE2/AtUXS3:AsUGT99D1) and WA5 (WAT11AtUGDH3:GuUXE1/GuUXS2:AsUGT99D1). (C) Extracted ion chromatograms (EICs) of m/z 588 obtained from LC-MS analysis of WA4 and WA5 after feeding with OA. (D) Relative peak areas of ion abundance were used for the comparisons of Ara-OA yield in WA4 and WA5 (mean ± s.d.; n = 3 technically independent experiments). (E) Mass spectra of the Ara-OA products produced by strains WA4 and WA5, indicating successful transfer of arabinose to the C-3 hydroxyl group of OA via AsUGT99D1.

The functional activity of these strains was assessed by feeding OA as a substrate, followed by LC-MS analysis. A peak at m/z 588 with a retention time of approximately 10.2 min was detected in both WA4 and WA5 (Fig. 3C–E), corresponding to the formation of Ara-OA. Notably, WA5, which expressed the GuUXS2/GuUXE1 gene pair, exhibited a higher signal intensity compared to WA4 (Fig. 3D), indicating greater catalytic efficiency of the G. uralensis derived enzymes. Based on these findings, the gene combination used in WA5 was selected for subsequent pathway construction and functional studies. Collectively, these results confirm the successful de novo assembly of a functional UDP-Ara biosynthesis pathway in S. cerevisiae. Furthermore, the intermediate formation of UDP-GlcA, and UDP-Xyl established a diverse sugar donor chassis, enabling the microbial synthesis of diverse glycosylated natural products beyond ASA VI.

3.3. Pathway reconstruction and flux optimization for OA production in S. cerevisiae

OA, a pentacyclic triterpenoid, serves as the aglycone scaffold in ASA VI biosynthesis. In S. cerevisiae, OA can be synthesized from 2,3-oxidosqualene via a two-step reaction: β-amyrin synthase catalyzes cyclization to β-amyrin, followed by C-28 oxidation by a cytochrome P450 monooxygenase. Among several candidates, CqbAS1 and CqCYP716A78 [29], derived from C. quinoa, were selected based on their previously demonstrated catalytic activity and expression compatibility in heterologous systems. These genes were introduced into the WAT11 yeast chassis to generate strain WA6 (Fig. 4A–D).

Fig. 4.

Construction of the OA biosynthetic pathway in S. cerevisiae. (A) Schematic diagram of the engineered biosynthetic pathway leading to OA production. (B) Extracted ion chromatograms (EICs) of m/z 455 from strains WA6 and WA7. (C) Relative peak areas of ion abundance were used for the comparisons of OA yield in WA6 and WA7 (mean ± s.d.; n = 3 technically independent experiments). (D) Genotypes of engineered yeast strains WA6 (WAT11CqbAS1/CqCYP716A78) and WA7 (WA6ERG1/ERG9). (E) Mass spectra of the OA standard and the OA product detected in strain WA7.

LC-MS analysis of WA6 revealed a characteristic OA peak at m/z 455 with a retention time of approximately 15 min (Fig. 4B), matching the mass spectrum of the OA standard (Fig. 4E). However, the observed signal was relatively weak, indicating suboptimal yield.

To improve OA production, we employed a metabolic flux enhancement strategy by overexpressing two key endogenous genes in S. cerevisiae, ERG9 (encoding squalene synthase) and ERG1 (encoding squalene epoxidase) [34,35], which increase the availability of 2,3-oxidosqualene from farnesyl pyrophosphate. These genes were integrated into the WA6 to construct strain WA7. As shown in Fig. 4C, WA6 produced OA at a titer of 3.6 mg/L, while WA7 reached 10.7 mg/L, demonstrating a clear improvement in yield after ERG pathway enhancement, confirming that precursor pathway optimization effectively enhanced OA biosynthesis in yeast. Based on these results, the genes involved in OA biosynthesis from strain WA7 were used for downstream construction of the full ASA VI biosynthetic pathway.

3.4. Construction of the oxidation and glycosylation synthesis module in S. cerevisiae

3.4.1. C-23 oxidation of OA to generate the HED intermediate

HED, also known as ivoryoside aglycone, is a pentacyclic triterpenoid that serves as a key intermediate in the biosynthesis of ASA VI [36]. In addition to its structural role, HED exhibits various pharmacological activities, including anti-tumor, anti-inflammatory, anti-depressive, and anti-viral effects [37]. Notably, its anti-cancer potential suggests utility in chemotherapeutic development.

To enable de novo production of HED, we gradually constructed strain WA9 using the S. cerevisiae WAT11 chassis. Specifically, we introduced the OA biosynthesis genes (ERG9, ERG1, CqbAS1, and CqCYP716A78), previously functionally validated in strain WA7, into the sugar donor chassis WA3, which already carried AtUGDH3, PsUGE2, AtUXS3, GuUXS2, and GuUXE1 for UDP-Ara production. WA9 thus served as a fully equipped chassis strain combining triterpenoid backbone synthesis and sugar donor supply. Subsequently, WA10, WA11, and WA12 were derived from WA9 by introducing different C-23 hydroxylase candidates (MtCYP72A68v2, CaCYP714E19, and AtCYP71A16 [38], respectively) (Fig. 5A–D), enabling evaluation of their catalytic activities for HED production.

Fig. 5.

Synthesis pathway of HED. (A) Synthetic pathway diagram of HED. (B) Ion chromatograms of m/z 471 extracted from strains WA9 (WA3ERG1/ERG9:CqbAS1/CqCYP716A78), WA10 (WA9MtCYP72A68v2), WA11 (WA9CaCYP714E19), WA12 (WA9AtCYP71A16), WA13 (WA10CaCYP714E19) and WA14 (WA11CaCYP714E19/CaCYP714E19). (C) Relative peak areas of ion abundance were used for the comparisons of HED yield in WA10, WA11, WA13 and WA14 (mean ± s.d.; n = 3 technically independent experiments). (D) Genotypes of strains WA10, WA11, WA12, WA13 and WA14. (E) Mass spectra of the HED standard and the HED product produced by WA14.

LC-MS analysis revealed that both WA10 and WA11 produced HED with a retention time and ion signal matching that of the HED standard (m/z 471) (Fig. 5B–E). Notably, WA11, which expressed CaCYP714E19, showed higher HED yield compared to WA10 expressing MtCYP72A68v2 (Fig. 5B), suggesting superior catalytic efficiency of CaCYP714E19 under the tested conditions. In contrast, WA12, carrying AtCYP71A16, failed to generate detectable levels of HED (Fig. 5B). To improve the C-23 hydroxylation efficiency of OA, we sequentially evaluated two strategies in the WA9 chassis. First, MtCYP72A68v2 and the more catalytically active CaCYP714E19 were co-expressed to construct strain WA13. LC-MS analysis (Fig. 5B) revealed a stronger HED peak in WA13 compared to the single-enzyme strains WA10 and WA11, indicating a synergistic effect between the two P450 enzymes. Building upon this, we next constructed strain WA14 by increasing the copy number of CaCYP714E19 alone. As shown in the quantitative results (Fig. 5C), WA13 achieved an HED titer of 1.3 mg/L, while WA14 reached the highest titer at 3.9 mg/L. These data demonstrate that while co-expression enhances hydroxylation efficiency, increasing the dosage of CaCYP714E19 is a more effective strategy. Therefore, WA14 was selected as the preferred strain for downstream ASA VI pathway reconstruction.

3.4.2. Glycosylation of HED and de novo synthesis of ASA VI

The C-28 glycosylation of HED is a critical step in the biosynthesis of ASA VI, involving the sequential attachment of two glucose moieties to form the intermediate HED-28-Glc-Glc (Fig. 6A). To identify suitable glycosyltransferases for this process, a set of candidate genes was selected through literature review. Specifically, CaUGT73AD1 [31], CaUGT73C8 and CaUGT94M2 all derived from C. asiatica, were tested for catalytic activity at the C-28 position [32,39]. The candidate genes were expressed using the WA14 chassis, which accumulated HED as established in section.

Fig. 6.

Synthesis of HED-28-Glc-Glc. (A) Synthetic pathway diagram of HED-Glc-Glc. (B) Genotypes of strains WA15 (WA14CaUGT73AD1/CaUGT73C8), WA16 (WA14CaUGT73AD1/CaUGT94M2). (C) Extracted ion chromatograms (EICs) of m/z 633 and m/z 795 from WA15, WA16. (D) Mass spectra of HED-28-Glc and HED-28-Glc-Glc detected in WA15.

Strain WA15 was constructed by introducing CaUGT73AD1 and CaUGT73C8 into the chassis WA14, while strain WA16 was constructed by introducing the genes carrying CaUGT73AD1 and CaUGT94M2 into the chassis WA14. LC-MS analysis showed that WA15 successfully produced both mono- and di-glucosylated forms of HED, consistent with the expected masses of m/z 633 and m/z 795 (Fig. 6C and D). In contrast, WA16 accumulated only the mono-glucosylated intermediate, indicating that CaUGT94M2 did not exhibit catalytic activity for the second glucose addition under the tested conditions. These findings demonstrate that CaUGT73AD1 and CaUGT73C8 function as an effective glycosyltransferase pair and were therefore selected for the complete biosynthetic reconstruction of ASA VI (Fig. 6B–D).

The reaction of HED-28-Glc-Glc catalyzed by arabinose glycosyltransferase constitutes the final step in ASA VI biosynthesis. Previously, when validating the sugar donor module, AsUGT99D1 from A. sativa was introduced into the sugar donor chassis to verify its ability to transfer an arabinose moiety to the 3-hydroxyl group of OA, forming Ara-OA. Based on these results, AsUGT99D1 was further introduced into strain WA16, which accumulates HED-28-Glc-Glc, to yield strain WA17 (Fig. 7A–C). After feeding and cultivation, LC-MS analysis revealed a new peak at m/z 927 with a retention time of 1.9 min, consistent with the ASA VI standard (Fig. 7B–D). This peak was detected in WA17 but not in the control strain WA16, confirming the de novo biosynthesis of ASA VI in S. cerevisiae. Although only trace amounts were produced, quantified at 395 ng/L, this result marks the first complete reconstruction of the ASA VI pathway in yeast.

Fig. 7.

Final biosynthetic step and confirmation of de novo ASA VI synthesis. (A) Synthetic pathway diagram of ASA VI biosynthesis from HED-28-Glc-Glc and UDP-Ara. (B) Extracted ion chromatograms (EICs) of m/z 927 from WA17. (C) Genotype of strain WA17 (WA16AsUGT99D1). (D) Mass spectra of the ASA VI standard and the ASA VI product detected in strain WA17.

To gain insight into potential pathway bottlenecks, we next performed targeted feeding experiments using WA17. Supplementation with either OA or HED did not lead to significant increases in ASA VI titer (Fig. S1C and F), nor did it noticeably improve the accumulation of intermediate glycosylation products such as HED-28-Glc (Fig. S1A and D) and HED-28-Glc-Glc (Fig. S1B and E). Moreover, in the fully synthetic system starting from glucose, residual OA and HED were still detected (Fig. S2), suggesting that upstream modules, including OA formation and C-23 hydroxylation, are not limiting under current conditions. These findings collectively indicate that the major bottlenecks reside in the downstream glycosylation steps and the non-native UDP-Ara sugar donor biosynthesis pathway.

4. Discussion

In this study, we successfully reconstructed the complete biosynthetic pathway of ASA VI, a complex oleanane-type triterpenoid saponin, in S. cerevisiae using a modular synthetic biology approach [40]. Unlike traditional strategies that rely on laborious gene mining and de novo pathway elucidation in native plants, we adopted a combinatorial design principle by assembling well-characterized enzyme modules from various plant species. This strategy significantly enhanced the efficiency of pathway reconstruction and demonstrated the feasibility of rapidly engineering microbial hosts to produce structurally complex natural products. Given the current low production titer (395 ng/L), a systematic analysis of pathway bottlenecks is essential for further strain improvement.

The first downstream tailoring step after OA biosynthesis is the C-23 hydroxylation to produce HED. Among the three candidate CYP450 enzymes tested, CaCYP714E19 exhibited the highest catalytic efficiency in yeast. Co-expression of CaCYP714E19 with MtCYP72A68v2 (WA13) modestly improved HED production, while further enhancement via multicopy expression of CaCYP714E19 (WA14) significantly boosted HED titers to 3.9 mg/L. These results confirm the functional viability of the oxidation step. Nonetheless, feeding experiments with OA or HED into the complete biosynthetic strain WA17 failed to improve ASA VI levels, and residual upstream intermediates were still observed. These findings indicate that the C-23 oxidation module is not currently a major bottleneck under the existing pathway configuration. Even so, this step remains a critical target for future optimization, especially as upstream flux increases. Potential strategies include redox partner engineering [41], P450-CPR fusion [42], and cofactor (e.g., NADPH) regeneration to further enhance catalytic efficiency [43,44].

The subsequent glycosylation module emerged as a more significant limiting factor. This module involves three successive sugar addition steps: two glucose moieties at C-28 catalyzed by CaUGT73AD1 and CaUGT73C8, and one arabinose moiety at C-3 catalyzed by AsUGT99D1. Although LC-MS analysis confirmed formation of the expected glycosylated intermediates, their low abundance and the lack of significant titer increase upon precursor feeding suggest limited in vivo catalytic efficiency. In particular, the final arabinosylation step catalyzed by AsUGT99D1 generated only trace levels of ASA VI, highlighting a major bottleneck at this stage. This may stem from poor enzyme expression, folding inefficiency, or suboptimal substrate compatibility. Addressing this issue may require codon or promoter optimization [45,46], directed evolution [47], or replacement with microbial orthologs better suited for yeast expression [48].

In addition to enzyme efficiency, the limited intracellular availability of the sugar donor UDP-Ara further restricts ASA VI synthesis. To overcome this, we introduced a de novo UDP-Ara module composed of AtUGDH3, GuUXS2, and GuUXE1, enabling in vivo conversion of UDP-Glc to UDP-Ara. While this pathway was validated by Ara-OA formation in feeding assays, the low final titer in WA17 indicates insufficient UDP-Ara supply under full-pathway conditions. Future improvements may involve overexpression of limiting enzymes [49], enhancing precursor and cofactor pools (e.g., UDP-Glc, NAD+, UTP) [50], or establishing salvage and recycling routes for arabinose nucleotide sugars [51].

In summary, we established a functional modular platform for the de novo biosynthesis of ASA VI in yeast, confirming the feasibility of engineering arabinosylated triterpenoids via microbial fermentation. While current production remains at a low level, our pathway design offers a solid foundation for future optimization through targeted metabolic engineering.

CRediT authorship contribution statement

Lin Hao: Writing – original draft, Project administration, Investigation, Formal analysis, Data curation. Guiru Dong: Validation, Software, Data curation. Tianzhen Sun: Formal analysis, Data curation. Jingyan Liu: Methodology, Investigation, Formal analysis. Hui Wu: Software, Resources, Investigation, Data curation. Fahui Li: Supervision, Resources, Conceptualization. Weiguo Song: Validation, Resources. Xiaozhou Luo: Writing – review & editing, Validation. Jian Zhang: Writing – review & editing, Resources. Yanan Qiao: Writing – original draft, Supervision, Funding acquisition, Formal analysis.

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. ASupplementary data

The following is the Supplementary data to this article.