Engineering Salvia miltiorrhiza hairy roots as a scalable platform for high-yield production of paclitaxel and ginsenoside precursors

Abstract

Plant-derived natural products offer a rich source of therapeutic agents. However, sustainable and high-yield production remains a grand challenge. We engineered Salvia miltiorrhiza hairy roots to produce taxadiene, a key precursor for the anticancer drug paclitaxel, and protopanaxadiol, a precursor for ginsenosides. The heterologous expression of two key biosynthetic genes, taxadiene synthase from Taxus wallichiana and protopanaxadiol synthase from Panax notoginseng, enabled the production of taxadiene and protopanaxadiol, respectively. Our strategy combined multiple approaches to enhance terpenoid production, including genome editing to redirect metabolic flux by eliminating a competing GGPP sink (via SmCPS1 disruption), transcriptional reprogramming through SmWRKY61 overexpression to enhance terpenoid precursor pathways (MVA/MEP), and optimization of cultivation conditions. This holistic approach yielded 65.17 ± 5.25 mg/kg fresh weight (FW) taxadiene in batch cultures, and the protopanaxadiol yield reached 50.04 ± 2.94 mg/kg dry weight (DW) without optimization. These results highlight the potential of this platform for industrial-scale production. Our findings demonstrate that S. miltiorrhiza hairy roots can serve as a robust and scalable platform to produce valuable plant-derived compounds. This work paves the way for future metabolic engineering efforts to achieve cost-effective and sustainable production of high-value natural products using medicinal plant systems, addressing critical supply bottlenecks for pharmaceutical compounds.

This study reports the engineering of Salvia miltiorrhiza hairy roots as a scalable platform for the high-yield production of paclitaxel and ginsenoside precursors, taxadiene and protopanaxadiol, respectively. Efficient production of these high-value diterpenoid precursors was achieved through the introduction of key enzyme genes, knockout of competing diterpenoid pathways, and enhancement of precursor supply, demonstrating the potential of this approach to provide a sustainable supply of pharmaceutical ingredients.

Introduction

Plant-derived natural products represent the largest and most diverse class of natural products, including diterpenoids such as paclitaxel, a clinically used anticancer drug, and triterpenoids such as ginsenosides, a group of compounds isolated from ginseng with reported pharmacological properties and exhibiting a wide range of biological activities (Goddard et al., 2024). Biosynthesis of these complex compounds involves a series of modifications, including cyclizations, oxidations, and glycosylations, catalyzed by a diverse array of enzymes, such as terpene synthases and cytochrome P450 monooxygenases (Singh et al., 2023). Although numerous attempts have been made to bioengineer the production of these natural compounds or their precursors in microbial or model plant systems, the achievement of high yields remains challenging owing to inherent limitations, including supply chain issues that hinder the reliable and sustainable supply of these valuable compounds (Hu et al., 2023). Consequently, the production of plant-derived bioactive compounds relies primarily on plant cell culture techniques or semi-synthesis (Lin et al., 2023). This necessitates the exploration of alternative, sustainable biomanufacturing platforms to ensure a consistent and adequate supply of plant-derived natural products for medicinal and other applications.

The rapid development of metabolic engineering and synthetic biology offers promising avenues for bioengineering the production of plant-derived natural compounds, intermediates, or precursors (Engler et al., 2014; Wu et al., 2021; Bradley et al., 2023; Jiang et al., 2024). Microbial hosts such as Escherichia coli have been successfully bioengineered to synthesize the paclitaxel precursors taxadiene and taxadien-5α-ol by integrating the upstream methylerythritol phosphate (MEP) pathway and terpenoid-forming pathway (Ajikumar et al., 2010). Similarly, yeast has been used as a host organism to biosynthesize the diterpene precursor geranylgeranyl pyrophosphate (GGPP) through manipulation of native mevalonate (MVA) pathways, leading to the production of taxadiene and taxadien-5α-ol (Malcı et al., 2023). Similarly, engineering the MVA pathway in yeast has enabled the efficient synthesis not only of protopanaxadiol but also of its downstream product ginsenoside Rh2 (Wang et al., 2019). Furthermore, strategies involving the overexpression of MVA pathway genes, plant-derived protopanaxadiol biosynthetic enzymes, and transcription factors (TFs) in yeast, coupled with the use of strong promoters and the inhibition of competing pathways, have achieved a protopanaxadiol titer of 93.1 mg/l (Zhou et al., 2023). These advances in metabolic engineering have propelled the production of essential natural products. However, these strategies necessitate intricate bioengineering techniques, including the introduction of upstream pathways such as GGPP biosynthesis and the modification of cytochrome P450 expression or structural modifications to address issues such as incorrect folding and incorporation in microbial biosynthesis of plant natural products (Lin et al., 2023). Model plants such as Arabidopsis thaliana and Nicotiana benthamiana have been exploited to harness the native pathway for GGPP production by expressing rate-limiting enzymes to produce paclitaxel precursors (Supplemental Table 1) (Barja and Rodriguez-Concepcion, 2021; Dong et al., 2022). In addition, there are reports on plant-based synthesis of ginsenosides. For instance, a hairy root system of Vietnamese ginseng has been used to generate dammarane-type saponins (Loan et al., 2016). Overexpression of the NAC72 TF in Panax ginseng callus significantly enhanced total saponin content, particularly that of dammarane-type ginsenosides (Ting et al., 2024). However, the levels achieved were suboptimal, and there were constraints related to substrate limitations.

Compared with microbial systems (E. coli and yeast) and model plants (A. thaliana and N. benthamiana), cell culture systems derived from medicinal plants present a compelling and often more advantageous platform for the biosynthesis of complex plant natural products. This advantage arises from two principal attributes: first, the capacity to generate essential precursors and substrates via endogenous high-flux metabolic pathways and second, the natural endowment of diverse, specialized enzyme systems—notably cytochrome P450 monooxygenases (P450s) and their associated reductases. These enzymes intrinsically contribute to the transformation of diterpenoid and triterpenoid core structures into a spectrum of high-value, typically bioactive metabolites (Bohlmann and Keeling, 2008; Guo et al., 2022; Lin et al., 2023; Perez-Matas et al., 2024). By contrast, microbial hosts often require extensive engineering for the functional incorporation of plant P450s—a process frequently challenged by protein misfolding, low stability, and inefficient electron transfer—and model plants generally lack a full complement of specialized modifying enzymes. Medicinal plant cells, however, inherently harbor these advanced catalytic systems. Salvia miltiorrhiza hairy roots are a representative example of such an ideal chassis. Recognized for the high-yield production of pharmaceutically relevant diterpenoids (tanshinones) and phenolic acids (salvianolic acids), these root cultures demonstrate intrinsically elevated flux through the endogenous diterpenoid biosynthetic pathway (Supplemental Figure 1). The hairy roots of S. miltiorrhiza therefore constitute an innovative and versatile platform for the synthesis of other valuable terpenoids, including paclitaxel and ginsenoside precursors. Importantly, in addition to its rich native metabolite profile, S. miltiorrhiza inherently possesses an array of P450s and reductases that facilitate regio- and stereospecific hydroxylations and other complex oxidative modifications. This native enzymatic repertoire alleviates a major bottleneck in heterologous biosynthesis, offering a distinct advantage over systems that require co-expression and optimization of multiple heterologous P450s (Bohlmann and Keeling, 2008; Wu et al., 2021). S. miltiorrhiza hairy roots are also supported by established, efficient genetic transformation protocols, advanced gene-editing tools (CRISPR–Cas9), and strategies for transcriptional metabolic reprogramming. Together, these features render them a tractable and promising chassis for targeted metabolic engineering and pathway refinement (Shi et al., 2021; Zhang et al., 2021; Xia et al., 2023). By harnessing their inherent high precursor flux, endogenous enzymatic diversity, and genetic plasticity, S. miltiorrhiza hairy roots can be used as a robust and sustainable platform for the production of complex plant-derived compounds. In recent years, our group and others have extensively explored this metabolic engineering system to enhance the production of native secondary metabolites through a multipronged approach encompassing genome editing, metabolic pathway optimization, elicitation, and transcriptional reprogramming (Kai et al., 2011; Zhao et al., 2015; Xu et al., 2016; Hao et al., 2020; Chen et al., 2022; Cheng et al., 2023; Hsu et al., 2024). By synergizing this interdisciplinary expertise, we aim to develop a novel, universal medicinal plant cell chassis capable of producing valuable bioactive compounds, exemplified by paclitaxel and ginsenosides. In this work, we successfully heterologously synthesized the key precursors of paclitaxel and ginsenosides, taxadiene and protopanaxadiol, and used a multi-dimensional approach to optimize taxadiene biosynthesis in S. miltiorrhiza hairy roots, achieving a yield of 65.17 ± 5.25 mg/kg fresh weight (FW). The content of protopanaxadiol reached 50.04 ± 2.94 mg/kg dry weight (DW) without further optimization strategies. Our findings underscore the potential for S. miltiorrhiza hairy roots to serve as a robust and sustainable platform for the production of complex terpenoid compounds. This work provides a foundation for future metabolic engineering endeavors focused on optimizing the production of valuable natural products.

Results

Establishment of S. miltiorrhiza hairy roots for heterologous production of paclitaxel and ginsenoside precursors

To evaluate the heterologous expression of natural compounds in the S. miltiorrhiza hairy root system, we initially established the pathway to the paclitaxel precursor, taxadiene. The taxadiene synthase gene (TwTS) from Taxus wallichiana, whose encoded enzyme converts GGPP to taxadiene, was cloned into the plant binary vector pK7WG2R to generate the pK7WG2R-TS construct (Figure 1A). Agrobacterium rhizogenes ATCC15834 harboring the pK7WG2R-TS vector, which contained a red fluorescent protein, was used to infect S. miltiorrhiza and generate TwTS-expressing hairy root lines. Two independent lines, TS(+)#1 and TS(+)#8, were established through multiple rounds of selection with the appropriate antibiotic (Figure 1B and Supplemental Figure 2).

Figure 1.

Heterologous expression of taxadiene in S. miltiorrhiza hairy roots.

(A) Schematic diagram of the TwTS overexpression vector pK7WG2R-TS. p35S and pNOS denote the promoters, and tNOS denotes the terminator. DsRed denotes red fluorescent protein. RB and LB are the right boundary and left boundary, respectively.

(B) Visualization and analysis of the heterologous expression of TwTS in S. miltiorrhiza hairy roots using a LUYOR-3415RG dual fluorescent protein viewing lamp. Scale bars, 1 cm.

(C) Biomass was measured in the indicated lines.

(D) GGPP concentration was measured in the indicated lines.

(E) Taxadiene concentration was measured in the indicated lines.

(F) Gas chromatogram information for taxadiene in recombinant S. miltiorrhiza TS(+) hairy root lines. WT, wild-type hairy roots; TS(+), TwTS overexpression.

Values represent the mean ± SD, n = 3. ∗p < 0.05; ns, no significant difference (Student’s t test).

Under controlled conditions for 30 days, the TS(+) lines displayed reduced biomass compared with the wild-type (WT) line (Figure 1C). We also observed a reduction in GGPP, the precursor of both taxadiene and other diterpenoids, in the TS(+) lines compared with the WT (Figure 1D). Although TS(+)#1 showed no statistically significant reduction in GGPP content, TS(+)#8 showed a significant reduction in GGPP (Figure 1D). Gas chromatography–mass spectrometry (GC–MS) analysis confirmed taxadiene production in the selected TS-expressing S. miltiorrhiza hairy root lines (Figure 1E and 1F and Supplemental Figure 3), with yields of 3.02 ± 0.78 mg/kg FW and 4.29 ± 0.28 mg/kg FW in the recombinant lines, and taxadiene levels were inversely correlated with intracellular GGPP accumulation (Figure 1D and 1E). We propose that the observed reduction in GGPP pool size was due primarily to its biosynthetic conversion into taxadiene. This interpretation is supported by the restoration of GGPP to WT levels in CPS(−)/TS(+) lines, which coincided with elevated taxadiene production, indicating effective diversion of GGPP toward taxadiene synthesis. Nonetheless, contributions from other GGPP-utilizing metabolic pathways, such as gibberellin biosynthesis, cannot be ruled out. Together, these results confirm the successful functional expression of TwTS in S. miltiorrhiza for the production of taxadiene.

To demonstrate the versatility of S. miltiorrhiza hairy roots as a platform for heterologous biosynthesis of other terpenoid precursors, we reconstituted the biosynthesis of protopanaxadiol, a precursor of ginsenosides, by expressing protopanaxadiol synthase (PPDS). The PPDS gene CYP716A47 (PnPPDS) from Panax notoginseng was cloned into the plant binary vector pK7WG2R, creating the pK7WG2R-PPDS recombinant vector (Supplemental Figure 4A). This recombinant vector was introduced into S. miltiorrhiza via A. rhizogenes ATCC15834, generating a hairy root line capable of expressing PnPPDS. After multiple rounds of antibiotic selection, three independent lines, PPDS(+)#4, PPDS(+)#8, and PPDS(+)#9, were obtained (Supplemental Figures 4B and 4C). The PPDS(+)#8 hairy root line accumulated the highest content of protopanaxadiol, reaching 50.04 ± 2.94 mg/kg DW, whereas no detectable product was produced in the WT (Supplemental Figure 4D). High-performance liquid chromatography–mass spectrometry (HPLC–MS) analysis confirmed the heterologous synthesis of protopanaxadiol in PnPPDS-expressing S. miltiorrhiza hairy roots (Supplemental Figure 4E). Collectively, our data demonstrate the feasibility of using S. miltiorrhiza hairy roots as a platform for heterologous production of terpenoid precursors.

Metabolic flux engineering via genomic editing to enhance precursor supply for taxadiene production

To increase the metabolic flux toward taxadiene biosynthesis and reduce competitive consumption of the shared precursor GGPP by native diterpenoid pathways, we targeted the recently reported enzyme copalyl diphosphate synthase (CPS) (Zheng et al., 2025). This enzyme catalyzes the committed cyclization of GGPP to copalyl diphosphate (CPP), which represents the entry step into the tanshinone biosynthetic pathway (Supplemental Figure 1). Disruption of CPS thus limits flux toward tanshinone synthesis and enriches the intracellular GGPP pool available for heterologous taxadiene production. Guided by this rationale, we used CRISPR-Cas9-mediated knockout of the CPS gene in S. miltiorrhiza hairy roots to redirect GGPP toward taxadiene synthesis, an approach conceptually aligned with earlier metabolic engineering efforts (Wang and Peters, 2022). Among the CPS family members in S. miltiorrhiza, SmCPS1 was selected as the target because it encodes the predominant isoform—highly expressed and enzymatically active in hairy roots—and is functionally essential for tanshinone biosynthesis (Ma et al., 2021). Owing to its central role in partitioning GGPP into the tanshinone pathway, SmCPS1 represents an ideal genomic editing target for efficient metabolic redirection with minimal off-target effects. First, we designed an sgRNA targeting the first exon of the SmCPS1 gene and generated the SmCPS1 gene-editing vector 2218A-CPS1 (Figure 2A). PCR amplification and sequencing of the target loci revealed various base changes in SmCPS1 in the recombinant lines CPS(−)#4 and CPS(−)#5 (Figure 2B). The two recombinant hairy root lines were noticeably whiter than the WT, possibly owing to reduced tanshinone production (Xia et al., 2023) (Figure 2C and Supplemental Figure 1). We next investigated the downstream biosynthesis of diterpenoids to confirm the biological consequences of SmCPS1 knockout. As shown in Figure 2D and 2E, the biosynthesis of tanshinones was compromised, and the expression levels of SmCYP76AH3 and SmCYP76AK1, which are involved in the native tanshinone biosynthetic pathway, were also reduced in CPS(−)#4 and CPS(−)#5 relative to the WT (Supplemental Figure 5A). These observations suggest that SmCPS1 is important for tanshinone biosynthesis in S. miltiorrhiza, consistent with previous findings (Gao et al., 2009). We then analyzed the accumulation of GGPP and biomass in these SmCPS1-edited hairy roots and found that SmCPS1(−) increased the GGPP pool in S. miltiorrhiza hairy roots and resulted in growth inhibition (Figure 2L and 2M). Our results suggest that SmCPS1 knockout can increase the GGPP pool for metabolic engineering of terpenoid biosynthesis.

Figure 2.

Gene editing of SmCPS1 enhances the GGPP pool and taxadiene biosynthesis in S. miltiorrhiza hairy roots.

(A) Schematic diagram of the SmCPS1 gene-editing vector 2218A-CPS1. pNOS, pU6, and pUBQ10 denote the promoters, and tNOS and tMAS denote the terminators. NPTII denotes a resistance gene. RB and LB are the right boundary and left boundary, respectively.

(B) Target sites for the SmCPS1 gene in hairy roots of S. miltiorrhiza. Solid arrows indicate base substitutions, and dashed arrows indicate base indels.

(C) Phenotypic analysis of SmCPS1-knockout S. miltiorrhiza hairy roots.

(D) Cryptotanshinone concentration was measured in the indicated lines.

(E) HPLC analysis of different tanshinones in WT and two types of recombinant S. miltiorrhiza hairy roots. (1), (2), (3), and (4) represent dihydrotanshinone I, cryptotanshinone, tanshinone I, and tanshinone IIA, respectively.

(F) Target sites for the SmCPS1 gene in hairy roots of S. miltiorrhiza. Solid arrows indicate base substitutions, and dashed arrows indicate base indels.

(G) Visualization and analysis of the heterologous expression of TwTS in S. miltiorrhiza hairy roots in the SmCPS1-knockout background using a LUYOR-3415RG dual fluorescent protein viewing lamp.

(H–K) Concentrations of different tanshinones were measured in the indicated lines.

(L) Biomass was measured in the indicated lines.

(M) GGPP concentration was measured in the indicated lines.

(N) Taxadiene concentration was measured in the indicated lines.

WT, wild-type hairy roots; CPS(−), CRISPR-Cas9-mediated knockout of SmCPS1; CPS(−)/TS(+), CRISPR-Cas9-mediated knockout of SmCPS1 and TwTS overexpression. CT, DT I, TA I, and TA IIA represent cryptotanshinone, dihydrotanshinone I, tanshinone I, and tanshinone IIA, respectively. Scale bars, 1 cm in (C) and (G). Values represent the mean ± SD, n = 3. ∗∗∗p < 0.001, ∗p < 0.05; ns, no significant difference (Student’s t test).

We next co-transformed the SmCPS1 gene-editing vector 2218A-CPS1 with the TwTS overexpression vector pK7WG2R-TS, resulting in CPS(−)/TS(+) recombinant hairy root lines that exhibited simultaneous SmCPS gene knockout and TwTS overexpression. PCR amplification and sequencing of the SmCPS1 loci in CPS(−)/TS(+)#2 and CPS(−)/TS(+)#3 confirmed knockout of CPS (Figure 2F). Red fluorescent protein expression was also observed in these lines, demonstrating successful integration and expression of the TwTS transgene (Figure 2G). HPLC analysis revealed significantly lower levels of diterpenoids, including dihydrotanshinone, cryptotanshinone, and tanshinone I, in the CPS(−)/TS(+)#2 and CPS(−)/TS(+)#3 lines and a lower level of tanshinone IIA in the CPS(−)/TS(+)#3 line compared with the WT (Figure 2E and 2H–2K), consistent with earlier findings (Wang and Peters, 2022). The expression of SmCYP76AH1, which is involved in the native tanshinone biosynthetic pathway, was also significantly reduced in the CPS(−)/TS(+)#2 and CPS(−)/TS(+)#3 lines compared with the WT, whereas a significant increase in the expression of SmKSL1 was observed owing to the feedback-regulatory effect (Supplemental Figure 5B). To examine whether metabolic flux was redirected toward taxadiene biosynthesis by inhibiting native terpenoid biosynthesis from GGPP, we performed a 30-day culture experiment and measured the biomass and relevant metabolite contents in different types of recombinant hairy roots (Figure 2L and 2M). GC–MS analysis revealed that the production of taxadiene in CPS(−)/TS(+)#2 and CPS(−)/TS(+)#3 lines was significantly increased to 4.53 ± 0.32 mg/kg FW and 7.69 ± 1.23 mg/kg FW, respectively, representing a 1.5-fold and 2.5-fold increase compared with the TS(+)#1 line (Figure 2N). Notably, the CPS(−)#4 and CPS(−)#5 lines exhibited reduced biomass and increased GGPP accumulation compared with the WT (Figure 2L and 2M). By contrast, the CPS(−)/TS(+)#2 and CPS(−)/TS(+)#3 lines displayed biomass and GGPP levels similar to those of the WT. We also determined the content of miltiradiene in the CPS(−)#4 and CPS(−)/TS(+)#3 lines (Supplemental Figure 6) and found that the recombinant lines contained significantly less miltiradiene than the WT and that the CPS(−)/TS(+)#3 line contained less miltiradiene than CPS(−)#4. These findings suggest that deletion of SmCPS1 can increase the GGPP pool and direct the metabolic flux of GGPP to the taxadiene synthesis pathway, thereby facilitating taxadiene production.

Transcriptional reprogramming of early terpenoid biosynthesis in S. miltiorrhiza hairy roots through overexpression of SmWRKY61 to boost taxadiene production

TF engineering is a promising strategy for enhancing metabolite production in plants (Butelli et al., 2008). To investigate the potential of TF engineering to enhance metabolite production in S. miltiorrhiza hairy roots, we genetically modified S. miltiorrhiza hairy roots by expressing the TF SmWRKY61, a previously characterized positive regulator of diterpene biosynthesis (Zhang et al., 2021). In a previous omics analysis of S. miltiorrhiza, we discovered that SmWRKY61 can significantly regulate the expression of genes in the precursor MEP pathway and the tanshinone pathway, thereby promoting the synthesis of downstream products (Chen et al., 2022). We cloned the SmWRKY61 gene into a plant expression vector driven by the UBQ10 promoter (Figure 3A) and introduced it into the CPS(−)/TS(+)#3 line (Supplemental Figures 7 and 8 and Supplemental Table 2) to generate W61(+)/CPS(−)/TS(+) recombinant transgenic lines. Two independent lines, W61(+)/CPS(−)/TS(+)#1 and W61(+)/CPS(−)/TS(+)#3, were selected for further analysis. We observed red fluorescent protein expression in both of these lines, similar to TS(+) and CPS(−)/TS(+) lines (Figure 3B), demonstrating the successful expression of TwTS. Quantitative reverse-transcription PCR (RT–qPCR) confirmed that the expression of SmWRKY61 was higher in these lines than in the WT (Supplemental Figure 9).

Figure 3.

Heterologous production of taxadiene in transgenic S. miltiorrhiza hairy roots overexpressing SmWRKY61.

(A) Schematic diagram of the SmWRKY61 overexpression vector. pUBQ10 and pNOS denote the promoters, and tNOS denotes the terminator. RB and LB are the right boundary and left boundary, respectively.

(B) Visualization and analysis of the heterologous expression of TwTS in SmWRKY61-overexpressing S. miltiorrhiza hairy root lines. Scale bars, 1 cm.

(C–E) Comparison of biomass (C), GGPP content (D), and taxadiene content (E) in the indicated S. miltiorrhiza hairy root lines.

(F) Gas chromatogram analysis of taxadiene in transgenic S. miltiorrhiza hairy roots. TS(+)#1 was used as a reference for the analysis of significance in (E).

W61(+)/CPS(−)/TS(+), SmWRKY61 overexpression, CRISPR-Cas9-mediated knockout of SmCPS1, and TwTS overexpression. Values represent the mean ± SD, n = 3. ∗∗∗p < 0.001, ∗p < 0.05; ns, no significant difference (Student’s t test).

To investigate the effect of WRKY61 overexpression on early terpenoid biosynthesis, we analyzed the mRNA levels of genes involved in terpenoid biosynthesis, such as SmHMGS, SmDXS2, SmDXR, and SmIPPI. Expression of these genes was elevated in the W61(+)/CPS(−)/TS(+) lines compared with the parental line CPS(−)/TS(+)#3 (Supplemental Figure 9), consistent with the known transcriptional activation of terpenoid biosynthesis by WRKY61 (Chen et al., 2022). Whereas biomass was reduced in the W61(+)/CPS(−)/TS(+) lines, GGPP content was increased compared with the parental line (Figure 3C and 3D). Taxadiene accumulation reached 45.53 ± 1.07 mg/kg FW in W61(+)/CPS(−)/TS(+)#1 and 41.84 ± 4.64 mg/kg FW in W61(+)/CPS(−)/TS(+)#3, representing a 15.1-fold and 13.9-fold increase, respectively, compared with the TS(+)#1 hairy root line (Figure 3E and 3F and Supplemental Figure 10). The overexpression of SmWRKY61 thus led to increased GGPP levels and a dramatic upregulation of taxadiene production.

Large-scale production of taxadiene in bioengineered S. miltiorrhiza hairy roots through elicitation

Elicitors, such as yeast extract (YE) and methyl jasmonate (MeJA), are well-established signaling molecules that can stimulate the biosynthesis of plant secondary metabolites (Zheng et al., 2021; Cao et al., 2024). To optimize taxadiene production in our bioengineered S. miltiorrhiza hairy roots, we subjected the CPS(−)/TS(+)#3 line to treatments with YE and MeJA. The elicitation treatments significantly increased taxadiene biosynthesis (Figure 4A) by 2.4-fold and 3.5-fold, respectively, although they also led to reduced biomass (Supplemental Figure 11A and 11B), likely due to the redirection of resources toward secondary metabolite production. Notably, this trade-off was mitigated by significantly higher accumulation of taxadiene in response to MeJA treatment relative to YE treatment (Figure 4). To determine the molecular basis of this phenomenon, we performed RT–qPCR to explore the expression of genes encoding enzymes of the terpenoid biosynthetic pathway in CPS(−)/TS(+)#3 with or without elicitation. Key genes involved in the early terpenoid biosynthesis pathway, including SmDXS2, SmDXR, SmHMGS, and SmGGPPS, were induced upon YE (Figure 4B–4F) or MeJA treatment (Figure 4G–4K). For example, the expression of SmDXS2 was increased 1.9-fold and 4.9-fold in response to YE and MeJA, respectively, and the expression of SmDXR was enhanced 2.5-fold and 3.9-fold by YE and MeJA under the same conditions. Notably, SmHMGS expression was dramatically induced by MeJA at 12 h post treatment (18.4-fold) but returned to a 1.5-fold increase after 6 days. Similarly, SmIPPI expression peaked at 12 h (8.9-fold) and then decreased to 5.8-fold after 3 days with MeJA treatment. SmGGPPS expression also responded more rapidly upon MeJA treatment. Overall, the majority of genes were expressed at higher levels with MeJA treatment than with YE treatment, correlating positively with taxadiene accumulation.

Figure 4.

Taxadiene accumulation in the transgenic S. miltiorrhiza hairy root line CPS(−)/TS(+)#3 under elicitor treatment.

(A) Accumulation and comparison of taxadiene in the indicated S. miltiorrhiza hairy root lines.

(B–F) RT–qPCR analysis of the expression of genes related to diterpenoid precursor synthesis in CPS(−)/TS(+)#3 hairy roots after treatment with yeast extract (YE, 25 mg/l).

(G–K) RT–qPCR analysis of the expression of genes related to diterpenoid precursor synthesis in CPS(−)/TS(+)#3 hairy roots after treatment with methyl jasmonate, (MeJA, 200 μM). TS(+)#1 was used as the reference for the analysis of significance in (A). β-Actin was used as the internal reference gene for RT–qPCR.

Values represent the mean ± SD, n = 3. ∗∗∗p < 0.001, ∗p < 0.05; ns, no significant difference (Student’s t test).

To further confirm this elicitation effect, we next examined whether elicitation would induce taxadiene biosynthesis in another bioengineered hairy root line, W61(+)/CPS(−)/TS(+)#1. Consistent with our previous observations, MeJA treatment of this engineered hairy root line led to a substantial reduction in biomass, accompanied by a dramatic increase in taxadiene content to 65.17 ± 5.25 mg/kg FW (Figure 5A and Supplemental Figure 11C). This represents a 21.6-fold enhancement over the TS(+)#1 line, underscoring the synergistic effect of MeJA and SmWRKY61 overexpression. Moreover, the combined treatment strongly upregulated the expression of early terpenoid biosynthesis genes and SmWRKY61 (Figure 5B–5G), indicating a coordinated transcriptional response.

Figure 5.

Taxadiene accumulation in S. miltiorrhiza W61(+)/CPS(−)/TS(+)#1 hairy roots upon elicitor treatment.

(A) Accumulation and comparison of taxadiene in the indicated recombinant S. miltiorrhiza hairy root lines.

(B–G) RT–qPCR analysis of the expression of SmWRKY61 and genes related to diterpenoid precursor synthesis in W61(+)/CPS(−)/TS(+)#1 hairy roots after MeJA treatment (200 μM). TS(+)#1 was used as the reference for the analysis of significance in (A). β-Actin was used as the internal reference gene.

Values represent the mean ± SD, n = 3. ∗∗∗p < 0.001, ∗p < 0.05; n.d., not detected; ns, no significant difference (Student’s t test).

Discussion

Over the past decade, significant advances have been made in engineering the biosynthesis of plant natural products through heterologous expression in microorganisms and model plants (Wu et al., 2022; Zhang et al., 2023; Jiang et al., 2024; Liu et al., 2024; Yang et al., 2024). However, these approaches often encounter challenges related to precursor supply and metabolic efficiency. This study leveraged the medicinal plant S. miltiorrhiza hairy root system to develop a novel platform for the heterologous biosynthesis of natural plant compounds. It also aimed to establish a biosynthetic platform for taxadiene and protopanaxadiol. We successfully achieved heterologous production of these two key precursors in engineered S. miltiorrhiza hairy roots. Detection of their accumulation, along with verification of TwTS and PnPPDS gene expression, confirmed the functionality of the heterologous pathway, providing preliminary evidence for the potential of this platform in natural product development. The further increase in taxadiene production achieved through elicitation, particularly with MeJA, demonstrated its synergistic effect with both genome editing and transcriptional reprogramming. This multipronged approach, which achieved 65.17 ± 5.25 mg/kg FW in batch cultivation (Figure 5A), illustrates the power of integrating various metabolic engineering strategies to achieve high-level production of complex natural products in S. miltiorrhiza hairy roots. We also performed preliminary scale-up experiments in a 5-l bioreactor (Supplemental Figure 12). During the 30-day cultivation period, the CPS(−)/TS(+)#3 strain was inoculated into 3 l of liquid medium, ultimately achieving taxadiene accumulation of 8.52 mg/kg FW (the taxadiene content in shake flasks was 7.69 mg/kg FW). This study thus establishes a robust framework for future metabolic engineering efforts, leveraging native precursor flux and transcriptional bioengineering to optimize the biosynthesis of complex natural products using medicinal plant chassis.

The heterologous expression of complex plant natural products faces several challenges. These include the intricate interplay among precursor supply, intermediate channeling from primary metabolism, functionalization of downstream products, and co-factor availability within the host organism. Thus, metabolic engineering is frequently used to optimize precursor supply and functional modification of downstream products for the bioengineered biosynthesis of complex plant products (Guo et al., 2013; Bureau et al., 2023; Lin et al., 2023; Jiao et al., 2024; Kant et al., 2024; Wang et al., 2024). Upstream pathways, such as the MVA/MEP pathways, in microorganisms like E. coli and S. cerevisiae often have insufficient capacity to produce the necessary isoprenoid building blocks, isopentenyl pyrophosphate (IPP) and dimethylallyl pyrophosphate (DMAPP), for biosynthesis of the diterpenoid precursor GGPP (Moser and Pichler, 2019). Molecular engineering strategies to modify the upstream biosynthetic pathway have been implemented to address this limitation, leading to increased production of the desired molecules. For example, overexpression of bottleneck enzymes in the MEP pathway (e.g., dxs, idi, ispD, and ispF) significantly enhanced taxadiene production in engineered E. coli (Ajikumar et al., 2010). Similarly, introduction of the isopentenol utilization pathway (IUP) into S. cerevisiae led to the accumulation of IPP/DMAPP, resulting in GGPP accumulation and increased terpenoid biosynthesis (Ma et al., 2023). In the model plant tobacco, co-expression of DXR, IspH, or IDI with TS led to some improvements in taxadiene production, although DXS overexpression had the most significant effect (Li et al., 2019). Here, we leveraged the diterpenoid-rich plant S. miltiorrhiza, with its abundance of diterpenoid precursors and diterpenoids, to serve as an innovative platform for taxadiene biosynthesis (Supplemental Figure 1). Our results demonstrate that disruption of the native diterpenoid pathway by SmCPS1 knockout effectively redirected GGPP flux toward heterologous taxadiene biosynthesis. This strategy not only enhanced taxadiene production but also mitigated the growth inhibition observed when GGPP accumulated in the SmCPS1 single-knockout lines (Figure 2I), suggesting a successful rebalancing of metabolic resources. Our study highlights the significant advantage of using a high-flux native plant system for engineering complex natural products.

Previous studies have demonstrated that TF engineering is a promising approach for enhancing metabolite production (Butelli et al., 2008). MeJA, a well-known elicitor of diterpenoid biosynthesis, has been extensively used to identify MeJA-responsive TFs for bioengineering diterpenoid pathways (Sharma et al., 2019; Sun et al., 2019; Wasternack and Strnad, 2019; Zheng et al., 2021; Li et al., 2024). For instance, the MeJA-responsive TF ERF106 can bind directly to the promoters of SmKSL1 and SmIDI1, key enzymes in tanshinone biosynthesis, upregulating their expression and enhancing diterpenoid production (Sharma et al., 2019). However, to date, no TF has been successfully used to engineer Taxol biosynthesis in a heterologous host. Our prior research identified a MeJA-responsive WRKY TF, SmWRKY61, capable of stimulating diterpenoid biosynthesis in S. miltiorrhiza (Chen et al., 2022). In this study, we bioengineered S. miltiorrhiza hairy roots by overexpressing SmWRKY61, and the substantial increase in taxadiene production (15.1-fold compared with TS(+)#1) upon SmWRKY61 overexpression demonstrated the efficacy of transcriptional reprogramming for increasing precursor supply and overall product yield. This highlights the potential of using TFs to fine-tune metabolic pathways for enhanced natural product biosynthesis.

In summary, a final taxadiene production of 65.17 ± 5.25 mg/kg FW was achieved in batch cultivation through a multi-dimensional bioengineering approach (Figure 6). There still remains a notable gap between the current yield and that achieved through heterologous synthesis in microbial chassis systems. This discrepancy may be attributable to the intricate regulatory networks governing plant metabolism, lower cultivation efficiency, and constraints on biomass accumulation. Future efforts will therefore focus on further optimization of the metabolic network through synthetic biology strategies, including the combinatorial regulation of multiple rate-limiting steps to enhance overall flux and the development of novel bioreactor systems to improve cultivation efficiency and scalability. Nonetheless, our experimental results demonstrate that S. miltiorrhiza hairy roots possess a high capacity for efficient integration of multi-step enzymatic pathways. They exhibit excellent compatibility with plant-specific P450 enzyme systems, providing a robust foundation chassis for the biosynthesis of high-value natural products that rely on complex oxidative modifications. This work paves the way for future metabolic engineering in medicinal plants, leveraging synthetic biology and genome-editing technologies to achieve low-cost, industrial-scale production of high-value plant-derived compounds. This platform holds the greatest promise for initial application in the small- to medium-scale production of high-value, structurally complex drug intermediates that are challenging to synthesize in microbial systems. It also serves as an excellent research tool for characterizing and reconstructing the biosynthetic pathways of plant natural products.

Figure 6.

Paradigm of multi-dimensional approaches to taxadiene biosynthesis in S. miltiorrhiza hairy roots.

Step I, TwTS overexpression; step II, CRISPR–Cas9 knockout of SmCPS1; step III, SmWRKY61 overexpression; and step IV, MeJA elicitation treatment. Upward arrows indicate increased levels of gene expression, downward arrows indicate reduced contents of products, and dashed arrows indicate multi-step enzymatic reactions. CT, DT I, TA I, and TA IIA represent cryptotanshinone, dihydrotanshinone I, tanshinone I, and tanshinone IIA, respectively. DXS, 1-deoxy-D-xylulose 5-phosphate synthase; DXR, 1-deoxy-D-xylulose 5-phosphate reductoisomerase; MCT, 2-C-methyl-D-erythritol 4-phosphate cytidylyltransferase; CMK, 4-(cytidine 5'-diphospho)-2-C-methyl-D-erythritol kinase; MDS, 2-C-methyl-D-erythritol 2,4-cyclodiphosphate synthase; HDS, 4-hydroxy-3-methylbut-2-enyl-diphosphate synthase; HDR, 4-hydroxy-3-methylbut-2-enyl diphosphate reductase; IPPI, isopentenyl diphosphate isomerase; AACT, acetyl-CoA acetyltransferase; HMGS, hydroxymethylglutaryl-CoA synthase; HMGR, 3-hydroxy-3-methylglutaryl-CoA reductase; MK, mevalonate kinase; PMK, phosphomevalonate kinase; MPDC, MVA diphosphate decarboxylase; SmCPS1, copalyl diphosphate synthase 1; and TwTS, taxadiene synthase.

Methods

Plant material and growing conditions

Seedlings were grown on MS (Murashige and Skoog) agar medium in a plant culture room, and S. miltiorrhiza and N. benthamiana plants were cultured in a greenhouse at 25°C with a 16-h light/8-h dark photoperiod. T. wallichiana plants were cultured outdoors from spring to summer.

Strains and vector construction

The TwTS gene was cloned from cDNA of T. wallichiana using the TS-F/R primer pair (Supplemental Table 3) and inserted into the Gateway vector pGP-B2E. This recombinant vector was transformed into E. coli DH5α (Weidi) for plasmid extraction. After verification, the pGP-B2E vector was subjected to an LR reaction with the overexpression vector pK7WG2R, resulting in the final TwTS gene overexpression plasmid.

To construct the SmCPS1 gene-editing plasmid, the sgRNA for the SmCPS1 gene was obtained from publicly available sources (Li et al., 2017). The target sgRNA was then introduced into the pICSL002218A gene-editing vector using the BsaI restriction endonuclease (NEB), resulting in the final SmCPS1 gene-editing plasmid.

The Golden Gate method was used to construct the SmWRKY61 gene overexpression plasmid (Weber et al., 2011). Specifically, the pUBQ10 promoter was cloned from the A. thaliana genome, the coding region of SmWRKY61 was amplified from S. miltiorrhiza cDNA, the tNOS terminator was amplified from the pK7WG2R vector, and then the pUBQ10 promoter, SmWRKY61 gene, and tNOS terminator were combined into a single expression cassette in the pAMGM4723 vector to generate the pUBQ10:SmWRKY61:tNOS cassette. The Hyg resistance gene was synthesized commercially (Tsingke). Similarly, the pNOS promoter, Hyg resistance gene, and tNOS terminator were assembled into a separate cassette (pNOS:Hyg:tNOS cassette). Finally, two level 1 vectors were assembled in the level 2 vector pAGM4723, resulting in the SmWRKY61 overexpression plasmid.

Hairy root induction and formation

The relevant plasmid was introduced into ATCC15834 using the electroporation method and plated on LB solid medium containing the appropriate antibiotics. Single colonies were then selected for molecular characterization of the target gene (Supplemental Table 3). Positive colonies were cultured until the optical density at 600 nm (OD₆₀₀) reached 0.5. The bacterial culture was centrifuged, and the pellet was resuspended in 1/2 MS liquid medium with 50 μM acetosyringone, then incubated in the dark for 1 h. Healthy S. miltiorrhiza leaves were cut with a razor blade to facilitate the introduction of the bacterial solution into the abaxial surface. The treated leaves were then placed on antibiotic-free 1/2 MS solid medium for co-cultivation. After 3 days, the antibiotic concentration was gradually reduced until Agrobacterium was completely eliminated. Finally, recombinant hairy root lines were selected for further positive screening.

Fluorescence visualization analysis

Recombinant hairy root monocultures containing the TwTS overexpression construct were screened for fluorescence using red fluorescent protein visualization. The vector includes the DsRed gene, which allows detection of fluorescence with the LUYOR-3415RG dual fluorescent protein viewing lamp (LUYOR). Phenotype pictures of the lines were obtained by incubating the positive hairy root monocultures in Petri dishes for a certain period of time and taking photographs (these were mainly used for phenotypic evaluation, such as determining the presence or absence of red fluorescence, the degree of albinism of the hairy roots, etc.).

Determination of hairy root biomass of S. miltiorrhiza

The fresh hairy roots were weighed, and 0.2 g of roots was placed in triangular flasks containing 50 ml of 6,7-V liquid medium, and the roots were incubated for 30 days under dark conditions at 25°C on a 110-rpm shaker.

Generation and genotyping of CRISPR–Cas9 mutants

Recombinant hairy roots containing the SmCPS1 gene-editing construct were screened by phenotyping and sequencing. Initially, monocultures that displayed a white phenotype were selected. Genomic DNA from these recombinant hairy roots was amplified using CPSyz-F/R primers (Supplemental Table 3), ligated into a TOPO vector, and transformed into E. coli. Single colonies were picked for sequencing. Finally, allelic mutations in the different monoculture lines were analyzed on the basis of the sequencing results.

UPLC–MS/MS analysis

To quantify GGPP levels in transgenic hairy roots, an appropriate amount of hairy root monoculture was precisely weighed, frozen in liquid nitrogen, and ground. To the ground tissue, 1 ml of 80% (v/v) acetonitrile was added, and the mixture was subjected to low-temperature ultrasonication for 10 min. It was then centrifuged at 4°C for 10 min at 13,400 g. This process was repeated three times. The extracts were combined and evaporated using a concentrator. To the residue, 200 μl of 80% (v/v) acetonitrile was added, and the mixture was extracted with low-temperature ultrasonication for 10 min, followed by centrifugation at 4°C and 13 400 g for 5 min. The supernatant was filtered through a 0.22-μm membrane and transferred to an injection bottle.

The liquid-phase conditions were as follows. The column was a Waters ACQUITY UPLC BEH C18 (1.7 μm, 2.1 mm × 100 mm). The mobile phases consisted of ultra-pure water with 0.1% ammonia (phase A) and acetonitrile with 0.1% ammonia (phase B). The flow rate was set to 0.35 ml/min, and the column temperature was maintained at 40°C. The injection volume was 5 μl, and elution was performed using gradient mode (0–1 min, 80% A; 1–3 min, 55% A; 3–5 min, 10% A; and 5–7.5 min, 80% A).

The MS conditions were as follows. Electrospray ionization (ESI) was performed at a temperature of 550°C with an MS voltage of 4,500 V. The ion source gas I was set to 50 psi, gas II to 60 psi, and curtain gas to 35 psi. The collision-activated dissociation parameter was set to 8. Each ion pair was scanned using a triple quadrupole mass spectrometer, with optimized declustering potential and collision energy. The GGPP reference standard was obtained from Yuanye.

GC–MS analysis

To quantify taxadiene levels in transgenic hairy roots, the hairy roots were weighed accurately. One portion was dried in an oven at 45°C for 3 days and then weighed again. The remaining portion was frozen in liquid nitrogen and ground. To this sample, 5 ml of hexane was added, and the mixture was subjected to ultrasonication for 20 min. It was then centrifuged at 4°C for 10 min at 13 400 g; this process was repeated five times. The extracts were combined and concentrated using a concentrator. A hexane solution containing 200 ng of nonadecane standard was added to the tube with the dried sample (standard from Yuanye). The solution was filtered through a 0.22-μm membrane and then transferred to an injection bottle using a syringe. GC was performed using an HP-5MS column (30 m × 250 μm × 0.25 μm). For MS, helium was used as the carrier gas with a flow rate of 1.0 ml/min. The temperature program involved an initial hold at 80°C for 2 min, followed by a temperature ramp to 200°C at a rate of 10°C/min and a final hold at 200°C for 10 min. The injector and transfer line temperatures were set to 200°C and 250°C, respectively. The column effluent was ionized using electron ionization at 70 eV, and mass spectra were obtained in the range of m/z 30–550. The retention time and relative content of the target compound, taxadiene, were determined by comparing the characteristic ions of paclitaxel (m/z 107, 122, and 272) and the peak areas of the standards.

For quantitative analysis of miltiradiene in transgenic hairy roots, samples were accurately weighed and extracted with 5 ml of n-hexane under ultrasonication for 20 min. The mixture was subsequently centrifuged at 13,400 g for 10 min at 4°C, and this procedure was repeated five times. The resulting supernatants were pooled, concentrated under a stream of nitrogen, and filtered through a 0.22-μm membrane before being transferred to a sample vial using a syringe. GC–MS analysis was performed on a Trace1310 series gas chromatograph interfaced with a TSQ8000 mass spectrometer (Thermo Fisher Scientific, Waltham, MA, USA).

Separation was achieved using a TR-5 ms capillary column (30 m × 0.25 mm i.d., 0.25 μm film thickness; Thermo Fisher Scientific) with helium as the carrier gas at a constant flow rate of 1 ml/min. The injector temperature was maintained at 280°C. The oven temperature began at 50°C (held for 1 min), ramped to 150°C at 5°C/min, to 230°C at 20°C/min, and to 300°C at 30°C/min, and finally was held at 300°C for 5 min. A miltiradiene standard was generously provided by Professor Juan Guo (China Academy of Chinese Medical Sciences).

Quantitative fluorescence analysis

RNA was extracted from hairy root lines using a commercially available RNA extraction kit (Vazyme), followed by cDNA synthesis with a reverse transcription kit (AG). The cDNA was then analyzed for gene expression using a fluorescence quantitative PCR instrument (QuantStudio 6 Flex, Thermo Fisher). RT–qPCR was performed using the comparative threshold cycle (2^−ΔΔCt) method, with β-actin as the internal reference gene. Primers used in this study are detailed in Supplemental Table 3. Each sample was analyzed in three independent biological replicates.

HPLC analysis

To quantify tanshinone levels in transgenic hairy roots, each sample was initially weighed and dried in an oven at 45°C for 3 days to remove moisture. After drying, the sample was reweighed and transferred to a grinding tube containing grinding beads for pulverization (20 s per cycle, repeated for 5 cycles). The ground dry sample was reweighed and transferred to an EP tube; 1 ml of 70% methanol was added, and the mixture was sonicated for 1 h, with inversion and mixing every 20 min. Following sonication, the mixture was centrifuged at 13,400 g for 10 min. The supernatant was collected with a syringe, filtered through a 0.22-μm membrane, and transferred to an injection bottle.

The HPLC analysis was performed using a Waters SunFire C18 column. The mobile phases consisted of 0.02% phosphoric acid in water (A) and acetonitrile (B), with a flow rate of 1 ml/min and a sampling volume of 10 μl. The column temperature was set to 30°C, and separation was achieved using gradient elution mode (0–10 min, 5%–20% B; 10–15 min, 20%–25% B; 15–20 min, 25% B; 20–25 min, 20%–30% B; 25–28 min, 30% B; 28–40 min, 30% B; 40–45 min, 45%–58% B; 45–58 min, 58%–60% B; 58–67 min, 58%–50% B; 67–70 min, 50%–60% B; 70–80 min, 60%–65% B; 80–85 min, 65%–95% B; 85–95 min, 95% B; and 95–96 min, 5% B). The contents and retention times of tanshinones—dihydrotanshinone I, cryptotanshinone, tanshinone I, and tanshinone IIA—in the hairy root monocultures were quantified and analyzed at 270 nm by comparison with a standard curve (standards from Yuanye).

Seedling generation from hairy roots and pot planting

The callus and adventitious shoot induction medium consisted of MS medium supplemented with 1 mg/l 6-benzyladenine, 0.1 mg/l naphthalene acetic acid, 30 g/l sucrose, and 8 g/l agar, adjusted to pH 5.8. Healthy hairy root segments of appropriate length were placed in the induction medium and cultured in the dark at 25°C for 30 days until callus formation. The calli were then transferred to a light/dark cycle of 16 h/8 h at 25°C for 7 days to induce the formation of green adventitious shoots. Several adventitious shoots were excised with a scalpel and transferred to rooting medium composed of 1/2 MS, 0.2 mg/l indolebutyric acid, 30 g/l sucrose, and 8 g/l agar (pH 5.8), and cultured at 25°C with a 16-h light/8-h dark cycle for 15 days. Seedlings were obtained when the root length reached 2–3 cm. Finally, fully expanded leaves of the recombinant histocultured seedlings were collected for molecular identification of the TwTS gene and the DsRed fluorescent protein gene.

Deionized water was added to the verified recombinant seedlings, and the caps were sealed tightly. The seedlings were incubated at 25°C under a 16-h light/8-h dark cycle for 3 days. The caps were then partially lifted, and incubation continued under the same conditions for an additional 3 days. After this period, the seedlings were removed from the MS medium, and any residual medium was washed from the hairy roots. The seedlings were transferred to soil (soil:perlite:vermiculite = 7:1:2) and covered with a moisture-retaining film for 10 days. Finally, the film was removed, and the seedlings were planted in the greenhouse under the same conditions.

Hairy root elicitor treatment

The recombinant hairy root line was cultured in 6,7-V liquid medium. After 18 days of growth, 200 μM MeJA or 25 mg/l YE was used for elicitation. Samples were collected at 12, 24, 72, and 144 h post elicitation, with three biological replicates per sampling point.

LC–MS–IT–TOF

To quantify protopanaxadiol levels in transgenic hairy roots, a precise amount of each hairy root line was weighed, frozen, and ground in liquid nitrogen. One milliliter of 70% methanol was added, and the mixture was ultrasonicated for 1 h, with inversion and mixing every 20 min. After ultrasonication, the sample was centrifuged at 13,400 g for 10 min. The supernatant was then aspirated using a syringe, filtered through a 0.22-μm membrane, and transferred to an injection vial for sampling.

Liquid chromatography (LC) was performed using a Waters ACQUITY UPLC BEH C18 column (1.7 μm, 2.1 mm × 100 mm). The mobile phases consisted of acetonitrile (phase A) and ultrapure water (phase B). The flow rate was set to 0.35 ml/min, and the column temperature was maintained at 40°C. The injection volume was 5 μl, and a gradient elution program was used: 0–5 min, 25%–55% A; 5–10 min, 55%–70% A; 10–13 min, 70%–90% A; 13–20 min, 90%–100% A; 20–25 min, 100%–75% A; and 25–27 min, 75% A.

The MS detector voltage was 1.60 kV, the nebulizing gas (N) flow rate was 1.5 l/min, and the drying gas (N₂) flow rate was 50 kPa. The time-of-flight (TOF) zone force was set to 1.5 × 10⁻⁹ Pa, and the sub-trap pressure was 1.7 × 10² Pa. The ion build-up time was 30 ms, and the precursor ion selection width was 3.0 amu. The scan range was 100–1000 for MS1 and 100–500 for MS2. Ultra-high-purity gas was used for cooling and collision. The protopanaxadiol reference standard was obtained from Yuanye.

Statistical analysis

We performed pairwise comparisons of each transgenic line and its corresponding control and assessed statistical significance using a two-tailed unpaired Student’s t test performed with GraphPad Prism 8 software; a p value less than 0.05 was considered statistically significant.

Data and code availability

Gene sequences used in this study were obtained from GenBank (https://www.ncbi.nlm.nih.gov/): TwTS (AY007207), SmCPS1 (EU003997), SmWRKY61 (KM823184), and PnPPDS (GU997665). Source data are also provided with this paper.

Acknowledgments

This work was supported by the National Key Research and Development Program of China (2023YFA0915800 and 2024YFC3506400), the Zhejiang Provincial Natural Science Foundation of China (LMS25C020001 and LD25H280001), the Key Project of the Central Government: Capacity Building of Sustainable Utilisation of Traditional Chinese Medicine Resources (2060302), the Scientific and Technological Innovation Project of CACMS (CI2023D002), and the National Natural Science Foundation of China (82574523).

Author contributions

J.L., W.L., Y.Z., and Y. Liu performed the biological experiments; L.L. and F.Y. performed sample collection and chromatographic processing; W.L., Z.X., C.E.F., J.G., Z.L., B.C., and D.Y. supervised and assisted the experiments; J.L., B.C., and D.Y. wrote the manuscript. J.L. and W.L. contributed equally to this study.

Published: March 19, 2026