Abstract

Cardiovascular and cerebrovascular diseases are among the leading causes of death worldwide, and current drugs are limited by side effects and adherence issues. Leonurine, with antioxidant, cardioprotective, and vascular repair properties, shows strong therapeutic promise, but its trace abundance in natural sources makes extraction unsuitable for industrial use. We are the first team to synthesize leonurine using synthetic biology, demonstrating clear innovation. In Cycle 1, we established a hairy root–based production system in Nicotiana tabacum, achieving 3.687 μg /g DW (≈1.47 μg per 200 mL batch)—a 17-fold increase compared with previous yeast-based production. This step not only validated the activities of LjUGT5 and LjSCPL12 but also explored whether a plant chassis could offer industrial advantages. Building on this, and guided by Human Practices advice from Professor Cai, we pursued Cycle 2, switching to E. coli as the chassis, purifying the enzymes, and performing in-vitro cascade catalysis. This yielded 8.65 μg per 200 mL batch, representing a further 5.9-fold increase over hairy roots. Together, these results highlight that combining chassis selection with process strategy innovation provides a viable route for scalable and cost-effective leonurine production. In the future, our product is envisioned as an active pharmaceutical ingredient (API) supplied to pharmaceutical companies for cardiovascular protection and adjunctive therapy, enabling broader public health impact.

Design-Build-Test-Learn (DBTL) cycle

This year, to achieve efficient production of leonurine, NAIS 2025 completed the following two DBTL cycles (Figure 1).

Figure 1. DBTL cycle of our project

Cycle 1

Production of leonurine using hairy root cultures

Design 1

1.1 Introduction

Cardiovascular and cerebrovascular diseases have become the leading cause of death worldwide, with both incidence and mortality steadily increasing [1,2]. These diseases seriously impair quality of life and place heavy burdens on healthcare systems. Developing new therapeutics is therefore an urgent global priority.

Leonurine, a major bioactive compound in Leonurus japonicus (yi mu cao), has shown great pharmacological potential in cardiovascular protection, including anti-oxidative, cardioprotective, and vascular repair functions [3,4]. However, its natural content in Leonurus japonicus is extremely low, making direct extraction inefficient and unsuitable for industrial application (Figure 2).

Recent multi-omics studies have preliminarily elucidated the biosynthetic pathway of leonurine [5]. Professor Ping Xu’s team identified two key enzymes responsible for the final biosynthetic steps—LjUGT5 and LjSCPL12—and validated their catalytic activities in tobacco leaves and yeast cells. Nevertheless, the observed yields were very low (approximately 140 ng/mL in yeast and even less in tobacco leaves), far from the threshold required for industrial production [5]. This highlights the significant challenges in heterologous synthesis of leonurine.

Against this backdrop, we aimed to explore synthetic biology strategies to enable efficient and controllable leonurine production. Our goal is to provide a sustainable route toward leonurine-based therapeutics, contributing to the treatment of cardiovascular diseases.

Figure 2. The photograph of Leonurus japonicus Houtt. and the structure of leonurine [3]

1.2 The chassis: Nicotiana tabacum

In designing our system, we adopted a “validate first, then transform” logic. For the first validation stage, we selected Nicotiana tabacum hairy root cultures as our plant chassis.

This system offers several advantages:

​ a) It is well-established for genetic manipulation and widely used as a plant biotechnology model.

​ b) Hairy root cultures can effectively simulate the metabolic environment of plants, making them ideal for testing plant-derived biosynthetic pathways.

​ c) They provide rapid growth and stable propagation, which are favorable for repeated experimentation.

By employing N. tabacum hairy roots, we established a reliable platform to validate the catalytic functions of LjUGT5 and LjSCPL12 and to monitor leonurine production under conditions closely resembling native plant metabolism.

1.3 The enzymes: LjUGT5 and LjSCPL12

The biosynthetic pathway of leonurine involves several upstream steps, but its final two reactions are catalyzed by UDP-glycosyltransferase (UGT) and serine carboxypeptidase-like acyltransferase (SCPL) (Figure 3).

​ ✓ LjUGT5 activates syringic acid (SA) by glycosylation to produce SA-Glc.

​ ✓ LjSCPL12 catalyzes the condensation of SA-Glc with 4-guanidinobutanol to form leonurine.

Figure 3. Construction of the Leonurine Biosynthetic Pathway

By focusing on these two enzymes, we could bypass the complex upstream metabolic processes and directly reconstruct the key bottleneck steps. To further improve efficiency, we supplied the inexpensive substrates syringic acid (≈¥300/kg) and 4-guanidinobutanol (≈¥500/kg) exogenously, thereby reducing metabolic burden on the chassis and offering cost advantages for potential industrial scale-up.

Thus, LjUGT5 and LjSCPL12 were chosen as the core enzymes of our synthetic design, representing the minimal but sufficient pathway to achieve controllable leonurine synthesis.

1.4 Plasmid selection: pAGM4723

We selected pAGM4723 as our plant expression vector (Figure 4). This plasmid is part of the MoClo modular cloning system, optimized for Golden Gate Assembly, and allows efficient assembly of promoters, CDSs, and terminators.

Reasons for selection:

​ a) MoClo compatibility, enabling rapid modular construction.

​ b) Multi-gene assembly capacity, suitable for reconstructing the UGT-SCPL pathway.

​ c) Agrobacterium-mediated hairy root transformation, ensuring stable expression in Nicotiana tabacum.

​ d) Contains the chloramphenicol resistance gene (cat) for selection.

Thus, pAGM4723 provided an ideal platform for validating the functions of LjUGT5 and LjSCPL12 in the hairy root system.

Figure 4. Plasmid map of pAGM4723

1.5 Plasmid Construct Designs

Building upon the key enzymes UGT5 and SCPL12 for leonurine biosynthesis, we introduced GFP as a reporter and XVE as an inducible regulator. GFP facilitates the screening of positive hairy root transformants, providing a visual marker to confirm successful integration. Meanwhile, XVE, an artificial estrogen-responsive system, can be activated by β-estradiol. Once activated, XVE binds to the PLEXA35S promoter and recruits the transcriptional machinery, thereby driving the expression of UGT5 and SCPL12 at specific time points. This design not only reduces the metabolic burden associated with constitutive expression but also provides precise temporal control over the biosynthetic pathway.

Figure 5. Control of Protein Expression

Our plasmid design followed a stepwise assembly strategy: at Level 0, basic modules such as promoters, CDSs, and terminators were first standardized; at Level 1, these modules were combined into complete expression cassettes; and finally, at Level 2, all cassettes were integrated into one construct (Figure 6). Leveraging the pAGM4723 MoClo system, this hierarchical approach ensures high efficiency, directional fidelity, and robust success rates, making it ideal for assembling multi-gene pathways. The final Level 2 plasmid thus supports the stable expression of GFP, XVE, and UGT5/SCPL12 in hairy roots, enabling a complete workflow from selection and regulation to leonurine biosynthesis.

Figure 6. Schematic diagram of plant expression vector construction

Build 1

2.1 Recombinant vector construction – level 1

Level 0 consists of standardized basic modules such as promoters, CDSs, and terminators, with some obtained from plasmids preserved in our laboratory and others chemically synthesized by GENEWIZ (Suzhou, Jiangsu, China). The expression of the GFP gene is driven by the constitutive promoter PAtUBI5 and terminated by Tmas. Both LjUGT5 and LjSCPL12 are controlled by the inducible promoter PLEXA35S, with Thsp18.2, Tnos, and T35S serving as commonly used terminators in plant expression systems. P35S is a strong constitutive promoter that drives the overexpression of XVE, thereby enabling regulation of the target genes. The Golden Gate Assembly system and procedure for constructing Level 1 vectors from these basic modules are illustrated in Figure 7.

Figure 7. System and Procedure of Level 1 Vector Construction

After assembly, the recombinant plasmids were transformed into E. coli competent cells DH5α. Following overnight incubation, positive colonies of GFP, LjUGT5, LjSCPL12, and XVE were obtained on antibiotic-containing (Amp+) plates (Figure 8).

Figure 8. Construction of Level 1 Vectors for Four Genes (Note: From left to right are GFP, LjUGT5, LjSCPL12, and XVE, respectively.)

Subsequently, colony PCR verification was performed on positive clones of LjUGT5 and LjSCPL12. The results showed that all eight randomly selected clones produced bands consistent with the positive control, while no bands were observed in the negative control (Figure 9). These findings indicate that the Level 1 vectors were successfully constructed with high assembly efficiency, providing a solid foundation for the subsequent multi-gene assembly at Level 2.

Figure 9. Colony PCR Identification of the LjUGT5 (A) and LjSCPL12 (B) Level 1 Vectors (Note: PC, positive control; NC, negative control; 1-8, eight randomly selected single colonies)

2.2 Recombinant vector construction – level 2

After constructing the Level 1 vectors, we needed to combine the expression cassettes of the four genes into a single construct to obtain the final multi-gene expression vector, Level 2, which enables simultaneous pathway co-expression in tobacco hairy roots. The Golden Gate Assembly system and procedure for constructing Level 2 vectors are illustrated in Figure 10.

Figure 10. System and Procedure of Level 2 Vector Construction

The recombinant plasmids were transformed into E. coli DH5α and plated on LB agar containing kanamycin, followed by overnight incubation at 37 °C. Since the vector carries the chromogenic marker gene crtW aa, non-recombinant colonies appeared yellow, while successful insertion of the target fragment replaced this marker, resulting in white colonies. Thus, white colonies indicate positive recombinants, whereas yellow colonies represent empty vectors. Six parallel replicates were performed, and the results are shown in Figure 11.

Figure 11. Construction of the plant expression vector (Level 2), six repeats.

We picked 96 white single colonies for colony PCR. First, we verified the GFP gene, obtaining 74 positive clones, which served as candidates for the next step of LjSCPL12 identification (Figure 12).

Figure 12. Colony PCR verification of the GFP gene in the Level 2 vector

Based on GFP-positive clones, PCR amplification of LjSCPL12 identified 46 positive clones, which were then used as candidates for LjUGT5 identification (Figure 13).

Figure 13. Colony PCR verification of the LjSCPL12 gene in the Level 2 vector

Based on GFP and LjSCPL12 identification, PCR amplification of LjUGT5 yielded 40 positive clones, which were used as candidates for XVE identification (Figure 14).

Figure 14. Colony PCR verification of the LjUGT5 gene in the Level 2 vector

Finally, after confirming GFP, LjSCPL12, and LjUGT5, PCR amplification of XVE identified 36 positive clones, from which 3 clones were selected for sequencing (Figure 15).

Figure 15. Colony PCR verification of the XVE gene in the Level 2 vector

2.3 Transgenic tobacco hairy root construction

After confirming the sequence accuracy of the Level 2 plasmid, we electroporated it into the hairy root Agrobacterium rhizogenes strain Ar.1193, and plated the cells on LB agar supplemented with rifampicin (200 mg/L), kanamycin (50 mg/L), and streptomycin (100 mg/L). The plates were incubated upside down at 28 °C for 2 days. Positive colonies were identified, expanded, and subsequently used to infect tobacco leaves.

Figure 16. Process of Induction and Culture of Tobacco Hairy Roots (Note: A, 0 days after Agrobacterium infection; B, Hairy root emergence at 10 days after Agrobacterium infection; C, Individual culture of hairy roots on solid medium; D, Transfer of hairy roots to liquid culture with the addition of an inducer and substrates.)

As shown in Figure 16, hairy roots began to emerge around 10 days after infection. Once the roots grew to 1–2 cm, GFP expression was examined, and GFP-positive roots were selected (Figure 17). Genomic DNA was then extracted from these positive roots for further verification.

Figure 17. Identification of transgenic positive hairy roots based on fluorescent signals

We selected three GFP-positive hairy roots and extracted their genomic DNA for genotypic analysis. The results confirmed that the GFP, LjUGT5, LjSCPL12, and XVE genes were successfully integrated into the tobacco hairy roots, indicating a high transformation efficiency (Figure 18).

Figure 18. PCR identification of transgenic positive hairy roots (Note: A, B, C, D panels represent the GFP, LjUGT5, LjSCPL12, and XVE genes, respectively. PC, positive control; EV, empty vector; lanes 1, 2, 3 represent three independent transgenic hairy root lines.)

Test 1

3.1 Expression Analysis of LjUGT5 and LjSCPL12 Genes

After obtaining PCR-positive hairy roots, they were transferred into liquid medium and induced with β-estradiol (10 μM) for 2 days, followed by supplementation with 4-guanidinobutanol, syringic acid, and UDPG (0.5 mM). The addition of UDPG was essential because the formation of SA-Glc requires UDP-glucose as a glycosyl donor, catalyzed by LjUGT5, which promotes efficient glycosylation of syringic acid and ensures smooth downstream synthesis of leonurine. After 3 more days, total RNA was extracted for qRT-PCR analysis.

As shown in Figure 19, both genes exhibited near-background expression in the control groups (CK1–3). In contrast, in the β-estradiol–treated samples (Sample 1–3), LjUGT5 expression was significantly upregulated (approximately 10-fold), while LjSCPL12 expression also increased markedly (about 4–6 fold). These results demonstrate that the constructed multi-gene expression vector was successfully integrated into the hairy root system and that the target genes could be robustly induced by β-estradiol, confirming the feasibility of our design.

Figure 19. Expression level of LjUGT5 (A) and LjSCPL12 (B) in transgenic hairy roots (Note: Samples 1-3 were treated with β-Estradiol to a final concentration of 10 μM; CK1-3 were treated with an equal volume of DMSO.)

3.2 Analysis of Leonurine Yield in Tobacco Hairy Roots

Leonurine production was induced under the same conditions as described above. After 3 days of substrate supplementation, hairy roots were harvested, freeze-dried for 2 days, and ground into powder. Metabolites were extracted with 70% methanol and analyzed by LC-MS (AB SCIEX 6500+) for leonurine quantification.

In metabolite analysis, we first performed qualitative detection by LC-MS retention time. A distinct peak was observed in the transgenic samples at the same retention time as the standard, while no peak was detected in the controls, indicating successful biosynthesis of leonurine in the transgenic hairy roots (Figure 20A). We then used a standard curve to correlate peak area with concentration, enabling quantitative analysis (Figure 20B). The results showed that all three transgenic samples contained stable levels of leonurine, with an average content of 3.687 μg/g DW, while no detectable signal (ND) was observed in the controls (Figure 20C). Based on biomass yield, approximately 0.4 g of freeze-dried hairy root powder was obtained from a 200 mL culture, corresponding to about 2.0 g DW per liter. With an average leonurine content of 3.687 μg/g DW, this translates into:

​ ➢ a specific yield of 3.687 μg/g DW,

​ ➢ a volumetric titer of approximately 7.37 μg/L, and

​ ➢ a batch yield of about 1.47 μg per 200 mL culture.

These values represent apparent yields, as no correction was applied for extraction recovery during metabolite analysis.

Figure 20. Analysis of leonurine production yield in transgenic hairy roots (Note: Samples 1-3 were supplemented with β-estradiol to a final concentration of 10 μM; control groups CK1-3 were supplemented with an equal volume of DMSO.)

Together, these results demonstrate that our multi-gene expression construct not only enabled effective inducible expression of the target genes in hairy roots but also successfully drove the stable biosynthesis of leonurine, thereby validating the functional effectiveness of the reconstructed pathway.

Learn 1

Through Cycle 1, we successfully validated the functions of LjUGT5 and LjSCPL12 in the tobacco hairy root system and detected stable biosynthesis of leonurine under substrate supplementation and β-estradiol induction. This confirmed the feasibility of our designed pathway in a plant chassis. However, the yield was only 3.687 μg /g DW (≈7.37 μg /L), which remains far below the threshold required for industrial application (≥60 mg/200 mL). The low yield may be attributed to limited overall metabolic flux in hairy roots, inefficient substrate transport, or restricted enzyme activity in the plant environment [6,7,8]. These findings suggest that while hairy roots provide a reliable platform for functional validation, a more efficient microbial chassis (e.g., E. coli) should be employed in Cycle 2 to further enhance production

Cycle 2

Production of leonurine using E. coli

Design 2

1.1 Introduction

In Cycle 1, we validated the feasibility of leonurine biosynthesis in the hairy root system and obtained an average yield of 3.687 μg/g DW (≈7.37 μg /L). Although this level was approximately 17 times higher than that reported in yeast (140 ng/mL, equivalent to 84 ng per 200 mL batch), it still fell far short of the requirements for industrial application [5]. During our Human Practices (HP), Professor Cai suggested an alternative strategy: to use E. coli for efficient expression of the key enzymes, followed by in vitro enzymatic catalysis to synthesize leonurine. This approach bypasses complex intracellular metabolic networks, enabling scalable, controllable, and standardized production, which makes it more suitable for industrial-scale applications. Therefore, in Cycle 2, we aimed to employ E. coli as the chassis for high-level enzyme expression and establish an in vitro catalytic system for leonurine synthesis.

1.2 The chassis: E. coli BL21

We selected E. coli BL21(DE3) as the chassis strain for expressing LjUGT5 and LjSCPL12. This strain is a classical host for recombinant protein production and offers several advantages:

a) It lacks Lon and OmpT proteases, reducing degradation of heterologous proteins and increasing yield.

b) It carries the T7 RNA polymerase gene, enabling strong protein expression under T7 promoters.

c) It is easy to culture with short growth cycles and low costs, making it highly suitable for large-scale fermentation.

d) It has been widely used in both laboratory and industrial protein production, providing robust scalability and process feasibility.

Thus, BL21(DE3) was chosen as the ideal chassis for Cycle 2 to achieve efficient heterologous protein expression and support subsequent in vitro enzymatic catalysis of leonurine.

1.3 Plasmid selection

For plasmid selection, we employed different expression systems for the two key enzymes. LjUGT5 was expressed using the pRSFDuet-sumo vector, while LjSCPL12 was expressed using the pETDuet-MBP vector. Both are part of the Duet series, designed to coexist stably in the same host cell and avoid plasmid incompatibility.

​ ✓ pRSFDuet-sumo: Equipped with a strong T7 promoter and a SUMO tag, enhancing protein solubility and stability for downstream purification.

​ ✓ pETDuet-MBP: Contains a T7 promoter and an MBP tag, significantly improving protein solubility and folding efficiency, while facilitating affinity purification.

This dual-plasmid system not only enables efficient co-expression of multiple proteins but also improves protein stability and purification efficiency through tag design.

1.4 Plasmid Construct Designs

For plasmid construction, we employed a double-digestion strategy combined with Gibson Assembly-based homologous recombination. Specifically, the vectors were first linearized by double digestion, effectively reducing empty-vector background caused by self-ligation. Subsequently, Gibson Assembly enabled seamless recombination of the target gene fragments (with homologous arms) into the digested backbones, yielding complete recombinant plasmids. This method offers high efficiency, directional fidelity, and robust assembly success rates, well-suited for rapid multi-gene construction. Ultimately, we generated pRSFDuet-sumo carrying LjUGT5 and pETDuet-MBP carrying LjSCPL12, enabling efficient expression in BL21(DE3) and providing a reliable source of enzymes for subsequent in vitro catalysis of leonurine (Figure 21).

Figure 21. Plasmid map of pETDuet-MBP-LjSCPL12 (A) and pRSFDuet-sumo-LjUGT5 (B)

Build 2

Using the successfully constructed plasmids from Cycle 1 as templates, PCR amplification of LjUGT5 (1392bp) and LjSCPL12 (1347bp) was performed. Clear single bands appeared between 1 kb and 2 kb, consistent with the expected sizes, indicating successful amplification of the target fragments (Figure 22).

Figure 22. PCR amplification of the LjUGT5 and LjSCPL12 genes.

The recombinant plasmids were transformed into E. coli DH5α competent cells and plated on LB agar supplemented with the corresponding antibiotics, followed by overnight incubation at 37 °C. Numerous single colonies were observed on both plates, corresponding to pRSFDuet-sumo-LjUGT5 (left) and pETDuet-MBP-LjSCPL12 (right), indicating successful plasmid uptake and stable replication in the host (Figure 23). Subsequently, selected colonies were subjected to PCR and sequencing to confirm the accuracy of the inserted fragments (Figure 24).

Figure 23. Plate containing single colonies after overnight culture (Left: pRSFDuet-sumo-LjUGT5; Right: pETDuet-MBP-LjSCPL12.

Figure 24. Colony PCR verification of the construction of the pRSFDuet-sumo-LjUGT5 (A) and pETDuet-MBP-LjSCPL12 (B) vectors. (Note: M, DNA marker; PC, positive control; NC, negative control; lanes 1-8, eight randomly selected single colonies.)

Test 2

We further validated the expression of LjUGT5 and LjSCPL12 in E. coli BL21(DE3). Sequenced-confirmed plasmids were transformed into the expression strain, and positive clones were identified by colony PCR. After IPTG induction, the bacterial cells were harvested, lysed, and subjected to Ni-NTA affinity purification. The principle of Ni-NTA purification relies on the specific binding of His-tags to nickel ions, allowing efficient capture and elution of the target proteins.

SDS-PAGE analysis of the purified proteins (Figure 25A) revealed distinct bands at approximately 65.1 kDa and 96.8 kDa, corresponding to LjUGT5 and LjSCPL12, respectively. Further confirmation was obtained by Western blotting (Figure 25B), where anti-His antibodies specifically recognized the His-tagged recombinant proteins.

Together, these results demonstrate that we successfully achieved heterologous high-level expression of LjUGT5 and LjSCPL12 in E. coli, obtaining purified enzymes suitable for subsequent in vitro catalytic assays.

Figure 25. Detection of target protein expression. (A) Electrophoretic analysis of purified LjUGT5 and LjSCPL12; (B) Western blotting identification of LjUGT5 and LjSCPL12. (Note: The molecular weights of the recombinant proteins are 65.1 kDa and 96.8 kDa, respectively.)

After obtaining purified LjUGT5 and LjSCPL12 proteins, we established an in vitro enzymatic system to test leonurine biosynthesis. The reaction mixture had a total volume of 100 μL (Figure 26), consisting of 10 μL each of LjUGT5 and LjSCPL12 proteins, 10 μL of substrates 4-guanidinobutanol (100 mM), syringic acid (50 mM), and UDPG (50 mM), 10 μL MgCl₂ (50 mM), and 40 μL PB buffer (pH 8.0). From each 200 mL bacterial culture, approximately 600 μL of purified protein solution was obtained, of which 10 μL was used per reaction. The reaction was incubated at 28 °C for 16 h, allowing sufficient time for substrate conversion and leonurine synthesis.

Figure 26. In vitro enzyme activity assay system

We first performed qualitative analysis by LC-MS retention time: a leonurine peak was detected only when LjUGT5 + LjSCPL12 were both present, while “no enzyme”, “UGT5 only”, and “SCPL12 only” showed no signal (Figure 27A), confirming that both enzymes are required. Quantification using the standard curve (Figure 27B) gave a concentration of 1.442 μg/mL in the 100 μL reaction (Figure 27C), corresponding to 0.144 μg per reaction by volume. From each 200 mL bacterial culture, ~600 μL of purified enzyme solution was obtained; using the same mixing ratio (10 μL enzyme per 100 μL reaction), this allows a 6 mL total reaction volume, yielding 1.442 μg/mL × 6 mL = 8.652 μg per batch. This is an apparent yield (not corrected for extraction recovery). Compared with the hairy-root system (1.47 μg per batch), the in vitro enzymatic route achieves ~5.9-fold higher batch yield, indicating better scalability and process controllability.

Figure 27. Yield of leonurine produced by the in vitro enzyme activity assay.

Learn 2

In Cycle 2, we successfully achieved high-level expression and purification of LjUGT5 and LjSCPL12 in E. coli BL21(DE3), and realized leonurine biosynthesis in an in vitro catalytic system. Compared with the hairy root system in Cycle 1, the batch yield of the in vitro enzymatic assay reached approximately 8.65 μg/200 mL, which is about 5.9 times higher than that of hairy roots (1.47 μg/200 mL), demonstrating greater efficiency and better controllability. This result indicates that switching the chassis and adopting an in vitro catalytic strategy is indeed an effective way to enhance leonurine production. At the same time, we recognize that the current yield still falls short of industrial requirements (≥60 mg/200 mL). As the next step, we plan to improve production through protein engineering (e.g., enzyme mutagenesis) and process optimization (e.g., reaction scale-up and substrate supply adjustment), as well as explore continuous in vitro synthesis systems, thereby providing a technological foundation for industrial application.

Summary

Through two iterative DBTL cycles, we established the first synthetic biology–based route to leonurine production, addressing the limitations of traditional extraction. In Cycle 1, we reconstructed the terminal pathway in Nicotiana tabacum hairy roots, achieving a 17-fold increase in yield compared with previous yeast-based production, and demonstrated the potential of plant chassis for scalable biosynthesis [5]. Building on this, Cycle 2 applied Human Practices insights to switch the chassis to E. coli, enabling high-level enzyme expression and in-vitro cascade catalysis, further boosting yield by 5.9-fold over hairy roots.

These results not only validate our pathway design but also highlight that chassis optimization and process innovation can synergistically overcome production bottlenecks. Looking ahead, we envision our system as the basis for producing leonurine as an active pharmaceutical ingredient (API), supplying pharmaceutical companies to develop low-cost, accessible protective and adjunctive therapies for cardiovascular diseases, ultimately expanding patient benefit.

References

1. Flora GD, Nayak MK. A Brief Review of Cardiovascular Diseases, Associated Risk Factors and Current Treatment Regimes. Curr Pharm Des. 2019;25(38):4063-4084. doi: 10.2174/1381612825666190925163827. PMID: 31553287.

2. National Center for Cardiovascular Diseases & The Writing Committee of the Report on Cardiovascular Health and Diseases in China. (2025). 中国心血管健康与疾病报告2024概要. 中国循环杂志, 40 (06), 521-559.

3. Huang L, Xu DQ, Chen YY, Yue SJ, Tang YP. Leonurine, a potential drug for the treatment of cardiovascular system and central nervous system diseases. Brain Behav. 2021 Feb;11(2):e01995. doi: 10.1002/brb3.1995. Epub 2020 Dec 10. PMID: 33300684; PMCID: PMC7882174.

4. Liu S, Sun C, Tang H, Peng C, Peng F. Leonurine: a comprehensive review of pharmacokinetics, pharmacodynamics, and toxicology. Front Pharmacol. 2024 Jul 19;15:1428406. doi: 10.3389/fphar.2024.1428406. PMID: 39101131; PMCID: PMC11294146.

5. Li P, Yan MX, Liu P, Yang DJ, He ZK, Gao Y, Jiang Y, Kong Y, Zhong X, Wu S, Yang J, Wang HX, Huang YB, Wang L, Chen XY, Hu YH, Zhao Q, Xu P. Multiomics analyses of two Leonurus species illuminate leonurine biosynthesis and its evolution. Mol Plant. 2024 Jan 1;17(1):158-177. doi: 10.1016/j.molp.2023.11.003. Epub 2023 Nov 10. PMID: 37950440.

6. Rao, X., & Liu, W. (2025). A Guide to Metabolic Network Modeling for Plant Biology. Plants, 14(3), 484. https://doi.org/10.3390/plants14030484

7. Tran TM, McCubbin TJ, Bihmidine S, et al. Maize Carbohydrate Partitioning Defective33 Encodes an MCTP Protein and Functions in Sucrose Export from Leaves. Molecular Plant. 2019 Sep;12(9):1278-1293. DOI: 10.1016/j.molp.2019.05.001. PMID: 31102785.

8. Youjun Zhang, Alisdair R. Fernie,Metabolons, enzyme–enzyme assemblies that mediate substrate channeling, and their roles in plant metabolism,Plant Communications, Volume 2, Issue 1, 2021, 100081, ISSN 2590-3462, https://doi.org/10.1016/j.xplc.2020.100081.