INTRODUCTION
Pyrethrin insecticides are derived from the plant
Tanacetum cinerariifolium (
T. cinerariifolium). In the early 1920s, natural pyrethrins were identified as the active ingredients of pyrethrum, enabling the complete structural characterization of all six natural compounds as monoterpenoid esters
1. As contact neurotoxins, they rapidly disrupt the insect nervous system, causing muscle paralysis, immobilization, and death. Natural pyrethrins exhibit low toxicity to mammals and undergo rapid photodegradation in the environment, with a half-life of only 2 to 48 h
2,3. Synthetic pyrethroids, including permethrin, tetramethrin, cypermethrin, and cyfluthrin, were developed to overcome the rapid photodegradation of natural pyrethrins
4. Pyrethroids are now among the most widely used insecticides in agriculture and household products worldwide, holding a substantial share of the modern insecticide market
5. However, synthetic pyrethroids can persist in soil for weeks or months, leading to environmental accumulation, ecological hazards, and growing resistance in pests. Notably, natural pyrethrins remain effective against certain pests that have developed resistance to synthetic pyrethroids
6-9.
Currently, producing natural pyrethrin insecticides relies on large-scale farming of pyrethrum flowers. This plant exhibits optimal growth and highest pyrethrin yields at high altitudes near the equator
10. As a result, commercial production is limited to only a few regions such as East Africa, Tasmania in Australia, and Yunnan in China. This limited supply makes natural pyrethrins rare and valuable, creating an urgent need to develop alternative production methods.
Over the past 20 years, genetic and molecular studies of
T. cinerariifolium have elucidated multiple steps in the pyrethrin biosynthetic pathway
11-21. Briefly, pyrethrin biosynthesis can be divided into three stages: the synthesis of the acid moiety, the synthesis of the alcohol moiety, and their subsequent condensation (Fig. 1). The biosynthesis of the acid moiety originates from the plastidial terpenoid pathway. Dimethylallyl diphosphate (DMAPP) is converted to chrysanthemyl diphosphate (CDP) by chrysanthemyl diphosphate synthase (TcCDS)
11. CDP is then dephosphorylated to chrysanthemol, a reaction that has been attributed to a secondary, low-efficiency activity of TcCDS, or is routed through chrysanthemyl monophosphate (CMP) by the Nudix-family phosphatase TcNudix1 and likely requires an additional phosphohydrolase to complete
12. Chrysanthemol is oxidized to chrysanthemic acid by alcohol dehydrogenase (TcADH2) and aldehyde dehydrogenase (TcALDH1)
13. As a branch point, chrysanthemol can also be hydroxylated by chrysanthemol 10-hydroxylase (TcCHH) and subsequently oxidized by TcADH2 and TcALDH1 to form 10-carboxychrysanthemol
14, which is then methylated by the 10-carboxychrysanthemic acid methyltransferase (TcCCMT) to yield pyrethric acid
14 (Fig. 1a).
In contrast, the biosynthesis of the alcohol moiety remains less clearly defined. Its precursor, α-linolenic acid (α-LeA), is thought to be cyclized to form 12-oxo-phytodienoic acid (
cis-OPDA), which may then undergo β-oxidation, isomerization, and decarboxylation to yield
cis-jasmone
15. However, the catalytic mechanisms underlying this transformation remain elusive.
cis-jasmone is hydroxylated by the jasmone hydroxylase (TcJMH) to form jasmolone
16, which is subsequently desaturated by pyrethrolone synthase (TcPYS) to generate pyrethrolone
17. The biosynthetic mechanism of cinerolone remains unreported (Fig. 1b). Ultimately, two acid moieties and three alcohol moieties are condensed by the Gly-Asp-Ser-Leu (GDSL) lipase-like protein (TcGLIP) to form the six natural pyrethrins
18,19 (Fig. 1c), although whether coenzyme A (CoA) activation of the acid moiety is required remains unclear.
Although the pyrethrin biosynthetic pathway remains incompletely elucidated, significant advances have been made in recent years, including production of the acid moiety in
Nicotiana benthamiana (
N. benthamiana) and tomato
14,20,21, the efficient biosynthesis of chrysanthemic acid in
Escherichia coli (
E. coli) through metabolic engineering strategies
22,23, and the discovery of transcription factors TcbZIP60 and TcMYC2 that regulate pyrethrin biosynthesis
24,25. However, critical gaps remain in our understanding of pyrethrin biosynthesis, including the fundamental mechanisms underlying the low efficiency of acid moiety biosynthesis, the biosynthesis of
cis-jasmone, and whether CoA activation is essential for the condensation reaction. The absence of key enzymes and unclear mechanisms hinder complete microbial heterologous synthesis, making systematic pathway elucidation essential for efficient production.
In this study, we combined high-quality genome sequencing of T. cinerariifolium with comprehensive transcriptome analysis to systematically identify previously uncharacterized enzymes of the pyrethrin biosynthetic pathway. Using multi-host screening and module-by-module pathway reconstruction, we functionally characterized three previously unannotated enzymes. Having established these enzyme functions and reconstituted routes to multiple pyrethrin congeners (pyrethrin I/II and jasmolin I/II) in N. benthamiana, we stepwise reconstructed the pathway modules in Saccharomyces cerevisiae (S. cerevisiae) to enable fermentative jasmolin I production, providing a starting point for strain optimization and scalable production.
RESULTS
Genome sequencing and pyrethrin accumulation in T. cinerariifolium
Natural pyrethrins consist of six esters formed by the conjugation of two monoterpenoids (chrysanthemic acid and pyrethric acid) and three rethrolones (pyrethrolone, jasmolone, and cinerolone) (Fig. 1c). Comparative profiling across various plant tissues and flower developmental stages revealed that chrysanthemic acid and its corresponding esters, jasmolin I and pyrethrin I, were most abundant in disk flowers, followed by jasmolin II (Fig. 2a). To identify genes involved in the pyrethrin biosynthetic pathway, we performed genomic and transcriptomic sequencing. Using 491.87 Gb of Illumina, 393.04 Gb of PacBio Sequel II HiFi long-read, and 693.50 Gb of Illumina Hi-C data (Supplementary Table S1), we assembled a high-quality genome of T. cinerariifolium. The final assembly had a total size of 7.71 Gb, consisting of 160 sequences (including 9 chromosome-level scaffolds and 151 supercontigs), and achieved an N50 of 866.94 Mb (Fig. 2b; Supplementary Tables S2, S3). Benchmarking Universal Single Copy Orthologs (BUSCO) analysis based on plant gene models showed that its completeness was 97.3%, indicating relatively high completeness of the assembled T. cinerariifolium genome (Supplementary Table S4). By combining transcriptomic data, protein sequences from closely related species, and de novo prediction, we predicted a total of 64,171 protein-coding genes. Functional annotation against multiple gene function databases successfully assigned putative functions to 60,832 genes (94.80%) (Supplementary Table S5). A comparative genomic analysis with the genomes of 11 other plant species (Fig. 2c; Supplementary Fig. S1) uncovered 8,527 species-specific gene families in T. cinerariifolium, encompassing 17,454 genes (Supplementary Table S6). Further Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment analysis highlighted metabolism-related adaptations (Supplementary Fig. S2). Phylogenetic reconstruction using 215 highly conserved single-copy orthologs dated the divergence of T. cinerariifolium from its closely related Asteraceae sister lineages (e.g., Artemisia annua, Chrysanthemum lavandulifolium) to approximately 7.9 million years ago (Mya) (Fig. 2c). Furthermore, we identified significant gene family dynamics, with 1,173 families (3,036 genes) undergoing expansion and 1,848 families (2,133 genes) undergoing contraction along the T. cinerariifolium lineage (Fig. 2c; Supplementary Table S7). Notably, these evolutionarily dynamic families were predominantly enriched in pathways such as “fatty acid degradation/biosynthesis” and “linoleic acid metabolism” (Supplementary Figs. S3 and S4), suggesting that T. cinerariifolium has evolved distinct regulatory or metabolic features in fatty acid-derived secondary metabolism (such as pyrethrin biosynthesis) compared to its relatives.
A terpene synthase acts as a dephosphatase in chrysanthemic acid biosynthesis
TcCDS catalyzes the first step in the biosynthesis of these monoterpenoids by condensing two molecules of DMAPP (
1). It has also been proposed to perform additional cleavage of the diphosphate group, thereby generating chrysanthemol (
3)—the branch point intermediate for the synthesis of both chrysanthemic acid (
4) and pyrethric acid (
6) (Fig. 3a). However, when
TcCDS was expressed in a previously constructed high-DMAPP-producing yeast chassis strain JCR27
26,27, no chrysanthemol (
3) was detected (Fig. 3b). The gene encoding a Nudix-family phosphatase,
TcNudix1, which was reported to dephosphorylate CDP (
2) to CMP
12, was co-expressed with
TcCDS in the JCR27 strain. This combination yielded only trace amounts of chrysanthemol (
3) (Fig. 3b). These findings indicate that although TcCDS can catalyze the formation of chrysanthemol (
3) from DMAPP (
1), the reaction proceeds with limited efficiency, suggesting that additional enzymes (e.g., phosphatase) might be required to promote this catalytic process
12,13.
In monoterpenoid biosynthesis, certain terpene synthases (TPS) have been reported to exhibit dephosphorylation activity
28. Firstly, using the Hidden Markov Models (HMMs) search based on the Pfam ID (PF01397 and PF03936)
29,30, we preliminarily identified eighty-three
TPS genes from the
T. cinerariifolium genome. Subsequently, we performed transcriptome sequencing of various pyrethrum tissues (leaf, root, stem, flower, etc.). Based on co-expression analysis, we screened for
TPS genes that showed significant co-expression (correlation coefficient > 0.7) with
TcCDS, the core gene responsible for chrysanthemol (
3) biosynthesis. This led to the final identification of eighteen high-confidence
TPS candidate genes (Supplementary Fig. S5 and Table S8). When co-expressed with
TcCDS in the yeast chassis strain JCR27, four of these
TPS genes (
TcTPS1,
TcTPS4,
TcTPS8, and
TcTPS10) led to the detection of chrysanthemol (
3), among which
TcTPS1 showed the highest efficiency (Fig. 3c).
To further validate the dephosphorylation activity of TcTPS1, we subsequently expressed and purified TcTPS1 in
E. coli, confirming its capacity to catalyze the dephosphorylation of CDP (
2) to chrysanthemol (
3)
in vitro, yielding 1.85-fold higher product levels than TcCDS alone (Fig. 3d; Supplementary Fig. S6a). Moreover, when DMAPP (
1) was used as the substrate, the combination of TcCDS and TcTPS1 increased chrysanthemol (
3) production by 1.66-fold compared to TcCDS alone (Fig. 3e; Supplementary Fig. S6a). Kinetic analysis revealed that the catalytic efficiency (
kcat/
Km) of TcTPS1 in converting CDP (
2) to chrysanthemol (
3) was 3.15 mmol⁻¹s⁻¹, which was 3.12-fold higher than the reported catalytic efficiency of TcCDS (1.01 mmol⁻¹s⁻¹)
31 (Supplementary Fig. S6b). Transient expression in
N. benthamiana leaves further confirmed that the inclusion of TcTPS1 significantly enhanced the catalytic efficiency of TcCDS, leading to 2.55-fold higher chrysanthemol (
3) yields (Fig. 3f). TcCDS exhibits low hydrolytic efficiency toward CDP (
2), despite its ability to efficiently synthesize CDP (
2) from DMAPP (
1). This dual activity profile combines efficient CDP (
2) synthesis with only limited dephosphorylation. As a result, cooperation with TcTPS1 enhanced catalytic efficiency for chrysanthemol (
3) production. Subcellular localization analysis confirmed co-localization of TcCDS and TcTPS1 in plastids (Supplementary Fig. S6c), which was further validated by luciferase complementation assays
in planta (Supplementary Fig. S7). The discovery of TcTPS1 as a functional partner that compensates for TcCDS’s hydrolytic limitation may have implications for metabolic engineering applications.
A phylogenetic analysis of terpene synthase gene families in pyrethrum,
Arabidopsis, rice, and lotus revealed that
TcTPS1 belongs to the TPS-g subfamily (Fig. 3g). The TPS-g subfamily constitutes an angiosperm-specific lineage of
TPS genes, characterized by functional diversification and serving as a key driver of the evolutionary expansion of plant secondary metabolic pathways
30. Inspired by the dephosphorylation activity of TcTPS1 toward CDP (
2), we investigated whether TcTPS1 could dephosphorylate other phosphorylated small-molecule substrates. For comparison, we also examined
NnTPS3, a TPS-g member from lotus
32. When co-expressed with canonical and non-canonical precursor synthases for monoterpenes and sesquiterpenes, which include
TcCDS,
LPPS33,
GPPS34, and
FPPS27, both
TcTPS1 and
NnTPS3 produced dephosphorylated and hydroxylated products, leading to the formation of chrysanthemol (
3), lavandulol, linalool, and nerolidol (Fig. 3h; Supplementary Figs. S8 and S9). These results demonstrate that
TcTPS1 effectively improves the catalytic efficiency of chrysanthemol (
3) production. Moreover, they support the possibility that the plant TPS-g subfamily may possess broad dephosphorylation activity beyond their established roles in terpenoid formation.
Chrysanthemic acid (
4) and pyrethric acid (
6) are two constituent monoterpenoids of pyrethrins, biosynthetically originating from chrysanthemol (
3) (Fig. 3a). A previously reported P450 (TcCHH) was proposed to hydroxylate the C-10 position of chrysanthemol (
3) to yield 10-OH chrysanthemol (
5)
14. Co-expression analysis identified two additional cytochrome P450 enzymes, TcCYP71BZ2 and TcCYP71BZ3, differing from the previously reported TcCHH by 10 and 3 amino acids, respectively (Supplementary Fig. S10 and Table S9). Genomic analysis revealed that
TcCYP71BZ2 is located adjacent to, but distinct from, the
TcCHH locus, while
TcCYP71BZ3 resides at a separate genomic location, indicating that these represent paralogous genes rather than allelic variants (Supplementary Fig. S11). Functional analysis in
N. benthamiana confirmed that both enzymes catalyze chrysanthemol (
3) to 10-OH chrysanthemol (
5), as verified by comparison with an in-house synthesized chemical standard (Fig. 3i). By co-expressing
TcCDS,
TcTPS1,
TcADH2, and
TcALDH1 in
N. benthamiana leaves, we achieved
de novo production of chrysanthemic acid (
4) (Fig. 3j; Supplementary Fig. S12a). Furthermore, upon introducing
TcCYP71BZ2 and the previously reported methyltransferase
TcCCMT14, we successfully produced pyrethric acid (
6) (Fig. 3k; Supplementary Fig. S12b).
cis-jasmone biosynthesis through specialized β-oxidation in pyrethrum
cis-jasmone (
17) serves as a critical volatile defense signal in many plants, being rapidly released upon mechanical wounding or herbivory to repel pests and attract natural enemies
16,35-37. Although its biosynthesis remains incompletely characterized, it is potentially linked to the jasmonic acid (JA) pathway due to structural similarity. Based on this, previous studies hypothesized that
cis-jasmone (
17) might be directly derived from
iso-OPDA (
14) via a proposed β
-oxidation step to form 3,7-ddh-JA (
16), followed by decarboxylation (Fig. 4a). In this hypothetical pathway,
iso-OPDA (
14) and 3,7-ddh-JA (
16) are posited as key intermediates
15,38. However, this mechanistic model lacks experimental validation
in planta.
In plants,
cis-OPDA (
13) is synthesized from α-LeA (
10) by the sequential action of 13-lipoxygenase (LOX), allene oxide synthase (AOS), and allene oxide cyclase (AOC)
39. To define this upstream branch in pyrethrum, we performed co-expression analysis with
TcJMH and
TcPYS, and obtained ten candidate genes (seven
TcLOX, two
TcAOS, and one single-copy
TcAOC). Subsequent
in vitro reconstitution assays revealed that only
TcLOX2,
TcLOX4,
TcLOX7,
TcAOS1, and
TcAOC catalyzed the formation of
cis-OPDA (
13) (Supplementary Fig. S13). Their activities were consistent with those of the previously characterized
GmLOX,
AtAOS, and
AtAOC from soybean and
Arabidopsis40 (Fig. 4b; Supplementary Fig. S14). By contrast, the formation of
iso-OPDA (
14) has only been reported via
cis-OPDA (
13) isomerization by a glutathione transferase 16 (GST16) cloned from the gut of cotton bollworm larvae
41,42, with no known plant pathway. We therefore prepared
cis-OPDA (
13) and
iso-OPDA (
14) by chemo-enzymatic synthesis for use as both enzymatic substrates and analytical standards (Supplementary Fig. S15a), and used them to screen the candidate genes in pyrethrum genome. However, no pyrethrum gene capable of directly producing
iso-OPDA (
14) was identified, and none of pyrethrum
GST genes displayed
cis- to
iso-OPDA isomerase activity (Supplementary Fig. S16).
β-oxidation represents a critical step in jasmonate biosynthesis, dependent on four core enzymes, namely acyl-CoA synthetase (ACS), acyl-CoA oxidase (ACX), multifunctional protein (MFP), and L-3-ketoacyl-CoA thiolase (KAT), which together catalyze three iterative cycles of oxidation to convert
cis-OPDA (
13) and OPC (derived from
cis-OPDA via OPDA reductase, OPR) into 4,5-ddh-JA (
15) and 7-
iso-JA, respectively, ultimately yielding JA
43,44 (Fig. 4a). Notably,
iso-OPDA (
14) has been shown to undergo β-oxidation in various plants to form 3,7-ddh-JA (
16), a direct precursor of
cis-jasmone (
17)
38. Based on co-expression network analysis with the identified pathway genes (
TcAOC,
TcCDS, and
TcJMH), we identified twenty-five β-oxidation-related genes from the pyrethrum genome, comprising eight
ACS, twelve
ACX, two
MFP, and three
KAT genes. To enable unambiguous assignment of their reaction products during functional validation, we synthesized 4,5-ddh-JA (
15) as a reference standard (Supplementary Fig. S17a–d).
To validate the pyrethrum β-oxidation genes, we adopted a combinatorial replacement strategy in which the four core Arabidopsis enzymes (AtACS, AtACX, AtMFP, and AtKAT) served as a reference backbone, and each component was swapped individually with its pyrethrum counterpart. When assaying the ACS/MFP/KAT candidate genes with cis-OPDA (13), the major product was 4,5-ddh-JA (15) (retention time [RT] = 21.5 min), which was consistent with the results from the full Arabidopsis control group. Among them, TcACS7, TcMFP1, and TcKAT1 exhibited the highest efficiency (Supplementary Fig. S18a, c, d). By contrast, ACX candidate assays (including six enzymes such as TcACX2, TcACX3, and TcACX5) produced a new peak at RT = 20.9 min with identical molecular mass to 4,5-ddh-JA (15) (Supplementary Fig. S18b), suggesting an isomeric product. To verify its identity, we enzymatically and chemically synthesized 3,7-ddh-JA (16) (Supplementary Fig. S17a–d), confirming that the 20.9-min peak corresponded to 3,7-ddh-JA (16). Among six pyrethrum ACX genes tested in N. benthamiana, TcACX3 exhibited the highest activity, converting cis-OPDA (13) specifically into 3,7-ddh-JA (16) (Supplementary Fig. S19). Stepwise replacement of Arabidopsis β-oxidation enzymes with their pyrethrum counterparts revealed that substituting only AtACX with TcACX3 was sufficient to produce 3,7-ddh-JA (16) as the sole product (Fig. 4c; Supplementary Fig. S20a), along with detectable cis-jasmone (17) (Supplementary Fig. S21).
When four pyrethrum β-oxidation genes (TcACS7, TcACX3, TcMFP1, and TcKAT1) were transiently expressed in N. benthamiana and incubated with cis-OPDA (13), they produced almost exclusively 3,7-ddh-JA (16), with negligible 4,5-ddh-JA (15). Only trace amounts of 3,7-ddh-JA (16) were derived from iso-OPDA (14) (Fig. 4d; Supplementary Fig. S20b). This indicates that β-oxidation genes in pyrethrum simultaneously oxidize cis-OPDA (13) and isomerize its double-bond position—a previously unrecognized mechanism. Feeding synthesized 13C-labeled cis-OPDA (13’) (Supplementary Fig. S15a, d) to N. benthamiana leaves co-expressing the four genes yielded 13C-labeled 3,7-ddh-JA (16’) and its decarboxylation product, 13C-labeled cis-jasmone (17’), but no labeled iso-OPDA (14’) or 4,5-ddh-JA (15’) (Fig. 4e, f). These data confirm that pyrethrum harbors a specialized β-oxidation system capable of converting cis-OPDA (13) directly into 3,7-ddh-JA (16).
Since
cis-jasmone (
17) can arise from spontaneous decarboxylation of 3,7-ddh-JA (
16)
15,45, we consistently detected
cis-jasmone (
17) alongside 3,7-ddh-JA (
16)
in vitro with recombinant β-oxidation enzymes, as well as in
N. benthamiana, and yeast systems (Supplementary Fig. S22). Further incubation of purified 3,7-ddh-JA (
16) at room temperature confirmed time-dependent decarboxylation to
cis-jasmone (
17), reaching approximately 75% conversion after 54 h (Fig. 4g). We further noted that 3,7-ddh-JA (
16) partially decarboxylates in the GC inlet during GC-MS analysis (Supplementary Fig. S17d); accordingly, 3,7-ddh-JA (
16) was quantified by LC-MS and the
cis-jasmone (
17) signal was used for qualitative assignment only. Collectively, these results indicate that
cis-jasmone (
17) is produced via a specialized pathway in pyrethrum (Fig. 4a): α-LeA (
10) is converted into
cis-OPDA (
13) in three enzymatic steps; the specialized β-oxidation cascade, with TcACX as the key determinant, directly transforms
cis-OPDA (
13) into 3,7-ddh-JA (
16) through a coupled oxidation-isomerization reaction; and 3,7-ddh-JA (
16) finally undergoes spontaneous decarboxylation to yield
cis-jasmone (
17).
Glycosylation of chrysanthemic acid is the key step for ester condensation
Ester condensation is the final step in the biosynthesis of the six pyrethrins, yet the activation mechanism enabling this condensation remained to be fully elucidated. A GDSL lipase-like protein (GLIP) isolated from pyrethrum flowers can catalyze,
in vitro, the condensation of either chrysanthemoyl-CoA or pyrethroyl-CoA with three rethrolones to produce all six natural pyrethrins (Fig. 5a)
18,19. Using chemically synthesized chrysanthemoyl-CoA (
7), we confirmed that only chrysanthemoyl-CoA (
7), but not free chrysanthemic acid (
4), undergoes condensation with jasmolone (
18) in the presence of TcGLIP (Fig. 5b; Supplementary Fig. S23a). However, the CoA ligase responsible for activating chrysanthemic acid (
4) to chrysanthemoyl-CoA (
7)
in vivo remains unknown, and the chrysanthemoyl-CoA (
7) used in prior TcGLIP studies was chemically synthesized
18,19. To identify this enzyme, we screened candidate CoA ligase genes from pyrethrum that showed co-expression with chrysanthemic acid (
4) biosynthetic genes and tested their activity
in vitro. Unfortunately, none of the eight candidates successfully produced chrysanthemoyl-CoA (
7) (Supplementary Fig. S23b).
During activity screening in
N. benthamiana leaves, we unexpectedly found that chrysanthemic acid (
4), when injected into
N. benthamiana leaves, was converted into metabolites that could serve as substrates for TcGLIP. When these metabolites were incubated with jasmolone (
18), they underwent enzymatic condensation to yield jasmolin I (
21) (Fig. 5b). This finding is consistent with prior reports demonstrating that transient co-expression of
TcJMH and
TcGLIP genes in
N. benthamiana, followed by feeding with chrysanthemic acid (
4) and
cis-jasmone (
17) alone (without the addition of an exogenous CoA ligase), is sufficient for the biosynthesis of jasmolin I (
21)
16,17. Together, these results indicate that
in planta modification of chrysanthemic acid (
4) in
N. benthamiana is essential for the esterification reaction. LC-MS/MS analysis revealed that chrysanthemic acid (
4) is further modified into chrysanthemoyl-glucoside (
8) and chrysanthemoyl-malonylglucoside (
9) in
N. benthamiana leaves (Fig. 5c, d; Supplementary Fig. S24a, b), consistent with previous reports
13. We also identified these glucosylated conjugates of chrysanthemic acid (
4) in pyrethrum disk flowers (Fig. 5c, d); their abundance changed dynamically during inflorescence development (Supplementary Fig. S24c, d) and correlated significantly with most pyrethrin congeners and with the expression of
TcJMH,
TcPYS,
TcUGT1/6, and
TcGLIP across tissues (Supplementary Fig. S25). By contrast, chrysanthemoyl-CoA (
7) was barely detectable in leaves, stems, roots, and inflorescence tissues at different growth stages (Supplementary Fig. S24e, f). These results suggest that chrysanthemoyl-glucoside (
8) is likely a physiologically relevant acyl donor in pyrethrin biosynthesis, although a role for chrysanthemoyl-CoA (
7) cannot be fully excluded.
We then selected ten glycosyltransferases that are highly expressed in flowers for functional screening through correlation analysis (Supplementary Fig. S26 and Tables S10, S11). After heterologous expression in E. coli, we found that TcUGT1 and TcUGT6 were able to glycosylate chrysanthemic acid (4), producing chrysanthemoyl-glucoside (8)—a product identical to the modified form of chrysanthemic acid (4) previously detected in N. benthamiana leaves (Fig. 5e; Supplementary Fig. S27). We then used purified chrysanthemoyl-glucoside (8) as the acyl donor in a TcGLIP-catalyzed condensation with jasmolone (18) (Supplementary Fig. S28). TcGLIP produced jasmolin I (21) when chrysanthemoyl-CoA (7) or chrysanthemoyl-glucoside (8) was supplied as donor (Fig. 5f), and accumulation of jasmolin I (21) tracked with consumption of chrysanthemoyl-glucoside (8) (Supplementary Fig. S29). No product was detected in controls lacking TcGLIP, lacking jasmolone (18), or containing free chrysanthemic acid (4) instead of an activated acyl donor. This confirms that trace residual chrysanthemic acid (4) co-purified with chrysanthemoyl-glucoside (8) did not contribute to ester formation (Fig. 5f).
The capacity of TcGLIP to catalyze ester condensation using both the bulky chrysanthemoyl-CoA (
7) and the smaller chrysanthemoyl-glucoside (
8) attracted our interest. Typically, the ester condensation of such structurally distinct substrates has been reported to require distinct enzyme families. For instance, in the biosynthesis of plant tropane alkaloids, the esterification of tigloyl-CoA and phenyllactyl glucoside is catalyzed by a BAHD acyltransferase (BAHD-AT, TS)
46 and a serine carboxypeptidase-like acyltransferase (SCPL-AT, LS)
47, respectively. However, we conducted functional characterization on the nine
TcSCPL genes selected through the co-expression analysis, and found that none of them had any functional significance (Supplementary Fig. S30 and Table S12). To investigate the underlying molecular basis, we predicted the protein structure of TcGLIP and performed molecular docking simulations with each substrate paired with jasmolone (
18). The results revealed that Ser40 within the conserved N-terminal GDSX motif acts as a nucleophile attacking the ester carbonyl of the substrate (Fig. 5g). The distances between Ser40 and the carbonyl carbon of chrysanthemoyl-CoA (
7) and chrysanthemoyl-glucoside (
8) were 3.1 Å and 4.3 Å, respectively. Meanwhile, jasmolone (
18) was positioned for a second nucleophilic attack, with its hydroxyl oxygen located 3.5 Å and 5.1 Å from the carbonyl carbons of the two substrates, respectively. Both distances fall within the effective nucleophilic attack radius (≤ 5.5 Å). In addition, the TS active site (804.16 Å
3) is significantly larger than that of LS (654.58 Å
3), consistent with their respective specificities for the bulky tigloyl-CoA and the smaller phenyllactyl glucoside. The volume of GLIP’s active pocket (744.91 Å
3) is intermediate between them (Supplementary Fig. S31). These results suggest that the broad substrate acceptance of GLIP is enabled by its characteristically open and plastic active pocket.
Reconstitution and combinatorial biosynthesis of multiple pyrethrins in N. benthamiana
To validate the functionality of the assembled pathway, we attempted heterologous biosynthesis of multiple pyrethrin congeners (pyrethrin I/II, jasmolin I/II) in N. benthamiana through different gene combinations (Fig. 6a). The biosynthetic steps were divided into modules: Group 1, comprising four genes responsible for chrysanthemic acid (4) synthesis (TcCDS, TcTPS1, TcADH2, and TcALDH1), was combined with the jasmolone hydroxylase gene TcJMH and the final esterification enzyme gene TcGLIP. Transient co-expression of these genes in N. benthamiana, followed by feeding with cis-jasmone (17), successfully yielded jasmolin I (21) (Fig. 6b). Feeding with 3,7-ddh-JA (16) also produced jasmolin I (21) (Fig. 6b), confirming that 3,7-ddh-JA (16) undergoes spontaneous decarboxylation to form cis-jasmone (17) in planta.
Next, we combined Group 2 (the β-oxidation module: TcACS7, TcACX3, TcMFP1, and TcKAT1) with Group 1, TcJMH, and TcGLIP. Transient expression of this set in N. benthamiana supplemented with cis-OPDA (13) successfully generated jasmolin I (21) (Fig. 6b). Further extending the pathway, we introduced a total of thirteen genes, including Group 3 (the cis-OPDA (13) synthesis module: TcLOX2, TcAOS1, and TcAOC) together with Groups 1 and 2, TcJMH, and TcGLIP into N. benthamiana leaves. Supplementation with α-LeA (10) enabled the heterologous production of jasmolin I (21) (Fig. 6b), demonstrating the functionality of the reconstructed route to jasmolin I (21).
Natural pyrethrins comprise six esters formed by conjugating structurally modified chrysanthemic acids with rethrolones (Fig. 1). To diversify the products, we introduced Group 4 genes (TcCYP71BZ2 and TcCCMT for acid modification, and TcPYS for rethrolone modification) alongside Group 1, TcJMH and TcGLIP. Feeding cis-jasmone (17) to this engineered background yielded jasmolin I (21) (Fig. 6c; Supplementary Fig. S32a). To enhance the supply of chrysanthemic acid (4) and its derivatives, we further supplemented the system with the intermediate chrysanthemol (3), which enabled the simultaneous production of jasmolin I (21) and pyrethrin I (23) (Fig. 6c; Supplementary Fig. S32c). Due to the low efficiency of pyrethric acid (6) formation, direct feeding of pyrethric acid (6) allowed the detection of jasmolin II (22) and pyrethrin II (24) in N. benthamiana leaves (Fig. 6c; Supplementary Fig. S32b, d). Collectively, by incorporating modification enzymes and supplementing key intermediates, we achieved the heterologous biosynthesis of four pyrethrins in N. benthamiana (Fig. 6a).
Modular engineering enables heterologous biosynthesis of jasmolin I in yeast
Having reconstituted the pyrethrin biosynthetic pathway and achieved heterologous production of jasmolin I (21) from α-LeA (10) in N. benthamiana, we next sought to establish the same route in a yeast chassis. The overall metabolic engineering strategy was divided into three modules: Module I for chrysanthemic acid (4) biosynthesis, Module II for jasmolone (18) biosynthesis, and Module III for the formation of chrysanthemoyl-glucoside (8) and the final ester condensation (Fig. 7a).
In Module I, we started with the mevalonate (MVA)-overexpressing
S. cerevisiae strain JCR27
26,27. Genomic integration of three copies of
TcCDS, one copy of
TcTPS1, and three copies of
TcALDH1 and
TcADH2 produced strain J1 (Supplementary Table S13), which produced detectable but very low levels of chrysanthemic acid (
4) (Fig. 7b; Supplementary Fig. S33). To enhance DMAPP precursor supply, we replaced the native promoter of
ERG20 (yeast FPPS) with the glucose-inducible promoter
PHXT1 to dynamically downregulate its expression, thereby reducing competitive flux toward endogenous terpenes
48. The resulting strain J2 showed a substantial increase in chrysanthemic acid (
4) production, reaching 28.54 mg/L in shake-flask culture (Fig. 7b).
For Module II, we began with a yeast strain 30000B and integrated the high-efficiency
cis-OPDA (
13) synthesis genes
GmLOX,
AtAOS, and
AtAOC as a cassette into the
LEU2 locus, generating strain S1 (Supplementary Table S13). Feeding α-LeA (
10) to S1 successfully resulted in the production of
cis-OPDA (
13) (Fig. 7c). Subsequently, the β-oxidation genes
TcACS7,
TcACX3,
TcMFP1, and
TcKAT1 were integrated as a second and third cassette into the
YPRCΔ15 and
URA3 loci, producing strain S2 (Supplementary Table S13). Supplementation of S2 with α-LeA (
10) enabled the production of 3,7-ddh-JA (
16) (Fig. 7d). Extending the fermentation to four days further allowed for the detection of
cis-jasmone (
17) in the culture (Fig. 7e). Due to the low endogenous
cis-jasmone (
17) levels in the yeast chassis, we supplied exogenous
cis-jasmone (
17) and proceeded to build the jasmolone (
18) synthesis module in strain J2. Jasmolone (
18) formation relies on the cytochrome P450 enzyme TcJMH, whose activity can be influenced by the choice of cytochrome P450 reductase (CPR). We therefore screened various plant-derived CPRs, including TcCPR, AaCPR
49, and SaCPR
49,50, using whole-cell biocatalysis assays. The combination of
TcJMH with
SaCPR showed the highest catalytic efficiency (Fig. 7f). Consequently, two copies of
SaCPR and
TcJMH were integrated into the genome of J2, yielding strain J3 (Supplementary Table S13), which efficiently converted fed
cis-jasmone (
17) into jasmolone (
18) (Fig. 7g).
In Module III, we integrated a single copy of
TcUGT6 and
TcGLIP into J3 to generate strain J4 (Fig. 7h; Supplementary Table S13). While this strain produced chrysanthemoyl-glucoside (
8), the majority of chrysanthemic acid (
4) remained unglycosylated (Supplementary Fig. S34). This low glycosylation efficiency of chrysanthemic acid (
4) likely accounts for the absence of the final product jasmolin I (
21) at this stage. To improve glycosylation efficiency for the subsequent condensation step, we introduced an additional copy of
TcUGT6 together with the yeast phosphoglucomutase gene
PGM1, which promotes UDP-glucose accumulation into the
SPR1 locus
51. Deletion of the endogenous glucosidase
SPR152 was also performed to potentially reduce product degradation. The resulting strain J5 exhibited markedly improved glycosylation efficiency, with chrysanthemoyl-glucoside (
8) accumulation increasing by approximately 4.6-fold relative to J4. Reduced product toxicity further elevated the titer of this glycosylated compound (Fig. 7h; Supplementary Fig. S34). Moreover, jasmolin I (
21) was successfully detected in J5, with a titer of 0.55 mg/L (Supplementary Figs. S35, S36 and Table S13). Furthermore, to lower fermentation costs, we knocked out
GAL80, a negative regulator of galactose metabolism
53 and added an extra copy of
TcGLIP, creating the final production strain J6 (Supplementary Table S13). Fed with
cis-jasmone (
17), J6 successfully produced jasmolin I (
21) at a titer of 1.22 mg/L in shake-flask fermentation, representing an approximately 2.2-fold improvement over J5 (Fig. 7i; Supplementary Fig. S35).
DISCUSSION
Over the past two decades, the biosynthetic pathways of several iconic plant terpenoids, paclitaxel
54,55, artemisinin
56, and cannabinol
57, have been elucidated, enabling synthetic-biology routes to plant-derived medicines. Here we extend this paradigm to botanical insecticides. We resolved three previously unidentified elements of the pyrethrin pathway and, on this basis, achieved heterologous production of jasmolin I (
21) both in
N. benthamiana and in
S. cerevisiae upon supplementation with specific precursors. A bio-based route to natural pyrethrins thus points toward a potentially more sustainable alternative to chemically synthesized pyrethroids, whose widespread use has accelerated pest resistance and harmed non-target organisms.
Our pathway elucidation began from a catalytic bottleneck. Previous studies enhanced DMAPP (
1) supply for chrysanthemic acid (
4) biosynthesis in
E. coli (15.30 mg/L in flasks, 141.78 mg/L in bioreactors)
22 and developed an isopentenol utilization pathway yielding 510 mg/L chrysanthemol
23. Yet in the high-DMAPP yeast chassis JCR27
26,27, TcCDS alone produced no detectable chrysanthemol (
3) (Fig. 3b), implicating inefficient dephosphorylation of the CDP (
2) intermediate. We therefore hypothesized that plant TPS may harbor unrecognized dephosphorylating activity and identified TcTPS1 (Fig. 3c). Co-expression of
TcCDS and
TcTPS1 substantially improved conversion of DMAPP (
1) to chrysanthemol (
3) in yeast (Fig. 3c),
E. coli (Fig. 3d, e) and
N. benthamiana (Fig. 3f). Although TcTPS1 may not be the optimal native enzyme in pyrethrum, it raises catalytic throughput and limits toxic intermediate accumulation. The “TcCDS + TcTPS1” module thus overcomes the core catalytic bottleneck in chrysanthemic acid (
4) biosynthesis and provides a broadly applicable strategy for heterologous production of chrysanthemol (
3), chrysanthemic acid (
4) and, ultimately, pyrethrins.
Although β-oxidation is highly conserved in eukaryotes including yeast
58, TcACX3 from pyrethrum exhibits distinctive catalytic properties. Whereas
cis-jasmone (
17) is reported to arise via conventional β-oxidation of
iso-OPDA (
14) in various plants and yeast
38, the expected products from
cis-OPDA (
13) and
iso-OPDA (
14) should be 4,5-ddh-JA (
15) and 3,7-ddh-JA (
16), respectively. Our screening showed that TcACX3 selectively acts on
cis-OPDA (
13) while simultaneously mediating double-bond isomerization, without acting directly on
iso-OPDA (
14) (Fig. 4d); isotope-labeling experiments confirmed this substrate preference (Fig. 4e). Systematic evolutionary analysis showed that most
ACX genes in pyrethrum (e.g.,
TcACX5,
TcACX7,
TcACX8, and
TcACX9) arose by tandem duplication (Supplementary Fig. S11), whereas
TcACX3 did not. Instead, it is specifically retained in the pyrethrum lineage but lost from other closely related Asteraceae species (Supplementary Fig. S37). This lineage-specific evolution history likely underlies the specialized activity of TcACX3 toward
cis-OPDA (
13), and structural characterization will be needed for mechanistic insight.
In this study, we found that the acid moiety is activated by glycosylation: TcUGT1 and TcUGT6 convert chrysanthemic acid (
4) into chrysanthemoyl-glucoside (
8) (Fig. 5e, f), whereas chrysanthemoyl-CoA (
7) was not detectable
in planta (Supplementary Fig. S24e, f). Since TcGLIP is also active with chrysanthemoyl-CoA (
7), however, we cannot formally exclude a CoA-dependent activation of the acid moiety, and a low-abundance or rapidly turned-over chrysanthemoyl-CoA (
7) pool in glandular trichomes remains possible. Glycosylation therefore appears to serve as a prerequisite activation step for condensation, analogous to the situations reported in nicotine and littorine biosynthesis
47,59,60. Notably, GLIP displays an unusual substrate promiscuity, accepting both chrysanthemoyl-glucoside (
8) and chrysanthemoyl-CoA (
7) for ester formation. Definitive assignment of the
in planta acyl donor will require loss- or gain-of-function analysis in
T. cinerariifolium, which remains an important direction for future studies.
Synteny analysis revealed the absence of conserved TcCDS homologs in several closely related Asteraceae species, and phylogenetic evidence indicated that TcCDS originated from a recent whole-genome duplication (WGD) event in T. cinerariifolium, evolving from a farnesyl diphosphate synthase gene (Supplementary Figs. S37, S38 and Table S17). Microsynteny analysis of six core genes in the pyrethrin pathway (TcTPS1, TcCDS, TcACX3, TcJMH, TcUGT6, and TcGLIP) further showed that only TcTPS1, TcUGT6, and TcGLIP likely evolved through gene duplication events shared among Asteraceae species (Supplementary Fig. S37). By contrast, TcCDS, TcACX3, and TcJMH appear specifically retained in pyrethrum and lost in the other four closely related Asteraceae lineages, suggesting lineage-specific recruitment (Supplementary Fig. S37). Collectively, these findings illustrate how specialized metabolic pathways in non-model plants often arise through lineage-specific gene recruitment and functional innovation, revealing a reservoir of enzymatic diversity beyond the canonical pathways characterized in model species.
In summary, by demonstrating that TcTPS1 enhances the catalytic efficiency of TcCDS, elucidating the pyrethrum-specific β-oxidation route from cis-OPDA (13) to cis-jasmone (17), and revealing glycosylation of the acid moiety prior to condensation, we achieved the biosynthesis of pyrethrins in both N. benthamiana and S. cerevisiae (Supplementary Fig. S39). Beyond providing a new paradigm for heterologous synthesis, our findings highlight how evolutionary innovation in specialized enzymes drives natural-product diversity and underscore that plant metabolism still harbors a vast repertoire of uncharacterized enzymes with specialized functions.
MATERIAL AND METHODS
Plant materials
T. cinerariifolium plant samples were collected from Yuxi City (24° 20′ 49″ N, 102° 31′ 37″ E), Yunnan Province, China. Fresh leaves, stems, roots, and flower tissues were harvested and immediately frozen in liquid nitrogen, then stored at −80 °C prior to genome sequencing, transcriptome sequencing, and metabolite analysis. N. benthamiana plants were cultivated in an artificial climate chamber at 23 °C with a 16-h light/8-h dark cycle. Before Agrobacterium infection, N. benthamiana plants grew for about 4 weeks with regular watering.
Chemicals
All solvents used for extractions, chemical synthesis and preparative HPLC were of HPLC grade and purchased from Fisher Scientific. Chrysanthemol (3, CAS: 5617-92-5), chrysanthemic acid (4, CAS: 4638-92-0), α-LeA (10, CAS: 463-40-1), cis-OPDA (13, CAS: 85551-10-6) and cis-jasmone (17, CAS: 488-10-8) standards were purchased from Sigma-Aldrich. Compounds that were not commercially available were synthesized according to the procedures described in the Chemical Synthesis section of the Supplementary Materials.
Genome sequencing, assembly, and annotation
High-quality genomic DNA was isolated from fresh leaves by a modified CTAB method. PacBio HiFi (SMRTbell Express Template Prep Kit 2.0, Revio), 150-bp paired-end Illumina (NovaSeq 6000), and Hi-C libraries were prepared and sequenced following standard protocols. Total RNA from three replicates of roots, stems, leaves and flowers (stages S2, S4, S6) was extracted with TRIzol (Invitrogen) and used for PacBio Iso-Seq (Sequel, P6C4) and Illumina RNA-seq (NEBNext Ultra II RNA Kit, HiSeq 6000).
Hifiasm (v0.16.1) was used for self-correction and assembly with default parameters
61. Assemblies were polished for two rounds with Pilon (v1.23)
62, and purge_dups (v1.2.5) removed duplications and reassigned allelic contigs. Scaffolds were anchored into chromosome-level scaffolds from Hi-C data with Juicer (v1.6)
63 and 3D-DNA (v180922)
64. Assembly quality was assessed with BUSCO (v5.4.7)
65 and QUAST
66, and overall alignment rates were calculated after mapping HiFi, Iso-Seq, Illumina, assembled-transcript and RNA-seq reads with BWA (v0.7.13)
67, minimap2 (v2.17)
68, and HISAT2 (v2.1.0)
69. Detailed procedures are provided in the Supplementary Materials and Methods.
Genome annotation
Gene prediction combined PacBio Iso-Seq data (Iso-Seq pipeline)
70 with open reading frames identified by TransDecoder (v5.5.0). For repeat annotation, both homology-based and
de novo approaches were used:
de novo identification with RepeatModeler (v2.0.3)
71, LTR retrotransposon prediction with LTR_Finder (v1.07)
72 and LTRharvest
73 and redundancy removal with LTR_retriever (v2.9.0)
74; these sequences were combined with RepBase
75 as a library for RepeatMasker (v4.1.2)
76, and tandem repeats were identified with Tandem Repeat Finder (v4.09)
77.
Gene structures were annotated by integrating homology-based (miniprot),
ab initio (GeneMark-ET v4.71
78, Augustus v3.3.3
79) and RNA-seq-based (HISAT2
69, StringTie v2.0.6
80, TransDecoder (v5.5.0) predictions with Evidence Modeler (v2.0.0)
81. Functional annotation was performed against Pfam, NR, SwissProt, TrEMBL
82 and KEGG with an
E-value threshold of 1 × 10⁻⁵; motifs, domains and Gene Ontology terms were assigned with InterProScan (v5.66-98.0)
83.
Gene family and evolutionary analysis
Genome-wide gene families were identified with OrthoFinder
84 for pyrethrum and eleven other plant species:
A. annua85,
A. thaliana86,
Amborella trichopoda87,
C. lavandulifolium88,
Chrysanthemum nankingense89,
Helianthus annuus90,
Mentha suaveolens91,
N. benthamiana92,
Taraxacum mongolicum93,
Jasminum sambac94 and
Vitis vinifera95. Genome and protein sequences were retrieved from public databases. A maximum-likelihood species tree was constructed with RAxML (PROTGAMMAWAG)
96 from single-copy orthologs shared by all twelve species. Gene-family expansion and contraction were analyzed with CAFE
97 under default settings using a birth–death model. KEGG enrichment was performed for pyrethrum-specific and expanded gene families, and divergence times were estimated with the mcmctree program in PAML
98.
WGD analysis
The synonymous substitution rate (
Ks) method was used to identify WGD events by plotting intra- and inter-species
Ks distributions. Protein sequences were aligned using BLAST and syntenic blocks detected with MCScanX
99. Paralogous and orthologous gene pairs within these blocks were used to calculate
Ks, the nonsynonymous substitution rate (
Ka), and the
Ka/
Ks ratio with the yn00 module in PAML
98, and density plots of
Ks values were generated.
Transcriptomic and co-expression analysis
Transcript expression was quantified with HISAT2
69 and FPKM (fragments per kilobase of transcript per million mapped reads) values were calculated from gene length and mapped read counts. A gene co-expression network was constructed with the R package WGCNA
100 from the expression of identified pathway genes. Functionally characterized genes (
TcCDS,
TcGLIP, and
TcJMH) and the tissue-specific acid-moiety and pyrethrin contents were used as targets for Pearson correlation screening (correlation coefficient ≥ 0.7,
p ≤ 0.05).
Metabolite analysis of T. cinerariifolium
To quantify pyrethrin-pathway intermediates, leaves, stems, roots and flowers were ground into powder in liquid nitrogen; 100 mg of powder was extracted with 1 mL ethyl acetate by sonication for 15 min. The homogenate was centrifuged at 13,000 × g for 10 min, and the supernatant was concentrated under a nitrogen stream to 200 µL. After a further centrifugation, the supernatant was used for GC-MS and LC-MS analysis.
For GC-MS, an Agilent 8890 GC coupled to a 7000E TQ-MS was used with a DB-5Q column (30 m × 0.25 mm × 0.25 μm). The temperature program was: 50 °C (1 min), 15 °C/min to 110 °C, 6 °C/min to 200 °C (hold 1 min), 10 °C/min to 280 °C (total 30.33 min). Injection volume: 1 μL; split ratio: 15:1; injection port: 240 °C; MS scan: 45–500 m/z. High-resolution GC-Orbitrap (Thermo Scientific GC 240 coupled to Orbitrap Exploris) was used for precise mass confirmation (resolution 60,000; mass tolerance 5 ppm).
For LC-MS, supernatants were filtered through a 0.22 μm PTFE syringe filter before injection on an Agilent 1290 Infinity LC coupled to a 6475 triple-quadrupole mass spectrometer with a C18 column (250 mm × 4.6 mm, 5 μm) and a 3 μL injection. The mobile phase: 0.1% formic acid in water (A) and acetonitrile with 0.1% formic acid (B); gradient: 10%→60% B (0–15 min), 60%→95% B (15–25 min), 95% B (25–32 min), 95%–10% B (32–32.1 min), 10% B (32.1–40 min). Flow rate: 0.4 mL/min; column temperature: 40 °C. TQ-MS operated in full-scan mode (50–500 m/z, 300 ms) in positive and negative ion modes. To detect chrysanthemyl-CoA compounds, tissues were ground in liquid nitrogen and extracted with 80% methanol/water, sonicated on ice for 30 min, and centrifuged at 10,000 × g for 10 min. The filtered extract was analyzed by high-resolution LC-MS/MS on a Thermo UPLC system coupled to a Q Exactive Plus with an ACQUITY UPLC BEH C18 column (100 mm × 2.1 mm, 1.7 μm) at 45 °C, using the same mobile phases with a linear gradient, flow rate 0.4 mL/min and 1 μL injection, in negative HESI mode (100–1500 m/z).
Identification and cloning of candidate genes
Candidate genes involved in pyrethrin biosynthesis were identified by Hidden Markov Model
29 searches using corresponding Pfam IDs (LOX: PF00305; AOC: PF06351; ACX: PF01756; KAT: PF02803/PF00108; ACS: PF00501; SCPL: PF00450; CYP450: PF00067; UGT: PF00201), and by BLAST against reported homologs; protein sequences were aligned with MAFFT (v7.475)
101. Phylogenetic trees were constructed using IQ-TREE
102 (see Supplementary Table S16 for the UGT tree). Full-length candidate genes were amplified from cDNA with Phanta Max Super-Fidelity (Vazyme) or PrimeSTAR GXL (Takara) using primers in Supplementary Table S15. For transient expression, PCR products were cloned into a modified pEAQ vector (
XhoI/
SalI) and transformed into
Agrobacterium tumefaciens GV3101; for heterologous expression in
E. coli, CDSs were cloned into pET28a and transformed into DH5α and then Rosetta cells. Genes codon-optimized for
S. cerevisiae were synthesized by GenScript Co., Ltd. (China).
Agrobacterium-mediated transient expression in N. benthamiana
Plasmids were transformed into A. tumefaciens GV3101 and grown on selective LB. Positive transformants were cultured, pelleted at 4,000 × g, resuspended in induction buffer (10 mmol/L MES, 10 mmol/L MgCl2, 100 µmol/L acetosyringone, pH 5.6) and incubated for 1–2 h; suspensions were adjusted to OD600 = 0.4–0.6 (mixed at equal concentrations for combinations) and infiltrated into the abaxial side of 4–5-week-old leaves (three leaves from three plants per mixture). For substrate co-infiltration, plants were maintained until 3 days post-infiltration; substrate (100 ppm in water with 1% DMSO) was infiltrated into the infected region. Leaves were harvested 1 day later, snap-frozen, and stored at −80 °C. This system was used throughout for in vivo pathway validation.
Heterologous expression and enzymatic assays in E. coli
Chemically competent Rosetta cells were transformed with pET28a constructs and selected on chloramphenicol and kanamycin. Cultures were subcultured until OD600 ≈ 0.6, and induced with 0.1 mmol/L IPTG at 18 °C overnight. Cells were harvested at 3,000 × g, resuspended in 50 mmol/L sodium phosphate (pH 8.0), and disrupted by sonication, then used directly or stored at −80 °C. For crude-enzyme reactions, 1 mmol/L MgCl2, 1 mmol/L substrate, and cofactors were added to 1 mL crude extract, incubated at 30 °C for 24 h, and quenched with 1 mL ethyl acetate. For purified TcCDS, TcTPS1, TcUGT6, and TcGLIP, lysates were purified on Ni2+-NTA resin (250 mmol/L imidazole elution) and checked by SDS-PAGE; standard 100 µL reactions contained 100 µg enzyme, 150 µmol/L substrate, 1 mmol/L MgCl2, 2 mmol/L DTT, 12.5% glycerol, 0.1% Tween-20 and 1 mmol/L ascorbic acid in 50 mmol/L Tris-HCl (pH 7.5) at 30 °C for 12 h with an empty-vector control, then quenched and analyzed by untargeted GC-MS, and kinetics were fitted with GraphPad Prism.
Chrysanthemoyl-glucoside (8) was synthesized enzymatically with purified recombinant TcUGT6. Chrysanthemic acid (2 mmol/L) was incubated with UDP-glucose (4 mmol/L) and purified TcUGT6 in 50 mmol/L Tris-HCl (pH 7.5) containing 1 mmol/L MgCl2 and 2 mmol/L DTT (total 1 mL) at 30 °C for 12 h. The reaction was quenched with methanol, desalted by C18 solid-phase extraction, and purified by semi-preparative reversed-phase HPLC (Eclipse XDB-C18, 9.4 mm × 250 mm, 5 μm). The target fraction was lyophilized and redissolved in methanol; the identity of 8 was confirmed by LC-MS/MS ([M+HCOOH−H]⁻, m/z = 375.1), and it was stored at −80 °C as the acyl donor for TcGLIP assays.
Protein crystal modeling
The three-dimensional structure of the target protein was predicted using AlphaFold3
103, and molecular docking was performed with AutoDock Vina 1.2.3
104. Generated poses were ranked by binding free energy (ΔG), and the top-ranked pose was selected. The binding mode of the protein-ligand complex was visualized in PyMOL 3.0.0 (pymol.org/pymol.html); the receptor was shown as a cartoon, key pocket residues and the ligand as sticks, and non-covalent interactions such as hydrogen bonds were annotated with PyMOL tools.
Modular engineering for heterologous biosynthesis in S. cerevisiae
The
S. cerevisiae strain JCR27
26, engineered to boost terpenoid production, was the parent for chrysanthemic acid, and strain 30000B for 3,7-ddh-JA (Supplementary Tables S13, S14). DNA was transformed by the LiAc/SS carrier DNA/PEG method
105, with mutants confirmed by colony PCR. For growth and heterologous production, colonies were grown in 4 mL synthetic drop-out medium (30 °C, 220 rpm) and transferred to 50 mL YPDG at OD
600 = 0.1 and incubated 3–5 days before extraction. For whole-cell biotransformation, cells grown to OD
600 ≈ 5.0 were pelleted at 3,000 ×
g, resuspended in 1 mL 50 mmol/L potassium phosphate (pH 7.4), and incubated with substrate at 30 °C, 220 rpm; products were extracted with 1 mL ethyl acetate and analyzed by GC-MS. Pathway genes were integrated module-by-module to achieve
de novo chrysanthemic acid synthesis, followed by precursor feeding with
cis-jasmone to produce jasmolin I, following established procedures
27.
The Author(s) 2026. Published by Higher Education Press. This is an Open Access article distributed under the terms of the CC BY license (https://creativecommons.org/licenses/by/4.0/).