Introduction

Natural products play pivotal roles in chemical biology and drug discovery due to their unique chemical structures and specific biological targets1. The construction of natural product analogs is important for exploring structure-activity relationships, developing drugs with enhanced bioactivity and bioavailability, and uncovering disease mechanisms as chemical biology tools2. With the advances in synthetic biology, combinatorial biosynthesis exploiting enzymatic and pathway promiscuity provides an alternative approach to generate novel natural product analogs2,3,4,5.

Meroterpenoids represent a class of natural products with conspicuous biological activities derived from the terpenoid biosynthetic pathway coupled with other biogenic pathways. Among these, cannabinoids stand out as unique meroterpenoids initially isolated from the plant Cannabis sativa, a herbal medicine with a history spanning over 3500 years6. More than 120 phytocannabinoids have been isolated from Cannabis, most of which are alkyl resorcinol derivatives with a monoterpenoid unit7. Notably, cannabinoids, such as Δ9-tetrahydrocannabinol (THC) and cannabidiol (CBD), exhibit potent pharmacological effects, including analgesic, antiemetic, anti-spasmodic, antimicrobial, and anti-inflammatory activities8,9,10, which renders them of great medicinal importance11. The pharmacological activities of cannabinoids principally occur through interacting with cannabinoid receptors CB1 and CB2, belonging to the superfamily of G-protein coupled receptors which mediate most of physiological responses and so have great potential as therapeutic targets for a broad spectrum of diseases12,13. The important influence of the C-3 side chain of cannabinoids on the pharmacological activities has been identified through chemical synthesis of a series of C-3 alkyl analogs14,15,16, however, the effect of the prenyl chain on pharmaceutical properties remains unexplored.

The biosynthesis of cannabinoids is derived from a monoterpenoid unit produced by 2-C-methyl-D-erythritol 4-phosphate (MEP) pathway and an aromatic moiety olivetolic acid (OA) catalyzed by an olivetol synthase and an OA cyclase using malonyl-CoA and hexanoyl-CoA as substrates17. Cannabigerolic acid (CBGA) is the committed biosynthetic precursor of cannabinoids, formed by a C-C Friedel-Craft alkylation of OA with geranyl diphosphate (C10, GDP)18. Subsequently, the geranyl moiety of CBGA undergoes cyclization reaction to form cannabinoids such as Δ9-tetrahydrocannabinolic acid (THCA), cannabidiolic acid (CBDA), or cannabichromenic acid (CBCA), which can be decarboxylated to their neutral forms (Fig. 1)19.

Fig. 1
figure 1

The biosynthetic pathway of cannabinoids.

Full size image

Heterologous biosynthesis of cannabinoids has been achieved in engineered yeasts, paving an alternative way for access to cannabinoids20,21,22. Based on enzyme promiscuity and engineered biosynthetic pathways, a series of unnatural cannabinoid analogs with distinct alkyl moieties at the C-3 position were obtained by combinatorial biosynthesis through feeding different fatty acids20,23. Given the diversity of terpenoids which includes hemiterpenoids, monoterpenoids, sesquiterpenoids, diterpenoids, and sesterterpenoids, we are particularly interested in the biosynthesis of CBGA analogs with distinct terpenoid chains. Several aromatic prenyltransferases have been reported, among which AscC, a UbiA-type prenyltransferase from Fusarium sp. NBRC 100844, could catalyze the conversion of 2-methyl-4,6-dihydroxy benzaldehyde into the corresponding farnesylated product24. In this work, we synthesized four CBGA analogs 1–4 with a C5, C10, or C15 prenyl unit using the engineered Escherichia coli harboring the AscC gene. Molecular docking and site-directed mutagenesis resulted in two mutants selectively producing the undescribed compoud 4. In addition, the anti-neuroinflammatory and antibacterial activities of these compounds were observed. 

Results

Heterologous expression of promiscuous aromatic prenyltransferases in engineered E. coli strains for cannabinoid analogs production

To biosynthesize cannabinoid analogs featuring diverse prenyl side chains, the engineered E. coli strains were recruited, which contain an exogenous MVA pathway along with a farnesyl diphosphate synthase (FDPS) or geranylgeranyl diphosphate (GGDPS) or geranylfarnesyl diphosphate synthase (GFDPS) previously constructed for terpenoid biosynthesis25. The aromatic prenyltransferase (CsPT4) from C. sativa and four fungal promiscuous aromatic prenyltransferases (aPTases), including AscC from Fusarium sp., Trt2 and AtaPT from Aspergillus terreus, and Orf2 from Streptomyces sp.24,26,27,28,29. were picked out to test their activity for transferring various prenyl groups to the aromatic ring of OA (Supplementary Fig. 1). The full-length cDNA of these aPTases was synthesized after codon optimization and then constructed into the corresponding expression vector.

The four fungal aromatic prenyltransferases could express substantial amounts of recombinant proteins with expected molecular mass in the soluble fraction. These aPTases were subsequently transformed into the engineered E. coli strains producing C5-C25 prenyl diphosphates. By feeding with OA, four specific peaks could be detected in the extract of engineered E. coli containing AscC using high-performance liquid chromatography (HPLC) and liquid chromatography-mass spectrometry (LC-MS) analysis (Fig. 2). The pseudo-molecular ion peaks ([M-H]–) of specific products were 291.1601, 359.2226, 427.2853, and 427.2851, indicating that AscC could catalyze the introduction of C5, C10, and C15 prenyl chains into OA (Fig. 2). The conversion rate of OA was 70.48% by AscC. By contrast, no obvious product was observed in the engineered bacteria harboring Trt2, AtaPT, and Orf2. In addition, we tested the catalytic activity of AscC towards a series of OA analogs (S1-S7), but no products were observed (Supplementary Fig. 2). Compared to OA, these analogs lack a hydroxyl group at the C-4 position, except S4 that has an ester group at the C-1 position instead of a carboxyl group. These results suggest that both the C-4 hydroxyl group and the C-1 carboxyl group in the substrate are important for enzyme activity of AscC.

Fig. 2: The conversion of olivetolic acid (OA) to prenylation products by engineered E. coli harboring AscC.
figure 2

A HPLC analysis of the extract of engineered bacteria harboring pET28a/AscC (a) and empty vector pET28a (b). B The mass spectrometry of peaks 1–4.

Full size image

Identification of cannabinoid analogs produced by AscC in engineered E. coli

The plasmid pET28a/AscC was transferred into E. coli strain BL21 harboring the pBbA5c-MevT-MBIS plasmid, which contains a complete gene set of the mevalonate (MVA) pathway and a FDPS gene30, resulting in an engineered E. coli strain BPA1. To elucidate the structure of cannabinoid analogs, a 10 L scale fermentation of the engineered BPA1 strain was performed and extracted with ethyl acetate. The E. coli extracts were further purified with a combination of repeated silica-gel and Sephadex LH-20 column chromatography. Afterward, semi-preparative HPLC was carried out using a mixture of acetonitrile and acid water as a mobile phase to afford compounds 1–4. Notably, compound 4 emerged as the predominant component (Fig. 2).

Compound 4 was obtained as a yellowish powder and characterized by a molecular formula of C27H40O4 with eight degrees of unsaturation based on high-resolution electron ionization mass spectroscopy (HR-EI-MS) (m/z 427.2851 [M-H]-, calcd. 427.2854) and NMR spectroscopic data (Supplementary Figs. 3–10). Compound 4 showed the same molecular weight as compound 3 but a different retention time, implying that they might be prenylated products of OA at different positions. Compound 3 was determined as a cannabinoid analog 2,4-dihydroxy-3-[(2E,6E)-3,7,11-trimethyl-2,6,10-dodecatrien-1-yl]-6-pentylbenzoic acid through careful inspection of the NMR data (Supplementary Figs. 11, 12) and comparison with those reported in the literature31. The 1D and 2D NMR spectra of 4 closely resembled those of 3 with a key difference being the upfield movement of the carbon at C-3 (δC 102.2) and C-1′′ (δC 31.5) in 4 compared to 3 (δC 111.1 for C-3 and δC 36.7 for C-1′′). Additionally, the carbon at C-5 (δC 119.9) and C-1′ (δC 25.1) in 4 moved downfield compared to 3 (δC 111.6 for C-5 and δC 22.2 for C-1′) indicating a C15 farnesyl chain attached to C-5. Confirmation of this attachment was provided by the 1H-1H correlation spectroscopy (COSY) correlations of H-1′/H-2′, H-4′/H-5′/H-6′ as well as heteronuclear multiple bond correlation (HMBC) correlations of H-1′ with C-4/C-5/C-6/C-2′/C-3′ (Fig. 3). The 1H-1H COSY correlations of H-1′′/H-2′′/H-3′′/H-4′′/H-5′′, combined with the HMBC correlations of H-1′′ with C-1/C-5/C-6, provided evidence for the attachment of the n-pentyl chain to C-6 and confirmed the planar structure of 4 (Fig. 3). Accordingly, the structure of 4 was determined as 2,4-dihydroxy-5-[(2E,6E)-3,7,11-trimethyl-2,6,10-dodecatrien-1-yl]-6-pentylbenzoic acid (Fig. 3). The structures of compounds 1 and 2 were identified as 2,4-dihydroxy-3-(3-methylbut-2-en-1-yl)-6-pentylbenzoic acid32 and CBGA33 (Supplementary Figs. 13–16), respectively, through comparison of their NMR data with those reported in the literature (Fig. 3). Compounds 1–4 are cannabinoid analogs, in which compound 4 is an undescribed compound. The results indicated that engineered E. coli harboring AscC could produce four cannabinoid analogs with a C5, C10, or C15 prenyl chain, likely attributed to the spacious hydrophobic substrate-binding pocket of promiscuous aPTases34.

Fig. 3
figure 3

Chemical structures of the products produced by engineered E. coli harboring AscC and key 1H-1H COSY and HMBC correlations of compound 4.

Full size image

Compounds 1–4 displayed potent anti-inflammatory and antibacterial activities

Cannabinoids are known for their potent anti-inflammatory and immunosuppressive properties, likely through interacting with G protein-coupled receptors that are closely related to immune regulation and neurodegeneration12,13,35. Therefore, four cannabinoid analogs were evaluated for their anti-neuroinflammatory activity by investigating the inhibition of TNF-α and IL-6 secretion in BV-2 microglial cells. As shown in Table 1, compound 4 showed potent inhibitory effects on TNF-α and IL-6 secretion, with IC50 values of 3.06 and 4.31 µM, respectively. Compounds 1–3 showed moderate activity against TNF-α and IL-6 secretion with 34.29–48.17% inhibition rates at 20 µM. By comparing the structures and activity of compounds 1–4, it can be concluded that the C15 farnesyl unit at the C-5 position of OA played an important role in anti-inflammatory activity.

Table 1 Inhibitory effects of compounds 1–4 on TNF-α and IL-6 secretion of BV-2 microglial cells induced by LPS
Full size table

Considering the reported broad-spectrum antimicrobial activity of many cannabinoids, we aimed to assess the antimicrobial potential of compounds 1–4. Three Gram-positive bacteria (Staphylococcus aureus, Bacillus subtilis, and Micrococcus luteus), one Gram-negative bacteria (E. coli), and one fungus (Trichophyton rubrum) were recruited. As shown in Table 2, compounds 1–4 exhibited specific inhibitory activities against Gram-positive bacteria but were inactive against E. coli and T. rubrum. The EC50 values of compounds 1–4 against S. aureus, B. subtilis, and M. luteus ranged from 0.87 to 35.13 µM. The substrate OA showed no activity, emphasizing the importance of prenylation for antibacterial activity. Notably, compound 4 exhibited the most potent antibacterial activity among these cannabinoid analogs, with EC50 values of 0.87–3.16 µM, suggesting that the C15 farnesyl group at the C-5 position of OA also contributed to the antibacterial property against Gram-positive bacteria.

Table 2 Antimicrobial effects of compounds 1–4
Full size table

Improved production of compounds 1–4 in E. coli

We analyzed the production of four compounds in E. coli strains carrying pBbA5c-MevT-MBIS and pET28a/AscC using an external standard method. HPLC analysis showed that the yields of compounds 1–4 were 0.12 ± 0.02 mg/L, 0.08 ± 0.01 mg/L, 0.38 ± 0.05 mg/L, and 7.31 ± 0.60 mg/L, respectively. Notably, a significant amount of unconsumed prenyl substrate FDP was observed in the engineered E. coli, prompting an attempt to enhance the yield of compounds 1–4 by increasing the copy numbers of the AscC gene. For this purpose, an additional plasmid pCold-TF/AscC was transformed into the E. coli harboring pBbA5c-MevT-MBIS and pET28a/AscC while the pCold-TF empty vector was co-transformed into the same strain as a negative control. HPLC analysis showed that the yields of compounds 1–4 significantly improved to 2.06 ± 0.10 mg/L, 1.63 ± 0.12 mg/L, 0.91 ± 0.09 mg/L and 14.85 ± 0.91 mg/L respectively in the strain containing two AscC plasmids, which were 17.85, 20.80, 2.42 and 2.03-fold of those in the strain containing plasmid pET28a/AscC, respectively (Fig. 4). The results indicated that increasing AscC copy numbers led to a substantial increase in the yields of compounds 1–4. Different yields of compounds 1–4 may be attributed to the availability of various prenyl diphosphates, as FDP was the predominant product in the engineered E. coli containing the pBbA5c-MevT-MBIS plasmid. However, the production of compounds 1 and 2 showed a significantly higher increase compared to compounds 3 and 4 after introducing an additional copy of AscC, suggesting that the enzymatic efficiency of AscC has a greater impact on the product profiles than the availability of prenyl diphosphates. Although prenylation at either the C-3 or C-5 position of OA occurs when using FDP as a prenyl donor, the yield of compound 4 was significantly higher than that of compound 3. These results suggest that prenylation reactions catalyzed by AscC exhibit a regioselective preference for the ortho position of the n-pentyl group over the two electron-donating hydroxyl groups on the aromatic ring.

Fig. 4: The yields of compounds 1–4 in engineered E. coli harboring AscC. Co-transformation pET28a/AscC and pCold-TF empty vector in engineered E. coli containing the pBbA5c-MevT-of MBIS plasmid served as a control for determining the effect of AscC gene copy on product yields. Four biologically independent experiments were performed.
figure 4

The error bars represent the standard deviation (S.D.) of the mean.

Full size image

Engineering AscC for selective production of compound 4

Given that four different cannabinoid analogs are produced by AscC, we are motivated to engineer the enzyme to improve its product specificity. The AscC protein model contained a sizable central cavity capable of accommodating diverse prenyl substrates (Fig. 5A). In the model of AscC docked with structurally geometry-optimized FDP and OA, we found 14 amino acid residues interacting with the bound substrates (Fig. 5B). The residues Y162, K166, and D226 located in the conserved motifs (Y162XXXK166 and D226XXXD230) and R110 interacted with the pyrophosphate group of FDP were excluded because of their crucial roles for enzyme activity36. Then, computational alanine scanning on the residues within 3 Å of OA or FDP was performed, and the amino acids with binding energy exceeding 0.4 were selected as candidates for further investigation (Supplementary Table 1). The in silico prediction showed that mutating V215 to alanine would result in the destabilization of the protein when FDP served as the lignan. Mutating D219 and L176 to alanine was predicted to increase protein stability and lead to substantial energy alterations, but mutation of D219A would disrupt the essential hydrogen bond between D219 and FDP. Mutating Y309 to alanine would lead to a significant change in binding energy and enhance protein stability, but the π-alkyl interaction between Y309 and FDP would be disrupted. Mutations L180A and Y222A exhibited a neutral effect and moderate energy changes. Therefore, we obtained six variants by site-directed mutagenesis, including L176F, L176Y, L180F, L180Y, Y222L, and Y222F, based on ΔΔG changes to improve the thermal stability of the protein. These mutants were individually transformed into E. coli harboring the pBbA5c-MevT-MBIS plasmid.

Fig. 5: Analysis of crucial residues in AscC.
figure 5

A Overview of AscC model docked with FDP and OA. B Interactions between the substrates and individual residues in AscC. Yellow, blue, red, green, orange, violet, and black dash lines represent hydrogen bonds, alkyl interactions, π-alkyl interactions, carbon interactions, charge interactions, π-cation interactions, and donor-donor interactions, respectively. The visualization was generated using Pymol. C Compound yields of wild-type AscC and its mutants expressed in the pET28a vector in engineered E. coli containing the pBbA5c-MevT-MBIS plasmid. **P < 0.01vs WT. Five biologically independent experiments were performed. The error bars represent the standard deviation (S.D.) of the mean. D–G Interactions between the substrates and residues in AscC and its mutant L180F.

Full size image

Interestingly, the variants L180Y and L180F selectively utilized FDP as a prenyl donor and specifically catalyzed the isopentenyl reaction of OA at the C-5 position to produce compound 4 (Fig. 5C). We constructed the protein model of L180F using AlphaFold 2 and docked the substrates OA and FDP into the binding pocket. OA was observed to exhibit different conformations in the wild-type and mutant enzymes (Fig. 5E). In L180F, there were more interactions between the residues R41, G47, and R90 and the carboxyl or 2-hydroxyl group of OA (Fig. 5F, G), which likely reduced the binding flexibility of OA within the pocket. Moreover, a π-alkyl interaction between Y222 and the alkyl side chain of OA in L180F replaced the alkyl-alkyl interaction between M50 and the alkyl side chain of OA in AscC. Additional π-alkyl interactions between OA and FDP were also observed. Accordingly, the positioning of OA was likely stabilized within the enzyme pocket, facilitating regioselective Friedel-Craft alkylation at C-5 (Fig. 5D). In addition, the yields of compound 4 in variants L180Y and L180F reached up to 11.71 and 12.95 mg/L, respectively. The switch from an alkyl-alkyl to a π-alkyl interaction between the amino acid at position 180 and FDP, coupled with the additional π-alkyl interaction between Y218 and FDP, may provide enhanced cation stabilization. Besides, the reduced distance between C-5 and C-1 of FDP (from 6.7 to 6.2 Å) might also contribute to the improved conversion efficiency in variants L180Y and L180F. Furthermore, the natural substrates 2-methyl-4,6-dihydroxy benzaldehyde and FDP were also docked into the wild-type and mutant enzymes, respectively. The results suggested that L180F might still catalyze Friedel-Craft alkylation at the C-3 position, probably attributed to the methyl group at the C-1 position facilitating substrate flexibility within the enzyme pocket (Supplementary Fig. 17). However, the experiment was not conducted due to the unavailability of the substrate.

The variants L176Y and L176F resulted in a substantial loss of enzymatic activity (Fig. 5C). Y222 interacted with FDP via π-alkyl interactions, and the mutants Y222L and Y222F exhibited a dramatic reduction in enzymatic activity, with 21% and 0.4% production of wild-type AscC, respectively. The residues R41, G47, and R90 were predicted to form hydrogen bonds or carbon interaction with OA, suggesting they play an important role in the binding and stabilizing of OA. This was confirmed by the substantial loss of enzymatic activity upon mutation R90E (Fig. 5C). The results highlighted the importance of these residues in enzyme catalysis.

Discussion

The development of natural product analogs holds significant promise in chemical biology and drug discovery. Cannabinoids represent a unique class of natural products with substantial pharmaceutical values. Notable efforts have focused on the generation of cannabinoid analogs with distinct alkyl moieties, such as combinatorial biosynthesis of unnatural cannabinoid analogs through feeding different fatty acids20,23. In this study, cannabinoid analogs with varied terpenoid chains were created by introducing a promiscuous fungal aromatic prenyltransferase AscC into engineered bacteria strains capable of terpenoid production. The most abundant product, compound 4, is an undescribed compound featuring a C15 prenyl chain at the C-5 position of olivetolic acid.

Considering the steric hindrance of the n-pentyl group and the electron-donating effect of the hydroxyl groups in the aromatic ring, prenylation of OA would theoretically favor the C-3 position over the C-5 position. However, when using farnesyl diphosphate (FDP) as the prenyl donor, the amount of compound 4 prenylated at the C-5 position was significantly higher than that of compound 3 at the C-3 position. Notably, C-5 substituted cannabinoids are rare, with the exception of 5-geranyl olivetolate isolated from Helichrysum umbraculigerum37. A number of aPTases responsible for catalyzing the attachment of prenyl groups to aromatic substrates have been reported. For instance, UbiA-type PTases CsPT1 and CsPT4 reported from Cannabis catalyze the attachment of a geranyl group to OA20, and NphB (formerly Orf2) identified from Streptomyces sp. strain CL190 could catalyze the attachment of a geranyl group to aromatic substrates28. Recently, the core cannabinoid biosynthetic enzymes have also been identified from H. umbraculigerum, including four aPTases HuPT1-4 among that HuPT4 (HuCBGAS4) geranylated the aromatic dihydrostilbenic acid38. Notably, these identified aPTases predominantly catalyze prenylation at the C-3 position of aromatic substrates20,31,39. By contrast, AscC can catalyze OA to produce farnesylated products at both the C-3 and C-5 positions, suggesting a distinct catalytic mechanism. The unusual regioselectivity of AscC warrants further investigation, and the rise of artificial intelligence and language modes will help enzyme engineering in future40. The unique structure of compound 4 presents significant challenges for regioselective chemical synthesis. Guided by structure-based enzyme engineering, we successfully improved the selectivity of AscC, and developed two variants, L180Y and L180F, which exclusively produce compound 4. These findings provide promising catalysts for manipulating the production of compound 4 using synthetic biology.

We found that compounds 1–4 exerted significant anti-neuroinflammatory and antimicrobial properties, with compound 4 being the most effective. This indicates that the prenyl side chains at the C-5 position of OA play crucial roles in these activities. Additionally, novel cannabinoids with remarkable bioactivities could potentially be generated via prenyl cyclization. For instance, CBD demonstrates enhanced anti-inflammatory and antimicrobial activity compared to CBGA41,42. Further efforts should focus on exploring more cyclases and expression systems to expand the chemical space of cannabinoids.

Methods

General experimental procedures

Column chromatography was performed on silica gel (200 − 300 mesh, Qingdao Dingkang Silicone Co., Ltd., P. R. China) and Sephadex LH-20 (20 − 100 µm, Amersham Pharmacia Biotech, Sweden). HPLC-DAD analysis was carried out on an Agilent 1260 Infinity II equipped with a Zorbax SB-C18 (5 µM, 4.6 × 250 mm, 1 mL/min). Semi-preparative HPLC analysis was performed on a HITACHI Primaide series instrument with a Shimadzu Shim-pack GIST-C18 column (5 µm, 10 × 250 mm, 3 mL/min). Mass spectra were obtained on a Waters Synapt XS spectrometer equipped with a Zorbax Extend-C18 column (1.8 µm, 4.6 × 50 mm, 0.3 mL/min). NMR spectra were recorded on a Bruker AV-700 (700 MHz) spectrometer, with TMS as an internal standard. IR spectra were recorded on a Bruker Tensor-27 spectrometer with KBr pellets. Genes were synthesized by Tsingke Biotechnology Co., Ltd.. OA was purchased from J&K Scientific Co., Ltd.. The ethyl acetate used for extraction was analytical grade and acetonitrile used for HPLC was chromatographic grade. Other reagents were analytical grade.

Strains, plasmids and media

All primers used in this study were listed in Supplementary Table 2. E. coli DH5α was used for plasmid maintenance and E. coli BL21 (DE3) was used for gene heterologous expression. Plasmid pBbA5c-MevT-MBIS was kindly provided by Prof. Tao Liu at Tianjin Institute of Industrial Biotechnology, Chinese Academy of Sciences. Codon-optimized CsPT4 from C. sativa, AscC from the fungus Fusarium sp., Trt2 and AtaPT from Aspergillus terreus, and Orf2 from Streptomyces sp. were subcloned into NdeI and XhoI sites of expression vector pET28a using ClonExpress® Ultra One Step Cloning Kit (Vazyme Bio.), generating corresponding plasmids. Similarly, the cDNA of AscC was inserted into SalI and XhoI sites of expression vector pCold-TF, resulting in the plasmid pCold-TF/AscC. The recombinant plasmids were individually transformed into E. coli expression strain BL21(DE3) that harbored the pBbA5c-MevT-MBIS vector to generate the corresponding strains. pET28a empty vector was transformed into the same strain as a negative control. The plasmid pCold-TF/AscC was co-transformed with pET28a/AscC into E. coli strain BL21 harboring pBbA5c-MevT-MBIS to generate the strain BPA3. pCold-TF empty vector was transformed into the same strain as a negative control (BPA4). All strains constructed in this study were listed in Supplementary Table 3. The culture media used for E. coli strains were Luria-Bertani (LB) medium (10 g/L tryptone, 5 g/L yeast extract, 10 g/L NaCl) or Terrific Broth (TB) medium (12 g/L tryptone, 24 g/L yeast extract, 4 mL/L glycerol, 2.31 g/L KH2PO4, 12.54 g/L K2HPO4). Ampicillin sodium (100 µg/mL), chloramphenicol (34 µg/mL), and kanamycin (50 µg/mL) were supplemented to media for screening positive recombinant strains.

Product assay of engineered E. coli

The colonies of strains constructed in this study were screened using LB solid plates containing corresponding antibiotics and further confirmed by colony-PCR. Seed cultures (10 mL) of confirmed transformants were grown overnight at 37 °C in TB medium supplemented with corresponding antibiotics. 1 mL of seed culture was inoculated into 50 mL fresh TB broth supplemented with the same antibiotics and cultured at 37 °C until the cultures reached the optical density (OD600) of 0.6–0.8. Subsequently, 50 µL of 0.3 mM isopropyl-β-D-thiogalactopyranoside (IPTG) was added to induce the expression of the target proteins at 16 °C for 16 h. Afterward, 0.25 mg substrates (OA, compound S1-S7: salicylic acid, acetylsalicylic acid, ginkgoneolic acid, 2,4-dihydroxy-6-pentylbenzoate, 2-hydroxy-6-methylbenzoic acid, 2-methoxy-4-methylbenzoic acid or 4-ethyl-3-hydroxybenzoic acid) dissolved in anhydrous ethanol (10 mg/mL) was added to the cultures which were incubated at 25 °C for 72 h. Subsequently, the cultures were extracted with ethyl acetate by ultrasound (100 mL× 3, 30 mins per time) and the solvents were then removed by rotary evaporators. The residues were dissolved in 1 mL methanol and filtered for HPLC and LC-MS analysis. The mobile phase was composed of acetonitrile (solvent A) and 0.2% formic acid in water (solvent B), at a flow rate of 1.0 mL/min and a column temperature of 30°C. The elution gradient was as follows: 0–40 min, 5–75% A; 40.01–50 min, 75–95% A; 50.01–55 min, 95% A (for OA); while 0–5 min, 5% A; 5.01–30 min, 5–100% A; 30.01–35 min, 100% A (for S1-S7). The injection volume was 20 µL. LC-MS was performed under the same condition of HPLC, while electrospray ionization mass spectrometry was used in the negative ion mode with a capillary voltage of 3500 V.

Isolation and structural elucidation of compounds produced by engineered E. coli

A total of 10 L engineered E. coli cells harboring pET28a-AscC and pBbA5c-MevT-MBIS were cultivated. The cultures were extracted three times with ethyl acetate and the organic phase was combined and concentrated in vacuo. The crude extract (7 g) was subjected to silica gel column chromatography and eluted with petroleum ether-EtOAc (100:0, 10:1, 5:1, 2:1 and 0:1, v/v) to obtain five fractions (Fr.1-Fr.5). Fr.3 (1.2 g) was subjected to Sephadex LH-20 and eluted with CH2Cl2/MeOH (1:1, v/v), and then purified by semipreparative HPLC using MeCN/H2O (1:1, v/v) as a mobile phase to yield compound 1 (6 mg, tR = 37.0 min), 2 (3 mg, tR = 44.8 min), 3 (2 mg, tR = 51.1 min) and 4 (32 mg, tR = 52.5 min). Afterward, the compounds were subjected to extensive NMR analysis.

Analytical data of compounds 1–4

Compound 1: Yellowish powder; 1H NMR (700 MHz, CD3OD): δ 6.18 (s, 1H), 5.21 (t, J = 6.9 Hz, 1H), 3.26 (d, J = 7.2 Hz, 2H), 2.88 (d, J = 7.7 Hz, 2H), 1.76 (s, 3H), 1.65 (s, 3H), 1.55 (m, 2H), 1.33 (m, 4H), 0.90 (t, J = 6.7 Hz, 3H); 13C NMR (150 MHz, CD3OD): δ 175.8, 164.2, 160.5, 146.7, 131.6, 124.3, 113.8, 110.6, 104.5, 37.4, 33.3, 32.9, 26.0, 23.6, 22.9, 17.9, 14.4.

Compound 2: Yellowish powder; 1H NMR (700 MHz, CDCl3): δ 6.27 (s, 1H), 5.28 (t, J = 6.6 Hz, 1H), 5.05 (t, J = 6.4 Hz, 1H), 3.43 (d, J = 7.1 Hz, 2H), 2.88 (t, J = 7.9 Hz, 2H), 2.11 (m, 2H), 2.06 (m, 2H), 1.81 (s, 3H), 1.67 (s, 3H), 1.59 (s, 3H), 1.56 (m, 2H), 1.33 (m, 4H), 0.89 (t, J = 6.7 Hz, 3H); 13C NMR (150 MHz, CDCl3): δ 175.4, 163.7, 160.6, 147.4, 139.4, 132.2, 123.9, 121.5, 111.6, 111.3, 103.4, 39.9, 36.7, 32.2, 31.6, 26.5, 25.8, 22.7, 22.2, 17.9, 16.4, 14.2.

Compound 3: Yellowish powder; 1H NMR (700 MHz, CDCl3): δ 6.24 (s, 1H), 5.27 (t, J = 7.2 Hz, 1H), 5.08 (m, 2H), 3.39 (m, 2H), 2.87 (m, 2H), 2.11 (m, 2H), 2.05 (m, 4H), 1.96 (m, 2H), 1.81 (s, 3H), 1.67 (s, 3H), 1.59 (s, 3H), 1.58 (s, 3H), 1.55 (m, 2H), 1.29 (m, 4H), 0.88 (t, J = 7.1 Hz, 3H); 13C NMR (150 MHz, CDCl3): δ 174.9, 163.6, 160.3, 147.2, 139.2, 135.7, 131.4, 124.5, 123.8, 121.5, 111.6, 111.1, 103.6, 39.9, 39.8, 36.7, 32.2, 31.6, 26.8, 26.5, 25.8, 22.7, 22.2, 17.8, 16.4, 16.2, 14.2.

Compound 4: Yellowish powder; [α]20D = - 5.13 (c 0.04 M in MeOH); 1H NMR (700 MHz, CDCl3): δ 6.33 (s, 1H), 5.08 (m, 3H), 3.35 (d, J = 6.6 Hz, 2H), 2.96 (t, J = 8.3 Hz, 2H), 2.11 (m, 2H), 2.04 (m, 4H), 1.96 (m, 2H), 1.81 (s, 3H), 1.67 (s, 3H), 1.59 (s, 6H), 1.54 (m, 2H), 1.37 (m, 4H), 0.92 (t, J = 7.1 Hz, 3H); 13C NMR (150 MHz, CDCl3): δ 176.1, 163.9, 161.1, 147.4, 137.9, 135.6, 131.5, 124.5, 123.8, 122.4, 119.9, 104.7, 102.2, 39.8, 39.8, 32.6, 31.5, 31.0, 26.8, 26.5, 25.8, 25.1, 22.6, 17.8, 16.5, 16.2, 14.2 (see Supplementary Table 4); IR (KBr) νmax: 3389, 2927, 1631, 1608, 1449, 1252, 1178 cm−1; UV (MeOH): λmax 200 nm; HR-ESI-MS (m/z): [M-H]- calcd. for C27H39O4, 427.2854; found, 427.2851.

Modeling and molecular docking simulations

The alphafold model of AscC and AscCL180F was generated using web service (https://wemol.wecomput.com/). Hydrogen atoms and atom charges were subsequently added. The substrates FDP and OA undergoing structural optimization were docked into the binding pocket of AscC and AscCL180F using the AutoDockTools-1.5.6. The docking results were then visualized and analyzed by PyMOL.

Site-directed mutagenesis of AscC

Alanine scanning was used to identify candidate residues with a Radius within 3 Å, while affinity maturation was employed to design mutants with a focus on those exhibiting a binding energy change exceeding 0.4. Mutants were constructed using overlap extension PCR and then subcloned into the pET28a vector. The primers were listed in Supplementary Table 2. Each mutant was confirmed by DNA sequencing. Afterward, the resultant variants were transformed into E. coli BL21 (DE3) harboring pBbA5c-MevT-MBIS.

Qualitative analysis of the products in engineered E. coli strains

Calibration curves of compounds 1, 2, and 4 were generated using HPLC following the method described above. Five concentration gradients of the isolated compound 1 (80, 40, 20, 10, 5 µg/mL), compound 2 (60, 30, 15, 7.5, 3.75 µg/mL) and compound 4 (500, 250, 125, 62.5, 31.25 µg/mL) were prepared. Each concentration was injected in triplicate, and the linearity of the standard curve was determined by plotting the peak area versus concentration. The resulting equation and correlation coefficient obtained from the linearity study were y = 19252x + 55.984 (R2 = 0.9981) for compound 1, y = 13688x + 43.024 (R2 = 0.9973) for compound 2, and y = 15153x + 192.93 (R2 = 0.9993) for compound 4. The yields of compound 3 were calculated according to the calibration curve of compound 4.

Anti-inflammatory activity assay of compounds 1–4

The assay was performed according to the literature with certain modifications43. Briefly, BV-2 microglia cells were cultured in DMEM high-glucose medium containing 10% fetal bovine serum, penicillin (100 U/mL), and streptomycin (100 µg/mL) at 37 °C in a humidified atmosphere with 5% CO2 and 95% air. BV-2 cells (2×105 cells/well) were seeded in 96-well plates and cultured overnight until reaching 70–90% confluence. Subsequently, the cells were pretreated with different concentrations of test compounds (80, 40, 20, 10, and 5 µM) for 1 h, followed by exposure to 1 µg/mL of LPS for 24 h. The levels of TNF-α and IL-6 in the culture supernatants were assessed using enzyme-linked immunosorbent assay (ELISA) kits following the manufacturer’s protocol (BD Biosciences). The absorbance value (A) of each well at 450 nm (570 nm calibration) was measured with a microplate reader (Molecular Devices). Inhibitory rates of TNF-α and IL-6 production were calculated according to the following formula: Inhibitory rate = (1 – (Asample – Ablank)/(Asolvent – Ablank)) × 100%. IC50 values were analyzed by log (inhibitor) vs. normalized response – Variable slope using Prism 9.0 software. Each experiment was conducted with three independent biological replicates.

Antimicrobial activity assay of compounds 1–4

T. rubrum ATCC4438 was bought from Hospital for Skin Diseases, Institute of Dermatology Chinese of Medical Sciences, Peking Union Medical College. E. coli ATCC25922 was bought from HuanKai Microbial Co., Ltd. S. aureus, B. subtilis, and M. luteus were bought from Institute of Resource Insects, Chinese Academy of Forestry Sciences. The antimicrobial assays of compounds 1–4 were conducted using a micro-broth dilution method in 96-well culture plates according to the Clinical and Laboratory Standards Institute (CLSI). Compounds 1–4 were prepared as a 20 mM stock solution in dimethyl sulfoxide (DMSO). Ampicillin sodium and cefotaxime sodium were used as the positive agents for the antibacterial assay of Gram-positive bacteria and Gram-negative bacteria, respectively, while terbinafine hydrochloride was used as the positive agent for antifungal activity. The four bacterial strains were cultured on LB agar plates, whereas the fungal strain was cultured on the potato-dextrose agar (PDA) plates. The colonies were taken from the agar plates and incubated in the corresponding medium. The microbe suspension was deliquated with fresh LB or potato-dextrose broth (PDB) to 105 cfu/mL. The diluted suspensions were then transferred to a 96-well plate (200 µL per well) containing different concentrations of the test compound. Serial dilutions of the stock solution were performed with final concentrations ranging from 50 to 0.39 µg/mL (including 50, 25, 12.5, 6.25, 3.125, 1.5625, 0.78125, 0.390625 µg/mL). The microplates were incubated at 37 °C for 24 h for the bacteria or at 28 °C for 96 h for the fungi. All assays were performed in quardruplicate. EC50 was defined as the minimum concentration inhibiting 50% of microbial growth and was calculated using GraphPad Prism 9.0 software.

Statistics and reproducibility

We calculated the compound yields synthesized by the engineered strains using peak areas and the calibration curves derived from purified compounds. The compound yields for each strain were measured in triplicate biological replicates, with each replicate originating from different positive clones.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.