Identification of the oxepinone-forming enzyme through the activity-guided fractionation. As vibralactone biosynthetic genes unlikely occur as a gene cluster16, we set out to conduct mycelia enzyme fractionation instead of the direct gene isolation. According to the activity-guided fractionation combined with proteomic analyses, which was successfully used to discover the vibralactone cyclase VibC16, we start with the B. vibrans mycelia that significantly accumulate 3 in the culture broth. The fractionation was guided by a specific enzyme assay using 6a (analogue of 6, Fig. 6a, Supplementary Fig. 1) that harbors an allyl rather than dimethylallyl group to recognize the assay-formed product from the native 3 in fungal mycelia, due to a small amount of endogenous 3 that may still exist in fractions even after ion-exchange chromatography. Sequential chromatography led to seven fractions enriched in the enzyme activity, which were individually subjected to LC-MS/MS for proteomic identification. This yielded 69 proteins common to all the seven active fractions, among which 19 candidates were selected according to the apparent molecular weight (~ 60 kDa) of the active protein on the size-exclusion chromatography (Supplementary Fig. 2, Source Data). Fourteen candidates were successfully cloned from complementary DNA (cDNA) of B. vibrans, expressed via pET28a(+) in Escherichia coli BL21(DE3), and assayed for enzyme activity by incubating the cleared cell lysate with 6a as substrate. The reaction was analyzed by liquid chromatography-mass spectrometry (LC-MS) for the extracted ion chromatogram (EIC) of mass-to-charge ratio (m/z). Among the 14 candidates screened, two sequences with 99% identity, encoding putative flavin-binding monooxygenases, were found to produce a compound with a [M + H]+ of 181 m/z, corresponding to the authentic 3a (Supplementary Fig. 3, Supplementary Table 2). Furthermore, assays using 6 as substrate revealed a product consistent with the authentic 3 with a [M + Na]+ of 231 m/z, in comparison to control assays with the lysate of E. coli cells transformed with an empty pET28a(+) vector (Supplementary Fig. 4). Similarly, their homologues (95% identity, accessions. XP_007301038.1 and XP_007301015.1) amplified from cDNA of the mushroom Stereum hirsutum (producing vibralactone as a minor metabolite)5 and expressed in E. coli can also produce 3 from 6 as substrate (Supplementary Fig. 4). The enzyme, which was purified to homogeneity as an N-terminal His6-tag protein, was shown to yield 3 as a significant product when incubating with 6 in the presence of the exogenous NADPH (nicotinamide adenine dinucleotide phosphate, reduced), while no detectable 3 can be observed in the incubation mixture of 6 with NADH or in control assays with boiled enzyme (Fig. 2a, Supplementary Figs. 3 and 5). To obtain enough product for structural elucidation, we conducted product extraction and purification from a whole-cell biotransformation by feeding 90 mg of 6 to the E. coli expressing the enzyme, which led to isolation of approximately 12 mg of the product. Its identity was validated to be the oxepin-2(3H)-one-containing (1S)-1,5-seco-vibralactone (3) by evidence of 1H NMR spectrum and optical rotation, matching the authentic 3 isolated from the B. vibrans mycelial cultures (Supplementary Fig. 6). Considering that 3 and its oxepin-2(7H)-one isomer 3′ can tautomerize from each other16, LC-MS detection was performed immediately after incubation without product extraction and drying overnights. Results confirmed 3 as the direct product of the enzymatic catalysis (Fig. 2b, Supplementary Fig. 7). Thus, the oxepinone-forming enzyme for conversion of 6 into 3 was identified and named VibO, in line with VibC in pathway to the biosynthesis of vibralactone.
Alongside a remarkable production of 3 from 6 by the catalysis of VibO, a very small peak (4.36 min, Fig. 2a) was found on the same EIC trace for +m/z 231 but corresponding neither to 3 nor 3′. The peak was more noticeable on EIC for −m/z 207 where no signals for 3 and 3′ were observed. In the reaction with 1 mM NADPH, slightly more production of this by-product was observed (Fig. 2a, Supplementary Fig. 5). To determine its chemical structure, a scale-up VibO-catalyzed reaction was conducted and followed by extraction and purification. Consequently, the by-product was identified by 1H NMR spectrum as 7 whose data are consistent with those of earlier publication24 (Fig. 2b, Supplementary Fig. 8). To examine the dependence of NADPH, reactions of VibO and 6 with NADPH ranging from 0.001 to 10 mM were detected by EIC +m/z 231 for both 3 and 7. Production of 3 was observed at 0.001 mM NADPH and increased to the highest yield at 0.1 mM NADPH, then drastically decreased by ∼12-fold at 10 mM NADPH. By contrast, NADPH up to 10 mM resulted in the highest production of 7 with a ∼40-fold increase compared to its initial formation at 0.05 mM NADPH (Supplementary Figs. 9 and 10). The increase in 7 can hardly be attributed to the decrease in 3, as proved by incubations of 3 with NADPH in which no trace of 7 was found (Supplementary Fig. 11). To verify if the compound 7 is an intermediate leading to 3 or a shunt product, 7 was used as a substrate and incubated with VibO, which gave no detectable 3 (Fig. 2a, b; Supplementary Fig. 5). Unlike the multifunctional GilOII catalyzing consecutive hydroxylation and BV oxidation to give a hydroxyoxepinone22, no conversion of 6 or 7 to the putative hydroxyoxepinone by VibO did happen (Supplementary Figs. 12 and 18). Finally, 7 was barely observed in the B. vibrans mycelial cultures that significantly accumulate 3 (Supplementary Fig. 13). These results clearly indicate that 7 is a minor shunt product from the in vitro VibO reaction and derived from selective hydroxylation of 6.
The freshly purified VibO has an apparent KM value at 0.407 ± 0.083 mM toward 6 and a turnover number (kcat) at 0.111 ± 0.010 min− 1 (Fig. 2c, Source Data). Concomitant release of H2O2 was hardly observed in VibO reactions (Supplementary Fig. 14). To determine the origin of the inserted oxygen atom and hydrogen exchange in 3, isotope-labeling experiments using 18O2, H218O, or D2O (deuterated H2O) in the VibO reaction were performed with comparison to the normal O2 /H2O system. When the assay was conducted in H218O (90% enriched), no 18O incorporation into 3 was detected (Fig. 2d). In contrast, isotopologue ions (m/z 233) for 3 shifted by 2 m/z were significantly observed when the VibO reaction was conducted under 18O2 atmospheric conditions, indicating that one 18O atom from molecular oxygen was incorporated into 3 (Fig. 2b, d). This 18O2 condition also gave 18O-labelled 7 at m/z 233 (Supplementary Figs. 15 and 16). Meanwhile, the occurrence of m/z 232 for 3 in D2O system suggested one hydrogen exchange with moderate deuterium incorporation during the VibO catalysis (Fig. 2d). However, 7 was barely labelled in D2O system (Supplementary Figs. 15 and 16), indicating catalytic differences in the formation of 3 and 7. The VibO solution displays a yellow color, indicating the presence of flavin. LC-MS analysis of the denatured enzyme confirmed the flavin as flavin adenine dinucleotide (FAD), which is consistent with the finding that VibO has almost the same level of activities with or without the exogenous FAD and requires no additional FAD for in vitro assays (Supplementary Fig. 17). The VibO enzyme is dependent exclusively on NADPH, as evidenced by no detectable product in the assay with NADH (Fig. 2a, Supplementary Fig. 5). Collectively, our data revealed that VibO is an NADPH/FAD-dependent monooxygenase catalyzing alone an oxygenation on the phenol ring to form the oxepinone skeleton of 3 in the mushroom B. vibrans. This is distinct from the multifunctional GilOII22 and MtmOIV25, both of which perform successive BV oxidation and decarboxylation (Supplementary Fig. 18).
Crystal structure of VibO. To gain molecular insights into the enzymatic mechanism of VibO, the crystal structure of VibO in complex with the cofactor FAD was determined to 2.43 Å resolution (PDB ID: 7YJ0, Supplementary Table 1). In the crystal structure, each asymmetric unit contains four VibO molecules, which form two symmetric dimers (Fig. 3b), in line with our analytical ultracentrifugation-based assay (Supplementary Fig. 19). In the dimeric VibO structure, each VibO monomer contains 25 β-strands (β1-β25) and 13 α-helices (α1-α13) (Fig. 3b, c), and is composed of three distinct sub-domains: the FAD-binding domain (residues 1-127, 141–245 and 336–419), the middle domain (residues 128–140, 246–335 and 420–472), and the C-terminal domain (residues 473–653) (Fig. 3a, c). In particular, the FAD-binding domain is mainly assembled by an N-terminal extension (residues 1–56) and a central three-layer (ββα) sandwich that is formed by a β-sandwich packing with an overlaying helical section (Fig. 3c). The middle domain is not continuous in amino acid sequence (Fig. 3a), and folds into an architecture featured with a seven-stranded β-sheet flanked by two α-helices (α5 and α8) together with a short antiparallel β-sheet (β14 and β18) (Fig. 3c). The C-terminal domain of VibO adopts a peroxiredoxin-like α/β-fold, which directly packs against the FAD-binding domain and is far away from the middle domain (Fig. 3c). The overall architecture of VibO is similar to that of the flavin-dependent phenol hydroxylase PHHY (PDB ID: 1PN0)26 (Supplementary Figs. 20a, b), which catalyzes the ortho-hydroxylation of phenol to catechol and belongs to the Class A flavoprotein monooxygenase family27, as revealed by a structural similarity search using the program Dali28. However, the dimer assembly mode of VibO is distinct from that of PHHY (Supplementary Fig. 20c). In particular, the dimerization of VibO is mediated by three different types of sub-domain/sub-domain interactions between the two monomeric VibO molecules (the middle domain/middle domain interaction, the FAD-binding domain/FAD-binding domain interaction and the FAD-binding domain/C-terminal domain interaction) (Fig. 3b, Supplementary Fig. 21), covering a total of ~ 3436 Å2 surface area. Further detailed structure analyses revealed that the dimerization interface of VibO is mainly mediated by extensive polar (charge-charge and hydrogen bonding) and hydrophobic interactions (Supplementary Fig. 21).
Across the middle and FAD-binding domains of VibO, the co-factor FAD is buried in a solvent-exposed and highly positive-charged pocket (Fig. 3d, e). Specifically, the ADP and ribityl groups of the bound FAD are embedded within the FAD-binding domain, whereas its isoalloxazine ring is located at the interface between the middle domain and the FAD-binding domain (Fig. 3e). Notably, the co-factor FAD in the crystal structure of the VibO/FAD complex adopts an “out”-conformation (Supplementary Fig. 20d), suggesting that the bound FAD is located outside of the active site of VibO. Further careful analyses of the VibO structure uncovered that hydrogen bonds and electrostatic interactions are primarily responsible for the specific binding of VibO with FAD, involving the main chains of G66, V68, S89, D356, A363 and M369 as well as the side chains of R88, R97, Q160, R287 and D356 of VibO (Supplementary Fig. 22).
Interestingly, a highly hydrophobic cavity was found adjacent to the isoalloxazine ring of the bound FAD in the VibO/FAD complex, and importantly, this pocket corresponds to the substrate-binding site of the phenol hydroxylase PHHY (Fig. 4a). Further structural comparison analyses uncovered that although the sizes and compositions of these pockets in VibO and PHHY are somewhat different, the critical D54, I279, Y289 residues in PHHY for the hydroxylation activity are shared by VibO (Fig. 4b). Based on these observations and the fact that the substrate 6 of VibO contains a phenol ring, we inferred that this pocket should be the substrate-binding site of VibO. Strikingly, in contrast to PHHY, VibO contains a unique channel, which is located on the opposite side of the substrate-binding pocket and away from the FAD-binding site (Fig. 4c, Supplementary Fig. 23). Detailed structural characterizations elucidated that the substrate-binding pocket and the channel of VibO are formed between the middle domain and the FAD-binding domain (Fig. 4c, Supplementary Fig. 23), and are both assembled by hydrophobic and polar residues from the two domains (Fig. 4c). Considering that the substrate-binding site of VibO is buried deep inside the structure (Fig. 4a), we speculated that this unique channel of VibO might be used for the substrate entrance or product leaving. Consistent with our hypothesis, the replacement of V459 that locates at the solvent-exposed site of the channel with a much larger Leu residue can nearly halve the enzymatic activity of VibO (Fig. 5a).
Site-directed mutations and mechanism of VibO catalysis. Based on the structural analyses of the active-site of VibO (Fig. 4b), the residues forming the substrate pocket including D99, F265, I279, Y289, and A366 were conducted for site-directed mutagenesis to reduce (D99N, Y289N) or even remove the hydrogen bond interaction (D99A, Y289A, and Y289F), or to narrow down the substrate-binding pocket (F265Y, I279F, A366L, and A366Q). Consequently, none of these mutants can give the product 3 except for F265Y with a half decrease in accumulation of 3 compared to the wild-type enzyme (Fig. 5a). On the other side, three mutants of VibO including D99A, D99N, and F265Y were observed for production of 7 with catalytic activities of 16%, 10%, and 22%, respectively relative to the wild-type enzyme (Fig. 5a, Supplementary Figs. 25 and 26), which is consistent with a ∼5-fold decrease in hydroxylation activity for the D54N mutant of PHHY29, as Asp54 in PHHY corresponds to the Asp99 residue in VibO (Fig. 4b). For the PHHY catalysis, the Asp54-assisted deprotonation of the substrate phenol to phenolate was postulated to be more favorable for the subsequent hydroxylation, thereby likely adopting an electrophilic aromatic substitution type of mechanism30. Unfortunately, our efforts to obtain the crystal structure of VibO in complex with 6 were not very successful, and we could not unambiguously identify good electron density for the bound substrate 6. However, a careful examination of the substrate-binding pocket of VibO in the presence of 6 uncovered weak extra electron density around Asp99 and Tyr289 that can be reasonably fitted with 6 (Supplementary Fig. 24). Importantly, further structural modeling analysis showed that the phenolic hydroxyl group of 6 is approaching FAD, and forms good hydrogen bonds with the side chains of Asp99 and Tyr289 (Supplementary Fig. 24). Collectively, these results suggested that the residues Asp99, Phe265, Ile279, Tyr289, and Ala366 may provide the active site geometry for VibO to perform its unique enzymatic function; Asp99 and Tyr289 appear to be crucial for the catalytic activity of VibO.
During VibO catalysis the oxygenation may occur via Baeyer-Villiger (BV) oxidation (Fig. 5b) and hydroxylation (Fig. 5c). Following the PHHY mechanism30, the hydroxylation logic to 7 as a minor shunt product by VibO from 6 may involve an electrophilic attack on the phenol ring at its ortho position via C4a-hydroperoxyflavin (flavin-C4a-OOH), resulting in the presumed dienone product 7′ which is converted nonenzymatically into the product 7 via keto-enol tautomerization (Fig. 5c). For the oxepinone formation in 3 as a major product, however, we favor a BV oxidation to 3 via the Criegee intermediate rather than a hydroxylation to 3 via the intermediate 7′ (Fig. 5b, c), based on the observations that (i) 7 cannot be transformed into 3 (Fig. 2a), (ii) 3 and 7 are unlikely to share the common intermediate 7′ since more NADPH led to the higher production of 7 but decrease in 3 where no conversion of 3 to 7 did occur (Supplementary Figs. 9−11). Combined with structural analyses and incorporation studies with 18O2 and D2O (Fig. 2d), we proposed that VibO may initiate deprotonation of the hydroxy group on the phenol unite of 6 presumably by Asp99 and protonation of the prenyl carbon to afford the putative intermediate 6′ (Fig. 5b), followed by a nucleophilic flavin-C4a-OO− attack on the carbonyl group to creating the oxepin-2-one skeleton of 3 via the Criegee intermediate as known in GilOII-mediated BV oxidation22. Future structural elucidation of VibO in complex with the substrate and relevant steady-state kinetics analyses may help identify key residues involved in the substrate binding and/or catalysis. To better understand the oxepinone mechanism of VibO, a detailed comparison of the catalytic features of the active sites of VibO and PHHY would be interesting for quantitative structure-activity relationship studies. As a first step, we synthesized the PHHY sequence26, expressed it in E. coli, and conducted enzyme assays using phenol or 6 as substrate. While significant production of catechol was observed for PHHY incubating with phenol, neither 3 nor 7 was detected in the reaction with 6 as substrate (Supplementary Fig. 25).
Verification of 4 and 5 as intermediates en route to vibralactone in B. vibrans. To complete the vibralactone biosynthesis, we investigated the conversion of 4-hydroxybenzoate (2) to 3-prenyl-4-hydroxybenzylalcohol (6). As a known biosynthetic precursor 4-hydroxybenzoate is incorporated into ubiquinones and other meroterpenoids (e.g., asperpentyn,31 biscognienyne B32), starting with prenylation on its aryl ring. Vibralactone has a dimethylallyl group and its bicyclic lactone core can be derived from 2, thus we hypothesized that the vibralactone biosynthesis may begin with prenylation of 2 to give 3-prenyl 4-hydroxybenzoate (4), which may undergo stepwise reductions via 3-prenyl-4-hydroxybenzaldehyde (5) to the alcohol 6. To test the hypothesis, we conducted the intermediate analogue feeding as used in our prior studies15. Compounds 4a and 5a (Fig. 6a), analogues of 4 and 5 that contain an allyl instead of a dimethylallyl group, were synthesized and fed (1 mM) respectively, to the growing cultures of B. vibrans once the accumulation of metabolite 3 can be observed by LC-MS, taking the broth immediately after feeding as control. The culture was incubated for additional 3∼7 days, then the culture broth was extracted for LC-MS analyses and compared directly with the synthetic reference 6a and authentic products (1a, 3a, 3a′) obtained from 6a-feeding cultures in our work before15. Production of sodium adduct ions [M + Na]+ at m/z 187 for 6a, m/z 203 for 1a, 3a, and 3a′ (an isomer of 3a) were exclusively revealed in both 4a- and 5a-feeding assays, and all the allyl metabolites were absent in controls (Fig. 6b). This was further supported by ESI–high-resolution MS (HRMS) analyses, revealing product ions at m/z 187.0731 as 6a (calcd for C10H12NaO2+, [M + Na]+: 187.0730), m/z 203.0680 as 3a, m/z 203.0678 as 3a′ and 1a (calcd for C10H12NaO3+, [M + Na]+: 203.0679) (Fig. 6c, Supplementary Figs. 27−29). Thus, feeding experiments demonstrate that 4 and 5 are on-pathway biosynthetic intermediates leading to vibralactone (1) in B. vibrans.
VibP1 and VibP2 function as membrane-bound UbiA prenyltransferases. Prenyltransferases that install dimethylallyl pyrophosphate (DMAPP) to 4-hydroxybenzoate (2) were recently reported in the biosynthesis of fungal meroterpenoids asperpentyn31 and biscognienyne32, and classified into the UbiA prenyltransferase family. We cloned the UbiA homologous sequences from mRNA of B. vibrans and expressed via pET28a(+) in E. coli BL21(DE3) (Supplementary Table 4). As UbiA prenyltransferases are known to be membrane bound, we prepared microsome extracts from the induced E. coli and stored them at -80°C for functional characterization. To identify the enzymatic products from incubation with DMAPP and 4-hydroxyphenyl substrates, especially the site for prenylation on the phenol ring, we synthesized 4, 5, 6 and their C4-O-dimethylallyl congener 4′, 5′, 6′ as authentic references (Supplementary Method 4 and Figs. 44−51). Reaction mixtures were analyzed by LC-MS and compared with the synthetic references. Out of eight candidates assayed, two proteins were shown to generate 4 as the only product from 2, and no existence of 4′ was observed in the reactions. Production of 5 rather than 5′ from 4-hydroxybenzaldehyde, and 6 rather than 6′ from 4-hydroxybenzylalcohol were also evident from LC-MS analyses (Fig. 7a, b; Supplementary Figs. 30−32). Further ESI-HRMS found the product ions at m/z 205.0870 (calcd for C12H13O3−, [M−H]−: 205.0870) as 4, m/z 189.0924 (calcd for C12H13O2−, [M−H]−: 189.0921) as 5, m/z 191.1077 (calcd for C12H15O2−, [M−H]−: 191.1078) as 6 (Fig. 7d, Supplementary Figs. 33–35). The two prenyltransferase sequences, coming from different scaffolds of the genome and sharing 68% amino acid identity, are named VibP1 and VibP2 thereafter. Their homologue ShPT1/ShPT2 (accession no. XP_007303462, XP_007299647) amplified from cDNA of Stereum hirsutum can also accept the above 4-hydroxyphenyl substrates and form C3-prenyl products (Fig. 7a, Supplementary Figs. 30−32). To examine enzymatic activities with different 4-hydroxyphenyl substrates, the yield of 4, 5, and 6 was quantified on account of the standard courves. Results revealed that all the four prenyltransferases exhibited highest activity to 4-hydroxybenzoic acid (2) (Fig. 7c, Source Data file). When incubating with 2 and a mixture of four prenyl pyrophosphates including DMAPP (C5), geranyl pyrophosphate (GPP, C10), farnesyl pyrophosphate (FPP, C15), and geranylgeranyl pyrophosphate (GGPP, C20), VibP1 was shown to form 4 as a major product in addition to trace amounts of C10 and C15 prenylated products (Supplementary Figs. 36 and 37). However, the soluble aromatic prenyltransferase formerly reported in B. vibrans gave 6′ (C4-O-prenylation) as a major product and 6 (C3-prenylation) as a minor product when incubating with 4-hydroxybenzyl alcohol and DMAPP14,33. It is closer (67% sequence identity) to the basidiomycete BypB34 (also soluble) for prenylation of orsellinic acid and has very low sequence identity (∼6%) with VibP1 and VibP2 (Supplementary Fig. 38), which are integral membrane proteins and clearly prefer 4-hydroxybenzoic acid (2) as substrate for C3-prenylation. These results demonstrate that VibP1/VibP2 catalyzes C3-prenylation on the 4-hydroxyphenyl ring of 2 to afford 3-prenyl 4-hydroxybenzoate (4), which is validated as an intermediate en route to vibralactone (1) in B. vibrans (Fig. 6, Supplementary Figs. 27−29). Collectively, VibP1 and VibP2 function as UbiA prenyltransferases and are most likely involved in the vibralactone biosynthesis.
Reductases catalyzing 4 to form 6. Since the reduction of carboxylic acids to the corresponding aldehydes and benzylic alcohols is well precedented in microbial systems, we set out to identify the relevant reductases by homology-based cloning from B. vibrans mRNA with primers (Supplementary Tables 5 and 6) and functional characterization using recombinant proteins obtained from overexpression via pET28a(+) in E. coli BL21(DE3). We amplified eight homologous sequences of the carboxylic acid reductase (CAR); among them only one protein, named BvCAR, was shown to catalyze reduction of 3-prenyl 4-hydroxybenzoate (4) to 3-prenyl-4-hydroxybenzaldehyde (5). The purified BvCAR can give 5 as a sole product when incubating with 4 in the presence of both ATP and NAD(P)H. Higher relative activity was observed for BvCAR incubating with NADPH in comparison to NADH, and no products can be observed without the exogenous ATP (Fig. 8a, b, Supplementary Figs. 39 and 40). The enzyme cannot reduce 4-hydroxybenzoate (2) to form 4- hydroxybenzaldehyde, indicating that BvCAR is specific for production of 5 from 4 and most likely dedicated to the vibralactone pathway. Meanwhile, three aldehyde reductases (AR)/alcohol dehydrogenases (ADH), sharing a sequence identity of 22%∼68%, were identified to produce 3-prenyl-4-hydroxybenzylalcohol (6) from 5 in the presence of NAD(P)H and thereby named BvAR1, BvAR2, and BvAR3 individually. Much higher activity was observed for BvAR1 incubating with NADPH than NADH (Fig. 8b, c, Supplementary Figs. 41 and 42). Considering reductases/dehydrogenases with relaxed substrate specificity, multiple enzymes including BvARs1-3 are likely to be involved in the reduction of 5 to 6 for the vibralactone biosynthesis.
Reconstruction of the vibralactone pathway in vitro and in E. coli. After clarifying all previously unknown steps, we propose the biosynthetic pathway for vibralactone (1) in the mushroom B. vibrans as depicted in Fig. 1. To further validate the biosynthetic logic underlying vibralactone formation, we attempt to reconstruct the pathway both in vitro and in vivo. For in vitro reconstruction by one-pot, each of the five enzymes (VibP1, BvCAR, BvAR1, VibO, VibC) was overexpressed and purified from E. coli BL21(DE3). In the one-pot reaction mixture supplied with 2, DMAPP, NADPH, and ATP, production of 1 and the oxepinone 3 was clearly observed by LC-MS analyses (Supplementary Fig. 43). For the complete pathway reconstructed in E. coli BL21(DE3), we co-expressed pBbA5C-MevT-MBIS35 for overproduction of DMAPP and pET28a(+)-VibP1-BvCAR-sfp-BvAR1-VibC-VibO for the vibralactone pathway. Here a phosphopantetheinyl transferase (sfp) was expressed for activation of the CAR enzyme36,37. As a result, significant accumulation of 1 was observed in the E. coli transformants fed with 2, and no detectable 1 existed in the control (Supplementary Fig. 43). Our efforts on the potential of utilizing E. coli as a cell factory to produce 1 are still ongoing.