The Biosynthesis of a Fungal Flavonoid, Chlorflavonin, Precisely Controlled by Two Types of Equilibrium States

Abstract

Flavonoids are polyphenolic natural products predominantly isolated from plants and exhibit a diverse array of biological activities. Because they serve as important pharmaceuticals and nutraceuticals, there is a strong demand for their sustainable supply. Chlorflavonin is a rare fungal flavonoid with potent antitubercular activity. While its biosynthetic enzymes are expected to be valuable tools for application in fungal production of flavonoids, the biosynthetic pathway remains unknown. Here, we elucidate the complete biosynthetic pathway for chlorflavonin through detailed functional analysis of each biosynthetic enzyme. Previously, stepwise and straightforward introduction of 3-, 7-, and 8-methoxy; 2′-hydroxy; and 3′-chloro functionalities have been proposed. In contrast to this proposal, we uncovered an intricate biosynthetic route involving a dynamic interconversion between the 6- and 8-methoxy forms of flavonoid skeletons mediated by the chalcone isomerase CfvF and the flavin-dependent oxygenase CfvI. CfvF interconverted the 6- and 8-methoxyflavanones, likely via a chalcone intermediate. CfvI oxidized the chemically inert 2,3-double bond of flavanone, giving the hemiacetal product. Although these enzymes generate products existing in equilibrium states, respective downstream enzymes catalyze selective conversion of one specific species, facilitating smooth progression to the final product. We also solved the crystal structure of CfvK, the dehydratase that selectively converts the 8-methoxy form of the CfvI product. Through analyzing the structure complexed with its substrate and product and site-directed mutagenesis, key residues determining the substrate selectivity were identified. Our comprehensive analysis established a rational framework for preparing 46 flavonoids, including unnatural ones, setting the stage for fungal production of structurally diverse flavonoids.

Introduction

Flavonoids are widely distributed natural products with diverse biological activities, and their biosynthetic pathways in plants have been extensively studied. − Given the high value of flavonoids in pharmaceuticals and nutraceuticals, synthetic biology approaches have increasingly focused on microbial production platforms. While current strategies predominantly rely on the expression of plant-derived enzymes in yeast or Escherichia coli, filamentous fungi such as Aspergillus oryzae and Aspergillus nidulans have emerged as attractive hosts for natural product biosynthesis, owing to their considerable versatility in reconstructing biosynthetic pathways for fungal natural products. , Moreover, their distinct metabolic capabilities, compared with those of conventional microbial hosts, highlight the importance of developing production systems across multiple organisms. ,

Among flavonoids, chlorflavonin (1; Figure A) and its congeners are natural flavonoids produced by filamentous fungi and are the only known flavonoids derived from nonplant sources. − In this manuscript, compounds are numbered in the order of their appearance. For compounds lacking a 7-O-methyl group, the corresponding compound number is followed by the lowercase letter “a” (e.g., 1a). Chlorflavonin (1) exhibits inhibitory activity against acetolactate synthase (ALS), which catalyzes the first common step in the biosynthesis of branched amino acids. , Given that ALS is known as a target of antimicrobial and antifungal agents, including those targeting Mycobacterium tuberculosis and Aspergillus fumigatus, 1 is considered a promising lead compound for pharmaceutical development. ,, Given the fungal origin of 1, its biosynthetic machinery represents an attractive platform for producing natural and nonnatural flavonoids in fungi. Therefore, elucidating the biosynthetic pathway of 1 not only provides insight into fungus-specific enzymatic strategies for flavonoid biosynthesis but also facilitates the expansion of microbial production platforms for flavonoids.

1.

Overview of the biosynthetic pathway of 1. (A) The biosynthetic gene cluster for 1 and chemical structures of 1 and its dechloroderivative 6. (B) Comparison of the biosynthetic pathway of 1 proposed in the previous and present studies. FDH: flavin-dependent halogenase; MTase: methyltransferase; FMO: flavin-dependent monooxygenase; NTF2: nuclear transport factor 2; TPP: thiamine pyrophosphate; SRE: self-resistance enzyme.

We previously identified a nonribosomal peptide synthetase–polyketide synthase (NRPS-PKS) hybrid enzyme, DiapA. This enzyme possesses an N-terminal NRPS module and functions as a novel chalcone synthase (CHS), distinct from plant CHSs classified as type III PKSs. Using DiapA as a probe, we discovered a candidate biosynthetic gene cluster (BGC) for highly elaborate flavonoids in the genome of Aspergillus candidus, designated as the cfv cluster (Figure A). In a preceding study, this same BGC was proposed to be involved in the biosynthesis of 1, although no experimental validation was provided. We demonstrated that CfvA, a fungal CHS, and CfvF, a fungal chalcone isomerase (CHI), are responsible for the biosynthesis of S-pinocembrin (2a). Independently, another study identified the same BGC using a self-resistance gene-guided genome mining approach and proposed a biosynthetic pathway based on systematic gene disruption and metabolite analysis (Figure B).

Currently, the complete biosynthetic pathway of 1 has remained unresolved. This is primarily due to uncertainties in the reaction sequence and the substrate specificity of individual enzymes as well as potential interference from host endogenous activities. A key unresolved issue is the function of CfvF. Disruption of cfvF resulted in a marked decrease in 1 production and accumulation of R-7-O-methyl-8-methoxypinocembrin (ent-3). If CfvF functions as a simple CHI, then 1 should be produced at levels comparable to those of ent-3, as pinocembrin chalcone undergoes nonenzymatic cyclization to form rac-2a. This discrepancy between the proposed function of CfvF and the metabolite profile of its knockout mutant suggests that the enzyme has an additional previously unrecognized role in the pathway. Furthermore, the downstream steps involved in the formation of flavonol have been proposed mainly on the basis of gene disruption experiments without biochemical characterization. Therefore, the identities and functions of the enzymes responsible for these transformations remain unclear.

In this study, we elucidated the complete biosynthetic pathway of the fungal flavonoid 1, featuring an unusual interconversion between 6- and 8-methoxy flavonoid skeletons. We discovered a previously unrecognized biosynthetic route for the conversion of flavones to flavonols that is distinct from the canonical plant pathway. In this pathway, the flavin-dependent enzyme CfvI catalyzes the oxidation of the chemically inert 2,3-double bond of flavone, generating a hemiacetal intermediate that exists in equilibrium among multiple interconverting species. The downstream enzyme CfvK selectively converts only one of these species to a flavonol. We further found that CfvF mediates reversible ring closure and ring opening, establishing an equilibrium between 8- and 6-methoxyflavanones, which is essential for providing the correct substrate to downstream enzymes. Finally, by combining heterologous expression with in vivo enzymatic conversion, we generated 46 flavonoids, including rare plant-derived and novel compounds, demonstrating the utility of filamentous fungi as versatile platforms for flavonoid synthetic biology.

Results

Elucidation of the Reaction Sequence in the Biosynthetic Pathway

In the present study, we assigned the functions of each enzyme based on the conversion of 2a and 4a using a series of A. oryzae transformants expressing one or more biosynthetic enzymes, together with bioinformatic analysis (Figure A, Figure S1; experimental details are explained in SI). The results obtained were generally consistent with those reported in a previous study. However, we were unable to elucidate the exact reaction sequence in the biosynthetic pathway by this analysis because most of the biosynthetic enzymes exhibited promiscuous substrate specificity.

To gain insights into the actual biosynthetic pathway operating in vivo, we generated an A. oryzae transformant harboring all cfv genes. In this transformant (TF1), we observed the production of 1 along with dechloroflavonin (5) and 3-O-desmethyldechloroflavonin (6) (Figure A, Figure S2). Next, we inactivated the cfvA in TF1 to generate AO-cfv-allΔA (Figure S3) and used this strain to assess whether 2a and 4a function as biosynthetic intermediates. Compound 2a was converted to 1 by AO-cfv-allΔA, indicating that it is a biosynthetic intermediate (Figure B). In contrast, 4a was converted to 5-hydroxy-7,8-dimethoxyflavone (7), indicating that it is a shunt metabolite. Interestingly, both 2a and 4a were previously proposed to be biosynthetic intermediates (Figure ). We also observed that ent-2a was converted to ent-3 (Figure B). Based on these results, we hypothesized that 2a is converted to 1 via a pathway distinct from that involving 4a and 7 proposed in the previous study.

2.

Summary of heterologous reconstitution of the biosynthetic pathway of 1 in A. oryzae. (A) Proposed pathway operating in A. oryzae. LC–MS analysis of the metabolites of selected transformants is shown on the right. The lower-right panel indicates the genes introduced into each transformant. In this panel, black circles indicate genes that were functional, whereas white circles indicate genes that were likely nonfunctional. See Figure S5 for other transformants. (B) Feeding of 2a, 4a, 3, 6, and 13 into AO-cfv-allΔA to evaluate whether these compounds are converted to 1.

We further analyzed the metabolites that accumulated in a series of transformants and examined their relevance as biosynthetic intermediates. AO-cfvAFJECGDIK (TF2), which lacks cfvH and cfvB, accumulated 6 as the sole product (Figure A). Considering 6 was converted to 1 by AO-cfv-allΔA, it is likely a biosynthetic intermediate, and CfvH-catalyzed 3-O-methylation and CfvB-catalyzed halogenation appear to represent the final steps of the pathway (Figure B).

To investigate the role of the 7-O-methyltransferase CfvC, we compared TF2 with AO-cfvAFJEGDIKHBL (TF3) and AO-cfvAFJEGDIK (TF4). AO-cfvAFJEGDIKHBL (TF3) accumulated 6a as the major product along with trace amounts of 5a and 1a, whereas TF4 and TF2 predominantly accumulated 6a and 6, respectively (Figure A). These results suggest that the 7-methoxy group facilitates substrate recognition by CfvH and CfvB but is not essential for the earlier steps in the pathway.

AO-cfvAFJEGD (TF5), which lacks cfvI and cfvK, accumulated flavone 8a, with 9a and 10a–12a as minor products. Given the absence of the 3-hydroxy group in these metabolites, CfvI and CfvK are likely involved in 3-hydroxylation (Figure A). AO-cfvAFJEIK (TF6) and AO-cfvAFJE (TF7) exhibited similar metabolite profiles, accumulating 3a, 13a, and 14a (Figure A). These results suggest that CfvI and CfvK function after the formation of flavone catalyzed by CfvD.

In AO-cfvAJECGDIKHBL (TF8), which lacks cfvF, production of 1 was abolished, and rac-3 accumulated (Figure A, Figure S4). Considering that 2a and ent-2a were converted to 3 and ent-3, respectively, by AO-cfvJEC (Figure S4), the accumulation of rac-3 in TF8 is likely attributable to the substrate promiscuity of CfvJ, CfvE, and CfvC. Assays using AO-cfv-allΔA indicate that 3 is likely a biosynthetic intermediate, whereas ent-3 is not (Figure B). The lack of further modifications to 3 in TF8 suggests that downstream enzymes, namely CfvG, CfvD, CfvI, CfvK, CfvH, and CfvB, do not effectively process this intermediate in the absence of CfvF. Together, these findings suggest that CfvF has a critical role not only in the stereoselective cyclization of pinocembrin chalcone to 2a but also in a later step of the pathway following the formation of 3. Notably, AO-cfvAFJEC (TF9) produced S-5-hydroxy-6,7-dimethoxyflavanone (13) together with 3. Considering 13 was converted to 1 by AO-cfv-allΔA (Figure B), it is likely a biosynthetic intermediate, potentially formed through a CfvF-mediated transformation of 3.

In summary, these analyses indicate that CfvF functions at two distinct stages of the pathway. After the formation of 2a by CfvA and CfvF, 8-methoxy and 7-methoxy groups are introduced by CfvJ, CfvE, and CfvC, followed by a transformation likely involving CfvF. Subsequent 2′-hydroxylation by CfvG, dehydrogenation by CfvD, flavonol formation by CfvI and CfvK, and finally 3-O-methylation and halogenation by CfvH and CfvB furnish 1 (Figure A). For the full results of the heterologous expression in A. oryzae and the evaluation of each metabolite by AO-cfv-allΔA, see Figures S5 and S6.

Although we were able to assign the reaction sequence in the pathway, reconstruction of the pathway in A. oryzae occasionally resulted in metabolites that were further modified by endogenous host enzymes. For example, 11a and 12a, which possess additional hydroxy groups on the B-ring, likely arise from such off-pathway modifications (Figure A). In addition, several biosynthetic enzymes appear to exhibit broad substrate specificity, accepting structurally related compounds, including shunt metabolites. These factors complicate the precise interpretation of the pathway. To address these limitations, we performed in vitro enzymatic assays and in vivo bioconversion experiments to dissect each biosynthetic step in greater detail, as described below.

Conversion of 2a to 3 by CfvC, CfvJ, and CfvE

To determine the reaction sequences of CfvC, CfvJ, and CfvE, we prepared recombinant proteins (Figure S7). CfvC catalyzed the 7-O-methylation of 2a to give 2, whereas CfvJ catalyzed the 8-hydroxylation of 2a to yield S-8-hydroxypinocembrin (15a) (Figure A and Figure S8). Owing to the instability of 15a, we were unable to isolate this compound; therefore, its structure was assigned on the basis of the proposed function of CfvJ and its mass spectral data (Figure S9). When we incubated 2 with CfvJ, no conversion was observed, suggesting that the substrate bearing a 7-methoxy group was not accepted by CfvJ (Figure A). We then conducted a competitive assay of CfvC and CfvJ with 2a. In the presence of CfvC and CfvJ, 2a was smoothly converted to 15a and 15 (Figure S10). Under these conditions, only a trace amount of 2 was observed. The observation that CfvJ could not accept 2 as a substrate indicates that 8-hydroxylation by CfvJ preceded 7-O-methylation by CfvC. Furthermore, upon addition of 2a to AO-cfvJEC, an initial accumulation of 3a was observed, followed by the exclusive formation of 3 as the final product (Figure B). Although the possibility that CfvE converts 15 to 3 cannot be excluded, we propose that the introduction of the 8-methoxy group by CfvJ and CfvE precedes 7-O-methylation by CfvC (Figure C). This proposal is supported by the observation that 15 was not detected during the conversion of 2a by AO-cfvJEC (Figure B). Consistently, 7-O-methylation of 3a by CfvC was also confirmed (Figure S8).

3.

Analysis of the pathway from 2a to 8. (A) LC–MS analysis of CfvJ reactions. (B) Bioconversion of 2a by AO-cfvJEC. (C) Proposed pathway from 2a to 8. Note that only 2S-flavonoids were described in A–C, although rac-2a was used in these experiments. (D) Bioconversion of 3 and 13 by TF8 (ΔF). (E) LC–MS analysis of CfvF reactions. (F) Bioconversion by AO-cfvG. See Figure S12 for comparison with the other substrates. (G) CfvD reaction. See Figure S13 for comparison with the other substrates. (H) Bioconversion of 13 by AO-cfvGD.

CfvF Generates an Equilibrium between 8- and 6-Methoxyflavanones

AO-cfvAJECGDIKHBL (TF8 = ΔF), which harbors all cfv genes except cfvF, did not produce 1 at detectable levels and instead accumulated rac-3 (Figure A). Because 3 and 13, both of which were observed as products of TF9, were confirmed to be biosynthetic intermediates, we hypothesized that CfvF has an additional function that converts 8-methoxyflavanone 3 to 6-methoxyflavanone 13 (Figure B). Notably, 13 was converted to 1 by TF8, whereas 3 was not (Figure D). This result further suggests that CfvF is responsible for the conversion of 3 to 13. To examine this additional function of CfvF, we introduced LjCHI, a CHI from Lotus japonicus, into TF8. The resultant transformant predominantly produced S-3 rather than 1, indicating that CfvF possesses a specialized function in the biosynthesis of 1 that cannot be complemented by a plant-derived CHI (Figures S4 and S5). Although CfvF is phylogenetically distinct from plant and bacterial CHIs, a structural model generated by AlphaFold3 revealed close similarity to a bacterial CHI (Figure S11). Inspired by reports that the bacterial CHI from Eubacterium ramulus is involved in flavanone degradation, we hypothesized that CfvF catalyzes the ring-opening of flavanone. To test this hypothesis, we conducted in vitro enzymatic assays of CfvF using the 8-methoxyflavanones 3a and 3 as substrates (Figure S7). In these reactions, 13a and 13 were formed, and the reaction reached a steady state with approximately a 1:1 ratio of 8- to 6-methoxy forms (Figure E). Conversely, incubation of 13a and 13 with CfvF also yielded both 8- and 6-methoxy forms at a similar ratio (Figure E). These observations suggest that CfvF catalyzes the interconversion between 8- and 6-methoxyflavanones via a chalcone intermediate. The formation of the chalcone intermediate likely permits rotation around the C4–C4a bond, after which CfvF catalyzes the S-selective cyclization of chalcones to regenerate flavanones. Through this process, an equilibrium between the 6- and 8-methoxy forms may be established. Notably, CfvF did not convert the R-enantiomer of 3, indicating that this enzyme exhibits S-stereoselectivity (Figure E).

Next, we investigated the downstream enzymes that potentially act after the CfvF-mediated equilibrium. CfvD was purified and subjected to in vitro enzymatic assays (Figure S7). In contrast, CfvG, a cytochrome P450 enzyme with an N-terminal transmembrane region, was analyzed using an in vivo system (AO-cfvG) as its purification proved challenging. In this system, 13a and 13 were converted to 9a and 9, respectively (Figure F, Figure S12). Recombinant CfvD subsequently accepted 9a and 9 as substrates to produce 8a and 8 (Figure G, Figure S13). In addition, CfvD converted 13 into 16, which was further transformed into 8 by AO-cfvG (Figure F and G). These results suggest that the biosynthetic pathway from 13 to 8 may bifurcate. Consistently, time-course feeding experiments using AO-cfvGD showed that both 9 and 16 were detected prior to the formation of 8 (Figure H). However, it is notable that 9 and 9a appear to serve as preferred substrates for CfvD over 13 and 13a, whereas CfvG exhibited the opposite substrate preference (Figure F and G, Figure S13).

Collectively, these analyses established the pathway from 3a and 3 to 8a and 8, both of which were converted to 1 by AO-cfv-allΔA (Figure C, Figure S6). Considering 3a and 3 serve as inefficient substrates for the downstream enzymes CfvG and CfvD, the CfvF-mediated interconversion between 8- and 6-methoxyflavanones is indispensable for the biosynthesis of 1. This observation further indicates that the conversion of 6-methoxyflavanone into an 8-methoxyflavonoid skeleton, as observed in 1, is required during the downstream steps.

Flavonol Synthesis Accompanied by Isomerization of the A-Ring Substitution Pattern

We investigated the steps following the formation of 8, focusing on CfvI and CfvK. When 8 was added to AO-cfvIK, it was efficiently converted to 6 (Figure A). A similar conversion of 8aa 7-O-desmethyl derivative of 8to 6a was also observed (Figure A). Interestingly, the transformation from flavone to flavonol appeared to be accompanied by a change in the A-ring substitution pattern, from a 6-methoxy to an 8-methoxy form.

4.

Flavonol formation by CfvI and CfvK. (A) Bioconversion of 8a and 8 by AO-cfvIK and AO-cfvI. (B) Enzymatic reactions of CfvK and its variants with 17a. (C) Proposed pathway from 8a to 6a. (D) Cartoon representation of the trimer of CfvK^Y50F complexed with the substrate. Phe121 of each protomer intrudes into the active site of neighboring protomers. Three protomers are colored in yellow, green, and cyan, respectively. Close-up view of the active site is also shown. (E) Schematic representation of active site residues of CfvK that interacted with the substrate 17a-iv. Residues involved in the substrate recognition are shown in blue, while those involved in the catalysis are in pink. Phe121′ from a neighboring protomer is shown in green. Trp71, which did not contact the substrate during the MD trajectory, is shown in gray. (F) Time course analysis of the formation of 6a and 18a by CfvK and its variants.

To analyze the roles of CfvI and CfvK in detail, we examined the functions of each enzyme individually. While the recombinant CfvK was successfully obtained (Figure S7), CfvI could not be expressed in a soluble form; therefore, AO-cfvI was used for subsequent experiments. On the basis of the observation that CfvI and CfvK are homologous to a flavin-dependent monooxygenase (FMO) and a scytalone dehydratase (SDH), respectively, we hypothesized that the oxidation of flavone substrate by CfvI precedes the CfvK-mediated transformation. Accordingly, 8 was added to the AO-cfvI. This bioconversion experiment resulted in the formation of 17, showing a 34 Da mass increase and UV absorption spectrum characteristic of flavanones (Figure A, Figure S14). As 8a underwent a similar transformation to 17a, we used 8a for further studies because of its availability (Figure A, Figure S14). Isolation and structural analysis of 17a revealed that it consisted of an inseparable mixture of five interconverting compounds, designated 17a-i to 17a-v (experimental details are explained in the SI). This equilibrium is likely attributable to the presence of a hemiacetal moiety in the chemical structures.

Next, the equilibrium mixture (17a-i to 17a-v) was incubated with CfvK. The reaction proceeded rapidly and predominantly produced 6a (Figure B). Notably, under nonenzymatic conditions, minor spontaneous dehydration occurred, producing 18a as the major product with a different selectivity from that of the enzymatic reaction. This spontaneous dehydration was accelerated by treatment with a Lewis acid in organic solvent (Figure S15). The minor formation of 6a under these nonenzymatic conditions likely accounts for the residual production of 1 in the strain AO-cfv-allΔK (Figure S16). These results suggest that CfvI converts the flavone substrate to a 2,3-dihydroxyflavanone intermediate. The resulting hemiacetal is in equilibrium with the corresponding keto form and exists as a mixture of five interconverting species (17a-i to 17a-v). CfvK likely catalyzes the dehydration of the 8-O-methoxy form(s) of these intermediates to produce 6a, which shares the same A-ring substitution pattern as 1 (Figure C). During the interconversion of the hemiacetal forms, the stereochemistry at C-2 was labile, whereas the stereochemistry at C-3 was retained. Soaking experiments using CfvK crystals subsequently confirmed that the C-3 configuration was 3S (as described below).

Structural and Functional Analysis of CfvK

To obtain detailed insights into the catalytic mechanism and substrate specificity of CfvK, the crystal structures of CfvK in an apo form and in complex with its reaction substrate were determined. We first solved the crystal structure of CfvK at a 1.3 Å resolution (Table S1). CfvK exhibits structural homology to SDH. CfvK forms a symmetric trimer composed of three single-domain subunits, consistent with the reported structure of SDH (Figure S17A). Trimer formation was also confirmed in solution (Figure S18). Each subunit aligns well with that of SDH with root-mean-square deviations of 1.78 Å (Figure S17B). Within each subunit, residues corresponding to the active site of SDHTyr30, Asp31, Tyr50, His85, and His110are conserved in CfvK as Gln30, Asp31, Tyr50, His86, and His111, respectively, suggesting functional conservation (Figure S19). Site-directed mutagenesis of these residues resulted in either loss or significant reduction of enzymatic activity, indicating that CfvK catalyzes dehydration via a mechanism similar to that of SDH (Figure S15). In the apo structure, a loop region spanning Ile48 to Lys58 (hereafter referred to as the Y loop) adopts two conformations, likely corresponding to open and closed forms induced by substrate binding (Figure S20). The apo crystal of CfvK was soaked with an equilibrium mixture of 17a-i to 17a-v, and X-ray crystallographic analyses were performed at time points ranging from 10 s to 30 min. At 10 s and 30 min, clear electron density corresponding to 6a was observed (Figure S21A), consistent with the structure obtained by the cocrystallization of CfvK with 6a (Figure S21B). Among the residues targeted in mutagenesis, all except Asp31 and Tyr50 were located at positions capable of interacting with product 6a. In contrast, CfvK adopts an open conformation in which the side chain of Tyr50 is oriented away from the active site (>7 Å from 6a), suggesting that this conformation did not represent a substrate-recognition state.

Among the variants, the Y50F variant retained a residual activity (Figure B). To further investigate substrate recognition, an apo-form crystal of the Y50F variant was soaked with a mixture of 17a-i to 17a-v. X-ray crystallographic analyses were conducted at time points ranging from 10 s to 60 min. At 2 min, clear electron density corresponding to substrate 17a-iv was observed (Figure S21C). At 5 and 10 min, the striction between rings B and C of the substrate became indistinct, and at 30 and 60 min, electron densities corresponding to both substrate 17a-iv and product 6a were detected (Figure S21D). Cocrystallization of the Y50F variant with the substrate mixture also yielded a structure with 6a bound in the active site (Figure S21E). Upon soaking, clear electron density corresponding to the bound substrate became sufficiently clear to assign an S configuration at C-3. Notably, 17a-iv binds in the same pocket as 6a but in the opposite orientation. The Y-loop can be modeled in both open and closed conformations, as observed in the apo structure of the wild-type enzyme.

Analysis of the substrate-bound structure revealed key interactions within the active site (Figure D, Figure S22). His86 and His111 interacted with the 2-hydroxy group via a hydrogen bonding network that involved Asp31. Gln30 formed a hydrogen bond with the 3-hydroxy group. Trp162 and Trp71 interacted with the B-ring through π–π interactions. Glu107 formed a hydrogen bond with the 5-hydroxy group. Notably, Phe121′ from a neighboring subunit extended into the hydrophobic pocket formed by Ile155, Met158, and Trp162 and was positioned to interact with the 8-methoxy group (Figure D, E).

To examine the interatomic interactions in wild-type CfvK with 17a-iv, 500 ns molecular dynamics (MD) simulations were conducted (Figure S23–S25). The simulations showed that the interactions around 17a-iv found in the enzyme–substrate complex were essentially maintained across all monomeric trajectories in both the wild type (WT) and the Y50F variant. When comparing the contact frequencies of the atomic pairs around 17a-iv (Figure S26) between the WT and Y50F variant, the interactions around 17a-iv found in the crystal structure of the Y50F variant (9WW0) were almost consistent. In particular, the interactions involving Gln30, Asp31, His86, and His111 were observed frequently. During the simulation of WT, Tyr50 was in contact with C-3 or carbonyl group atom at C-4 in approximately 50% of the monomeric trajectories, with the interactions around 17a-iv maintained. For both the results of WT and Y50F, Trp162 was frequently observed near the B-ring of 17a-iv, while Trp71 was away from that ring and sometimes close to Gln30. Glu107 was in contact with not only 5-hydroxy group but also with C-6. Phe121 was close to the terminal methyl group at 8-methoxy in approximately 40%. No significant difference in the structural fluctuations of the Y-loop was observed between the WT and Y50F variant (Figure S27), suggesting that the motion of residues around the active site is also similar between them.

To evaluate the functional importance of these interactions, site-directed mutagenesis was performed. In both the W162A and W71A variants, production of 6a decreased with the substrate 17a-i–v remaining in the solution, suggesting the reduction in enzymatic activity (Figure S15). In the E107A variant, the production of the 8-methoxy compound slightly decreased, whereas the amount of the 6-methoxy product increased (Figure B, F). Notably, the rate of 6-methoxy product formation was greater than that observed under nonenzymatic conditions, indicating that its formation was enzyme-mediated (Figure F). Together with the observation that Glu107, as well as the water molecule that forms a hydrogen bond with the 7-hydroxy group of the substrate, occupies the available space and could potentially interfere with the 6-methoxy group (Figure E), we propose that Glu107 plays a key role in excluding 6-methoxy-substituted substrates from the active site. Furthermore, we successfully altered the selectivity of CfvK through the E107A/F121A mutations, suggesting that Phe121 contributes to selective recognition of the 8-methoxy group, likely via van der Waals interactions (Figure D, E).

Collectively, these results provide detailed insights into the substrate recognition and catalytic mechanism of CfvK based on structural analysis, substrate soaking experiments, MD simulations, and mutational studies.

3-O-Methylation and Halogenation by CfvH and CfvB

To examine the final steps in the biosynthesis of 1, we conducted feeding experiments by employing AO-cfvB. Among the substrates tested, AO-cfvB converted only 5a and 5 and did not process early intermediates in the pathway (Figure S28). These findings indicate that the final steps of the biosynthetic pathway of 1 proceed through 3-O-methylation followed by halogenation and that the 7-methoxy group has a crucial role in substrate recognition by both CfvH and CfvB (Figure A). Consistently, conversions of 6a and 6 by AO-cfvH were also confirmed (Figure S29).

Analysis of Bypass and Shunt Pathways

Upstream enzymes in the biosynthetic pathway, such as CfvG and CfvD, exhibited broad substrate tolerance, suggesting the possible existence of bypass routes to 1. A comprehensive summary of each enzyme substrate specificity is provided in SI. When S-2′-hydroxypinocembrin (19a) and 2′-hydroxychrysine (20a)both identified in feeding experiments and not belonging to the main pathwaywere fed to AO-cfv-allΔA, both compounds were converted to 1, supporting the presence of a bypass pathway (Figure S6).

To trace the pathway originating from 19a, we incubated 19a with CfvD, which was rapidly converted to 20a (Figure S13J). Compounds 21a and 21 were also converted into 22a and 22, respectively, by CfvD (Figure S13H,I). When 22a and 22 were fed to AO-cfvIK, 22 was converted to 6, whereas 22a was not converted (Figure S30). This finding suggests that the 7-O-methyl group is important for substrate recognition by CfvI, particularly for compounds bearing an 8-methoxy group. Moreover, 22a was methylated by CfvC to form 22, thereby entering the bypass pathway (Figure S8). In addition, 21a/22a and 21/22 were obtained by bioconversion of 19a and 20a using AO-cfvJE and AO-cfvJEC, respectively (Figure S31). Together, these findings support the existence of a bypass route from 19a to 1 that proceeds without the two A-ring isomerization steps present in the main pathway.

In addition to the above bypath route, shunt metabolites that were not converted to 1 were also observed. For example, 7 and 7a were not converted into 1 (Figure S6). Interestingly, heterologous expression and metabolite analysis revealed that compounds bearing methoxy groups at both the C-6 and C-8 positions accumulated alongside products of the main pathway in strains expressing CfvC, CfvJ, CfvE, and CfvF. To examine whether CfvJ and CfvE can introduce a second methoxy group, compounds 8a, 9a, 13a, and 16a were fed to AO-cfvJE. The corresponding 6,8-dimethoxy derivatives10a, 23a, 14a, and 24awere obtained, confirming that CfvJ and CfvE are capable of introducing a second methoxy group (Figure S31). These results indicate that enzyme promiscuity contributes not only to the formation of bypass routes but also to the generation of shunt metabolites in the biosynthetic network (Figure S32).

Discussion

The biosynthetic pathway of 1 involves two distinct equilibrium states generated by the multitasking isomerase CfvF and the hemiacetal-forming oxygenase CfvI. The latter isomeric equilibrium of 2,3-dihydroxyflavanones, 17a-i–17a-v, is reminiscent of that known in the reactions of plant C-glucosyltransferases, suggesting that similar flavonoid biosynthetic strategies have convergently evolved in plants and fungi. Selective conversions of specific chemical species existing in equilibrium by downstream enzymes enabled the biosynthetic pathway to proceed toward the final product 1 (Figure ).

5.

Complete biosynthetic pathway for 1.

CfvF was found to catalyze both the stereoselective cyclization of chalcones and the ring opening of flavanones (Figure S11). Considering that chalcones were not detected during the interconversion of 8- and 6-methoxyflavanones, the equilibrium of this reaction is likely biased toward the cyclization direction, as observed for bacterial and plant CHIs. Notably, this enzyme appears not to distinguish between the 8- and 6-methoxyflavanones. As a result, an apparent equilibrium is established between these flavanones. The downstream enzymes selectively accept 6-methoxyflavanones, thereby driving the pathway toward the formation of 1. Although we observed the formation of both 9 and 16 during the conversion of 13 to 8 by AO-cfvGD (Figure H), CfvG likely functions as a gatekeeper of the pathway because this enzyme prefers flavanones over flavones as substrates (Figure F). In contrast, the 2′-hydroxyflavanones produced by CfvG serve as better substrates for CfvD than do 13 and 13a (Figure G). This preference is even more pronounced in the absence of the 7-O-methyl group installed by CfvC.

To our knowledge, CfvI was identified as the first enzyme capable of oxidizing the chemically inert C2–C3 double bond of flavones. The resulting 2,3-dihydroxyflavanone then underwent dehydration catalyzed by CfvK. While carboxylesterase-family enzymes catalyze a similar dehydration reaction in plant flavonoid biosynthesis, CfvK belongs to the NTF2-like domain superfamily. The observation that plants and fungi employ distinct protein families to catalyze a similar transformation further supports the possibility of convergent evolution of flavonoid biosynthesis across biological kingdoms. Based on the interactions observed in the structure complexed with the substrate, we currently hypothesize that dehydration proceeds through a mechanism similar to that of SDH: His86 or His111 likely functions as a general acid that facilitates the elimination of a water molecule, while Tyr50 is likely involved in substrate enolization. By analogy with SDH, His86 may also function as a general base involved in the enolization step. CfvK exhibited remarkable precision because it selectively recognized only one species (17a-iv) among the five equilibrium intermediates generated by CfvI, resulting in the production of a single defined product. We successfully obtained the cocrystal structure of the enzyme–substrate complex, revealing the structural basis for substrate recognition. Notably, Phe121 of one subunit invaded into the active site of the neighboring protomer consisting of the CfvK trimer and formed a hydrophobic pocket with Ile155, Met158, and Trp162 to accommodate the 8-methoxy group. Such intersubunit amino acid invasion into the active site for substrate recognition was not observed in structural homologues such as SDH. The construction of the flavonol scaffold by CfvI and CfvK via the oxidation of flavone and subsequent dehydration of the resultant 2,3-dihydroxyflavanone constitutes a novel biosynthetic route, distinct from the plant-derived pathway involving the 3-hydroxylation of flavanone followed by the dehydrogenation of the resultant flavanol.

In the biosynthesis of 1, the A-ring undergoes a conformational switch from the 8-methoxy form to the 6-methoxy form and is subsequently converted back to the 8-methoxy form in the late stage of the pathway. In A. oryzae, the 8-methoxy form is subject to demethylation; however, the substrate specificity of the endogenous demethylase in A. oryzae closely resembles that of CfvC (Figure S33), allowing CfvC to outcompete the host enzyme and thereby contribute to the heterologous production of 1. Interestingly, although the 6-methoxy intermediate is not demethylated in A. oryzae, it tends to undergo B-ring oxidation by endogenous enzymes. This susceptibility to B-ring oxidation in the 6-methoxy form was also observed in the 1-producing fungus A. candidus MEFC1001. These findings suggest that the dynamic equilibrium between the 6-methoxy and 8-methoxy forms may serve as a strategic mechanism to circumvent interference from endogenous enzymes during biosynthesis. Further investigation of the biological significance of this pathway modulation would be of considerable interest.

A comprehensive analysis of substrate specificity across all the enzymes revealed that those acting in the upstream steps of the biosynthetic pathway tend to exhibit broad substrate tolerance. These findings enabled the construction of cryptic bypass and shunt pathways that are not utilized in native biosynthesis, thereby demonstrating the potential of biosynthetic machinery for 1 as a promising toolbox for synthetic biology approaches to flavonoid production. As filamentous fungi such as A. oryzae are attracting increasing attention due to their capacity to accommodate biosynthetic enzymes from diverse sources, including plants, − the integration of fungal and plant-derived enzymes into artificial pathways for the production of unnatural flavonoids may also be feasible. By combining heterologous expression strains developed in this study with both in vivo and in vitro enzymatic conversion, we successfully established a rational scheme for the preparation of 46 flavonoids, including 8 novel derivatives (namely 6a, 11a, 12a, 17a, 17, 18a, 23a, and 25) and rare phytochemical flavonoids, such as 9a, 9, 14a, 21a, and 26. Most of them are not commercially available or are available from only a limited number of suppliers. Notably, several of these compounds possess rare structural features such as 2′-hydroxylationuncommon in plant-derived flavonoidsand fully substituted A-rings, indicating their utility for the production of rare flavonoids. These results underscore the value of precise biosynthetic pathway elucidation for molecular design and diversification.

Conclusion

In summary, we elucidated the complete biosynthetic pathway of 1 through a comprehensive approach, including stepwise reconstruction of the biosynthetic pathway, in vivo and in vitro enzymatic conversion, and evaluation of biosynthetic intermediates using a transformant expressing all tailoring enzymes. Biosynthetic investigations have become increasingly vital for the synthetic biology-driven production of high-value compounds. For the reconstitution of pathways leading to desired molecules, it is essential to accumulate precise knowledge not only of the main routes but also of latent shunt and bypass pathways, even though such branching routes often receive limited attention in biosynthetic research. Although the structural resemblance of 1 to plant-derived flavonoids may obscure its biosynthetic origin, 1 is synthesized via a remarkably intricate sequence of enzymatic transformations. The comprehensive elucidation of its biosynthetic pathway highlights the critical importance of dissecting each stage of biosynthesis in detail. Such in-depth investigations not only elucidate biosynthetic mechanisms but also facilitate the rational design of novel biosynthetic routes.

Crystal structure data that support the findings of this study have been deposited in the Protein Data Bank with the PDB IDs: 9WVP (CfvKWT apo), 9WVQ (CfvKWT 17a-i to 17a-v 30 min), 9WVR (CfvKWT/6a), 9WW0 (CfvKY50F 17a-i to 17a-v 2 min), 9WW2 (CfvKY50F 17a-i to 17a-v 30 min), 9WVS (CfvKY50F/17a-i to 17a-v). The initial coordinates, simulation input files, and the final output trajectory files of MD simulations have been submitted to the Biological Structure Model Archive (BSM-Arc) under BSM-ID BSM00091 [https://bsma.pdbj.org/entry/91].

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/jacs.6c13856.

• Experimental details, materials, methods, supporting tables and figures, and NMR spectra (PDF)

All authors have given approval to the final version of the manuscript.

This work was supported by JSPS KAKENHI (23K24038 [T.A.], 25H01566 [T.O.], JP25KJ0585 [S.F.], JP23K07918 [H.H.], and JP25K01957 [E.K.]); JST FOREST Program (JPMJFR205W [T.A.]); JST-ALCA-Next Japan Grant Number JPMJAN23D4 (T.A.) from the Japan Science and Technology Agency; JST SPRING, Grant Number JPMJSP2114 [S.F.]; AMED-CREST (JP22gm1610007 [T.A.]); and Platform Project for Supporting Drug Discovery and Life Science Research (Basis for Supporting Innovative Drug Discovery and Life Science Research [BINDS]) from AMED [JP25ama121038 (T.A.), JP25ama121040 (T.O.), JP25ama121023 for Mi.I. and JP25ama121001 (support numbers 6105 and 7347)]. This study used the computational resources of the Yokohama City University, Tsurumi campus, and of Fugaku provided by RIKEN through the HPCI System Research Project (Project ID: hp240282 [M.In.]).

The authors declare no competing financial interest.