Biosynthesis of Terpenoid-Pyrrolobenzoxazine Hybrid Natural Product CJ-12662

Abstract

The fungal natural product CJ-12662 is a structurally complex terpene-amino acid hybrid, and is a potent anthelmintic compound. The biosynthetic pathway of CJ-12662 is elucidated based on metabolite analysis from heterologous expression. We demonstrate the terpene portion is derived from successive P450-catalyzed oxidations of amorpha-4,11-diene, while three flavin-dependent enzymes are involved in morphing the esterified tryptophan into a chlorinated pyrrolobenzoxazine, utilizing a cascaded [1,2]-Meisenheimer rearrangement.

Graphical Abstract

We have uncovered the biosynthetic pathway of the terpene-amino acid hybrid CJ-12662. In particular, the biosynthesis involves two different oxidation strategies (one-electron or two-electron oxidation routes) for the modifications of each unit, reflecting its chemical reactivity. Finally, ThmM catalyzes the formation of the N-oxide, which undergoes a [1,2]-Meisenheimer rearrangement to form the unique pyrrolobenzoxazine structure.

The diverse chemical structures of natural products have inspired generations of synthetic chemists to match nature’s ability to generate complexity.¹ Natural product biosynthetic pathways are concise and efficient,² and employ many of the powerful concepts adopted by chemists, including convergent synthesis,^3,4 redox tuning of substrates⁵ and cascaded reactions.⁶ Uncovering the enzymatic basis of biosynthesis can therefore lead to improved strategies for complexity generation and new biocatalysis candidates. Towards this goal, we have focused on the biosynthesis of CJ-12662 (1),^7,8 a potent anthelmintic and ectoparasiticidal natural product produced by Aspergillus fischeri var. thermomutatus ATCC 18618 (Figure 1A).^9–12 First isolated by Pfizer in 1991, CJ-12662 and the nonhalogenated derivative CJ-12663 (2), are pyrrolobenzoxazine terpenoids.⁹ The unique scaffold of 1 has led to several total synthesis efforts, and the pyrrolobenzoxazine core was constructed by consecutive oxidations of N-methyl-tryptophan.^9,13–15

Figure 1.

(A) Proposed retrobiosynthesis of 1 and 2. (B) Structures of representative terpene-polyketide and terpene-amino acid hybrids.

The structure of 1 is exemplary of the aforementioned complexity generation strategies used by nature. First, 1 is a hybrid terpene-amino acid natural product, of which the sesquiterpene component is esterified to the carboxylate group of a modified tryptophan. Such fusion of different families of fragments is a powerful method to access new chemical space and biological activity. Terpene-polyketide hybrid molecules, known as meroterpenoids, are well-documented in biosynthesis (e.g. fumagillin, Figure 1B).¹⁶ The convergence of amino acids and terpene scaffolds is however rare, with two recent examples in aculene A¹⁷ and flavunoidine.¹⁸ The clear chemical benefit of combining terpene and amino acid biosynthesis is enrichment of compounds with nitrogen atoms, which are rare among hydrocarbon-derived terpenoids.¹⁹ Second, both the terpene and the amino acid portions of 1 are heavily oxidized by redox enzymes. The sesquiterpene is a triply-oxygenated derivative of amorpha-4,11-diene, which is famously known to be the precursor of the plant natural product artemisinin.¹⁹ Oxidized amorpha-4,11-diene has not been documented in microbial natural products. The pyrrolobenzoxazine fragment of 1 is derived from oxidative cyclization and chlorination of tryptophan, and is considerably more complex compared to the amino acid pendants of flavunoidine and aculene A. Understanding the timing of the redox reactions relative to esterification can lead to enzymatic strategies for further structural diversification. Lastly, the defining benzoxazine ring in 1 is likely the result of a cascade [1,2]-Meisenheimer rearrangement reaction²⁰ resulting from N-oxygenation of the pyrroloindole precursor. While synthetic efforts have demonstrated the oxygen migration to form 1 can occur spontaneously after N-hydroxylation,⁹ a [1,2]-Meisenheimer rearrangement has not been reported in nature. Therefore, uncovering nature’s chemical logic to setup and trigger such cascade reaction is biosynthetically interesting.

Guided by the retrobiosynthetic proposal shown in Figure 1A, we located a putative biosynthesis cluster of CJ-12662 (1), thm, from the genome of 1-producing strain A. fischeri var. thermomutatus ATCC 18618. The cluster contains genes encoding a single-module nonribosomal peptide synthetase (NRPS, ThmA) and a terpene cyclase (TC, ThmB). The cluster also contains a plethora of genes encoding redox enzymes, including three cytochrome P450 monooxygenases (P450s, ThmF, ThmG and ThmI) and five flavin-dependent monooxygenases (FMOs, ThmC, ThmE, ThmH, ThmL and ThmM). The remaining enzymes include a methyltransferase (MT, ThmD) and an acyltransferase (AT, ThmJ). The cluster also encodes a noncanonical polyketide synthase (PKS, ThmK) with the domain arrangement of KS^o-AT-ACP (Figure 2A). However, the KS domain is likely inactive since the histidine residue of the catalytic triad²¹ is mutated to aspartic acid (Figure S3). RT-PCR analysis showed that all genes in the thm cluster are expressed under growth on Potato Dextrose Agar (PDA) (Figure S1A), of which 1 (16.5 mg/L) and 2 (13.5 mg/L) can be isolated (Tables S4–S5, Figures S1, S9–S13).⁹ Also isolated and characterized from the host strain are 7-chlorofischerindoline (3) (2.7 mg/L) and fischerindoline (4) (5.5 mg/L) (Tables S6–S7, Figures S1, S14–S25).¹²

Figure 2.

Biosynthesis of terpenoid core of 1. (A) The thm gene cluster; (B) GC-MS analysis of pathway intermediates via expressing different combinations of thm genes in A. nidulans; (C) LC-MS analysis of metabolites produced by A. nidulans transformed with different combinations of thm genes. Traces shown are selective-ion monitoring. Trace iv shows the biotransformation result when 9 is fed to A. nidualns expressing ThmA. ^acompounds 8 and 9 showed highest ion abundance at m/z 217, which were assigned to the [M+H-2H₂O]⁺ ion of 8 and [M+H-H₂O-CH₃COOH]⁺ ion of 9. (D) Proposed biosynthetic pathway from FPP (farnesyl diphosphate) to 10.

We used heterologous expression hosts to reconstitute the biosynthetic pathway. Expression of ThmB in either Saccharomyces cerevisiae RC01²² or Aspergillus nidulans A1145 ΔSTΔEM²³ resulted in the formation of a new compound 5 (m/z 204, 31.0 mg/L) as detected by GC-MS (Figure 2B, i, and Figure S4). Structure of 5 and its absolute configuration were confirmed as (−)-amorpha-4,11-diene by comparison of NMR and optical rotation data with those reported in the literature (Table S8, Figures S26–S27).²⁴ ThmB produces 5 exclusively in either host, and represents the first example of microbial amorphadiene synthase. ThmB is a typical class I terpene cyclase, with two metal-binding motifs conserved (DDVIE and NDLFSFNKE).²⁵ The closest homologues are other fungal terpene cyclases, such as fusicoccadiene synthase and ophiobolin synthase.^25,26 ThmB is predicted to adopt a class I fold with an α domain. Its alignment with the plant amorphadiene synthase (class II fold with an αβ domain architecture) showed an approximately 50% sequence coverage and less than 20% sequence identity, suggesting a divergent evolutionary origin.

Whereas the artemisinin pathway performs six-electron oxidation of C12 in 5 to give artemisinic acid,²⁷ triple hydroxylation at C1, C2 and C3 by P450s take place in biosynthesis of 1. To elucidate the sequence of reactions, we coexpressed ThmB with the P450 candidates in A. nidulans. Only coexpression of ThmB with ThmI afforded a new product with 16 mu increase in molecular weight (MW) (Figure 2B, ii). This product was isolated to be amorpha-4,11-diene-2-ol (6) (20.0 mg/L). The hydroxy group is located at C2 based on the HMBC correlations (Table S9, Figures S28–S34). The absolute configuration of C2 was determined to be S using Mosher’s method (Table S10, Figures S35–S38).²⁸ Further expression of ThmG gave 7 (18.0 mg/L), of which the structure was determined as amorpha-4,11-diene-2,3-diol (Table S11, Figures S39–S45). The relative configuration of 7 was determined from NOE interactions, and the absolute configuration was elucidated as 1R,2R by CD exciton chirality method²⁹ on its dibenzoyl derivative (Table S12, Figures S46–S48). This indicates that ThmG can regio- and stereoselectively hydroxylate C3 in 6 to produce 7 (Figure 2B, iii). Adding the remaining P450 ThmF led to hydroxylation of the bridgehead C1 and formation of amorpha-4,11-diene-1,2,3-triol (8) (Figure 2C, i) (61.0 mg/L), which was confirmed by comparison of the spectral data with those reported for 8 (Table S13, Figures S49- S54).³⁰

The next proposed step in preparation of the terpene fragment is acetylation of the C3 hydroxyl group. This is a possible protection step to shield the C3-OH group from the C2-OH esterification reaction. Coexpression of the acyltransferase candidate ThmJ, however, did not lead to acylation of 8 (Figure 2C, ii). This led us to consider the role of the noncanonical PKS ThmK (KS^o-AT-ACP) as a potential donor of the acetyl group. To test this hypothesis, we coexpressed ThmJ and ThmK together in the host that produced 8. This led to the complete conversion of 8 to 9 (5.0 mg/L). Structure of 9 was confirmed to be the desired 3-O-acetyl-8 (Table S14, Figures S55–S56).²⁹ While this shows ThmBFGIJK complete the biosynthesis of the terpene portion of 1, the seemingly circuitous route to install an acetyl group is unexpected. We propose the role of the AT domain in ThmK is to load the ACP phosphopantetheinyl thiol with an acetyl group, which is then transferred to 8 by ThmJ (Figure 2D). The requirement of a dedicated ThmK enzyme may be to control the flux of the substrate 9 for the subsequent steps; or to prevent untimely hydrolysis of acetyl-CoA by ThmJ in the absence of 8.

Having established the pathway to the fully modified terpene fragment 9, we next used the same reconstitution approach to identify pathways leading to the pyrrolobenzoxazine fragment. However, no tryptophan-derived metabolite can be detected when all combinations of the remaining genes encoding FMO and MT were expressed. This suggested that the esterification of 9 may occur with l-tryptophan as a substrate to form 10, while the subsequent modifications occur on the esterified product. To understand the role of the NRPS-like ThmA, we coexpressed the enzyme with those that produced 9, as well as singly expressed ThmA in A. nidulans followed with the feeding of 9 (100 μM final concentration). Both approaches led to the formation of 10, of which the molecular weight is consistent with the adduct between 9 and L-tryptophan. Full NMR characterization established the structure of 10 as shown in Figure 2D (Table S15, Figures S57–S63). When the partially hydroxylated terpene intermediates 6–8 were fed to A. nidulans expressing ThmA, we observed adducts that correspond to esterification of tryptophan (Figure S6). These results indirectly support the notion that acetylation of C3-OH in 9 serves to direct tryptophan esterification to the specific secondary alcohol at C2.

With 10 in hand, we next investigated the formation of the N-methylpyrroloindoline scaffold. Based on the biosynthesis of other indole alkaloid natural products such as asperlicin E,³¹ fumiquinazoline A,³² tryptoquialanine³³ and notoamide D,³⁴ an FMO-catalyzed 2,3-β-face epoxidation to facilitate α-amino attack at electrophilic C2 of the epoxyindole ring is a likely route (Figure 3B). The FMO ThmH shows moderate sequence identity (37%) to the asperlicin pathway enzyme AspB,³¹ and was first chosen to be coexpressed in the A. nidulans transformant that produces 10. The strain produced new compound 11 (Figure 3A, iii), which was structurally verified with NMR (Table S16, Figures S64–S70). An alternative mechanism of ThmH following epoxidation could be stereoselective epoxide opening driven by the indole nitrogen, followed by attack of the α-amino group on the C2 iminium to forge the pyrroloindoline ring system. Next, coexpression of the putative N-methyltransferase ThmD in the above strain led to the formation of fisherindoline (4) (Figure 3A, ii), which was characterized from the native host. ThmD function requires the pyrroloindoline ring, as direct coexpression of the enzyme in host that produced 10 did not lead to any methylated indole products.

Figure 3.

Biosynthesis of alkaloid core of 1. (A) LC-MS analysis of extracts from A. nidulans transformants expressing different combinations of thm genes. Traces shown are selective-ion monitoring. Traces iv-vii are results of biotransformation experiments in which the indicates compounds were supplemented to A. nidulans expressing the enzyme of interest; (B) Proposed biosynthetic pathway from 10 to 1 and 2.

The thm cluster encodes four additional flavin-dependent enzymes, ThmC, ThmE, ThmL and ThmM. One of these enzymes is proposed to be responsible for installing the C7 halogenation observed in 1. The halogenation step, however, likely takes place with 4 as substrate to give 7-chlorofischerindoline (3) since an electron-rich nitrogen is required for the aromatic halogenation at C7 in the indole ring.³⁵ Aromatic halogenation is generally catalyzed by flavin-dependent halogenases (FDHs) that contain a characteristic WxWxIP motif that is absent in FMOs.³⁶ While none of the four enzymes contains such motif, ThmC shows moderate sequence identity (57%) to PtmN³⁷, which was reported to catalyze chlorination on C7 of the indole ring during penitrem biosynthesis. To test the function of ThmC, we coexpressed ThmC with ThmABDFGHIJK in A. nidulans, which led to formation of 3 (Figure S8, v). The role of ThmC as a halogenase^37,38 was further confirmed from the bioconversion of fed 4 to 3 in A. nidulans expressing ThmC (Figure 3A, vi). However, feeding the pyrrolobenzoxazine 2 into the same strain did not lead to chlorination (Figure 3A, vii). Further substrate selectivity was assayed by feeding 10 or 11 to the transformant. 11 but not 10 was chlorinated upon expression of ThmC (Figure S7A, i, ii), confirming that the pyrroloindoline skeleton is required. Given that all related products (1–4) isolated from the native host are N-methylated, while both 2 and 4 are unchlorinated, we assigned chlorination to take place after N-methylation as shown in Figure 3B.

The last step in the biosynthesis of either 1 or 2 from 3 or 4, respectively, is the oxidation of the pyrroloindoline nitrogen to the corresponding N-oxide.³⁹ Formation of the N-oxide reverses the nucleophilic character of the nitrogen to a cationic and electrophilic one, which can drive the N-C bond breaking step.⁹ In the total synthesis of 1, chemical oxidation of the tertiary amine to form the polarized N-oxide led to a cascaded [1,2]-Meisenheimer rearrangement in a diastereospecific manner.⁹ Hence, one of the remaining FMO-like enzyme would be involved in generation of the N-oxide to trigger the same rearrangement. With this chemical logic, no additional enzyme would be required to form the six-member benzoxazine ring. To identify the enzyme, we coexpressed and screened the remaining FMOs including ThmE, ThmL, and ThmM using the A. nidulans that expresses ThmABCDFGHIJK and produces 3. Gratifyingly, coexpression of ThmM, which shows moderate sequence identity (41%) to AspB³¹, led to the formation of the pyrrolobenzoxazine 1 (Figure S8). To confirm this activity of ThmM, we performed feeding of either 3 or 4 to A. nidulans expressing ThmM, which led to the formation of 1 and 2, respectively (Figure 3A, iv and v). Furthermore, we expressed and purified ThmM as a recombinant protein from E. coli and directly assayed its activity in the presence of FAD and NADPH. The purified ThmM can indeed convert 3 and 4 to 1 and 2, respectively (Figure S5, A–C). These results therefore confirmed ThmM to be the last enzyme in the biosynthetic pathway of CJ-12662 and is responsible for triggering the [1,2]-Meisenheimer rearrangement that forms the unique pyrrolobenzoxazine structure (Figure 3B). The roles of the remaining FMO-like enzymes, ThmL and ThmE are unknown. Both enzymes appear not to be essential for reconstitution of the pathway.

Our biosynthetic analysis of CJ-12662 (1) showcased the different strategies nature employs to generate remarkable structure complexity. Formation of the terpenoid-alkaloid molecules involves early-stage functionalization of the sesquiterpene and late-stage modification of the acylated tryptophan, sandwiching the NRPS-dependent esterification step. It is notable that different redox strategies are used on the two stages of biosynthesis. In the maturation of amorphadiene portion, only one-electron, radical-based oxidation is employed, reflecting the relatively unactivated carbon atoms in the terpene precursor. In contrast, two-electron chemistry using the flavin-cofactor is used in modification of the tryptophan portion, consistent with its electron-rich feature. Finally, the cascading strategy to form the 1,2-benzoxazine is fully consistent with synthetic routes,⁹ and includes N-methylation to establish the tertiary nitrogen, followed by N-hydroxylation to reverse the polarity of the nucleophilic nitrogen. Full elucidation of the biosynthesis of CJ-12662 (1) therefore sets up discovery of additional hybrid natural products and establishes new enzymatic modifications to two familiar natural product building blocks, amorphadiene and tryptophan.