Genome Mining of Alkaloidal Terpenoids from a Hybrid Terpene and Nonribosomal Peptide Biosynthetic Pathway

Abstract

Biosynthetic pathways containing multiple core enzymes have potential to produce structurally complex natural products. Here we mined a fungal gene cluster that contains two predicted terpene cyclases (TCs) and a nonribosomal peptide synthetase (NRPS). We showed the flv pathway produces flavunoidine 1, an alkaloidal terpenoid. The core of 1 is a tetracyclic, cage-like and oxygenated sesquiterpene that is connected to dimethylcadaverine via a C-N bond, and is acylated with 5,5-dimethyl-l-pipecolate. The roles of all flv enzymes are established based on metabolite analysis from heterologous expression.

Graphical Abstract

Structural complexities of natural products (NPs) are generated by enzymes in the biosynthetic pathways.¹ Scaffolds assembled by core enzymes such as polyketide synthases (PKSs), nonribosomal peptide synthetases (NRPSs) or terpene cyclase (TCs), etc., can be morphed into complex NPs by accessory enzymes including transferases² and oxidoreductases,³ etc. In fungi, the combinations of different core enzymes in the same biosynthetic pathway, such as PKS/PKS,⁴ PKS/NRPS,⁵ PKS/TC (Figure 1A),⁶ can result in complex hybrid NPs unachievable with a single core enzyme. In contrast, biosynthetic pathways containing the combination of NRPS and TC have not been well-studied. While many metabolites are derived from prenylation of peptidyl cores via prenyltransferases,⁷ the use of a TC to generate a terpene core that is decorated by a NRPS is rare. However, bioinformatic scanning of sequenced fungal genomes suggests TC/NRPS hybrid clusters are common (Figure S2). Isolation of fungal aminoacylated terpenoids also suggests such hybrid molecules can be synthesized by fungi (Figure 1B).⁸ Recent characterization of the aculene A biosynthetic pathway demonstrates such collaboration between the TC and a single module NRPS.⁹ Based on these evidences, we believe there is significant potential in mining fungal TC/NRPS pathways for new NPs.

Figure 1.

Structures of fungal polyketides synthesized by collaborative efforts of core enzymes. (A) meroterpenoids derived from TC/PKS,^6b–d etc.; (B) compounds derived from TC and NRPS enzymes include ergokonin A^8a (proposed), aculene A,⁹ and flavunoidine 1 (this study).

Among the predicted gene clusters containing both TC and NRPS, we selected an uncharacterized nine-gene cluster conserved in several well-studied fungal species (Figure 2A, Figure S3, Table S1). The flv cluster in Aspergillus flavus encodes two TCs: FlvE and FlvF, with sequence similarity to trichodiene synthase¹⁰ and ophiobolin synthase,¹¹ respectively; and a single-module NRPS FlvI. This cluster also contains redox enzymes, including FlvB (short chain reductase, SDR), FlvC and FlvD (P450s), and a didomain enzyme FlvA. The N-terminal half of FlvA is predicted to be a PLP-dependent lyase,¹² while the C-terminal half is predicted to be a non-heme Fe, α-ketoglutarate dependent oxygenase (α–KG) (Figures S15–16).¹³ The remaining enzymes encoded in the gene cluster include FlvG, which is a homolog of ornithine decarboxylase; and FlvH, which has sequence similarity to histone lysine N-methyltransferase (N-MT).¹⁴ To investigate the metabolite that can be biosynthesized from the flv cluster, we used Aspergillus nidulans as a heterologous host to mine the pathway.

Figure 2.

Heterologous expression of flv pathway. (A) The flv gene cluster. (B) LC/MS analysis of extracts from A. nidulans expressing different combinations of flv genes. (C) The structure of hispidospermidin 4. (D) Proposed biosynthesis of 1 and 2 (For details see Figure S4).

When the entire flv gene cluster was introduced into A. nidulans, we detected and isolated a new metabolite 1 with molecular weight (MW) of 503 (Figure 2B, i) (1.2 mg/L). The structure of 1 was solved to be flavunoidine shown in Figure 1B (Table S5 and Figures S17–22). The tetracyclic cage was only previously found in the phospholipase C inhibitor hispidospermidin 4 (Figure 2C).¹⁵ The C7 axial trimethyl-spermidine substituent in 4 is replaced by N,N-dimethylcadaverine in 1, while the C10 position is hydroxylated and acylated with 5,5-dimethyl-l-pipecolate. 4 was the only NP with the same tetracyclic core, and has been the subject of total synthesis by Danishefsky,^16a–b Overman^16c and Sorenson,^16d–e etc. 1 does not display notable cytotoxicity, nor is it antifungal or antibacterial.

We first investigated formation of the dimethylpipecolate in 1. To probe the role of the NRPS FlvI, we expressed the other eight genes FlvA-H. This led to the absence of 1 but the emergence of 2 (MW 157) (0.7 mg/L) and 3 (MW 364) (0.9 mg/L) (Figure 2B, ii). NMR analysis showed that 2 is 5,5-dimethyl-l-pipecolic acid (Table S6 and Figures S23–27), while 3 is the unacylated precursor of 1 (Table S7 and Figures S28–33). The accumulation of these separate building blocks suggests the NRPS FlvI is responsible for esterifying 2 and 3 (Figure 2D). Indeed, feeding of 2 and 3 to A. nidulans expressing FlvI led to the biotransformation to 1 (Figure 2B, iii).

Pipecolate biosynthesis from lysine has been shown to involve a PLP-dependent enzyme and reductase,¹⁷ which led us to propose that the didomain enzyme FlvA and SDR FlvB may be involved in biosynthesis of 2. When FlvC-H were expressed, we only observed accumulation of 3 (Figure 2B, iv), while coexpression of FlvA and FlvB separately resulted in formation of 2 (Figure 2B, v). Individual expression of either FlvA or FlvB did not result in formation of 2. In addition, coexpression of FlvA, FlvB and FlvI, accompanied by feeding of 3, led to the production of 1 (Figure 2B, vi). These results implicate FlvA and FlvB in the biosynthesis of 2. We propose the PLP-dependent lyase¹² domain of FlvA catalyzes a γ-replacement reaction as shown in Figure 2D (detailed mechanism shown in Figure S4). l-O-acetyl-homoserine can bind and form the aldimine, which can undergo two proton abstraction steps to eliminate acetate and form the vinyl glycine quinonoid. This species can be attacked by α-keto-isovalerate to form a new C-C bond, which upon protonation and transaldimination can lead to release of the ketone 5 that can cyclize intramolecularly to yield the imine. The imine can be reduced by FlvB to yield the 6-carboxylated pipecolate 6. The C-terminal α-KG dependent oxygenase domain of FlvA is then proposed to catalyze the decarboxylation of 6 to yield 2. The mechanism, shown in Figure S4, can involve a radical decarboxylative route facilitated by an active site tyrosyl^{18, 19} or thiyl²⁰ radical generated from the Fe(IV)=O. Upon decarboxylation to generate the C4 radical, hydrogen delivery from the active residue forms 2, and reductive quenching of the radical carrier regenerates the enzyme.

We next investigated formation of the tetracyclic core and the unusual trans diaxial nitrogen and oxygen functionality in 3. We have established that coexpression of six enzymes FlvC-H can synthesize 3 (Figure 2A, iv). To examine the roles of the TCs (FlvE and FlvF), we removed either gene and analyzed the resulting metabolic profiles. Removing flvE abolished all related metabolites (Figure 3A, iii), suggesting its involvement in core synthesis; while removing flvF led to accumulation of a pair of metabolites 7a and 7b with the same MW (Figure 3, ii). 7a was purified and structurally determined (Figure 3C, Tables S8, Figures S34–39) to contain the same core as 3, but substituted at C7 by ethanolamine via an axial C-N bond. Based on data below for the stereoisomer pair 9a and 9b, we propose 7b is the C7 equatorial stereoisomer of 7a. Formation of both axial and equatorial isomers indicates the C-N bonds formed in 7a and 7b may be uncatalyzed. This also hints FlvF may be responsible for stereoselectively forming the C-N bond in 3.

Figure 3.

Biosynthesis of the core of 1. (A) LC-MS analysis of metabolites in vivo and in vitro assays; (B) Crystal structure of 10a; (C) Proposed biosynthetic pathway of 3. Observed shunt products are shown in the dashed box.

To analyze the function of core TC FlvE, we expressed the enzyme in Saccharomyces cerevisiae JHY651.²¹ GCMS analysis revealed a sesquiterpene product 8 (Figure S5). Purified FlvE also synthesized 8 using farnesyl diphosphate (FPP) (Figures S6 and S14). Isolation and characterization confirmed 8 to be (1R,4R,5S)-(+)-acoradiene.²² The (−)-enantiomer was previously isolated from plants.²³ We then coexpressed the P450s, FlvD and FlvC, with FlvE to determine if oxidative modifications of 8 can generate the core in 3. Coexpression of both P450s with FlvE in A. nidulans resulted in formation of 7a and 7b (Figure 3A, v), while coexpression of only FlvD and FlvE led to the C7-stereoisomers 9a (axial) and 9b (equatorial) (Figure 3a, iv) (Table S10 and Figures S42–53), which do not contain the C10 hydroxyl. Coexpression of FlvD-H without FlvC in A. nidulans resulted in the formation of 12 (Figure 3a, vi), which is the C10-deshydroxy version of 3. Collectively, these results implicate FlvC as the C10 hydroxylase (Figure 3C), while FlvD alone can oxidatively convert 8 into the tetracyclic cage, which nonenzymatically forms a C-N bond with ethanolamine in A. nidulans.

Feeding 8 to A. nidulans expressing FlvD-H, but without the TC FlvE, restored the otherwise abolished production of 12 (Figure S7), suggesting 8 is a precursor in the pathway. To analyze the function of FlvD in morphing 8, we overexpressed the enzyme in yeast and fed 8 to analyze biotransformation products. In addition to 9a and 9b, new metabolites 10a, 10b and 11 were detected (Figures 3a, vii and S8). 11 retains the carbon scaffold in 8, but with the C12, C13 diol (Table S12, Figures S60–65). 10a is substituted with an axial hydroxyl group at C7 (Figure 3C) (Table S11 and Figure S54–59). We obtained an X-ray crystal structure of 10a (Figure 3B), which confirmed the tetracyclic structure and allowed assignment of absolute stereochemistry of compounds. Based on the structures of 9a and 9b, we propose 10b is the equatorial stereoisomer at C7. Purified microsomes from yeast expressing FlvD converted 8 to 10a and 10b (Figure S9). No 9a or 9b was detected since no ethanolamine was present in the reaction. Performing the microsomal assay in H₂¹⁸O led to incorporation of labeled ¹⁸O into 10a and 10b, indicating the C7 hydroxyl groups are derived from water (Figure S10). We verified 9 and 10 are shunt products, as feeding these compounds to A. nidulans expressing FlvD-H without FlvE did not restore biosynthesis of 12 (Figure S7).

We propose 11 is derived from the nonenzymatic epoxide opening of 13, which can be formed stereoselectively from 8 by FlvD. 13 can undergo intramolecular [3+2]-cycloaddition between the olefin and epoxide to directly forge the tetracyclic core 14. This reaction may be assisted by an active site Lewis acid in FlvD and proceed in a step-wise mechanism.²⁴ This represents a very concise way to morph the terpene 8 into the caged core. The intermediate 14, which was not isolated in vivo nor in vitro, may be further oxidized at C7 by FlvD to yield a carbocation 15. 15 can then be quenched by nucleophiles such as water to yield 10a and 10b. Ethanolamine, which is biosynthesized by both yeast and A. nidulans,²⁵ may enter the active site of FlvD and quench 15 to yield 9a and 9b.

When the TC homolog FlvF is coexpressed, the nonenzymatic quenching is suppressed and dimethylcadaverine 16 can stereoselectively quench 15 to afford 12 (Figure 3A, vi). To test this, we coexpressed FlvD and FlvE and fed the strain with 16. This strain still synthesized only 9a and 9b, and no 12 was formed (Figure 3A, ix). Upon coexpression of FlvF with FlvD and FlvE, feeding of 16 led to formation of 12 (Figure 3, ix). We then performed an in vitro assay using yeast microsomes containing FlvD and purified FlvF (Figure S14) in the presence of 8 and 16. This reaction produced 12 and no shunt products were detected (Figure 3A, x). Excluding FlvF from this assay led to formation of only 10a and 10b, even in the presence of 16 (Figure S11). Directly adding 10 and 16 to FlvDF did not lead to formation of 12, confirming 10 is a shunt product from nonenzymatic quenching. The mechanism by which FlvF can enable the stereoselective C-N bond formation in 12 is unexpected for an enzyme annotated as TC. Moore reported an algal TC that can catalyze N-geranylation of l-Glu.²⁶ However, the mechanism here is different since the terpene substrate is not pyrophosphorylated. We propose FlvF may form a complex with FlvD and deliver 16 to the active site where 15 is generated. The mechanism of this reaction is under investigation.

The two remaining enzymes in the pathway, FlvH (N-MT) and FlvG (decarboxylase) are proposed to synthesize 16 from l-lysine in a two-step reaction, in which FlvH performs methylation to give 17, and is decarboxylated by FlvG to afford 16 (Figure 3C). When flvH was removed from A. nidulans that produces 3, trace amounts of 3 were formed (Figure S12). The titer of 3 can be restored upon feeding of 17 (Figure 3, viii). When 17 was fed to the same strain without flvG, biosynthesis of 3 was abolished and only 7a and 7b were observed. Feeding of 16 restored production of 3, establishing FlvG catalyzes decarboxylation of 17 to 16 (Figure S13).

In summary, we mined a fungal biosynthetic gene cluster that contained both TC and NRPS core enzymes, and discovered a new alkaloidal terpenoid 1. The tetracyclic core of 1 is synthesized by the TC FlvE and P450 FlvD, while a second TC FlvF is required for attachment of the C7 axial dimethylcadaverine. The NRPS acylates the terpenoid core with dimethylpipecolate. The unexpected structural features of 1 highlight potential of fungal genome mining using combinations of core biosynthetic enzymes as a criterion.