Enzyme-Catalyzed Azepinoindole Formation in Clavine Alkaloid Biosynthesis

Abstract

(−)-Aurantioclavine (1), which contains a characteristic seven-membered ring fused to an indole ring, belongs to the azepinoindole class of fungal clavine alkaloids. Here we show that, starting from a 4-dimethylallyl-l-tryptophan precursor, a flavin adenine dinucleotide (FAD)-binding oxidase and a catalase-like heme-containing protein, are involved in the biosynthesis of 1. The function of these two enzymes were characterized by heterologous expression, in vitro characterization and deuterium-labeling experiments.

GRAPHICAL ABSTRACT

The indole-containing clavine alkaloids are produced by the Clavicipitaceae and Trichocomaceae families of fungi.¹ Their natural and semisynthetic derivatives show potent pharmacological activities and have applications in central nervous system disorders, cancer and infectious diseases.² Based on their structures, they are grouped into the tetracyclic ergolines, tricyclic secoergolines (lacking a cyclized D-ring) and rearranged (herein referred to as azepinoindole) classes.^1a (−)-Aurantioclavine (1) is a representative clavine alkaloid that belongs to the azepinoindole class (Figure 1).³ The characteristic seven-membered ring fused to indole in 1 is the precursor to many other alkaloids, which are produced by diverse biosynthetic pathways. For example, 1 can couple heterodimerically to tryptamine to form the complex polycyclic scaffold of the communesins, anti-insecticidal alkaloids.^3–4 Many total synthesis efforts of 1, and indole alkaloids derived from it, have been inspired by their complex structures.⁵

Figure 1.

Structure of (−)-aurantioclavine (1) and the diverse biosynthetic pathways in clavine alkaloid biosynthesis.

Biosynthetically, compound 1 is proposed to be derived from l-tryptophan after an initial C4-prenylation step to give 4-dimethylallyl-l-tryptophan (4-l-DMAT, 2). N-methylation of the amino group in 2 forms N-methyl-dimethylallyl-l-tryptophan (N-Me-l-DMAT, 3). These species are precursors to both 1 and tricyclic secoergolines such as chanoclavine-I (Figure 1). The transformation of 3 to chanoclavine-I has been shown to involve oxidation of 3, decarboxylation and cyclization to form the six-membered C ring.⁶

The transformation of 2 to 1, involving the formation of a seven membered C ring, therefore requires a different cyclization strategy. In the communesin biosynthetic gene cluster, a FAD-binding oxidase (CnsA) and a catalase-like heme-containing protein (CnsD), were proposed to be involved in the generation of 1.^4a The CnsA oxidase contains a FAD-binding domain and a berberine bridge enzyme (BBE)-like domain, while the CnsD enzyme has a catalase-like domain that was initially proposed to remove hydrogen peroxide generated from CnsA catalyzed oxidation. Genetic deletion of cnsA led to the accumulation of 2,^4a indicating the essential role of CnsA in the formation of the C–N bond between the α-amino group and the benzylic carbon. However, the biochemical basis for the formation of the azepinoindole ring system remains unresolved. We report here the characterization of the Cns enzymes that form the azepinoindole framework by heterologous expression and in vitro characterization. Furthermore, deuterium-labeling experiments for CnsD were performed to investigate the mechanism of this catalase-like heme-containing protein.

To clarify the role of the catalase-like heme-containing protein CnsD, we inactivated cnsD in Penicillium expansum NRRL 976 using double-crossover recombination with a hygromycin-resistant gene hyg as a marker (Figure S1). The ΔcnsD mutant strain showed the complete abolishment of all communesin products and the accumulation of trans-clavicipitic acid 4 (Scheme 1, Figure 2A, Table S6 and Figures S25–S29).^{5d, 7} The retention of the carboxylic acid group in 4 suggests the role of CnsD is likely to perform decarboxylation to yield 1. CnsA may be responsible for the oxidative cyclization of 2 to generate 4 (Scheme 1). Hence, CnsD does not, as previously suggested,^4a only function as a catalase to remove hydrogen peroxide generated by CnsA, but also is an enzyme in the biosynthetic pathway.

Scheme 1.

The biosynthetic pathway of (−)-aurantioclavine (1) and (−)-methyl-aurantioclavine (6).

Figure 2.

(A) LC-MS analysis of metabolites extracted from P. expansum NRRL 976 wild type, which produces communesin B (ComB) and the ∆cnsD mutant. (B) LC-MS analysis of the co-expression of different combinations of dmaW, cnsA, cnsD and easF in S. cerevisiae.

To identify the functions of CnsD and the FAD-binding oxidase CnsA, heterologous reconstitution in Saccharomyces cerevisiae was performed. First, we expressed CnsA and CnsD in yeast by cloning the cDNA from P. expansum. The 4-dimethylallyl-l-tryptophan transferase DmaW from P. expansum was introduced to generate the substrate 4-l-DMAT (2). Compared to S. cerevisiae expressing dmaW, which only produces 2, coexpressing dmaW and cnsA led to the production of 4 (Figure 2B, ii & iv). Furthermore, upon feeding 2 to S. cerevisiae expressing cnsA, we observed the conversion of 2 into 4 (Figure 3A, i & iii). Recombinant CnsA with an N-terminal FLAG-tag was purified from S. cerevisiae (Figure S3). Incubation of the purified enzyme with FAD and 2 led to the formation of 4 (Figure S6). These results demonstrate that CnsA is a (−)-trans-clavicipitic acid synthase.

Figure 3.

Verification of the function of the FAD-binding oxidase CnsA. (A) LC-MS analysis of S. cerevisiae expressing cnsA supplemented with 2 or 3; (B) Proposed mechanism of CnsA.

We propose that the reaction catalyzed by CnsA involves the base-catalyzed removal of a benzylic C-10 hydrogen and capture of the N-1 hydrogen as a hydride by the FAD cofactor following double-bond isomerization (Figure 3B).⁸ The resulting electrophilic C-10 is then attacked by the amino group to yield the azepinoindole scaffold.

To characterize the function of CnsD, we co-expressed dmaW/cnsA/cnsD in S. cerevisiae, and the production of 1 was observed, suggesting a role for CnsD in catalyzing the decarboxylation of 4 (Figure 2B, v) as indicated from the knockout studies. Recombinant CnsD was purified from E. coli BL21 supplemented with δ-aminolevulinic acid. The enzyme had a dark brown color and a UV-Vis absorbance at 405 nm, suggesting the presence of heme in the protein (Figure S5).⁹ Incubation of 4 with CnsD resulted in the formation of 1, confirming the role of CnsD as a trans-clavicipitic acid decarboxylase (Figure 4, i & ii).

Figure 4.

Verification of the function of the catalase-like heme-containing protein CnsD. LC-MS analysis of in vitro assay of 10 µM CnsD and 40 µM substrates 4 and 5 (100 µM ascorbate).

Next, we tested whether CnsA and CnsD can accept N-methylated substrates. The easF cDNA encoding an l-DMAT methyltransferase¹⁰ was cloned from Aspergillus fumigatus Af293. Co-expression of dmaW/easF/cnsA in S. cerevisiae led to the production of trans-N-methyl-clavicipitic acid 5 (m/z 285 [M+H]⁺), a new compound (Scheme 1; Figure 2B, vii; Table S7 and Figures S30–S35). Likewise, upon feeding 3 to S. cerevisiae expressing cnsA, the conversion of 3 to 5 was observed (Figure 3A, ii & iv). Furthermore, co-expression of CnsD with dmaW/easF/cnsA led to the production of (−)-methyl-aurantioclavine (6, m/z 241 [M+H]⁺) (Figure 2B, viii; Table S8 and Figures S36–S41). As further confirmation, direct incubation of purified CnsD with 5 led to the formation of 6 (Figure 4, iv & v). These results indicated that both CnsA and CnsD can accept the N-methylated substrates, 3 and 5, respectively.

We also tested the selectivity of CnsD for cis versus trans substrates. To do this, the cis-clavicipitic acid (7) and cis-methyl clavicipitic acid (8) (Scheme 1), obtained from a ∆cnsD and Aspergillus nidulans in vivo system (this information can be found in the general methods section of the Supporting Information), were incubated with CnsD. No activity of CnsD was observed with either 7 or 8 (Figure S8B). Indeed, 7 and 8 were confirmed to be shunt products, which were generated when feeding the corresponding trans forms (4 and 5) to the A. nidulans A1145 strain (Figure S11). The results indicated that CnsD only accepts the trans forms of clavicipitic acid derivatives as substrates.

To understand the decarboxylation mechanism catalyzed by CnsD, 4 was incubated with CnsD in deuterated water. The mass signal of product 1a (m/z 228 [M + H]⁺) showed the addition of 1 Da, suggesting that one deuterium atom was incorporated from D₂O into 1 (Figure S8A). To verify the position of the deuterium in the product, 1a was purified from a 10 mL scale in vitro reaction with D₂O. The ¹H NMR spectrum of 1a showed the disappearance of H-5α and decoupled signals of H-5β and H-4 (Figure 5A & S18), indicating that the deuterium is in the α-orientation, which is the same stereochemistry as the carboxylic acid in 4. Next, to test if H-7 is involved in the CnsD catalyzed C-5 decarboxylation (Figure S15), we prepared [1,1-²H₂]-DMAPP¹¹ as a substrate and performed a one-pot reaction in the presence of DmaW, CnsA and CnsD. The mass signal (m/z 228 [M + H]⁺) of the product, 1b, showed a mass 1 Da larger than 1 indicating the retention of one deuterium from the labeled DMAPP. Thus, CnsD does not abstract a C-7 hydrogen from 4a during the decarboxylation (Figure 5B).

Figure 5.

(A) Comparison of the ¹H NMR spectrum of 1a, purified from a 10 mL scale in vitro reaction of 4 with CnsD incubated in deuterated water, and 1. In the ¹H NMR spectrum of 1a, the ratio of 1a to 1 is 5:1, based on integrations; (B) LC-MS analysis of the in vitro assay of [1,1-²H₂]-DMAPP and l-tryptophan with Cns enzyme combinations; (C) Proposed mechanism of CnsD.

Therefore, we propose that CnsD first abstracts an electron from the C-5 carboxylate, which triggers decarboxylation to generate a C-5 radical intermediate a (Figure 5C). Hydrogen atom delivery from the enzyme to the C-5 radical forms 1, and reductive quenching of the protein radical regenerates the enzyme. With regard to the amino acid acting as the reducing agent, tyrosine or cysteine are possible candidates. Both have a water-exchangeable proton, which could serve as the reductive quencher. This enzyme-mediated reduction is analogous to that proposed for the epimerization of carbapenem by CarC.¹²

We further characterized the biochemical properties of CnsD. The enzyme showed weak catalase activity (9.9 × 10² µmol/min/mL) compared to the typical catalase from bovine liver (5.3 × 10⁴ µmol/min/mL) (Supporting Information). A heme group was confirmed to be necessary for the activity of CnsD, as the protein purified from E. coli BL-21 cultured in minimal medium without δ-aminolevulinic acid showed no activity with 4 (Figures S8A, iii). To test for the need for O₂, CnsD was incubated with 4 under anaerobic conditions. No reaction was observed (Figure S13), suggesting the participation of oxygen in heme activation. The addition of hydrogen peroxide to the reaction under anaerobic conditions recovered the formation of 1, indicating that CnsD can be activated by hydrogen peroxide, which is consistent with a catalase mechanism.¹³ We also observed that the addition of reducing agents, such as ascorbate, NADPH and glutathione, facilitated the reaction to form 1 by CnsD (Figure 4, iii & S14). This finding suggests the requirement of reducing agents in the ferrous state of the iron in the heme for the activation of O₂.¹⁴

In the biosynthesis of ergoline alkaloids (Figure 1), the conversion of 3 to chanoclavine-I is catalyzed by two enzymes, chanoclavine-I synthase (EasE, a FAD-binding oxidase) and a catalase-like heme-containing protein (EasC).^{6, 15} CnsA and CnsD share moderate sequence similarities to EasE (identity/similarity, 51%/66%) and EasC (identity/similarity, 60%/72%), respectively. The role of these Eas enzymes and their catalytic mechanisms were reported in a recent biochemical study showing that EasE generates a 1,3-diene product from 3 and EasC performs oxidative cyclization to form C ring in the ergoline scaffold.¹⁶ Here, we demonstrate that the CnsA oxidase performs a regiospecific dehydrogenation of 2, while the CnsD catalase-like heme-containing protein functions as a decarboxylase to form the azepinoindole ring system in 1. Thus, despite the sequence similarities between these two pairs of enzymes, they carry out significantly different transformations.

In conclusion, we have elucidated the role of CnsA, a FAD-binding oxidase, and CnsD, a catalase-like heme-containing protein, which are involved in the generation of the azepinoindole scaffold in clavine alkaloid biosynthesis. This work provides enzymatic tools for synthesizing 1 for derivatization and to expand chemical diversity of clavine alkaloids.