The spe gene cluster governs the biosynthesis of notoamides. The whole genome of A. ochraceus was sequenced, followed by bioinformatic analysis with antiSMASH,26 and a genomic region encoding homologs of the known fungal notoamide biosynthetic enzymes (not/not′) was discovered, which was designated the spe cluster (Fig. 2a and Table S1). Protein sequence alignment analysis showed that the Spes share a high amino acid sequence identity with known Nots, such as SpeA (NotE, 61%), SpeB (NotF, 69%), SpeC (NotG, 71%), SpeD (NotC, 75%), SpeE (NotD, 67%), SpeF (NotB, 71%), and SpeG (NotH, 68%).27 Then, the speC gene encoding P450 monooxygenase was deleted by homologous recombination from the genome to generate the A. ochraceus ΔspeC strain. Metabolic analysis of the AO-ΔspeC strain showed that alkaloids 1–2 were abolished, which confirmed that the spe cluster is responsible for notoamides biosynthesis (Fig. 2bi, iv). Next, we coexpressed seven genes, speABCDEFG, in A. nidulans A1145, and the cotransformant AN-speABCDEFG could produce alkaloids 1–2, indicating that SpeA-G are sufficient for the biosynthesis of both of the alkaloids (Fig. 2ci).
On the basis of the structures of and previous work on notoamides, the partial biosynthetic pathway of compounds 1–2 were speculated, as shown in Fig. 2D. The NRPS SpeA first synthesizes cyclo-L-tryptophan-L-proline. Subsequently, SpeB and SpeD catalyze the reverse prenyl transfer reaction on cyclo-L-tryptophan-L-proline and the normal prenyl transfer reaction on deoxybrevianamide E (7), respectively.14 SpeF epoxidizes 10, followed by semipinacol rearrangement to obtain notoamides C (11) and D (12).28 To ascertain the roles of the other enzymes in the biosynthesis of 1–2, we also individually knocked out the other functional genes (speE and speG) to obtain the corresponding mutants (Fig. 2b). For the AO-ΔspeC strain, compound 7 accumulated (Fig. S46 and S47; Table S10). Thus, SpeC acts as a 6-deoxybrevianamide E 6-hydroxylase.29 For the AO-ΔspeE and AO-ΔspeG strains, compounds 9 and 11/12 accumulated, respectively (Fig. 1a, S48, S49, S52-S55 and Tables S11, S13, and S14). However, none of the intermediates, such as notoamide T* and stephacidin A* proposed before, accumulated in the ΔspeCEG mutant strains (Fig. 1c). This further confirmed the presence of a new biosynthesis pathway for the notoamides.
FPMO SpeE is involved in the construction of the 2 H -pyran moiety of 10. SpeE, featuring both BBE and GlcD domains, was predicted to be a flavoprotein monooxygenase (FPMO) (Fig. S7).2, 30, 31 Tetrahydrocannabinolic acid (THCA) synthase, belonging to this family, could catalyze the oxidative cyclization of the monoterpene moiety in cannabigerolic acid (CBGA).32 Therefore, we hypothesized that SpeE might be involved in the construction of the 2H-pyran ring in notoamides. Then, four genes (speABCD) were introduced into the A. nidulans A1145 strain (Fig. 2cv).33, 34 The transformant generated one major product, which was identified as 9 (Table S11).23 Next, we integrated speE into the AN-speABCD strain, resulting in the production of a new major metabolite, notoamide E (10) (m/z 434 [M + H]+) (Fig. 2civ). Its structure was determined by NMR and X-ray diffraction analysis (CCDC 2034887, Fig. 3a, S50, S51, Tables S12 and S25).35 Therefore, SpeE can construct the 2,2-gem-dimethyl-dihydropyran moiety in 10. This finding was further confirmed by the fact that 9 was converted to 10 in cell-free extracts containing SpeE from A. nidulans (Fig. 3b). Then, the mechanism for the formation of the 2H-pyran ring in 10 could be rationalized as follows: concerted deprotonation of 6-OH in 9 and capture of the C-25 hydride ion by the FAD cofactor produce an ortho-quinomethide moiety in 10a.6, 36 Then, C-27 in 10a approaches O1′ via the rotation of the C-25-C-26 single bond, followed by 6π-electrocyclization to construct the 2H-pyran ring in 10 (Fig. 3c, S8, S9). However, we could not ensure that SpeE involved in the 6π-electrocyclization of 10a, since its activation barrier is only 10.2 kcal/mol (Table S19). In addition, when we fed 10 to the ΔspeC strain, the production of 1–2 was recovered, indicating that 10 is an on-pathway precursor for notoamide biosynthesis (Fig. S6); in turn, the formation of 2H-pyran occurs prior to the construction of the bicyclo[2.2.2]diazaoctane core.
FPMO SpeF dearomatized the pyrrole ring, producing β-epoxide 1 (15). Bioinformatically, SpeF belongs to group A FPMO.37 As predicted by the TMHMM program,38 SpeF is likely a membrane-associated protein that is thought to catalyze 2,3-β-epoxidation of 10, similar to its analog NotB.28 Indeed, the AN-speABCDEF cotransformant produced two new compounds (Fig. 2cii), which were readily identified as 11 and 12, respectively, via spectroscopic analysis (Fig. 1a, 4a, S52-S55; Tables S13, S14, S17; CCDC 2034889). Then, in the in vitro assay, 10 was also converted to 11–12 (Fig. 4b, S10).28 However, feeding 11–12 to the mutant AO-ΔsepC did not lead to the emergence of 1–2, which confirmed that 11–12 are shutter compounds28 and that their common precursor, 2R,3R-epoxide 1 (15) was proposed to be the real intermediate to 1–2 (Fig. 2d, S6).
To further elucidate the facial selectivity of the epoxidation of 10, a structure of SpeF complexed with FAD was generated using homology modeling for binding mode and function prediction. The crystal structure of CtdE (PDB: 7KPQ) sharing a relatively high (36.6%) sequence identity with SpeF was selected as a template for homology modeling.39 The major residues surrounding the active site of substrate 10 are shown in Fig. 4c. Among them, residues D54 and R118, corresponding to D60 and R122 in CtdE, respectively, are conserved and active residues. Similar to CtdE, FAD was well positioned on the β-face of 10, indicating that SpeF can catalyze the β-epoxidation of 10. In addition, residues I55, F223, M232, H245, and L401, which contribute to hydrophobic interactions with 10, might facilitate FlOOH attack of the β-face of C-3 of 10 to form 15. Moreover, residue H245 forms a π-π interaction with the ethylene group in 10 instead of a typical hydrogen bonding interaction between H229 and the carboxylate oxygen of the substrate in CtrE.39 The SepF variant H245A retained 25% activity with 10, indicating that the π-π interaction has a significant effect on the stabilization of 10 (Fig. 4dii). In addition, the conserved Arg196 residue was located near the isoalloxazine moiety of FAD and stabilized FAD, which is consistent with those reported FPMOs, such as CtdE, PhqK24 (PDB: 6pvi),15 3HB6H (PDB: 4bk3),40 and PhzS42 (PDB: 2rgj).41 Then, we mutated the R196 residue to alanine, which completely abolished the enzymatic activity of SpeF (Fig. 4di).
P450 SpeG hydroxylated C-17 α of epoxide 15. SpeG, containing heme group, exhibited a 42% amino acid identity with FtmG which could hydroxylate dioxopiperazine of fumitremorgin C at C-11.29 Coexpressing speG with speABCDEF in A. nidulans could produce eight new compounds, including six with m/z 448 [M + H]+ (1–6) and two ones with m/z 466 [M + H]+ (13 and 14) (Fig. 2cv). Besides 1–2, these compounds were identified as taichunamides H (3), A (4),42, 43 E (5),42 amoenamide B (6),20 notoamide M (13) and speramide B (14)44, 45 based on their identical MS and NMR data with references (Fig. 1a, 4e, S30-S45 and S56-S59; Tables S4 − S9, S15, and S16). Their absolute configurations were further established by CD (Fig. S60), specific rotations (Table S17), X-ray diffraction analysis (1 (CCDC 2034890), 2 (CCDC 2034888), 3 (CCDC 2034894), 5 (CCDC 2034891, Fig. 4F) and 6 (CCDC 2034886), and chemical computations (4 (Fig. S21-S22, S61, Tables S27-S29)). Among them, (−)-(4) and (+)-5 were identified as two new antipodes.20, 42, 43 Further analysis of the structures of compounds 1–6 indicated the coexistence of 4 bicyclo[2.2.2]diazaoctane isomers, specifically, S(11)S(17)R(21) in 1, RRR in 2, SSS in 3, and RRS in 4, 5, and 6, and three 2R,3R-epoxy-ring-opening moieties, namely, 3S-spiro-oxindole (1, 2, 5), 3R-hydroxyindolenine (3, 4) and 2R-spiro-ψ-indoxyl (6). Then, when we fed 10 into the transformants AN-speG and AN-speFG, respectively, 1–6 appeared only in the latter, which further confirmed that the intermediate 10 is not the substrate of SpeG (Fig. 4g, S6).
As observed, compounds 13–14 are the 17-hydroxylated versions of 11–12, therefore, SpeG should be involved in the 17-hydroxylation. Then, when we fed 13–14 to the transformants AN-speG or AN-speFG, alkaloids 1–6 were not produced (Fig. S11B), indicating that SPR products 13–14 could not proceed the construction of the dienophile, in turn, the HDA reaction should precede the SPR reaction. Therefore, the biosynthesis pathway of 1–6 from 10 was proposed as follows: SpeF epoxidizes 10 to 15, followed by 17α-hydroxylation to 16, which could give an azadiene intermediate via spontaneous dehydration and ketone-enol tautomerization, followed by HDA and SPR to construct compounds 1–6 (Fig. 2d, 5a, S9).
Then, a homolog of SpeG complexed with heme was generated. The crystal structure of retinoic acid-bound cyanobacterial cytochrome P450 CYP120A1 (PDB: 2VE3_A), which shared 29% sequence identity with SpeG, was used as a reference for modeling.46 SpeG substrate 15 was too labile to be assayed enzymatically. We alternatively confirmed the attribution of the conserved tetrad to the catalytic activity of SpeG by coexpressing the mutant SpeG and wild-type SpeF in A. nidulans. Mutation of the T299 and R291 residues to alanine completely abolished the enzymatic activity of SpeG (Fig. 4diii, iv). This observation was consistent with the catalytic mechanism proposed for P450 monooxygenase.
DFT calculations implied the enzymatic HDA reaction for the formation of the RRS bicyclo[2.2.2]diazaoctane core. As shown in Fig. 5a intermediate 16 was putatively converted to a cationic azadiene intermediate int-c1, which could isomerize to a neutral azadiene intermediate int-n via int-c2. To understand the HDA mechanism, we performed density functional theory (DFT) calculations (Fig. 6b S14 and S15).47, 48, 49, 50, 51, 52 As a result, the activation barriers via the cation transition state TSc1s are much lower than those via the TSc2 and TSns, respectively, supporting that the biosynthetic strategy for the bicyclo[2.2.2]diazaoctane moiety in notoamides should be an inverse electronic demanded HDA reaction (IEDHDA), (Fig. 5A, 5B and S1). As expected, frontier molecular orbital analysis also supported this result (Fig. S16).53 Moreover, the relative uncatalyzed barrier of TSc1-SSR and TSc1-RRS are 2.2 kcal/mol (Fig. 5b). However, in fact, the ratio of these two forms in products (compounds 1 and 4–6) was almost 1:1.4, which suggested that the formation of the 17-RRS might be an enzymatic step.
DFT calculations implied enzymatic C3-O cleavage of the epoxy moiety for the formation of 3 S chiral center in 1 and 2. Protonated guanidine was selected as a general acid catalyst to model C-O bond cleavage in this study (Fig. 6 and S17).15 For 17-SSS and 17-RRS, H-21 is upward and axial. The relative barrier for the cleavage of C2-O over that of C3-O is 2.6 kcal/mol (Fig. 6a, Table S19) and calculated ratio of their products was approximately 1:80, which is roughly consistent with the population of 5 and 4/6 in the metabolites profile. Thus, the regioselectivity of the epoxy ring-opening of 17-RRS is nonenzymatic (Fig. 6a). On the other hand, for 17-SSR and 17-RRR, 21-H is downward and axial. Their relative barrier for the cleavage of C3-O over that of C2-O is only 0.6 kcal/mol. However, the C3-O cleavage products, 1/2, were obtained as unique metabolites (Fig. 6b, S19, Table S19). Therefore, the partitioning of the putative indole 2,3-epoxides in 17-RRR and 17-SSR to form 1/2 should be enzyme-controlled.
Colocalization experiments of SpeF and SpeG indicated that both of the proteins could work together to convert 10 to 1–6. In our proposed biosynthetic pathway of notoamides, the 2,3-epoxidation precedes HDA reaction, followed by epoxy-ring-opening reaction, which is significantly different from that of brevianamides A/B. Specifically, several intermediates with unstable epoxy moiety were proposed. Therefore, we conjectured that SpeF and SpeG close to each other to avoid the opening of their epoxy moiety. To demonstrate this hypothesis, we fused SpeF and SpeG with green fluorescent protein (GFP) and the red fluorescent protein mCherry, respectively, and used compounds 1–6 as indicators. Feeding experiments of compound 10 indicated that the fluorescently labeled proteins SpeF and SpeG could convert 10 to 1–6 (Fig. 7a, ci-ii). Simultaneously, the colocalization of SpeF and SpeG was observed via confocal experiments (Fig. 7a). Since both of proteins are membrane-bound, we truncated the membrane-bound region of SpeF and reconstructed the AN-speF*-gfp-speG-mCherry transformant, in which SpeF* was observed to be diffused throughout the cell (Fig. 7b), and feeding experiments showed that 10 could be converted to 11–12 rather than compounds 1–6 (Fig. 7ciii, S12). Thus, the colocalization of SpeF and SpeG is essential for the conversion of 10 to 1–6. Moreover, colocalization of SpeF with SpeC, SpeG with GFP, SpeF with mCherry, SpeG with SpeB, and SpeF with SpeB was not observed (Fig. S13).