Identification of key enzymes responsible for protolimonoid biosynthesis in plants: Opening the door to azadirachtin production

Abstract

Limonoids are natural products made by plants belonging to the Meliaceae (Mahogany) and Rutaceae (Citrus) families. They are well known for their insecticidal activity, contribution to bitterness in citrus fruits, and potential pharmaceutical properties. The best known limonoid insecticide is azadirachtin, produced by the neem tree (Azadirachta indica). Despite intensive investigation of limonoids over the last half century, the route of limonoid biosynthesis remains unknown. Limonoids are classified as tetranortriterpenes because the prototypical 26-carbon limonoid scaffold is postulated to be formed from a 30-carbon triterpene scaffold by loss of 4 carbons with associated furan ring formation, by an as yet unknown mechanism. Here we have mined genome and transcriptome sequence resources for 3 diverse limonoid-producing species (A. indica, Melia azedarach, and Citrus sinensis) to elucidate the early steps in limonoid biosynthesis. We identify an oxidosqualene cyclase able to produce the potential 30-carbon triterpene scaffold precursor tirucalla-7,24-dien-3β-ol from each of the 3 species. We further identify coexpressed cytochrome P450 enzymes from M. azedarach (MaCYP71CD2 and MaCYP71BQ5) and C. sinensis (CsCYP71CD1 and CsCYP71BQ4) that are capable of 3 oxidations of tirucalla-7,24-dien-3β-ol, resulting in spontaneous hemiacetal ring formation and the production of the protolimonoid melianol. Our work reports the characterization of protolimonoid biosynthetic enzymes from different plant species and supports the notion of pathway conservation between both plant families. It further paves the way for engineering crop plants with enhanced insect resistance and producing high-value limonoids for pharmaceutical and other applications by expression in heterologous hosts.

Keywords: insecticides; limonoids; natural products; neem; terpenes.

Fig. 1.

Hypothetical route of limonoid biosynthesis. Predictions of the major biosynthetic steps required for the biosynthesis of limonoids. The triterpene precursor 2,3-oxidosqualene is proposed to be cyclized to an unconfirmed tetracyclic triterpene scaffold. The structure of ring-intact limonoids implicates a tetracyclic triterpene precursor of either the euphane (20R) or tirucallane (20S) type. Retrosynthetic discrimination between these 2 side chains is impossible based on limonoid structures, because the formation of the furan ring eradicates any remnants of the precursor’s C20 stereochemistry. However, predictions can be made based on the immediate precursors of limonoids (protolimonoids); for instance, the C20 carbon of melianol (4) has been assigned (although not yet confirmed by X-ray crystallography) as the S configuration which implies a tirucallane precursor (60). Further, the C7-8 alkene of certain protolimonoids suggests the most likely triterpene precursor is in fact tirucalla-7,24-dien-3β-ol (1), as indicated by the retrosynthetic arrow (*), rather than tirucallol itself. Biosynthesis of limonoids from triterpene scaffolds is predicted to occur through protolimonoid structures such as melianol (4) and requires 2 major biosynthetic steps: scaffold rearrangements and furan ring formation accompanied by loss of 4 carbons. Scaffold rearrangement is proposed to be initiated by epoxidation of the C7 double bond (C7-8 epoxide) and furan ring formation could feasibly be initiated through oxidation and cyclization of the C20 tail (melianol) (4). The diversity of isolated protolimonoid structures has led to different predictions of the order of these 2 events (2, 61). The hemiacetal side chains of isolated protolimonoids such as melianol (4) suggest a Paal-Knorr-like (51, 52) route to the furan ring of limonoids. Isolation of nimbocinone from A. indica (SI Appendix, Fig. S1) (62), a feasible degradation product of this route (**), further supports conversion of protolimonoids to ring-intact limonoids through this mechanism. The ring-intact 7-deacetylazadirone has been isolated from both Meliaceae (63) and Rutaceae (64) species. Numerous further chemical transformations are required for the formation of seco-ring limonoid derivatives. In the Rutaceae, radioactive [¹⁴C]-labeling experiments in C. limon (lemon) have helped to delineate the late stages of the pathway, proving that nomilin can be biosynthesised in the stem and converted into other limonoids such as limonin and obacunone (SI Appendix, Fig. S1) elsewhere in the plant (–67).

Fig. 2.

Identification and characterization of OSCs from limonoid-producing species. (A) Phylogenetic tree of candidate OSCs from A. indica (blue), M. azedarach (green), and C. sinensis (orange). Functionally characterized OSCs from other plant species (20) are included, with the 2 previously characterized tirucalla-7,24-dien-3β-ol synthases from A. thaliana (AtLUP5 and AtPEN3) highlighted (yellow). Human and prokaryotic OSC sequences used as an outgroup are represented by the gray triangle. Candidate OSCs chosen for further analysis are indicated (circles). The phylogenetic tree was constructed by FastTree V2.1.7 (68) and formatted using iTOL (69). Local support values from FastTree Shimodaira-Hasegawa (SH) test (between 0.6 and 1.0) are indicated at nodes, and scale bar depicts estimated number of amino acid substitutions per site. (B) GC-MS total ion chromatograms of derivatized extracts from yeast strains expressing candidate OSCs. Traces for the empty vector (pYES2) and strains expressing the candidates AiOSC1 (blue), MaOSC1 (green), CsOSC1 (orange), and the previously characterized AtLUP5 (yellow) are shown. (C) GC-MS mass spectra of TMS-tirucalla-7,24-dien-3β-ol (1). (D) Confirmation of the structure of the cyclization product generated by AiOSC1 as tirucalla-7,24-dien-3β-ol (1) by NMR (SI Appendix, Table S3). Traces, mass spectra, and product structure for CsOSC2 and CsOSC3 are given (SI Appendix, Figs. S2 and S3).

Fig. 3.

Expression patterns of AiOSC1 and other coexpressed genes in A. indica. A heatmap of a subset of differentially expressed (P < 0.05) genes with similar expression patterns to AiOSC1 (blue circle) across flower, root, fruit, and leaf tissues of A. indica are shown. Raw RNAseq reads (28, 31) were aligned to a Trinity-assembled transcriptome of the same dataset. Read counts were normalized to library size and log₂ transformed. Values depicted are scaled by row (gene) to emphasize differences across tissues. The Pfam identifier for relevant predicted gene is included next to the contig number. Genes with no structural (Augustus) or functional (Pfam) annotations have been excluded. CYP candidates AiCYP71BQ5, AiCYP72A721, and AiCYP88A108 (blue triangles) are indicated with the latter 2 being considered gene fragments (<300 amino acids).

Fig. 4.

Accumulation of melianol and salannin and expression of MaOSC1, MaCYP71CD2, and MaCYP71BQ5 in M. azedarach. (A) Estimated concentrations (mg/g DW, n = 4 ± SE) of the protolimonoid melianol (4) and the seco-C-ring limonoid salannin in extracts from M. azedarach leaf, root, and petiole tissue. (B) Normalized expression of MaOSC1, MaCYP71CD2, and MaCYP71BQ5 relative to Maß-actin in RNA from leaves, roots, and petioles of M. azedarach by qRT-PCR. Relative expression levels were calculated using the ΔΔCq method (58) (n = 4 ± SE). T-test significance values are indicated: not significant (NS), P value ≤ 0.05 (*), 0.01 (**), or ≤ 0.001 (***).

Fig. 5.

Identification and functional analysis of cytochrome P450 enzymes capable of melianol biosynthesis. (A) A subset of a larger phylogenetic tree (SI Appendix, Fig. S4) showing the CYP71 family. Candidate CYPs from M. azedarach (green) and previously identified CYPs from A. thaliana (http://www.p450.kvl.dk) and C. sativus (http://drnelson.uthsc.edu/cytochromeP450.html) (black) are included. Candidate CYPs selected for cloning (SI Appendix, Table S4) were identified by homology to A. indica candidate CYPs identified as coexpressed with AiOSC1 (triangle) or occurrence in a unique CYP71 subclade lacking close homologs from A. thaliana or C. sativus (squares). The phylogenetic tree was constructed by FastTree V2.1.7 (68) and formatted using iTOL (69). Local support values from FastTree Shimodaira-Hasegawa (SH) test (between 0.6 and 1.0) are indicated at nodes, and scale bar depicts estimated number of amino acid substitutions per site. Data from the Arabidopsis Cytochrome P450, Cytochrome b₅, P450 Reductase, β-Glucosidase, and Glycosyltransferase Site, and from ref. . GC-MS total ion chromatograms (B) and LC-MS electrospray ionization (ESI) extracted ion chromatograms (C) of triterpene extracts from agroinfiltrated N. benthamiana leaves expressing A. indica and M. azedarach candidate genes in the pEAQ-HT-DEST1 vector. (D) LC-MS multimode ionization (MMI) extracted ion chromatograms of triterpene extracts from agroinfiltrated N. benthamiana leaves expressing C. sinensis candidate genes in the pEAQ-HT-DEST1 vector. Mass spectra for products are shown in SI Appendix, Figs. S5–S7. The peak marked with an asterisk is an endogenous N. benthamiana peak, not tirucalla-7,24-dien-3β-ol (1) (SI Appendix, Fig. S8). (E) Proposed pathway of melianol (4) biosynthesis in M. azedarach. NMR confirmation of all structures can be found in SI Appendix, Tables S3 and S5–S7.