Insights into stereoselective ring formation in canonical strigolactone: Identification of a dirigent domain-containing enzyme catalyzing orobanchol synthesis

Significance

Strigolactones (SLs) serve diverse functions as hormones and rhizosphere signals. Canonical SLs, characterized by a BC-ring, are divided into two types based on C-ring configurations. The functional implications of this structural diversity are poorly understood, and the molecular underpinnings of biosynthetic diversity, especially the stereoselective BC-ring closure, remain unclear. We identified a pivotal dirigent domain-containing enzyme for the stereoselective biosynthesis of SL from 18-oxocarlactonoic acid to orobanchol. The connection between this type of enzyme and SL biosynthesis had not been revealed. This finding led us to propose a C-ring regulatory mechanism using computational methods, including AlphaFold2-based protein structure prediction and molecular dynamics simulations. This study highlights how plants produce SLs in a stereospecific manner, enhancing our understanding of distinct SL functions.

Abstract

Strigolactones (SLs) are plant apocarotenoids with diverse roles and structures. Canonical SLs, widespread and characterized by structural variations in their tricyclic lactone (ABC-ring), are classified into two types based on C-ring configurations. The steric C-ring configuration emerges during the BC-ring closure, downstream of the biosynthetic intermediate, carlactonoic acid (CLA). Most plants produce either type of canonical SLs stereoselectively, e.g., tomato (Solanum lycopersicum) yields orobanchol with an α-oriented C-ring. The mechanisms driving SL structural diversification are partially understood, with limited insight into functional implications. Furthermore, the exact molecular mechanism for the stereoselective BC-ring closure reaction is yet to be known. We identified an enzyme, the stereoselective BC-ring-forming factor (SRF), from the dirigent protein (DIR) family, specifically the DIR-f subfamily, whose biochemical function had not been characterized, making it a key enzyme in stereoselective canonical SL biosynthesis with the α-oriented C-ring. We first confirm the precise catalytic function of the tomato cytochrome P450 SlCYP722C, previously shown to be involved in orobanchol biosynthesis [T. Wakabayashi et al., Sci. Adv. 5, eaax9067 (2019)], to convert CLA to 18-oxocarlactonoic acid. We then show that SRF catalyzes the stereoselective BC-ring closure reaction of 18-oxocarlactonoic acid, forming orobanchol. Our methodology combines experimental and computational techniques, including SRF structure prediction and conducting molecular dynamics simulations, suggesting a catalytic mechanism based on the conrotatory 4π-electrocyclic reaction for the stereoselective BC-ring formation in orobanchol. This study sheds light on the molecular basis of how plants produce SLs with specific stereochemistry in a controlled manner.

Strigolactones (SLs) comprise a group of plant apocarotenoids with diverse functions (1). They function as endogenous plant hormones inhibiting shoot branching/tillering (2, 3), as rhizosphere signaling molecules promoting hyphal branching in arbuscular mycorrhizal fungi (4), and as germination stimulants for root parasitic weeds (5). SLs are structurally diverse and more than 30 structures have been reported thus far (6, 7). Most plants utilize a common upstream biosynthesis pathway for SLs, converting all-trans-β-carotene to carlactone through the sequential action of three enzymes. The cytochrome P450 CYP711A subfamily further catalyzes the conversion of carlactone to carlactonoic acid (CLA) (8–12). Canonical SLs, which contain the ABC-ring system, are classified into two subgroups based on their C-ring configuration: the orobanchol type and the strigol type. The orobanchol-type SLs have α-oriented C-rings, whereas the strigol-type SLs have β-oriented C-rings. Conversely, noncanonical SLs possess incomplete ABC-ring systems (Fig. 1). The mechanisms underlying the structural diversification of SLs are only partially elucidated, and the functional consequences of this structural diversity remain largely unknown.

Fig. 1.

Canonical SL biosynthesis pathway downstream of CLA. The sequential reactions catalyzed by the DWARF27 (D27), CAROTENOID CLEAVAGE DIOXYGENASE 7 (CCD7), and CCD8 enzymes produce carlactone from all-trans-β-carotene. Carlactone is further oxidized by the CYP711A subfamily to produce CLA. Methyl carlactonoate is shown as a representative example of noncanonical SL. A variety of cytochrome P450 enzymes are involved in the production of canonical SLs from CLA. These include OsCYP711A2, catalyzing 4-deoxyorobancol formation in rice (Oryza sativa), and GaCYP722C, catalyzing 5-deoxystrigol formation in cotton (Gossypium arboreum). Enzymes from the CYP722C subfamily, common in dicots, have different functions based on the plant species. For instance, they participate in orobanchol biosynthesis in tomato. The tomato orobanchol biosynthesis pathway, highlighted in the solid red box, involves the SRF. This regulates the C-ring stereochemistry, as demonstrated in this study. 18-Oxocarlactonoic acid (18-oxo-CLA) is derived from CLA via 18-hydroxycarlactonoic acid (18-OH-CLA) through the reaction of SlCYP722C, followed by a series of reactions involving SlSRF to produce orobancol. The black dashed box shows that the acid-mediated cascade cyclization of 18-oxo-CLA based on 4π-electrocyclic reaction (4π-ECR) in flask generates orobanchol and ent-2′-epi-orobanchol, whose relative stereochemistry is controlled to be trans (13). In contrast, the conformation of 18-oxo-CLA is controlled within SlSRF, producing only orobanchol.

The steric configuration of the C-ring of canonical SLs is determined during the BC-ring closure reaction downstream of CLA (Fig. 1). In rice (Oryza sativa), the enzyme OsCYP711A2 catalyzes the stereoselective BC-ring closure of CLA, resulting in the formation of 4-deoxyorobanchol. This compound is an orobanchol-type SL characterized by an α-oriented C-ring (11, 12). The conversion of 4-deoxyorobanchol to orobanchol, which involves the introduction of a hydroxy group at the C-4 position, is catalyzed by OsCYP711A3 (Fig. 1). Initially, the role of the rice CYP711A subfamily in the stereoselective BC-ring formation suggested a potential for broader application in the formation of canonical SLs across different plant species. However, the process turned out to be more intricate than anticipated. The concept that the CYP711A subfamily stereospecifically governs the BC-ring closure reaction is not universally applicable and appears to be limited to certain plant species, including rice (11). The complexity of the canonical SL biosynthesis pathway is further increased in some plant species, such as sorghum (Sorghum bicolor), where the production of 5-deoxystrigol, a strigol-type SL, involves a combination of three distinct types of enzymes (14). This complexity is highlighted in a mutant deficient in one of these enzymes, known as lgs1, which synthesizes orobanchol instead of 5-deoxystrigol (15).

We previously pinpointed the cytochrome P450 subfamily CYP722C, conserved in dicots but absent in monocots, as a key player in canonical SL synthesis. The activity of this subfamily varies among plant species, leading to the production of different canonical SLs, such as orobanchol-type or strigol-type SLs. In dicot plants that produce strigol-type SLs, the CYP722C subfamily enzymes catalyze the stereoselective conversion of CLA to 5-deoxystrigol (16–19). Our research identified GaCYP722C and GhCYP722C in the cotton species Gossypium arboreum and G. hirsutum, respectively, as catalysts for this conversion reaction (Fig. 1) (17, 19). The CYP722Cs in other 5-deoxystrigol-producing plants, such as birdsfoot trefoil (Lotus japonicus) and woodland strawberry (Fragaria vesca), also participate in this stereoselective conversion (16, 18), suggesting a conserved function of the CYP722C subfamily in 5-deoxystrigol-producing dicot plants. However, in orobanchol-producing dicot plants like tomato (Solanum lycopersicum) and cowpea (Vigna unguiculata), the CYP722C enzymes participate in a different orobanchol biosynthesis pathway compared to rice, bypassing 4-deoxyorobanchol (Fig. 1) (20). Unlike the stereoselective synthesis of 5-deoxystrigol, CYP722C enzymes in orobanchol-producing dicot plants seem incapable of controlling the C-ring stereochemistry. In vitro enzyme assays using the substrate CLA for SlCYP722C and VuCYP722C from tomato and cowpea showed that their enzyme products are the orobanchol diastereomers, orobanchol and ent-2′-epi-orobanchol, the latter having the opposite C-ring configuration to that of orobanchol (Fig. 1). This suggests that the BC-ring formation occurred without strict control of the C-ring stereochemistry (20). Knockout of the SlCYP722C gene in tomato resulted in the loss of orobanchol and the alternative accumulation of its substrate CLA, indicating that this enzyme is crucial for the initial step in the BC-ring closure reaction (20). These findings suggest that SlCYP722C or VuCYP722C alone is not sufficient for achieving the BC-ring closure with stereoselective control of the C-ring, implying additional factors regulate the BC-ring stereochemistry in CYP722C-mediated orobanchol biosynthesis (20, 21). Despite these advancements in identifying key enzymes for canonical SL formation, the precise molecular mechanism of the BC-ring closure reaction leading to the tricyclic lactone moiety of canonical SLs remains partially understood.

Deciphering the molecular mechanism that governs the stereoselective BC-ring closure reaction is key to understanding how plants produce structurally diverse SLs by accurately controlling the C-ring configuration. In this study, we delve into the molecular mechanism of BC-ring formation leading to orobanchol, using tomato as a representative model. We elucidate the precise function of SlCYP722C and identify a biosynthetic enzyme that catalyzes the stereoselective BC-ring closure reaction subsequent to the SlCYP722C reaction. We suggest a potential mechanism for the regulation of the C-ring configuration, employing a computational structural biology approach.

Results

Conversion of CLA to 18-Oxocarlactonoic Acid, the Exact Function of SlCYP722C.

To understand the mechanism of BC-ring closure, we conducted an in-depth functional analysis of SlCYP722C at the biochemical level. Recombinant SlCYP722C protein was produced in Escherichia coli, truncating the N-terminal transmembrane domain for better solubility and tagging the C-terminal with His₆-tagged for purification (SI Appendix, Figs. S1 and S2 A and B). The enzyme reaction mixture of the recombinant SlCYP722C with CLA as a substrate was analyzed using multiple reaction monitoring (MRM) liquid chromatography–tandem mass spectroscopy (LC–MS/MS). The recombinant SlCYP722C produced 18-hydroxycarlactonoic acid (18-OH-CLA) (SI Appendix, Fig. S2C) and a compound with a molecular mass 14 Da larger than CLA (CLA+14), along with orobanchol diastereomers (Fig. 2A). Given its molecular weight and involvement in orobanchol biosynthesis, we considered CLA+14 as a further oxidized derivative of 18-OH-CLA. The methyl esterified CLA+14 by (trimethylsilyl)diazomethane was consistent with the authentic methyl 18-oxocarlactonoate (13) in LC–MS/MS analysis, identifying CLA+14 as 18-oxocarlactonoic acid (18-oxo-CLA) (Fig. 2B). Furthermore, SlCYP722C recognized the natural-type (11R)-CLA as a substrate but not (11S)-CLA. This confirms that the enzyme products have the D-ring in the R configuration (SI Appendix, Fig. S2 D and E). The addition of HCl to the enzyme solution at the end of the reaction facilitated the formation of orobanchol diastereomers, which is accompanied by a decrease in 18-oxo-CLA (Fig. 2A). In a separate experiment, adding 18-oxo-CLA to a mildly acidic buffer (pH 5.8) led to its conversion to orobanchol diastereomers over time (SI Appendix, Fig. S3). These findings suggest that 18-oxo-CLA spontaneously cyclizes in a nonenzymatic manner under acidic conditions. This supports our previous study, where we proposed that the BC-ring closure reaction could be an acid-mediated cascade cyclization based on the conrotatory 4π-electrocyclic reaction (ECR) (13, 22). We therefore conclude that SlCYP722C functions in a two-step oxidation process at the C18 position of CLA to generate 18-oxo-CLA via 18-OH-CLA (Fig. 1).

Fig. 2.

Orobanchol biosynthesis catalyzed by the sequential reaction of SlCYP722C and SlSRF. (A) In vitro enzyme assay of recombinant SlCYP722C; MRM chromatograms of reaction mixtures of recombinant SlCYP722C with CLA as a substrate. For the negative control, the assay was performed with heat-denatured recombinant SlCYP722C protein. (B) Identification of 18-oxo-CLA as a SlCYP722C product, whose molecular mass is 14 Da larger than CLA, by methyl ester derivatization with (trimethylsilyl)diazomethane followed by LC–MS/MS analysis. The retention time and MRM transition of the compound were identical to those of authentic methyl 18-oxocarlactonoate (18-oxo-MeCLA). (C) In vitro enzyme assay of recombinant SlSRF; MRM chromatograms of reaction mixtures of recombinant SlSRF with 18-oxo-CLA as a substrate. The negative control experiment was performed with heat-denatured recombinant SlSRF protein.

Stereoselective Conversion of 18-oxo-CLA to Orobanchol by a Dirigent Domain-Containing Enzyme.

The SRF, which governs the stereoselective cyclization of 18-oxo-CLA to generate orobanchol, is posited to operate downstream of SlCYP722C. In tomato and cowpea, as in many other plant species, phosphate deficiency triggers an increase in the expression of SL biosynthesis genes (20, 23, 24). The gene expression of SRF is also anticipated to up-regulate in response to phosphate deficiency. A previous study by Wang et al. reported time-course transcriptional changes in tomato roots under conditions of phosphate deficiency and replenishment (25). The SRF investigation in this study focused on 48 differentially expressed genes at the core of the phosphate response. These genes were up-regulated at all timepoints during phosphate deficiency and down-regulated upon phosphate replenishment (25) (SI Appendix, Fig. S4). From this list, the Solyc01g059900 gene was selected, which encodes the only dirigent domain-containing protein (DIR; Pfam PF03018) among the 48 genes. We further investigated the SRF candidates through a gene coexpression network analysis using the weighted gene coexpression network analysis method (26). This analysis utilized our previous cowpea transcriptome dataset (DDBJ Sequence Read Archive, accession no. DRA008222), leading to the identification of VuCYP722C (20). The analysis revealed that Vigun03g413800 and Vigun03g413900, which are homologs of Solyc01g059900, along with all known orobanchol biosynthesis genes were included in the same coexpression gene module as the hub genes associated with SL biosynthesis (SI Appendix, Fig. S5 and Dataset S1).

Some members of the DIR family have been shown to mediate the enantioselective bimolecular coupling, leading to the stereoselective coupling of coniferyl alcohol radicals in the formation of (+)- or (−)-pinoresinol in lignan biosynthesis (27, 28). In another instance, pterocarpan synthase with a dirigent domain enantiospecifically catalyzed the intramolecular cyclization of 2′-hydroxyisoflavanol to form pterocarpan (29, 30). The DIR family is classified into six subfamilies (DIR-a, b/d, c, e, f, and g) based on phylogenetic analyses (31). Pinoresinol-forming DIRs and pterocarpan synthase belong to the DIR-a and DIR-b/d subfamilies, respectively. On the other hand, Solyc01g059900, Vigun03g413800, and Vigun03g413900 belong to the DIR-f subfamily, and the biochemical functions of this subfamily remain unknown (SI Appendix, Fig. S6). The function of enantioselective substrate conversions by DIR members suggested the potential involvement of DIR as an SRF in the stereoselective formation of orobanchol through the intramolecular BC-ring formation of 18-oxo-CLA.

We conducted biochemical characterization of Solyc01g059900 (SlSRF). It has been reported that most members of the DIR family possess a putative signal peptide sequence at the N-terminal and are processed into the mature active forms following the truncation of this sequence (30, 32). For in vitro biochemical analysis, recombinant SlSRF was expressed in E. coli with the putative N-terminal transmembrane domain truncated (replacing the 24 N-terminal residues with Met) and the C-terminal His₆-tagged (30) (SI Appendix, Figs. S1 and S7A). Incubation of the purified recombinant SlSRF with 18-oxo-CLA, generated through the reaction of the recombinant SlCYP722C with CLA, resulted in the stereoselective formation of orobanchol with the utilization of 18-oxo-CLA (Fig. 2C and SI Appendix, Fig. S8). This suggests that Solyc01g059900 functions as an SRF, catalyzing the stereoselective BC-ring closure reaction of 18-oxo-CLA (Fig. 1). Biochemical analyses were also performed on cowpea SRF candidates (Vigun03g413800 and Vigun03g413900), using 18-oxo-CLA as a substrate. These analyses indicated that Vigun03g413900 acts as an SRF, catalyzing the same reaction as SlSRF to form orobanchol (SI Appendix, Fig. S9).

In this study, we conducted in vitro enzyme assays using the N-terminal truncated recombinant proteins. To evaluate the similarity between the truncated N-terminal and full-length forms, we transiently expressed the full-length tomato SlCYP722C and SlSRF with the cytochrome P450 reductase SlCPR2 in the leaves of Nicotiana benthamiana, which were then incubated with CLA. In the absence of SlSRF, the expression of SlCYP722C resulted in the production of 18-oxo-CLA and orobanchol diastereomers. Whereas, when SlSRF was coexpressed, we achieved stereoselective production of orobanchol from CLA (SI Appendix, Fig. S10). These outcomes align with the findings from the in vitro enzyme assay.

Functional Verification of SlSRF in planta.

To verify the role of SlSRF in orobanchol biosynthesis in planta, we generated two biallelic homozygote SlSRF-knockout tomato lines using CRISPR/Cas9-mediated genome editing. These lines are slsrf-1 (with 349 bp deletions in the target regions) and slsrf-2 (with 13 bp deletions and a 1 bp insertion in the target regions) (Fig. 3A). In the root exudates of slsrf-1 and slsrf-2, we detected approximately equal amounts of orobanchol and ent-2′-epi-orobanchol (Fig. 3B). This suggests that the loss of SlSRF function led to the spontaneous generation of orobanchol diastereomers from 18-oxo-CLA. Interestingly, the amounts of these detected orobanchol diastereomers exceeded those of orobanchol in the wild-type (WT) plants. We also examined solanacol and didehydro-orobanchol isomers, including 6,7-didehydroorobanchol, phelipanchol, and epiphelipanchol, which are downstream metabolites of orobanchol produced by tomato (33, 34). However, solanacol was undetectable, and only negligible amounts of didehydro-orobanchol isomers were found in the root exudates of the slsrf plants (SI Appendix, Fig. S11). These results indicate that SlSRF deficiency impairs both orobanchol biosynthesis and its metabolism.

Fig. 3.

Characterization of SlSRF in planta. (A) Generation of the SlSRF-knockout tomato lines (slsrf-1 and slsrf-2) by CRISPR/Cas9 system. The structure of SlSRF with the CRISPR/Cas9 target sites (gRNA Target-1 and Target-2) and sequences are shown. Each target region is shown in bold letters followed by the protospacer adjacent motif (PAM). The number of deleted (D) and inserted (+) nucleotides is indicated on the Right side of the sequences. (B) Analysis of the orobanchol diastereomers, orobanchol and ent-2′-epi-orobanchol, in the root exudates of WT and slsrf-lines grown under phosphate-deficient conditions. (C and D) Subcellular localizations of signal peptide of SlSRF (SlSRF-SP) and SlCYP722C. SlSRF-SP fused with mCherry (C) or SlCYP722C fused with enhanced green fluorescent protein (EGFP) (D) along with endoplasmic reticulum (ER) markers are transiently expressed in the leaves of N. benthamiana. The results show that SlSRF-SP and SlCYP722 colocalize with the ER markers. (Scale bars, 5 μm.)

Additionally, we investigated the subcellular localization of SlSRF. We found that the predicted transmembrane domain signal peptide of SlSRF (SI Appendix, Fig. S1), when fused with mCherry at the C-terminal (SlSRF-SP:mCherry) and transiently expressed in N. benthamiana leaves, showed overlapping fluorescence with the enhanced green fluorescent protein from the endoplasmic reticulum (ER) marker (Fig. 3C). This suggests that SlSRF and SlCYP722C, an ER-localized cytochrome P450 (Fig. 3D), colocalizes to the ER, facilitating swift substrate transfer between the two. Without this, the 18-oxo-CLA produced by SlCYP722C could cyclize to generate orobanchol diastereomers, as observed in the in vitro conversion (Fig. 2 and SI Appendix, Fig. S3) and root exudates of slsrf plants (Fig. 3B). As a result, we identified SlSRF, an enzyme from the DIR-f subfamily, that catalyzes the stereoselective BC-ring formation in orobanchol biosynthesis.

Computational Analysis of the Stereoselective Ring Formation Mechanism Using the Predicted Structure of SlSRF.

In the molecular mechanism of DIR, a general stereoselective mechanism involving the generation and stabilization of quinone methides has been proposed for pinoresinol-forming DIRs and pterocarpan synthase belonging to the DIR-a and DIR-b/d subfamilies, respectively (29, 35). Prior to this study, the biochemical function and molecular mechanism of the DIR-f subfamily were unknown. We proposed an inventive hypothesis using a synthetic organic chemistry approach, where the BC-ring is formed through an acid-mediated sequential cyclization process via the conrotatory 4π-ECR mechanism. We further hypothesized that the BC-ring formation in planta proceeds similarly to this acid-mediated cascade cyclization reaction as it occurs “in flask” (13, 22).

To investigate the BC-ring closure reaction of 18-oxo-CLA in vacuo, we utilized quantum mechanics (QM) calculations with Gaussian 16 Rev. C01. In these calculations, the 18-aldehyde group of 18-oxo-CLA was protonated, as indicated in our previous study (13). Subsequently, orobanchol and ent-2′-epi-orobanchol were generated when the initial C18–C5–C6–C8 dihedral angles (ϕ) of the protonated 18-oxo-CLA were −44.00° and 43.88°, respectively (Fig. 4A). The calculated activation energies were 9.82 and 9.61 kcal/mol for orobanchol and ent-2′-epi-orobanchol, respectively, and the reaction energies were −39.69 and −40.09 kcal/mol, indicating nearly equal progression of these reactions (SI Appendix, Fig. S12). However, when the 18-aldehyde group was not protonated, the activation energies increased to 33 kcal/mol for both products, highlighting the crucial role of protonation in the BC-ring closure reaction. These findings suggest that SlSRF employs a molecular mechanism that efficiently protonates the 18-aldehyde group of 18-oxo-CLA, selectively favoring the conformation with ϕ = −44.00° for orobanchol production.

Fig. 4.

QM and molecular dynamics (MD) simulation analysis on a SlSRF model predicted by AlphaFold2. (A) Two cyclization pathways from protonated 18-oxo-CLA. The C8 and C18 atoms are depicted in cyan, while the C5 and C6 atoms are shown in green. The distance between the C8 and C18 atoms is denoted as d. 18-Oxo-CLA can transform into two different compounds: orobanchol (Upper) and ent-2′-epi-orobanchol (Bottom). The Left, Middle, and Right panels display the reactant, transition state, and product structures, respectively. These structures were optimized in vacuo at the level of B3LYP/6-31++G(d,p) theory using Gaussian 16 Rev. C01. (B) Predicted monomeric SlSRF structure. The structure is colored by the per-residue predicted local distance difference test (pLDDT) confidence of AlphaFold2. (C) Conserved amino acids of SlSRF. The scores were calculated by the ConSurf webserver (36–38) on 24 April 2023. The dark red arrow represents the secondary structure based on its tertiary structure model as predicted by AlphaFold v2.1.0. (D) An equilibrated model of SlSRF (orange) in complex with 18-oxo-CLA (black), obtained using MD simulations. Hydrogen bonds are shown as dashed lines. (E and F) Differences in the conformational distribution of 18-oxo-CLA in complex with SlSRF (E) and in water (F) observed in each 500-ns MD simulation. The horizontal and vertical axes represent the distance (d) and the dihedral angle (ϕ), respectively. The scatter plots are colored according to the density obtained through a Gaussian kernel density estimation. Their representative conformations are shown in the Right panels.

Determining the crystal structure of the SlSRF complex with the substrate 18-oxo-CLA is vital for understanding the molecular mechanism, a feat that remains unaccomplished. As an alternative, we used AlphaFold2 (AF2) to model the tertiary structure of SlSRF, which allowed us to speculate on the catalytic mechanism (39). The monomeric structure of SlSRF was generated with high pLDDT scores, excluding the signal peptide region (Fig. 4B). This indicates that the functional domain is located within residues 26 to 174. However, it is important to note that a protein model predicted by AF2 may not accurately represent the conformation in the ligand-bound state (i.e., holo form) (40). Furthermore, conventional docking simulations using the AF2-predicted model pose challenges as the pLDDT value guarantees only the Cα coordinate and not the orientation of the side chain (40, 41). To deal with this difficulty, we performed a manual docking simulation and a 500-ns MD simulation to obtain a stable complex model.

The docking of 18-oxo-CLA into the predicted SlSRF structure proceeded through the following steps. First, considering the predicted structure and the widely conserved residues among DIR family proteins, we predicted the catalytic residues as D37, D70, Y90, D124, and R131 (Fig. 4C). Second, from the standpoint of chemical properties, the positions of the hydrophobic A-ring and hydrophilic D-ring of 18-oxo-CLA were positioned inside and outside the SlSRF model, respectively (Fig. 4D). Subsequently, we conducted a 30-ns MD simulation with distance restraints (SI Appendix, Table S1) to refine the complex model. During the MD simulation, we assumed D37 and D70 to be protonated because their carboxy groups are located in close proximity to each other and the pKa values were estimated to 6.71 and 8.74, respectively, using PROPKA3 (42). Notably, the formation of a salt bridge between Y90, R131, and the carboxy group of 18-oxo-CLA was observed, while V35, V68, L165, and L167 formed a hydrophobic pocket inside SlSRF to accommodate the A-ring (Fig. 4D). According to the simulation, the 18-aldehyde group was located near the protonated residues D37 and D70, which are considered proton donors and are conserved in related DIR proteins (29).

Following this, we conducted an additional 500-ns unrestrained MD production run starting from the final snapshot of the 30-ns simulation to assess the stability of the complex model. The salt bridge observed in the restrained simulation was well preserved. It strictly controlled the conformation of 18-oxo-CLA in SlSRF to approximate a ϕ of −30°, with a C8–C18 distance (denoted as d) of less than 3.5 Å throughout the simulation (Fig. 4E and Movie S1). An independent 500-ns simulation for 18-oxo-CLA in the absence of SlSRF exhibited a bimodal distribution with peaks of approximately −30° (62.3%) and 30° (37.7%) for ϕ (Fig. 4F and Movie S2). The latter included the reactant conformation of ent-2′-epi-orobanchol. These two simulations collectively suggest that SlSRF selectively recognizes only one of the two possible conformations of 18-oxo-CLA. This recognized conformation is similar to the reactant conformation of orobanchol.

Several studies using MD simulations have demonstrated that the correct protonation state for titrable residues significantly impacts the stability of the structure and the understanding of catalytic function (43–45). Hence, we conducted another 500-ns unrestrained MD simulation to evaluate the impact of nonprotonated D37 and D70 on the stability of the complex. Despite starting the simulation from the same snapshot as used in the protonated one, 18-oxo-CLA immediately adopted a different conformation with a ϕ of 20°~30° and a d of >3.5 Å. Although the most frequently observed dihedral ϕ of 18-oxo-CLA in this simulation approximated the reactant conformation of ent-2′-epi-orobanchol, the conformation was not suitable to initiate the BC-ring closure reaction due to the elongated distance d (SI Appendix, Fig. S13A). The destabilization was recovered when we used a protonated 18-oxo-CLA model for the MD simulation with nonprotonated D37 and D70 residues (SI Appendix, Fig. S13B). These additional simulation results showed that hydrogen bonds formed in the microenvironment around D37, D70, and the 18-aldehyde group were essential for 18-oxo-CLA to stably bind to SlSRF. Furthermore, 18-oxo-CLA could then form the desired conformation for conversion to orobanchol in the complex.

We further investigated the catalytic mechanism of SlSRF involved in the stereoselective BC-ring closure reaction using density functional theory and the Our own N-layered Integrated molecular Orbital and Molecular mechanics (ONIOM) calculations (46–49) on a snapshot taken from the MD trajectory at 100 ns. In the ONIOM optimization for the reactant state, a proton initially bonded to D37 was spontaneously transferred to the 18-aldehyde group of 18-oxo-CLA via a water molecule (Fig. 5A). Subsequently, the distance between the C8 and C18 atoms was shortened from 2.75 Å to 1.68 Å to yield an intermediate through the transition state, TS1 (transitionstate 1) (Fig. 5 A–C). The activation energy (ΔE^‡) of TS1 from the reactant state was 11.5 kcal mol⁻¹. The carboxy group of 18-oxo-CLA was then bonded to the C7 atom to yield the product state through another transition state, TS2 (Fig. 5 D and E). The activation energy of TS2 from the intermediate state was relatively small (3.9 kcal mol⁻¹) but required the removal of the salt bridge between the carboxy group and R131, which was not observed in the reaction pathway in vacuo (SI Appendix, Fig. S12). The calculated reaction energy (ΔE) of the product state was −10.9 kcal mol⁻¹, and the activation energy from the reactant state was 14.6 kcal mol⁻¹ (Fig. 5F). These calculations demonstrated that SlSRF can produce orobanchol in a conrotatory 4π-ECR-like manner (13, 22) (Fig. 1).

Fig. 5.

Quantum mechanics/molecular mechanics (QM/MM) analysis of the reaction mechanisms of the BC-ring closure reaction catalyzed by SlSRF. (A–E) Structures of the reactant (A), TS1 (B), intermediate (C), TS2 (D), and product states (E) optimized for the SlSRF in complex with protonated 18-oxo-CLA using the ONIOM method. The H-bond network formed by Y90, D124, and R131 is illustrated with purple dashed lines in panel (A). Relative energies from the reactant state (ΔE) are shown in kcal/mol. (F) Energy profile of the reaction in SlSRF. Calculations were performed using the ONIOM(B3LYP/6-31G +(d,p):AMBER) method with an electronic embedding scheme, and energies are expressed in relative values from the reactant in kcal/mol. Key structures of the BC-ring closure reaction catalyzed by SlSRF are available at Zenodo (DOI: 10.5281/zenodo.10633290).

Experimental Validation of the Predicted Catalytic Residues.

Based on the computational results, we performed experimental site-directed mutagenesis to validate the significance of the catalytic residues of SlSRF. Out of seven mutants, six exhibited decreased activities, yielding less orobanchol than the WT (Fig. 6A). None of them showed spontaneous cyclization of 18-oxo-CLA to the diastereomer, ent-2′-epi-orobanchol (SI Appendix, Fig. S14). The decreases in the enzymatic activities for the D70A, Y90F, D124A, and R131A mutants were in good agreement with the complex model because they impaired hydrogen bond formation to stabilize 18-oxo-CLA and the cyclization activity. The D124A and Y90F/R131A mutations drastically reduced the activities to less than 1%, indicating that the H-bond network formed by these three residues to stabilize the carboxy group of 18-oxo-CLA is crucial for the enzymatic activity. Moreover, the complete loss of activity of the D37A mutant underscored the critical role of this residue in initiating the reaction as a proton donor. Interestingly, the D37E mutant displayed 1.37-fold higher activity than the WT. Subsequent calculations were performed to explore the impact of this mutation on catalytic activity. The D37E mutant formed a hydrogen bond with D70 by folding the elongated carbon chain, and the conformation of 18-oxo-CLA was stabilized similarly to the WT during the 500-ns simulation (SI Appendix, Fig. S15). In contrast to the WT, the QM/MM calculation for the D37E mutant showed a transition state with an activation energy of 11.9 kcal/mol and a reaction energy of −9.3 kcal/mol (Fig. 6 B–D). These calculations suggest that the D37E mutant enhances activity by slightly reducing the activation energy while maintaining its catalytic function. While we acknowledge the absence of a crystal structure for the complex, our proposed complex model aligns with the mutagenesis results and suggests a plausible cyclization mechanism for the formation of orobanchol from 18-oxo-CLA by SlSRF.

Fig. 6.

Site-directed mutagenesis and QM/MM analysis for the D37E mutant. (A) Relative activity of the SlSRF mutants. Error bars represent the SEM (n = 3 replicates). (B–D) Structures of the reactant (B), TS1 (C), and product states (D) optimized for the D37E mutant of SlSRF in complex with protonated 18-oxo-CLA. Relative energies from the reactant state (ΔE) are indicated in kcal/mol.

Discussion

The process of BC-ring formation in SL biosynthesis is a critical step for constructing the basic skeleton of the canonical SL. However, limited information was available related to this process. This study has successfully filled this knowledge gap by elucidating an orobanchol biosynthesis pathway in dicots. This pathway is distinct from the one identified in rice. Our research led to the identification of SRF, a member of the DIR-f subfamily. The enzyme plays a pivotal role in catalyzing the stereoselective cyclization of 18-oxo-CLA to orobanchol (Fig. 1). By employing a blend of experimental and computational methodologies, we are able to propose a potential catalytic mechanism of stereoselective BC-ring formation for canonical SL by SlSRF in tomato.

All members of the DIR family with previously identified biochemical functions in plants utilize phenols as substrates to produce plant specialized phenolic metabolites, with quinone methides as the reaction intermediates. SlSRF, implicated in the biosynthesis of orobanchol, exhibits divergent characteristics compared to previously proposed DIRs. This is because it uses 18-oxo-CLA as a substrate rather than phenolic compounds. Despite this divergence, the presence of conserved amino acid residues within the DIR family proteins, which are also found in SlSRF, suggests a shared functional characteristic responsible for the reaction as a DIR. A crucial requirement for DIR-mediated catalysis is the protonation of the substrate by the aspartate residue, specifically D37 in SlSRF. This residue is widely conserved among the DIR family proteins (Fig. 4C). In our proposed catalytic mechanism, D37 acts as a proton donor to protonate the 18-aldehyde group of 18-oxo-CLA through a water molecule. This initiates the sequential cyclization of the B- and C-rings in a 4π-ECR-like manner (Fig. 5). D50 in pterocarpan synthase GePTS1 of Glycyrrhiza echinata, a homologous residue of D37 in SlSRF, potentially functions as a proton donor for the hydroxy group of the substrate which leaves as water and eventually forms the quinone methide intermediate (29). In the case of the pinoresinol-forming DIR AtDIR6 of Arabidopsis, a proposed mechanism involves the 8-8′ coupling of two coniferyl alcohol quinone methide radicals to form a bis-quinone methide intermediate. In this mechanism, a residue that is homologous to D37 in SlSRF, specifically D49 in AtDIR6, acts as a proton donor. It protonates the carbonyl oxygen at one end of the bis-quinone methide, which enhances the electrophilicity of the neighboring methide carbon and promotes the cyclization at that half of the bis-quinone methide (35). These conserved aspartate residues, therefore, play a pivotal role as proton donors, facilitating both intra- and intermolecular cyclization reactions. Furthermore, Y90, D124, and R131 in SlSRF are suggested to be important for enzyme activity by forming a H-bond network to stabilize the carboxy group of 18-oxo-CLA. In the GePTS1 mechanism, the conserved R131 (R145 in GePTS1) may also be important for the quinone methide stabilization of the partial negative charge at the carbonyl oxygen (29), suggesting a similar role for these conserved residues. The function of DIR-f subfamily proteins remains largely unexplored. However, it seems plausible that DIR members of the SRF clade contribute to orobanchol synthesis. This is especially likely given that phylogenetic analysis shows that this clade contains DIRs from soybean (Glycine max), which are capable of orobanchol production (SI Appendix, Fig. S6). The identification of the DIR-f protein as SRF in this research represents a significant advancement in our understanding of the relevance of the DIR family in plants. While the biochemical function of most DIR family proteins remains unknown, it is predictable that they participate in the stereoselective biosynthesis of other plant specialized metabolites through a common DIR biochemical mechanism.

Certain cytochrome P450 enzymes, such as OsCYP711A2 in rice and GaCYP722C/GhCYP722C in cotton species, are involved in the canonical SL biosynthesis process. They catalyze the stereoselective formation of the BC-ring. In contrast, we found that SRF is necessary for the generation of the BC-ring in the subsequent reaction catalyzed by the CYP722C subfamily enzymes that convert CLA to 18-oxo-CLA in tomato and potentially in other orobanchol-producing dicot plants, including cowpea. In rice and cotton, where the ring formation is exclusively driven by cytochrome P450, previous phylogenetic analysis indicates that rice lacks the DIR-f subfamily (50). Among the cotton DIRs, even though some belong to the DIR-f subfamily, none of them fall within the SRF clade, which includes SlSRF and VuSRF (SI Appendix, Fig. S6). The stereoselective ring closure reactions catalyzed by cytochrome P450s in rice and cotton may also occur through the 4π-ECR via a molecular mechanism which controls the conformation of the substrate as in SlSRF (12, 22). Monocots typically have duplicate enzymes from the CYP711A subfamily, unlike dicots, which generally have a single enzyme from the CYP711A subfamily capable of converting carlactone to CLA. The neofunctionalization of the CYP711A subfamily in monocots has likely been crucial in acquiring the ability to synthesize canonical SLs. In dicots, the acquisition of the CYP722C subfamily is considered vital for canonical SL synthesis. The detailed molecular mechanisms of the different catalytic activities of the CYP722C subfamily in dicots are an emerging area of research. Recent studies on 5-deoxystrigol biosynthesis in sorghum have revealed a concerted catalytic mechanism involving three different types of enzymes: SbCYP711A31, a cytochrome P450; Sb3500, a 2-oxoglutarate-dependent dioxygenase; and LOW GERMINATION STIMULANT 1 (LGS1), a sulfotransferase (14). The lgs1 mutant in sorghum stereoselectively produces orobanchol (14, 51), suggesting alternative mechanisms for BC-ring formation. The involvement of sorghum DIRs in the conversion of 18-oxo-CLA to orobanchol is plausible due to the activity of SbCYP711A31 in forming putative 18-oxo-CLA (14). However, a comprehensive functional validation of sorghum DIRs is needed. Previous reports have established the absence of the DIR-f subfamily in sorghum (52). Correspondingly, our research did not identify any DIRs in sorghum with amino acid sequences showing significant homology to SlSRF. These findings indicate that diverse stereoselective canonical SL biosynthesis pathway exist in different plant species and highlight the acquisition of various biosynthetic mechanisms that facilitate the development of such pathways.

The loss of SRF function in tomato led to the formation of orobanchol and ent-2′-epi-orobanchol (Fig. 3 A and B), suggesting that SRF exhibits stereoselectivity toward the production of orobanchol, excluding the formation of ent-2′-epi-orobanchol. Although it is unclear why plants such as tomato and cowpea avoid the formation of ent-2′-epi-orobanchol, the elucidation of the functionality of DIRs belonging to the SRF-clade across various plant species and a comprehensive analysis of SRF gene knockout plants will improve the understanding of the significance of the C-ring stereochemistry in canonical SLs. The C-ring stereochemistry of orobanchol metabolites, solanacol and didehydro-orobanchol isomers, may be affected in SlSRF knockout tomato plants; however, solanacol was not detected and the amount of detected didehydro-orobanchol isomers was negligible (SI Appendix, Fig. S11), suggesting that the orobanchol metabolic pathway is no longer functional due to the loss of SlSRF. Didehydro-orobanchol isomers are believed to arise through the conversion of orobanchol mediated by SlCYP712G1, while solanacol is thought to derive from either structure of didehydro-orobanchol isomers catalyzed by an unidentified enzyme (33, 34). The lack of these orobanchol metabolites in SlSRF knockout plants could be due to a potential disruption in substrate transfer to SlCYP712G1. The stepwise conversion of didehydro-orobanchol isomers from CLA through orobanchol by SlCYP722C, SlSRF, and SlCYP712G1 may require efficient intermediate channeling between the enzymes. This could be facilitated by the formation of enzyme complexes, known as metabolons. Given the subcellular localization of SlSRF to the ER (Fig. 3 C and D), it is plausible that SlSRF plays a pivotal role in coordinating comprehensive SL synthesis in tomato through the formation of a metabolon with cytochrome P450s within the same subcellular component. The potential formation of a metabolon in the biosynthesis of 5-deoxystrigol in sorghum has also been discussed (14), prompting further exploration of this mechanism across various plant systems. Studying the role of metabolons in SL biosynthesis could shed light on their contribution to the regulated and efficient synthesis of SLs.

Our proposed mechanism for controlling the C-ring configuration, which is the principal determinant of SL diversification, provides a potential molecular basis for generating plants with a tailored SL composition. This is subject to experimental validation. Structurally different SLs influence the relative abundance of microbial taxa (53, 54) and the germination-stimulating activity of root parasitic weeds. For some root parasitic weeds, a mix of SLs with different C-ring configurations significantly reduces their germination rate (21, 55). Tailored manipulation of SLs could greatly enhance our understanding of the relationship between distinct SLs and rhizosphere biology. It could also enable the generation of crops designed to attract beneficial microbes or reduce the incidence of root parasitic weeds by suppressing their germination, as suggested in recent research (56). For instance, a mixture of orobanchol and ent-2′-epi-orobanchol inhibits the germination of Striga gesnerioides, a root parasitic weed affecting cowpea (21). Therefore, the loss of VuSRF function in cowpea could lead to the production of ent-2′-epi-orobanchol, as shown in tomato in this study. This could potentially confer resistance to the parasite by preventing germination. Finally, we demonstrate the usefulness of these complementary approaches between biochemistry and computational structural biology. These approaches address the molecular mechanism of the stereoselective formation of SL and its applicability to other plant biosynthetic systems where enzyme structures have not been experimentally determined.

Materials and Methods

Biochemical Characterization of SlCYP722C and SlSRF.

Recombinant SlCYP722C and SlSRF proteins were heterologously expressed in E. coli with the N-terminal transmembrane region truncated and the C-terminal His₆-tag added. Each recombinant protein was purified using His-tag affinity chromatography, followed by additional purification steps including ion-exchange chromatography. The activities of SlCYP722C and SlSRF were assessed in vitro using CLA and 18-oxo-CLA, respectively, as substrates. The substrate 18-oxo-CLA was prepared at a time of use as follows: The SlCYP722C enzyme reaction products, extracted with ethyl acetate and dried, were dissolved with a small volume of acetonitrile and dispensed with 50 mM potassium phosphate buffer (pH 7.4). The reaction products were extracted with ethyl acetate and analyzed by LC–MS/MS.

The details and procedures of chemical preparations, analysis of SLs, production of recombinant proteins, biochemical analyses (in vitro enzyme assay and transient expression assay), generation of SlSRF knockout tomato plants, subcellular localization, and computational calculations are provided in SI Appendix, Materials and Methods. Primers used in this study are provided in SI Appendix, Table S4.

Supplementary Material

Appendix 01 (PDF)

Dataset S01 (XLSX)

Movie S1.

A 500-ns MD simulation for the 18-oxo-CLA in complex with SlSRF.

Movie S2.

A 500-ns MD simulation for the 18-oxo-CLA in water solvent.

Data, Materials, and Software Availability

1. Gene sequence data, 2. Pymol session file data have been deposited in 1. DNA Data Bank of Japan (DDBJ) and 2. Zenodo (1. LC798354 and LC798355 and 2. DOI: 10.5281/zenodo.10633290). Previously published data were used for this work (20, 25). (DDBJ Sequence Read Archive, accession no. DRA008222). All other data are included in the article and/or supporting information.