¹H and ¹³C NMR spectra were recorded in C₆D₆ on a Bruker Biospin AC400M spectrometer (400 MHz for ¹H and 100 MHz for ¹³C). Standard pulse sequence and phase cycling were used for HSQC, HMBC, COSY, and NOESY spectra analyses. Chemical shifts were referenced to C₆D₆ (δ _H 7.16, δ _C 128.06). UV and CD spectra were obtained using a V-630 spectrophotometer and a J-805 spectropolarimeter (JASCO, Tokyo, Japan), respectively. High resolution mass spectra were recorded using an Orbitrap Exploris 240 mass spectrometer equipped with an ESI source (HR-ESI-MS), connected to an Ultimate 3000 HPLC system (Thermo Fisher Scientific). LC-MS/MS analyses were performed using an LC-MS/MS system (Waters) comprising Acquity UPLC H-Class and an Acquity TQD tandem mass spectrometer. Data acquisition and analyses were performed using MassLynx 4.1 software (Waters).

Rac-orobanchol was prepared as reported previously (Matsui et al., 1999). Aromatic CLA was synthesized from 2,3,6-trimethylphenol based on the previously reported methods (Wakabayashi et al., 2019; Shiotani et al., 2021). The Supplementary Methods provide more information.

Tomato cultivars obtained from a local market in Japan were grown hydroponically according to the following procedure. Surface sterilized tomato seeds were sown on MS agar plates and grown for 11 days at 23°C with a 16-h light/8-h dark photoperiod. Then, the plants were transferred to test tubes containing 50 ml half-strength Hoagland nutrient solution without phosphate. After three days, the nutrient solution was replaced with a fresh one and each plant was grown for a week. Each filtrate was subjected to purification with a solid phase extraction using an Oasis HLB cartridge (Waters, USA) as described previously (Moriyama et al., 2022). An aliquot (5 µL) of the extract dissolved in 100 µL acetonitrile was subjected to an LC-MS/MS analysis.

The selected tomato cultivar, Yellow-Pico, tomato seedlings were grown hydroponically in a tank (245×365×140 mm) filled with phosphate-limited half-strength Hoagland nutrient solution (phosphorus concentration reduced to one-tenth of half-strength Hoagland nutrient solution) at 25°C with a 16-h light/8-h dark photoperiod. The solution was continuously circulated using a pump, and root exudates of 3–16 week old plants were adsorbed on XAD-4 (2.1 L in total) as described previously (Ueno et al., 2014). The XAD-4, replaced every week with a fresh one, was soaked in acetone (16.8 L in total). SLs were extracted by sonication and filtered. After evaporating the acetone, the residual aqueous solution was treated three times with the same volume of EtOAc. The organic layer was dried over Na₂SO₄ and concentrated in vacuo.

The crude extracts of tomato root exudate (4 g) were subjected to silica gel column chromatography (Silica Gel 60, spherical, neutral 120 g). Stepwise elution with a solvent system of n-hexane–EtOAc (100:0–0:100, 10% step, 827 mL each) produced 11 fractions (Frs. 1–11). A mixture of 6,7-didehydroorobanchol/A1, phelipanchol/A2, and epiphelipanchol/A3, eluted mostly in Fr. 8 (n-hexane–EtOAc, 30:70, v/v), was further purified using reversed-phase HPLC on a COSMOSIL 5C18-MS-II column (250 × 10 mm i.d., 5 µm) (Nacalai Tesque Inc., Kyoto, Japan) with a CH₃CN/H₂O (30:70, v/v) isocratic system at 5.0 mL/min flow rate. Eluents were monitored at 254 nm for three peaks at 39, 40.5, and 41.5 min, yielding 6,7-didehydroorobanchol/A1 (140 μg) phelipanchol/A2 (930 μg) and epiphelipanchol/A3 (130 μg), respectively.

Tables 1 , 2 ; Supplementary Tables S1, S2 and Supplementary Figures S5-S16 show NMR spectroscopic data on 6,7-didehydroorobanchol/A1, phelipanchol/A2, and epiphelipanchol/A3. HR-ESI-MS m/z calcd for [C₁₉H₂₀O₆ + H]⁺: 345.13326, found 345.1331 (6,7-didehydroorobanchol/A1); 345.1329 (phelipanchol/A2); and 345.1330 (epiphelipanchol/A3). Mass spectra ( Supplementary Figure S1 ). CD (6,7-didehydroorobanchol/A1: CH₃CN, c 0.000016) λ _ext (Δϵ) nm: 227 (6.05); (phelipanchol/A2: CH₃CN, c 0.000017) λ _ext (Δϵ) nm: 223 (4.78); (epiphelipanchol/A3: CH₃CN, c 0.000013) λ _ext (Δϵ) nm: 225 (7.10) ( Supplementary Figure S2 ).

The SLs were analyzed using LC-MS/MS under analytical conditions as previously described (Wakabayashi et al., 2019). Chromatographic separation was performed with an ODS column (COSMOSIL 2.5C18-MS-II, 100 × 2.0 mm i.d., 2.5 µm; Nacalai Tesque) at a column oven temperature of 30°C. The elution was performed in a linear gradient system of MeOH–H₂O with 0.1% formic acid (50:50–100:0 in 20 min) at a flow rate of 0.2 mL min⁻¹. The MRM transitions selected were at m/z 347.1 > 233.1 for orobanchol, 345.2 > 97 for DDHs, and 343.1 > 97 for solanacol, in the positive ESI mode. The cone voltage and the collision energy were 25 V and 18 eV, respectively.

The germination assay was conducted as reported previously (Sugimoto et al., 2003). Seeds of P. ramosa and S. hermonthica were collected from mature plants of tomato in Khatroum and sorghum in Gedarif, Sudan, respectively. The surface sterilized seeds were pretreated (conditioned) for 10 days at 30°C for S. hermonthica, and 23°C for P. ramosa on 8-mm glass-fiber filter paper disks (ca. 50 seeds each) placed on distilled water-saturated filter paper. Aliquots (20 μL) of diluted test solutions were assayed by applying them to conditioned seeds on 8 mm disks. The treated seeds were incubated at the same temperature as that used for the conditioning and were microscopically examined after 1 day (S. hermonthica) and 5 days (P. ramosa).

In a previous study, we generated SlCYP722C- knocked out (KO) tomato plants (cv. Micro-Tom) by CRISPR/Cas9-mediated genome editing to investigate its function in planta. Orobanchol and solanacol levels in root exudates of SlCYP722C-KO tomato plants decreased to below the detection limit, but CLA, a substrate of the SlCYP722C enzyme, was abundant (Wakabayashi et al., 2019). At least three DDHs were found in the tomato hydroponic solutions (Kohlen et al., 2013; Zhang et al., 2018). In our LC-MS/MS analytical conditions with an ODS column, SL-like compounds with the same molecular mass as DDHs were aggregated as a single peak on the chromatogram. However, the peak disappeared in the SlCYP722C-KO tomato exudates, indicating that the SL-like compounds are metabolites downstream of orobanchol (Wakabayashi et al., 2019) ( Supplementary Figure S3 ). We qualitatively and quantitatively analyzed SLs in root exudates of more than 20 tomato cultivars available in the local market in Japan. As a result, the cultivar Yellow-Pico (YP) was selected as the highest producer of the SL-like compounds for further isolation and structural elucidation ( Supplementary Figure S4 ).

YP was grown hydroponically in a phosphate-limited half-strength Hoagland nutrient solution, and root exudates were obtained by absorption on XAD-4 as previously described (Ueno et al., 2014; Moriyama et al., 2022). Purification of YP root exudates by silica gel column chromatography guided by LC-MS/MS analysis resulted in enriched fractions of SL-like compounds eluted with 30% EtOAc in n-hexane. We separated the SL-like compounds using HPLC on a 25-cm long ODS column, with three SL-like compounds eluting at 39, 40.5, and 41.5 min ( Figure 2 ). These novel SLs were tentatively called A1, A2 and A3 in the order of their elution from reversed-phase chromatography because their identity to the previously reported DDHs (Kohlen et al., 2013; Zhang et al., 2018) was unknown. The molecular formula of A1, A2 and A3 was established to be C₁₉H₂₀O₆ based on proton adduct ion [M + H]⁺ at m/z 345.1331, 345.1329, and 345.1330, respectively, obtained using HR-ESI-MS, which is consistent with that of DDH (calcd. for [C₁₉H₂₀O₆ + H]⁺, 345.13326). The mass spectra of A1, A2, and A3 were almost identical and exhibited major common fragment ions at m/z 97, indicating that they have the D-ring of SL ( Supplementary Figure S1 ). A1, A2, and A3 exhibited slightly different CD spectra, with A2 and A3 having a positive cotton effect around 250 nm, which is not observed in A1 ( Supplementary Figure S2 ).

Table 1 lists ¹H NMR spectroscopic data of A1, A2, and A3 aided with 2D NMR experiments, including ¹H-¹H COSY, and NOESY. Compared with the orobanchol data (Tokunaga et al., 2015), remarkable differences were observed in methylene protons in the A-ring in A1. ¹H NMR spectra of A2 and A3 closely resembled each other. Typical canonical SLs have 19 carbons with two geminal methyl groups at C-8 in the A-ring and an olefin methyl group at C-4′ in the D-ring. In contrast to A1 and orobanchol, A2 and A3 have only one methyl group in the A-ring. Instead, two mutually coupled broad ¹H singlets corresponding to a methylidene group were observed.

Table 2 shows ¹³C NMR, COSY, HMBC, and NOESY spectroscopic data from A2, indicating two methyl groups (H-9 and H-7′), two methylene groups (H-5a,b, and H6a,b), five methine groups (H-3a, H-4, H-7, H-8b, and H-2′), two olefin protons (H-3′ and H-6′) and two methylidene protons (H-10a,b). The ¹H-¹H COSY and HMBC correlations from H-2′, H-3′, H-6′, and H-7′ together with the mass fragment ions at m/z 97 and 231 showed the conserved enol ether bridged methylbutenolide (the D-ring). The ¹H-NMR spectroscopic data exhibited mutually coupled methine protons (H-3a at δ3.17 and H-8b at δ5.24), which are characteristic signals observed in canonical SLs. A broad methine singlet at δ4.32, corresponding to H-4 in orobanchol, was also observed. The HMBC experiment confirmed that the H-3a signal correlated with the C-2, C-3, C-4, C-4a, and C-6′ resonances, and the H-8b signal correlated with the C-2 and C-8a resonances. Correlations of an olefin proton (H-6′ at δ7.36) with the C-2, C-3, C-3a, and C-2′ resonances were also observed. These results indicated a γ-lactone moiety and cyclopentene moiety fused to form the BC-ring. Protons H5a,b, and H6a,b were mutually coupled. The ¹H-¹H COSY correlations of H-7 with protons H-6a,b, H-9, and H-10a,b were observed. The HMBC correlations of H-5a,b to C-4a, C-6, C-8, and C-8a, those of H-6a,b to C-5, C-7, C-8, C-8a, and C-9, those of H-9 to C-5, C-6, C-8, and those of H-10a,b to C-4a, C-5, C-7, and C-8 revealed a cyclohexene A-ring with a methyl group at C-7 and a methylidene group at C-8. H-10a and H-10b were distinguishable by the NOESY correlations of H-10a with H-7 and H-9, and that of H10b with H-8b. Thus, the unique structure of A2 having a methyl group and a methylidene group at C-7 and C-8 in the A-ring, respectively, was established ( Figures 1 , 3 ). The NOESY correlations of H-5b with H-4 and H-9 indicated that the absolute configuration at C-7 is S ( Figure 3 ).

The chemical shift and coupling patterns of A3’s ¹H NMR spectrum were very similar to those of A2. Little but critical difference was observed in the chemical shift of H-5b, which was shifted in an up-filed by 0.09 ppm compared with that of A2 ( Table 1 ). A comprehensive ¹³C NMR spectrum of A3 was not recorded because of the limited amount. However, significant correlations were observed in COSY and NOESY spectroscopic data and all the proton signals were assigned with reference to those in A2. The NOESY correlations of H-5a and H-9 ( Supplementary Table S1 ) strongly indicate that A3 is the epimer of A2 at C-7 ( Figure 1 ).

The ¹H NMR spectroscopic data of A1 revealed three methyl groups, which are typical of canonical SLs ( Table 1 ). Remarkable differences between A1 and orobanchol were observed in the methylene proton signals in the A-ring. The methylene protons at δ2.12 and δ2.47 in A1 resonated at a downfield compared with those in orobanchol. Mutually coupled olefinic protons correlated with the methylene protons were observed at δ5.40 and δ5.34 in the ¹H-¹H COSY spectrum. The coupling constants of the methylene protons with the former (3.0 and 3.2 Hz) were larger than those with the latter (1.5 and 1.7 Hz). The NOESY correlations of the methine proton at δ5.22 (H-8b) with the methyl protons at δ1.06 (H-9) and those of the olefinic proton at δ5.34 (H-7) with methyl protons at δ1.06 (H-9) and δ1.30 (H-10) were observed ( Supplementary Table S2 ). These data indicate a double bond between C-6 and C-7 in the A-ring of A1. Thus, the novel SL is identified as 6,7-didehydroorobanchol ( Figure 1 ).

Figure 4 shows the germination-inducing activities of 6,7-didehydroorobanchol/A1, A2, and A3 for P. ramosa and S. hermonthica. All of these SLs elicited the germination of P. ramosa, which is causing damage to tomato with significant yield losses (Parker, 2013). The 6,7-didehydroorobanchol/A1 showed lower germination-inducing activity than A2 and A3, and the different activity may reflect structural differences in the A-ring. A2 and A3, which structurally do not fall into the DDH category, were named phelipanchol and epiphelipanchol, respectively, because of their potent germination-inducing activities for the germination of P. romosa.

Three novel canonical SLs with the same molecular mass as DDHs, 6,7-didehydroorobanchol/A1, phelipanchol/A2, and epiphelipanchol/A3, were isolated from root exudates of tomato. Based on detailed analyses of NMR, mass, and CD spectra, their structures were established ( Figure 1 ). Note that 6,7-didehydroorobanchol/A1 is a DDH isomer but phelipanchol/A2 and epiphelipanchol/A3 do not fall into the category of DDH isomers. Solanacol was reported to be effective in inducing germination of P. ramosa seeds (Xie et al., 2007). These novel SLs induced germination of root parasitic weeds and showed potent germination-inducing activity for P. ramose similar to that of solanacol, but weak activity for S. hermonthica ( Figure 4 ), which was consistent with the finding that orobanchol-type SLs showed weak germination-inducing activity for S. hermonthica (Ueno et al., 2011; Nomura et al., 2013; Wakabayashi et al., 2022). Root parasitic weeds are likely to use these SLs as germination stimulants in the tomato rhizosphere. Furthermore, natural canonical SLs are known to be highly active in inducing hyphal branching in arbuscular mycorrhizal fungi (Akiyama et al., 2010), indicating that these SLs may function similarly.

The conversion of orobanchol to solanacol and DDH(s) was demonstrated in tomato (Zhang et al., 2018) in accordance with our observation that these SLs decreased to below the detection limit in root exudates of SlCYP722C-KO tomato plants (Wakabayashi et al., 2019) ( Supplementary Figure S3 ). In addition, 7-α-hydroxyorobanchol and 7-β-hydroxyorobanchol were previously isolated from cucumber (Cucumis sativus) as orobanchol-related compounds (Khetkam et al., 2014) and identified in tomato root exudates (Kohlen et al., 2013). Considering the structures of novel SLs elucidated in this study and the possible presence of hydroxyorobanchol isomers, a plausible biosynthetic pathway from orobanchol to solanacol can be proposed ( Figure 1 ). A hydroxy group is introduced into orobanchol at C-7, probably in α and β orientations. Dehydration of the 7-hydroxyorobanchol isomers could yield 6,7-didehydroorobanchol/A1 irrespective of the orientation of the hydroxyl group at C-7. Alternatively, introduction of a hydroxy group at C-6, followed by dehydration, could also yield 6,7-didehydroorobanchol as described by Wang et al. (2022). The migration of a methyl group at C-8 of the 7-hydroxyorobanchol isomers to C-7 assisted by elimination of the hydroxy group could result in carbocation intermediates. The formation of a double bond in the exocyclic position at C-8 of the intermediates could result in phelipanchol/A2 and its 7-epimer epiphelipanchol/A3. The orientation of the methyl group at C-7 is assumed to depend on the migration mechanism of the methyl group at C-8 in the 7-hydroxyorobanchol isomers. When the double bond is formed in the endocyclic position in the A-ring of the carbocation intermediates, they lose chirality at C-7 to yield 5,6-dihydrosolanacol. In this study, 5,6-dihydrosolanacol could not be isolated and is suspected to be present only in trace amounts as a biosynthetic intermediate because of its chemical instability. Furthermore, SlCYP712G1 in tomato was recently identified as an enzyme responsible for the conversion of orobanchol to three DDH structures, which remain unknown but include 5,6-dihydrosolanacol as a putative structure (Wang et al., 2022). Different from our proposed biosynthetic pathway, the authors hypothesized that DDHs and solanacol are derived from one of the hydroxyorobanchol isomers. The structural determination of SlCYP712G1 reaction products, including DDHs and hydroxyorobanchol isomers, is a future challenge, but if SlCYP712G1 is responsible for the reactions to the three novel SLs identified in this study, it would catalyze the above series of reactions ( Figure 1 ).

In addition to the above proposed biosynthetic pathways, we hypothesized a biosynthetic route to solanacol from the aromatic A-ring analog of CLA (aromatic CLA), which has the same structure in the A-ring as that of solanacol. If the A-ring of CLA is aromatized before the BC-ring formation, SlCYP722C may convert aromatic CLA to solanacol ( Figure 1 ). To examine this hypothesis, aromatic CLA was synthesized and used as a substrate in an enzyme reaction for SlCYP722C in vitro. However, the enzyme reaction of SlCYP722C with aromatic CLA did not result in the formation of solanacol (data not shown). Therefore, it is likely that the A-ring of solanacol is formed after the BC-ring closure, probably via 5,6-dihydrosolanacol ( Figure 1 ).

Future studies are required to elucidate the biosynthetic pathway from orobanchol to solanacol and the role of the structurally diverse SLs produced by tomato in the rhizosphere. This study provides new insights that will contribute to the entire understanding of the SL biosynthetic pathway in tomato.