Optical rotations were acquired on a JASCOP-1020 polarimeter. ECD data were measured on a Chirascan spectropolarimeter. IR spectra were measured on PerkinElmer infrared spectrophotometer. 1D and 2D NMR spectra were recorded on Bruker Avance 400 or 600 DRX spectrometers in acetone-d₆, methanol-d₄, DMSO-d₆ and chloroform-d. Column chromatography (CC) was performed on silica gel (200–300 mesh; Qingdao Marine Chemical Plant Branch., China), RP-C18 (ODS-A, 50 μm, YMC, Kyoto, Japan), or Sephadex LH-20 (100–200 mesh; Beijing Solarbio Technology Co., Ltd., China). Plates precoated with silica gel GF254 (Rushan, Shandong Sun Desiccant Co., Ltd.) were used for thin layer chromatography (TLC). An Agilent HPLC series 1260 and Shimadzu LC-20AR were used for analysis and isolation. For analysis, an Agilent Eclipse XDB-C18 column (4.6 × 150 mm, 5 μm) was used. The isolation was achieved on an Agilent semi-preparative Eclipse XDB-C18 column (9.4 × 250 mm, 5 μm). HPLC-MS data were acquired on an Agilent 1260 Series system coupled with an Agilent Accurate-Mass-Q-TOF MS 6520 system equipped with an Electrospray ionization (ESI) source.

The endophytic fungus F. decemcellulare F25 was isolated from the stem of the Chinese medicinal plant M. fortunei collected from Qingdao, People’s Republic of China. The fungal strain was deposited in 20% glycerol at −80°C in the school of Pharmaceutical Sciences, Guizhou University, Guizhou, China. The endophytic fungus was identified as F. decemcellulare by the analysis of internal transcribed spacer (ITS) region of the rDNA (GenBank No. OQ001346).

The fungal strain F. decemcellulare F25 was cultured on potato dextrose agar (PDA) media for a week at 28 ± 2°C. A week-old culture plate was cut into small pieces under aseptic conditions, and were then inoculated into 394 flasks (300 mL) each containing 40 g of rice, 0.12 g of peptone, and 60 mL of water. The cultures were incubated at 28 ± 2°C for 40 days. Afterward, the whole cultures were extracted with ethyl acetate by sonication under ice bath conditions for three times. Then the EtOAc solution was collected and evaporated to dryness, affording 352.2 g of brown extracts. After suspension of the crude extract in water, petroleum ether and EtOAc were used to extract 294.6 g and 50.0 g of the corresponding organic phase, respectively.

The EtOAc extract was fractionated by column chromatography (CC) on ODS eluting with a gradient of acetonitrile (CH₃CN)/H₂O (0:100; 3:7; 5:5; 7:3; 100:0, v/v, each 8 L) to give eight fractions (Fr. A–Fr. H).

Fraction A was applied to semi-preparative HPLC to yield compounds 11 (t_R 22.1 min, 53.3 mg), 4 (t_R 38.2 min, 8.9 mg), 5 (t_R 39.1 min, 16.5 mg), 6 (t_R 40.3 min, 29.1 mg), and a mixture of 8 and 9 (t_R 29.2 min, 16.2 mg), eluting with a gradient of CH₃CN in H₂O from 40 to 65%. The above mixture was subjected to isolation on a chiral HPLC column to afford compounds 8 (t_R 25.1 min, 6.2 mg) and 9 (t_R 28.9 min, 5.9 mg). Fraction A (14.4 g) was also separated into 12 subfractions (A1 − A12) by CC on silica gel eluted by CH₂Cl₂/CH₃OH (1:0, 40:1, 30:1, 20:1, 10:1, 5:1 and 0:1, v/v, each 7 L). Subfraction A3 was purified by semi-preparative HPLC with a gradient elution from 30 to 85% CH₃CN in H₂O to afford compounds 1 (t_R 33.0 min, 11.5 mg) and 18 (t_R 12.6 min, 16.4 mg). Subfraction A6 was then applied to semi-preparative HPLC with a gradient from 30 to 55% CH₃CN in H₂O as eluent to obtain compound 19 (t_R 14.2 min, 25.3 mg). Compound 10 (t_R 33 min, 27.8 mg) was purified by semi-preparative HPLC, eluting with a gradient of CH₃CN in H₂O from 40 to 65% (v/v) as eluent from subfraction A7. Compound 2 (t_R 17.5 min, 10.1 mg) was obtained from subfraction A8 using semi-preparative HPLC with a gradient elution from 30 to 85% CH₃CN in H₂O. Subfrcation A9 was separated by semi-preparative HPLC with a gradient elution from 30 to 75% CH₃CN in H₂O to yield compound 3 (t_R 27.0 min, 10.6 mg).

Fraction C was fractionated by CC on Sephadex LH-20 eluting with CH₃OH/CH₂Cl₂ (1:1, v/v) to give five subfractions (C1 ̶ C5). Subfraction C3 was purified by semi-preparative HPLC with a gradient elution from 45 to 50% CH₃CN in H₂O with 0.1% trifluoroacetic acid (TFA) to afford compounds 15 (t_R 38.1 min, 7.5 mg) and 16 (t_R 37.5 min, 10.3 mg). Fraction C4 was subjected to semi-preparative HPLC with a gradient elution from 35 to 65% CH₃CN in H₂O to yield compound 7 (t_R 16.0 min, 2.0 mg).

Fraction D was purified by semi-preparative HPLC eluting with gradient from 60 to 70% CH₃CN in H₂O with 0.1% TFA to offer compound 17 (t_R 33.5 min, 6.3 mg). Compound 12 (t_R 27.2 min, 6.2 mg) was obtained from fraction F by semi-preparative HPLC eluting with a gradient elution from 50 to 100% CH₃CN in H₂O. Compounds 13 (t_R 19.3 min, 17.5 mg), 14 (t_R 22.0 min, 17.7 mg), and 20 (t_R 28.5 min, 13.1 mg) were obtained from fraction G using semi-preparative HPLC with a gradient elution from 75 to 100% CH₃CN in H₂O.

Compound (1), yellowish solid; LC-UV (CH₃CN in H₂O) λ_max: 248, 330 nm; IR ν_max: 3319, 2942, 2832, 1677, 1020 cm⁻¹; ¹H NMR (DMSO-d₆, 400 MHz); and ¹³C NMR (DMSO-d₆, 100 MHz) data, see Table 1; HRESIMS m/z 285.0762 [M + H]⁺ (calcd. For C₁₆H₁₃O₅, 285.0757).

Compound (2), yellowish solid; [ α ] D 23 +51.6 (c 0.68, CH₃OH); LC-UV (CH₃CN in H₂O) λ_max: 244, 276, 330 nm; IR ν_max: 3330, 2925, 1681, 1642, 1625, 1572, 1357, 1237, 1163, 1018, 990 cm⁻¹; ¹H NMR (DMSO-d₆, 400 MHz); and ¹³C NMR (DMSO-d₆,100 MHz) data, see Table 1; HRESIMS m/z 411.1289 [M + H]⁺ (calcd. For C₁₉H₂₃O₁₀, 411.1286).

Compound (3), yellowish solid; [ α ] D 23 +63.2 (c 0.56, CH₃OH); LC-UV (CH₃CN in H₂O) λ_max: 204, 236, 336 nm; IR ν_max: 3310, 2917, 1678, 1616, 1570, 1505, 1470, 1400, 1158, 1024, 999 cm⁻¹; ¹H NMR (DMSO-d₆, 400 MHz); and ¹³C NMR (DMSO-d₆,100 MHz) data, see Table 1; HRESIMS m/z 433.1133 [M + H]⁺ (calcd. For C₂₁H₂₁O₁₀, 433.1129).

Compound (8), colorless oil; [ α ] D 24 -22.7 (c 0.59, CH₃OH); LC-UV (CH₃CN in H₂O) λ_max: 200, 230, 286 nm; IR ν_max: 3329, 2946, 2836, 1661, 1451, 1408, 1114, 1017 cm^{−1 1}H NMR (DMSO-d₆, 400 MHz) and ¹³C NMR (DMSO-d₆,100 MHz) data, see Supplementary Table 1; HRESIMS m/z 334,1,649 [M + H]⁺ (calcd. For C₁₈H₂₄NO₅, 334.1649).

Compound (9), colorless oil; [ α ] D 24 +8.23 (c 0.61, CH₃OH); LC-UV (CH₃CN in H₂O) λ_max: 200, 230, 286 nm; IR ν_max: 3329, 2946, 2836, 1661, 1451, 1408, 1114, 1017 cm⁻¹;¹H NMR (DMSO-d₆, 400 MHz); and ¹³C NMR (DMSO-d₆,100 MHz) data, see Supplementary Table 1; HRESIMS m/z 334, 1,652 [M + H]⁺ (calcd. For C₁₈H₂₄NO₅, 334.1649).

Compound (10), colorless oil; [ α ] D 23 -24.1 (c 2.25, CH₃OH); LC-UV (CH₃CN in H₂O) λ_max: 204, 228, 286 nm; IR ν_max: 3336, 3286, 2905, 1644, 1427, 1367, 1334, 1314, 1164, 1053, 1030 cm⁻¹; ¹H NMR (DMSO-d₆, 400 MHz); and ¹³C NMR (DMSO-d₆,100 MHz) data, see Supplementary Table 1; HRESIMS m/z 336,1808 [M + H]⁺ (calcd. For C₁₈H₂₆NO₅, 336.1805).

Compound (15), yellow powder; LC-UV (CH₃CN in H₂O) λ_max: 220, 280, 360 nm; IR ν_max: 3387, 2833, 1698, 1475, 1391, 1017 cm⁻¹; ¹H NMR (DMSO-d₆, 400 MHz); and ¹³C NMR (DMSO-d₆,100 MHz) data, see Supplementary Table 2; HRESIMS m/z 263.1277 [M + H]⁺ (calcd. For C₁₅H₁₉O₄, 263.1278).

The crystal structure of compound 9 was obtained from the solution of CH₃OH. A suitable crystal were collected on a Bruker APEX-II CCD Venture diffractometer using graphite-monochromated Mo Kα radiation (λ = 0.710 73 Å) at 297 K. Absorption correction using equivalent reflctions was performed with the SADABS program. Crystallographic tables were constructed using Olex2 (Dolomanov et al., 2010). The structure was solved with the Shelxt software package

(Sheldrick, 2015), and refined with the Shelxt refinement package using Least Squares minimization.

Crystal data for compound 9: C₁₈H₂₅NO₆ (M = 351.39 g/mol): triclinic, space group P1, a = 6.4597(7) Å, b = 7.5258(7) Å, c = 9.4129(7) Å, α = 95°, β = 100°, γ = 93°, V = 93.981(8) ų, Z = 1, T = 297 K, μ(Mo Kα) = 0.098 mm⁻¹, F(000) = 188, ρ_calc = 1.311 g/cm³; of the 8,205 reflections measured (4.44° ≤ 2Θ ≤ 50.01°), 2,823 were unique (R_int = 0.0922, R_sigma = 0.0810) which used in all calculations. The final R₁ was 0.0596 (I > 2σ(I)), and wR₂ was 0.1706 (all data).

Following our previously established methods (Wang et al., 2019; Tian et al., 2021), the crude extract of F. decemcellulare F25 was firstly evaluated for antifungal activity against five plant pathogens (Colletotrichum musae ACCC 31244, Alternaria solani, Fusarium foetens, Fusarium mangiferae, and Lasiodiplodia pseudotheobromae) by agar diffusion assay. The crude extract showed inhibitory activity against C. musae ACCC 31244, indicating the production of antifungal molecules. Further antifungal evaluation of pure compounds against C. musae ACCC 31244 was determined with the broth dilution method, and provided minimum inhibitory concentration (MIC) values. The cycloheximide was used as a positive control in parallel to reveal the comparative antifungal efficacy of compounds 1–20.

The OSMAC (One Strain Many Compounds) approach refers to the activation of many silent gene clusters in microorganisms by altering the culture environment of the strain. This strategy maximizes the biosynthetic capacity of a microorganism that produces structurally diverse and biologically active secondary metabolites. The F. decemcellulare F25 was cultured on four different solid media including rice-based, soybean-based, corn-based, and czapek-dox agar (CDA) culture. Remarkably, HPLC chromatograms showed a number of peaks in rice-based culture, suggesting that the rice medium strongly triggered the production of secondary metabolites (Supplementary Figure 1).

The antifungal activities of crude extracts were evaluated against five plant pathogens. Compared with the positive control drug cycloheximide, it was found that, at the concentration of 40 μg/paper disk, the crude extract of F. decemcellulare F25 showed antifungal activity against the fungal C. musae (Supplementary Figure 2). Considering the abundant secondary metabolites and the antifungal activity of F. decemcellulare F25, this strain F25 was further subjected to chemical investigation.

Seven new compounds, including three isocoumarins (1–3), three pyrrolidinone derivatives (8–10), and one pentaene diacid (15), together with 13 known compounds, were isolated from the rice culture of the endophytic fungus Fusarium decemcellulare F25.

Compound 1 was obtained as a yellowish solid. Its molecular formula C₁₆H₁₂O₅ was established by the HRESIMS at m/z 285.0762 [M + H]⁺ (calcd. For 285.0757), implying eleven degrees of unsaturation. The presence of hydroxyl and carbonyl groups were implied by IR absorption bands at 3319 and 1677 cm⁻¹, respectively. The ¹H and ¹³C NMR data of 1 (Table 1) were highly similar to those of pleosporalone A (Cao et al., 2016), excepted for a proton at C-7 in 1 rather than a methyl group in pleosporalone A. In addition, the coupling constant between H-5 and H-7 (J = 2.1 Hz) proved that there is no methyl substitution at C-7 of 1. The HMBC correlations (Figure 2) from H-5 (δ_H 6.47) to C-6 (δ_C 165.7), C-7 (δ_C 101.9), and C-8a (δ_C 98.4), and from H-7 (δ_H 6.38) to C-5 (δ_C 103.3), C-6 (δ_C 165.7), C-8 (δ_C 162.7), and C-8a (δ_C 98.4), together with the HSQC spectrum (δ_C/δ_H 101.9/6.38 and 103.3/6.47 ppm) further confirmed that there is a proton attached to C-7. Final detailed analysis of the HSQC and HMBC spectra allowed the assignment for all proton and carbon resonances of 1. Thus, the structure of compound 1 was assigned completely.

Compound 2 was obtained as a yellowish solid. Its molecular formula C₁₉H₂₂O₁₀ was established by the HRESIMS at m/z 411.1289 [M + H]⁺ (calcd for 411.1286), implying nine degrees of unsaturation. The presence of hydroxyl and carbonyl groups were shown by IR absorption bands at 3330, 1681 cm⁻¹, respectively. The attachment of sugar unit was determined to be ribose by comparison of ¹H and ¹³C NMR data of compound 2 with those of daldiniside C. The coupling constant of the anomeric proton at δ_H 5.74 (1H, d, J = 4.4 Hz) indicated the ribose unit should be α-configured (Hu et al., 2014). Additionally, the pentose moiety was linked to C-6 proved by the correlation of H-1’/C-6 observed in the HMBC (Figure 2) experiment and H-1’/H-5, H-1’/H-7 in NOESY spectrum (Figure 3). The ¹H NMR and ¹³C NMR data (Table 1) showed the presence of an isocoumarin unit in 2, whose structure was the same as that of compound 6 ((−)-citreoisocoumarin)) (Yamamura et al., 1991) by comparison with spectroscopic data. Thus, the structure of compound 2 was established.

Compound 3 was obtained as a yellowish solid. Its molecular formula C₂₁H₂₀O₁₀ was established by the HRESIMS at m/z 433.1133 [M + H]⁺ (calcd for 433.1129), implying 12 degrees of unsaturation. The presence of hydroxyl and carbonyl groups were shown by IR absorption bands at 3310, 1678 cm⁻¹, respectively. The ¹H and ¹³C NMR data of aglycone in 3 (Table 1) were highly similar to those of polyisocoumarin (Jiang et al., 2020), which was previously isolated from Polygonum cuspidatum. However, an α-ribose attached at the C-6 of 3 rather than a β-d-glucopyrancose at C-6 in polyisocoumarin. And the NMR data showed that compound 3 and compound 2 have the same ribose at C-6. Thus, the structure of compound 3 was assigned completely.

Compounds 8 and 9 were obtained as colorless oil. Their molecular formula C₁₈H₂₃NO₅ were established by the HRESIMS at m/z 334.1649 and 334.1652 [M + H] ⁺, respectively (calcd for 334.1649), implying eight degrees of unsaturation. The presence of hydroxyl and carbonyl groups were implied by IR absorption bands at 3329 and 1661 cm⁻¹, respectively. Detailed analysis of the ¹H and ¹³C NMR data suggested that 8 possessed the same planar structure as that of rigidiusculamide C (Li et al., 2009). A comparison of the NMR data (Supplementary Table 1) of 8 and 9 suggested that they differed only in the substituent at C-16. Detailed analysis of the HSQC and HMBC spectra allowed the assignment for all proton and carbon resonances of 9. The relative configuration of 9 was assigned by single-crystal X-ray diffraction as shown in Figure 4. To clarify the absolute configurations of 8 and 9, ECD calculations were performed by the time dependent density functional theory-predicted curve calculated at the quantum mechanical level. The calculated electronic circular dichroism (ECD) curve of (3S, 5S, 16S)-8 matched well with the 8 experimental ECD data, and the ECD curve of (3S, 5S, 16R)-9 is in good agreement with the experimental ECD data of 9. Therefore, the absolute configuration of compound 8 was determined as 3S, 5S, 16S, and the absolute configuration of compound 9 was 3S, 5S, 16R (Figure 5).

Compound 10 was obtained as a yellow oil. Its molecular formula C₁₈H₂₅NO₅ was established by the HRESIMS at m/z 336.1808 [M + H]⁺ (calcd for 336.1805), implying seven degrees of unsaturation. The presence of hydroxyl and carbonyl groups were implied by IR absorption bands at 3336 and 1644 cm⁻¹, respectively. The NMR data (Supplementary Table 1) of 10 revealed highly similar identical structural features to those found in 8, except that the C-4 ketone carbon was replaced by an oxygenated methine in 10. And 10 has the same planar structure as rigidiusculamide D (Li et al., 2009). The NOESY correlation of CH₃ with H-4 indicates that 3-OH and 4-OH are located on the same face, whereas H-5 and H-4 placed on the opposite face of the ring supported by the absence of a NOESY correlation between H-5 and H-4. To determine the absolute configuration, the simulated electronic circular dichroism (ECD) spectra of 10 were obtained from the calculation of Gaussian 16 based on time-dependent density functional theory. The absolute configurations of C-3, C-4, C-5, and C-16 in 10 were deduced as 3R, 4R, 5R, and 16S, respectively, by comparing calculated ECD spectra with the experimental ECD spectrum (Figure 5).

Compound 15 was isolated as a yellow powder. Its molecular formula C₁₅H₁₈O₄ was established by the HRESIMS at m/z 263.1277 [M + H]⁺ (calcd for 263.1278), implying seven degrees of unsaturation. The presence of hydroxyl and carbonyl groups were implied by IR absorption bands at 3387 and 1698 cm⁻¹, respectively. A comparison of NMR data (Supplementary Table 2) with those of nectriacid C (Cui et al., 2016) suggested that 15 possessed a closely similar structure as nectriacid C, except that there is no methoxy group (δ_H 3.74, δ_C 51.7) in 15. Its NMR data suggested that 15 belonged to the pentaene diacid derivative. Thus, the structure of compound 15 was assigned completely.

The remaining 13 known compounds from the F. decemcellulare F25 were identified as 12-epicitreoisocoumarinol (4) (Cui et al., 2016), eoisocoumarinol (5) (Cui et al., 2016), (−)-citreoisocoumarin (6) (Yamamura et al., 1991; Ola et al., 2013), trichophenol A (7) (Liu et al., 2020), rigidiusculamide B (11) (Li et al., 2009), fusaristatins A (12) (Shiono et al., 2007), enniatin H (13) (Nilanonta et al., 2003), enniatin I (14) (Nilanonta et al., 2003), nectriacid A (16) (Cui et al., 2016), nectriacid B (17) (Cui et al., 2016), 4-hydroxy-3,6-dimethyl-2 H-pyrane-2-one (18) (Hirota et al., 1999), macrocarpon C (19), (Ola et al., 2013), and α-linoleic acid (20) (Zeng et al., 2017) by comparison of their MS and NMR data with those reported in the literature.

Compounds 1–20 were assayed for their antifungal activities. The results showed that compounds 13, 14, and 17 exhibited inhibitory activities against the plant-pathogenic fungus C. musae ACCC31244 with MICs of 256, 64, and 128 μg/mL, respectively. The MIC of the positive control cycliheximide was 32 μg/mL.

The biosynthetic pathways of compounds 1 and 3 (Scheme 1) start with condensation catalyzed by nonreducing polyketide synthase (nrPKS) of two acetyl-coenzyme A molecules and six malonyl-CoA molecules resulting in the formation of the intermediate i (Liu and Begley, 2020). Intermediate i was catalyzed by TE domains or spontaneous C-O bond closure to form intermediate ii. The isocoumarin 1 was derived from intermediate ii through methylation and dehydration, while, intermediate ii will create 3 by methylation and ribosylation. The biosynthetic pathway of 2 (Scheme 1) starts with condensation catalyzed by a modular (nrPKS) of one acetyl-coenzyme A molecule and six malonyl-CoA molecules resulting in the formation of the intermediate iv. Intermediate iv was catalyzed by thioesterase (TE) domains or spontaneous C − O bond closure to form intermediate v. A series of reductive modifications for this intermediate led to intermediate vi. Intermediate vi then underwent a ribosylation reaction to afford 2.

As shown in Scheme 2, pyrrolidones originate from the cyclization of an amino acid and a polyketide (Royles, 1995; Li et al., 2009), leading to the formation of 11. The pyrrolidones 8 and 9 were likely to be biogenetically derived from 11 through prenylation, oxidation, cyclization, and hydroxylation. The derivative 10 was formed from 8 or 9 through one-step hydrogenation.