UV spectra were recorded using a UV-2450 visible spectrophotometer (Shimadzu, Japan). IR spectra (KBr disks) were obtained using a Shimadzu IRPrestige-21 instrument (Shimadzu, Japan). A JASCO J-815 spectropolarimeter was used to measure electronic circular dichroism (ECD) spectra (JASCO, Japan). Nuclear magnetic resonance (NMR) spectra were recorded on a Bruker Advance III 400 spectrometer (Bruker, Germany), with tetramethylsilane (TMS) as the internal standard. High-resolution electrospray ionization mass spectra (HRESIMS) were acquired using a Bruker impact II Q-TOF mass spectrometer (Bruker, Germany) and an Agilent 6520B mass spectrometer (Agilent, American). Analytical high-performance liquid chromatography (HPLC) was conducted with a Shimadzu LC-20 AD pump and a SPD-M20A UV detector (Shimadzu, Japan), using a YMC RP-C18 column (5 μm, 4.6 × 250 mm). Preparative high-performance liquid chromatography (HPLC) was performed on a Separation LC-UV system (Separation, China) using a YMC RP-C18 column (5 μm, 10 × 250 mm). The flow rate was set at 3.0 ml/min, and detection was carried out at wavelengths of 210 nm and 254 nm using a dual-channel UV detector. Column chromatography was performed using silica gel (100–200 and 200–300 mesh, Qingdao Marine Chemical Inc., China), MCI (50 μm, Mitsubishi, Japan), and Sephadex LH-20 (Pharmacia Fine Chemical Co., Ltd.).

The endophytic fungus was obtained from the traditional Chinese medicinal herb Coptis chinensis Franch. through the plate coating method, which were collected from Shizhu, Chongqing, China. The isolated strain was identified as Trichoderma citrinoviride based on morphological characteristics, and this identification was further supported by 18S rDNA and internal transcribed spacer (ITS) sequences, which showed 100% identity to the known Trichoderma citrinoviride (GenBank Accession KY750459.1). The basic characteristics of Trichoderma citrinoviride growth are as follows (Supplementary Figures S29, S30): When Trichoderma citrinoviride is inoculated onto a potato dextrose agar (PDA) medium at 28°C, the colony grows rapidly and exhibits aerial mycelium. During the first 1–3 days, the colony appears light green on the surface and yellowish-green on the reverse side, with a transparent mycelium. By the 4th day, extensive areas of white, fluffy colonies begin to appear. By the 5th day, the colonies become a cottony, olive-green mass. The mycelium is septate, and the conidiophores are characterized by a long main axis with shorter secondary branches, which are alternately arranged with unequal spacing and branching at acute or nearly right angles. Some of the terminal branches are flask-shaped, and the phialides bear smooth-walled spores. The conidia are colorless to green, ellipsoidal, and relatively small. The sequence information of this fungus is as follows:

GGATCACCTGATCCGAGGTCACATTTCAGAGTTTGGGGTGTTTTACGGCTGTGGCCGCGCCGCGCTCCCGGTGCGAGTGTGCAAACTACTGCGCAGGAGAGGCTGCGGCGAGACCGCCACTGTATTTCGGGGGCGGCCCGGTGAGGGGCCGATC CCCAACGCCGACCCCCCGGAGGGGTTCGAGGGTTGAAATGACG CTCGGACAGGCATGCCCGCCAGAATACTGGCGGGCGC AATGTGCGTTCAAAGATTCGATGATTCACTGAATTCTGCAATTCACATTACTTATCGCATTTCGCTGCGTTCTTCATCGATGCCAGAACCAAGAGATCCGTTGTTGAAAGTTTTGATTCATTTTCGAGACGCCCGCTAGGGTCGCCGAGAAAGGCTCAGAGCAAAAATAAAACAGAGCCGCGACGTAGGCCGCGACGGAGAGAAAAAAGAGTTTGAGTTGGTCCTCCGGCGGGCGCCATGGGATCCGGGGCTGCGACGCGCCCGGGGCAGAGAATCCCGCCGA GGCAACAGATTGGTAACGTTCACATTGGGTTTGGGAGTTGTAAACTCGGTAATGATCCCTCCGCTGGTTCACCAACGGAGACCTTGTT CCCTT.

The strain was cultured on potato dextrose agar (PDA) at 28°C for 7 days. Then, two pieces of the agar (about 1.0 cm³) were added to an Erlenmeyer flask (250 ml) containing 100 ml of potato dextrose liquid medium. The flask was then incubated on a rotary shaker at 28°C and 150 rpm for 5 days to prepare the seed culture. Solid fermentation was carried out in 400 Erlenmeyer flasks (1 L each). The flasks were sterilized by autoclaving prior to use and contained 200 g of rice, 1.0 g of glucose, 0.5 g of CuSO₄•5H₂0, and 200 ml of distilled water. All flasks were incubated at room temperature for 30 days. The solid cultures were extracted with ethyl acetate (EtOAc) three times at room temperature. The solvent was removed under reduced pressure to yield 1,000 g of crude extract.

The crude extract (1,000 g) was subjected to silica gel column chromatography (CC), and elution was performed using a mixture of petroleum ether (boiling point, 60–90°C) and EtOAc in ratios ranging from 15:1 to 0:1, resulting in increasing polarity. This process yielded eight fractions (Fr.1–8), as determined by thin layer chromatography (TLC) analysis. Fr. 5 (150 g) was further fractionated by repeated CC on silica gel, again eluting with petroleum ether and EtOAc (15:1 to 0:1), to produce seven subfractions (Fr.5.1–5.7). Fr.5.3 (45 g) was subjected to additional silica gel CC, eluting with the same solvent system to generate five fractions (Fr.5.3.1–5.3.5). Fr.5.3.2 (9.7 g) was then separated using MCI CC with gradient elution (MeOH/H₂O, 30:70 to 100:0), resulting in 154 subfractions (Fr.5.3.2.1–5.3.2.154). Specific subfractions were purified by semi-preparative HPLC as follows: Fr.5.3.2.12 (37.8 mg) was purified using ACN/H₂O (38:62, 3 ml/min) to obtain compound 3 (12.1 mg, t_R = 19.3 min; proportion of total extract, 0.00121%). Fr.5.3.2.26 (25.6 mg) yielded compound 2 (12.3 mg, t_R = 55.2 min; proportion of total extract, 0.00123%) after purification with MeOH/H₂O (63:37, 3 ml/min). Compound 6 (9.0 mg, t_R = 19.9 min; proportion of total extract, 0.0009%) was purified from Fr.5.3.2.28 (17.5 mg) using MeOH/H₂O (43:57, 3 ml/min). Fr.5.3.2.48 (25.7 mg) resulted in compound 5 (11.9 mg, t_R = 22.2 min; proportion of total extract, 0.00119%) after purification with ACN/H₂O (80:20, 3 ml/min). Separately, Fr.5.4 (7.6 g) was fractionated using Sephadex LH-20 with CH₂Cl₂-MeOH (1:1) to produce four subfractions (Fr.5.4.1–5.4.4). Fr.5.4.3 (3.1 g) was further separated by MCI CC through gradient elution with MeOH/H₂O (30:70 to 100:0) to yield 21 subfractions (Fr.5.4.3.1–5.4.3.21). Compound 1 (15.3 mg, t_R = 67.4 min; proportion of total extract, 0.00153%) was purified from Fr.5.4.3.12 (35.2 mg) using MeOH/H₂O (40:60, 3 ml/min). Finally, compound 4 (9.1 mg, t_R = 39.1 min; proportion of total extract, 0.00091%) was obtained from Fr.5.4.3.15 (31.2 mg) after purification with MeOH/H₂O (50:50, 3 ml/min).

Citrinsorbicillinol A (1): yellow powder; UV (MeOH) (log ε) λ_max 229 (3.06), 261 (3.61), 282 (2.85), 366 (3.38) nm; IR (KBr) ν_max/cm⁻¹ 3,752, 3,689, 2,372, 1701, 1,655, 1,544, 1,386, 1,155, 1,107, 1,026; HRESIMS m/z 237.0799 [M + H]⁺ (calcd for C₁₂H₁₃O₅⁺ 237.0763); ¹H NMR and ¹³C NMR data, see Table 1.

Citrinsorbicillinol B (2): yellow oil; [α]25 D -26.5 (c 0.24, ACN); ECD (ACN) λ_max (Δε) 218 (−4.17), 297 (+1.55), 344 (−1.95) nm; UV (MeOH) (log ε) λ_max 216 (3.63), 249 (2.81), 286 (3.55), 331 (2.96) nm; IR (KBr) ν_max/cm⁻¹ 3,751, 3,689, 3,651, 2,931, 2,376, 1739, 1718, 1,621, 1,560, 1,523, 1,388, 1,305, 1,232, 1,161, 1,111, 1,022; HRESIMS m/z 301.1415 [M + Na]⁺ (calcd for C₁₆H₂₂O₄Na⁺ 301.1416); ¹H NMR and ¹³C NMR data, see Table 1.

Citrinsorbicillinol C (3): yellow powder; [α]25 D -20.4 (c 0.22, ACN); ECD (ACN) λ_max (Δε) 217 (−2.54), 284 (+1.70), 324 (−1.11) nm; UV (MeOH) (log ε) λ_max 217 (3.45), 249 (2.36), 385 (3.36), 330 (2.98) nm; IR (KBr) ν_max/cm⁻¹ 3,398, 2,927, 2,378, 1712, 1,678, 1,662, 1,625, 1,537, 1,519, 1,487, 1,452, 1,384, 1,230, 1,166, 1,026; HRESIMS m/z 247.0946 [M + Na]⁺ (calcd for C₁₂H₁₆O₄Na⁺ 247.0955); ¹H NMR and ¹³C NMR data, see Table 1.

The initial conformational analysis of compounds 1–3 was carried out using the SPARTAN’ 14 program, which employed the Monte Carlo searching algorithm with the MMFF94 molecular mechanics force field (Halgren, 1999). This resulted in a set of relatively stable conformations within an energy range of 3 kcal/mol above the global minimum. The conformers with minimum energy from the force field were then further optimized using density functional theory (DFT) at the B3LYP/6-31G(d) level, as implemented in the Gaussian 16 program. To ensure the reliability of the optimized conformers, harmonic vibrational frequency calculations were performed, confirming the absence of imaginary frequencies. The conformers accounting for over 99% of the population were then subjected to subsequent calculations.

The predominant conformers of compounds 2 and 3 were subjected to theoretical ECD calculations using time-dependent density functional theory (TDDFT) at the M062X/def2SVP level in acetonitrile, employing the conductor-like polarizable continuum model (CPCM) for the solvent model. For each conformer, 30 excited states were calculated. The energies, oscillator strengths, and rotational strengths of each conformer were computed using the Gaussian 16 program. The ECD spectra for each conformer were approximated using Gaussian distributions. Finally, the overall ECD spectrum was determined by summing the spectra of individual conformers, weighted according to their Boltzmann populations, using the SpecDis v1.71 program (Bruhn et al., 2013).

The NMR shielding constants were calculated using the gauge-independent atomic orbital (GIAO) method at the mPW1PW91-SCRF/6–31 + G(2d,p) level with the PCM solvent model in dimethyl sulfoxide (DMSO). The shielding constants obtained were converted into chemical shifts by referencing TMS at 0 ppm (δ_cal = σ_TMS − σ_cal), where σTMS was the shielding constant of TMS calculated at the same level (Willoughby et al., 2014). The DP4+ probabilities for each possible candidate were calculated using the Excel spreadsheet provided by Grimblat et al. (2015). For each possible candidate, we calculated the parameters a and b of the linear regression δ_cal = aδ_exp + b. Additionally, we computed the correlation coefficient, R², the mean absolute error (MAE), which is defined as Σn|δ_cal − δ_exp|/n, and the corrected mean absolute error (CMAE), which is defined as Σn|δ_corr − δ_exp|/n, where δ_corr = (δ_cal − b)/a.

The antioxidant activity of compounds 1–6 was assessed based on their scavenging activity against the stable 1,1-diphenyl-2-picrylhydrazyl (DPPH) radical (Zhao et al., 2010).

Samples in methanol, with concentrations ranging from 2 to 200 μg/ml, were mixed with freshly prepared 0.1 mM DPPH in ethanol. Absorbance at 517 nm was measured after 30 min at room temperature. Anhydrous ethanol served as the blank control, and ascorbic acid was used as the positive control. The radical-scavenging activity was expressed as the percentage inhibition, calculated using the following formula: Inhibition (%) = [A_control − (_Asample − A_blank)]/A_control × 100%. Three parallel experiments were performed.

The inhibition of lipopolysaccharide-induced nitric oxide (NO) production in RAW 264.7 mouse macrophage cells was evaluated as follows (Kumar et al., 2024; Yin et al., 2021b): The cells were evaluated using 96-well plates (1 × 10⁵ cells/well) and allowed to adhere for 2 h at 37°C in 5% CO₂ in air. Then, the cells were treated with 1 μg/ml lipopolysaccharide (LPS) for 24 h, with or without the test compound (5 μg/ml). DMSO was used as the solvent. All compounds were tested at a final concentration of 0.2% (v/v) in the cell-culture supernatant. The NO production was determined by measuring the accumulation of nitrite in the culture supernatant using the Griess reagent. The absorbance of the mixture was read at 540 nm using a microplate reader. L-NAME (N^ω-nitro-L-arginine methyl ester) was used as the positive control. Data were presented based on three parallel experiments.

The data were expressed as mean ± SD. Error bars represented the three independent experiments. p-values were calculated using ordinary one-way ANOVA.

Citrinsorbicillinol A (1) was isolated as a yellow amorphous powder with a purity of over 95%, as determined by the HPLC analysis (Supplementary Figure S26). The molecule formula C₁₂H₁₂O₅ was established by HRESIMS (Supplementary Figure S5) at m/z 237.0799 [M + H]⁺ (calcd. for C₁₂H₁₃O₅, 237.0763), corresponding to seven degrees of unsaturation. The ¹H NMR data (Supplementary Figure S1 and Table 1) clearly exhibited the presence of 11 protons, including 2 singlet methyl groups at δ_H 1.83 and 3.92, a conjugated diene moiety with 4 olefinic protons at δ_H 6.25 (d, J = 15.0 Hz), 6.72 (d, J = 15.0 Hz), 7.11 (dd, J = 15.0, 11.4 Hz), and 7.25 (dd, J = 15.0, 11.4 Hz), and a singlet aromatic proton at δ_H 6.80. The ¹³C NMR data (Supplementary Figure S2 and Table 1) revealed a total of 12 resonances. These resonances were assigned with the help of HSQC data (Supplementary Figure S3) to two methyl groups at δ_C 9.4 and 57.3, five sp² methine groups at δ_C 99.5, 128.2, 129.6, 131.9, and 141.5, and five non-protonated carbons at 102.8, 156.0, 163.6, 165.9, and 168.0.

The structure of compound 1 was further elucidated through a comprehensive analysis of the 2D NMR data (Figure 1A). The key HMBC correlations (Supplementary Figure S4) from H-9 to C-10 and C-7; from H-8 to C-6, C-7, C-9, and C-10; and from H-6 to C-7 and C-8, along with the coupling constant of the related protons, support the construction of an (E)-penta-2,4-dienoic acid moiety. Moreover, the HMBC correlations from H₃-2-Me to C-1, C-2, and C-3; from H₃-3-OMe to C-3; from H-4 to C-2, C-3, C-5, and C-6; and from H-6 to C-4 and C-5, along with the chemical shift of C-5 at δ_C 156.0, suggested that the (E)-3-methoxy-2-methylpenta-2,4-dienoic acid moiety was connected to the (E)-penta-2,4-dienoic acid moiety via the oxygenated olefinic carbon at C-5. Considering the remaining unsaturation and the chemical shifts of C-1 at δ_C 163.6 and C-5 at δ_C 156.0, a lactone between C-1 and C-5 was deduced to establish the unsaturated six-membered lactone ring (Figure 1A).

To further verify the structure, the computer-assisted structure elucidation (CASE) algorithm was employed. Specifically, ACD/Structure Elucidator (ACD/SE), an advanced CASE expert system, was utilized to automatically and efficiently determine the most probable structure based on chemical rules and common knowledge. In recent years, CASE analysis has increasingly been applied to structural revisions and identifications (Fan et al., 2022; Chen et al., 2024). The molecular formula and NMR data of compound 1, which were analyzed using ACD/SE, led to the automatic generation of a molecular connectivity diagram (MCD), as shown in Figure 2. In the MCD (Figure 2), atom hybridization states are color-coded: sp3 is represented in blue, sp2 in violet, and undefined hybridization in black. The connecting lines are also color-coded to indicate different types of connectivity: green lines represent HMBC connectivity, and violet lines indicate non-standard connectivity for ⁿJ_HH and ⁿJ_CH (n > 3).

For each candidate structure, ¹H and ¹³C chemical shifts were predicted using three empirical methods provided by ACD/SE: HOSE (d_A), the incremental approach (d_I), and neural networks (d_N). The structures that did not meet the threshold criteria of an average C deviation greater than 4 ppm or a maximum C deviation exceeding 20 ppm were discarded. The remaining structures were ranked based on the average deviation (d_A) of the ¹³C chemical shifts between experimental and calculated values. The accuracy of the ¹³C chemical shift predictions was indicated by color-coded circles: green for the deviations ≤3 ppm and yellow for the deviations between 3 and 15 ppm. Ultimately, seven candidate structures were generated using ACD/SE from a total of 41,537 structures, and after removing the duplicates, six distinct structures remained, as shown in Figure 3. Among these, structure ^#1, listed at the top of the output file from the ACD-Lab, exhibited the highest match factor (MF) value (d_A = 1.605, d_N = 1.824, and d_I = 1.697), which further confirmed its structure.

Furthermore, the ¹³C NMR calculation (Figure 1B) provided strong evidence for the assignment of structure ^#1. This analysis was performed within the GIAO framework at the MPW1PW91/6–311 + G (2d, p) level, and geometries were optimized at the B3LYP/6-31G (d) level in chloroform. The correlation coefficient (R²) obtained from linear regression analysis between the calculated and experimental ¹³C NMR data for structure ^#1 was 0.9990, and the mean absolute error (MAE) was 1.53 ppm. Thus, the structure of compound 1 was constructed (Figure 4).

Citrinsorbicillinol B (2) was obtained as a yellow oil with a purity of over 95%, as determined by the HPLC analysis (Supplementary Figure S27). The molecular formula was determined as C₁₆H₂₂O₄ based on a positive HRESIMS (Supplementary Figure S13) peak at m/z 301.1415 [M + Na]⁺ (calcd. for C₁₆H₂₂O₄Na, 301.1416), which required six degrees of unsaturation. The ¹H NMR data (Supplementary Figure S8 and Table 1) for compound 2 showed signals for two singlet methyl groups at δ_H 2.13 and 2.21, one doublet methyl group at δ_H 1.66 (d, J = 4.9 Hz), two olefinic protons at δ_H 5.50 and 5.45, one aromatic proton at δ_H 7.36 (s), and one typical hydroxyl proton at δ_H 12.73 (s). The ¹³C NMR and HSQC data (Supplementary Figures S9, S10 and Table 1) displayed 16 carbon resonances, accounting for three methyl groups at δ_C 7.4, 15.6, and 17.9, three sp³ methylene groups at δ_C 28.7, 36.4, and 44.3, three sp² methine groups at δ_C 125.6, 129.2, and 130.6, one sp³ methine group at δ_C 67.4, and six non-protonated carbons at δ_C 110.5, 113.3, 114.9, 159.2, 161.6, and 204.4. In the 2D NMR spectra (Figure 5), the ¹H-¹H COSY correlations (Supplementary Figure S12) of H₃-14/H-13/H-12/H₂-11/H₂-10/H-9/H₂-8, along with the key HMBC correlations (Supplementary Figure S11) from H₃-14 to C-12 and C-13; from H₂-10 to C-11, C-12, C-8, and C-9; and from H₂-8 to C-7, C-9, and C-10, supported the construction of an (E)-3-hydroxynon-6-en-1-one moiety. Then, a 2,4-dimethylbenzene-1,3-diol moiety was confirmed by the key HMBC correlations from H₃-3-Me to C-2, C-3, and C-4; from H₃-5-Me to C-4, C-5, and C-6; from H-2 to C-1 and C-6; and from OH-6 to C-1, C-5, and C-6. Moreover, the key HMBC correlations from H-2 to C-7 indicated that the (E)-3-hydroxynon-6-en-1-one moiety at C-7 was connected to the 2,4-dimethylbenzene-1,3-diol moiety at C-1 by a C-C single bond. Hence, the planar structure of compound 2 was established (Figure 4).

The absolute configuration of compound 2 was determined using TDDFT calculations of the ECD spectrum with the Gaussian 16 program. The ECD spectrum calculated for the 9R configuration was consistent with the experimental ECD spectrum (Figure 6). Hence, the absolute configuration of compound 2 was assigned as 9R (Figure 4).

Citrinsorbicillinol C (3) was obtained as a yellow powder with a purity of over 95%, as determined by the HPLC analysis (Supplementary Figure S28). The molecular formula of compound 3 was established as C₁₂H₁₆O₄ based on its HREIMS data (Supplementary Figure S21), indicating that it was the unit of C₄H₆ less than compound 2. Detailed analysis of its NMR spectroscopic (Supplementary Figures S17, S18 and Figure 4) features implied that its chemical structure was very similar to that of compound 2. The main difference was that compound 3 was missing two aliphatic methylene groups and two olefinic methine groups compared to compound 2. Further analysis of the 2D NMR data (Supplementary Figures S19, S20 and Figure 5) showed that the side chain of compound 3 lacked an n-butene unit compared to that of compound 2. This was supported by the key HMBC correlations from H₃-3-Me to C-2, C-3, and C-4; from H₃-5-Me to C-4, C-5, and C-6; from H₃-10 to C-8 and C-9; from H₂-8 to C-1 and C-7; from H-2 to C-1 and C-7; and from OH-6 to C-1, C-5, and C-6. Finally, the absolute configuration was established as 9R by comparing the experimental ECD spectrum with the calculated one (Figure 7).

In addition, three known compounds were finally characterized as trichosorbicillin G (4) (Zhang et al., 2019), dibutyl phthalate (5) (Li et al., 2009), and 3-(4-Methoxyphenyl) propanoic acid (6) (Zou et al., 2021), by comparing their spectroscopic data with the literature.

The isolated compounds 1–6 were tested for their antioxidant properties using the DPPH assay (Table 2). As a result, compounds 1–4 exhibited radical scavenging activity, with IC₅₀ values ranging from 27.8 to 89.6 μM. Notably, compounds 2 and 3, with IC₅₀ values of 27.8 and 31.2 μM, respectively, exhibited significant activity comparable to that of ascorbic acid (IC₅₀ = 39.4 μM). Structurally, compounds 2 and 3 each possessed two identical phenolic hydroxyl groups and one similar alcohol hydroxyl group. Compound 4 had a similar structure but with different substituents on the benzene ring and a different alcohol hydroxyl group. In contrast, compound 1 contained a carboxylic acid group with a long conjugated system. These results indicated that the number and position of phenolic and alcohol hydroxyl groups are crucial for antioxidant activity, with the substituents on the benzene ring also playing a significant role. Furthermore, the presence of a long conjugated carboxylic acid group could be a potent fragment contributing to antioxidant activity.

In addition, all compounds were tested for their inhibitory activities toward nitric oxide (NO) production in lipopolysaccharide (LPS)-induced RAW 264.7 macrophages (Table 2). None of the compounds showed significant cytotoxicity at the concentrations of 50 μM for the inhibition of the NO production. Among them, only compound 1 showed moderate inhibitory activity, with an IC₅₀ value of 52.7 μM.