Aflatoxin B₁ was purchased from Pribolab (Qingdao, China). Chromatographic-grade methanol and acetone were obtained from Tianjin Saifu Rui Technology Company (Tianjin, China). Analytical-grade methanol, acetone, and DMSO were obtained from Chron Chemicals (Chengdu, China). For NMR analysis, all deuterium reagents were purchased from Sigma (St. Louis, MO, USA).

Standard solutions of AFB₁ were placed in a 2-ml sealed centrifugal tube, prepared in methanol–DMSO (9:1 v/v), and fully dissolved in methanol using an ultrasound device from Beijing Tianlin Hengtai Technology Company (Beijing, China), and then, it was submitted to be degraded.

To investigate the degradation of AFB₁, a UV lamp (20 W, 72 μws/cm², GGZ250-1, Shanghai Jiming Special Lighting Appliance Factory) at 365 nm wavelength was used to perform the irradiation experiments. 18 mg of pure AFB₁ was added to acetone solvent, and 10 mg of pure AFB₁ was added to methanol solvent, and both of them were placed in a sealed centrifugal tube and illuminated at room temperature for 45 h (Liu et al., 2010).

The degraded products were analyzed and isolated by semipreparative HPLC on SEP LC-52 with an MWD UV detector (Separation (Beijing) Technology Co. Ltd., Beijing, China) using a 250 mm × 10 mm i. d., 5 μm, ODS-A column (YMC, Kyoto, Japan). The mixture in methanol was purified by semipreparative HPLC (55–60% CH₃OH in H₂O, v/v, 2 ml/min, 30 min) and yielded 4 (0.6 mg, t _R = 18.6 min), 3 (0.5 mg, t _R = 22.3 min), 1 (0.5 mg, t _R = 23.8 min), and 2 (0.4 mg, t _R = 27.0 min), respectively. The mixture in acetone was isolated by semipreparative HPLC (40% CH₃OH in H₂O, v/v, 2 ml/min, 3 min; 40–100% CH₃OH in H₂O, v/v, 2 ml/min, 20 min) and yielded 5, 6 (7.0 mg, t _R = 18.7 min), and 7 (1.5 mg, t _R = 21.2 min).

The degraded products were identified by NMR experiments. Compounds were analyzed by UPLC-Q-TOF-MS/MS in positive ion mode. 1D and 2D-NMR spectra were acquired using solvent signals (CD₃OD: δ _H 3.31/δ _C 49.9; C₃D₆O: δ _H 2.05/δ _C 49.9; Pyridine-d ₅: δ _H 8.74, 7.58, 7.22/δ _C 150.4, 135.9, and 123.9) on a Bruker 600 spectrometer (¹H: 600 MHz) and a Bruker Avance III 500 spectrometer (¹H: 500 MHz; ¹³C: 125 MHz) (Bruker, Rheinstetten, Germany).

AFB₁ and degraded products were analyzed using a UPLC-Q-TOF-MS/MS system (Waters, United States). Chromatographic analysis was carried out with a Waters Acquity UPLC-PDA system equipped with an analytical reverse-phase C-18 column (2.1 × 100 mm, 1.7 μm, Acquity BEH, Waters, United States) with an absorbance range of 200–400 nm. The column temperature was maintained at 40°C. 0.1% formic acid in water (A) and 0.1% formic acid in acetonitrile (B) were used as the mobile phase. The gradient conditions were as follows: 0–10 min, 10 %–60% B; 10–12.5 min, 60 %–95% B; and 12.6–15 min, 10% B. The flow rate from the UPLC system into the ESI-Q-TOF-MS detector was 0.3 ml/min. The auto-injected volume was 3 μl. Time-of-flight MS detection was performed with a Waters SYNAPT G2 HDMS (Waters Corp., Manchester, United Kingdom) TOF mass spectrometer combined with an ESI source in the positive ion scan mode. The desolvation temperature was set at 400°C with desolvation gas flow at 600 L/h, and the source temperature was 100°C. The lock mass in all analyses was leucine–enkephalin ([M + H]⁺ = 556.2771), used at a concentration of 0.5 g/ml and infused at a flow rate of 10 L/min. Raw data were acquired using the centroid mode, and the mass range was set from m/z 50 to 1200. The capillary voltage was set at 3.0 kV with 40 and 4.0 V of the sample and extraction cone voltage. The collision energy was set as 20 eV for low-energy scan and a ramp from 20 to 30 eV for high-energy scan. The instrument was controlled by MassLynx 4.1 software.

All the degraded products and AFB₁ were tested for their cytotoxicity against human normal hepatocytes LO-2 and cancer cell lines Hep-G2 and MCF-7. Cells were incubated in a DMEM high glucose medium (Gibco, USA), added with 10% fetal bovine serum (Gibco, United States) and cultured in a 5% CO₂ incubator at 37°C. The cytotoxicity tests were performed using the MTS (Promega, United States) (Ahmed et al., 2019).

The UPLC-Q-TOF-MS/MS BPI of AFB₁ and its degraded products in methanol–H₂O and acetone–H₂O solvents are shown in Figures 1 and 2. The retention time and molecular weight of AFB₁ were 6.09 min and m/z 313 ([M+1]), respectively. A series of degraded products with different retention times (RTs) and molecular weights are shown in Table 1. Some ion peaks as (stereo) isomers possessed the same molecular weights (such as m/z 345) but with different RTs.

Different free radicals such as reactive hydroxyl (OH^•), hydrated electrons (eaq⁻), hydrogen atoms (H^•), and methoxy species (OCH₃ ^•) were produced when methanol–H₂O and acetone–H₂O solvents were irradiated under UV (White, 2001; Azrague et al., 2005). These free radicals could attack the AFB₁ structure to form different degraded products. The double bond C₈-C₉ in AFB₁ was broken easily by these free radicals via addition reactions. Ten and seven main degraded products in methanol–H₂O and acetone–H₂O solvents were characterized based on molecular weights and fragment ions of compounds (Supplementary Tables S1, S2). The other degraded product fragmentation rules are provided in supporting information, considering a similar fragmentation pathway with AFB₁ (Figure 3 and Supplementary Figures S2–S9).

Four ions as (stereo) isomers (m/z 345, C₁₈H₁₆O₇) appeared at t _R = 4.70, 5.40, 5.84 and 5.99 min in methanol–H₂O solvent with 32 Da (CH₄O) more than that of AFB₁ (Supplementary Figure S2). After a neutral loss of CH₃OH from ion (m/z 345), the fragmentation pathways of these four ions were nearly the same as those of AFB₁. It is suggested that these four degraded compounds might be addition products of CH₃OH with AFB₁ at C-8/C-9. The possible fragmentation pathways of these four (stereo) isomers are depicted in Supplementary Figure S2.

The molecular formula of the ion at m/z 361 ([M+1], t _R = 4.47 and 4.82 min) was determined to be C₁₈H₁₆O₈ based on HR-ESI-MS with 16 Da (O) more than that of degraded products (m/z 345) (Supplementary Figure S3), suggesting one more oxygen atom connected on C₈ or C₉. Both of them were suggested to be the addition products from free radical hydrogen atoms (OH^•) and methoxy species (OCH₃ ^•) with C₈/C₉ or C₉/C₈ of AFB₁ under UV irradiation. The possible fragmentation pathway of the ion (m/z 361) is shown in Supplementary Figure S3.

Three ions at m/z 359 ([M+1], t _R = 4.94, 7.08 and 7.30 min) gave the molecular formula as C₁₉H₁₈O₇ based on HR-ESI-MS. The neutral loss of -CO, -CH₃OH, and -C₂H₂ was observed in the MS/MS profiles (Supplementary Figure S4). The possible structure and fragmentation pathway of these three ions is suggested in Supplementary Figure S4.

The molecular formula of the degraded product at m/z 391 ([M+1], t _R = 6.41 min) was determined to be C₂₀H₂₂O₈ based on HR-ESI-MS (Supplementary Figure S5). Sequential losses of two -CH₃OH (391→359→327), one -CH₂O (327→297), and one -CH₂ (297→283) implied that four methoxyls might be present in degraded products. Two methoxyls might be connected on C-8/C-9 and the keto-carboxyl group might be transformed to another methoxyl through reduction and addition reactions, and the remaining -OMe was anchored on the aromatic ring. The possible fragmentation pathway of these three ions is suggested in Supplementary Figure S5.

The molecular formula of ions at m/z 331 ([M+1, C₁₇H₁₄O₇] in acetone–H₂O at 4.06 and 4.18 min) possessed 18 Da (H₂O) more than that of AFB₁, which indicated these two degraded products were (stereo) isomers (Supplementary Figure S6). The fragmentation pathways of these two ions were nearly the same as those of AFB₁ after the loss of a molecule of H₂O, which implied that two degraded products were the adducts of H₂O with the double bond C-8/C-9. Though the molecular formulas and fragmentation pathways of these two degraded products were the same, the retention time and abundance of fragment ions were different. A higher abundance of ion at m/z 313 (t _R = 4.18 min) was observed than the other (t _R = 4.06 min). This suggested that the position of OH on the furan ring was different in two degraded products. The possible fragmentation pathway of two ions is suggested in Supplementary Figure S6.

The molecular formula of the ion at m/z 347 ([M+1], t _R = 3.48 min) was determined to be C₁₈H₁₄O₈ based on HR-ESI-MS with 16 Da (O) more than that of degraded products (m/z 331), indicating two hydroxyl groups connected on C₈ and C₉, respectively (Supplementary Figure S7). The possible fragmentation pathway is suggested in Supplementary Figure S7.

The molecular formula of the ion at m/z 371 ([M+1], t _R = 4.34 and 5.63 min) was determined to be C₂₀H₁₈O₇ based on HR-ESI-MS with 58 Da (CH₃COCH₃) more than that of AFB₁ (m/z 313) (Supplementary Figure S8), which implied that one molecule of acetone attacked on C-8 or C-9 under UV irradiation. The possible fragmentation pathway of them is suggested in Supplementary Figure S8.

The molecular formulas of ions at m/z 401 ([M+1], t _R = 4.86 and 6.54 min) were determined to be C₂₁H₂₀O₈ based on HR-ESI-MS. The loss of 32 Da from m/z 401 to m/z 369 and the loss of 58 Da from m/z 401 to m/z 343 suggested that a methoxyl and acetone were connected on C-8/C-9 or C-9/C-8 (Supplementary Figure S9). The possible fragmentation pathway of these two ions is suggested in Supplementary Figure S9.

Though seventeen degraded products were characterized by molecular formula and fragment ions, the planar structures and configurations of some degraded products could not be determined only based on UPLC-Q-TOF-MS/MS analysis. Thus, further purification and NMR experiments are needed to elucidate their structures and stereochemistry.

Seven main degraded products with limited amounts were purified by HPLC and then elucidated by NMR spectra (Figure 4). Compounds 1–3 were isolated as the photochemical adducts of 6-methoxydifurocoumarone, which were analyzed based on the ¹H-NMR spectrum (Waiss and Wiley, 1969). In this study, the structures of these three degraded products were elucidated in detail by analyzing ¹H, ¹³C, and 2D-NMR spectra (Figure 5). The ¹H-NMR data of 1–3 and ¹³C-NMR data of 1 and 3 are shown in Table 2 and Table 3. The relative configurations of 1 and 3 were determined by NOESY correlations (Figure 5). Compound 4 was a new degraded product isolated from methanol solution. The molecular formula of 4 was determined to be C₁₈H₁₇O₈ on the basis of HR-ESI-MS with 16 more daltons than that of 1, implying that an additional hydroxyl group was present in 4, which was supported by the NMR spectra (Table 2 and Table 3). The ¹H–¹H COSY and HMBC correlations confirmed that the additional hydroxyl group was connected on C-9 (Figure 5). The NOESY correlations determined the relative configuration of 8-OMe and 9-OH to be β and α configuration, respectively (Figure 5).

Compounds 5 and 6 were obtained as an inseparable mixture through HPLC with various stationary and mobile phases, whereas well-resolved NMR spectra determined the structures of 5 and 6 as isomers. The ¹H and ¹³C spectra data of 5 were reported, and 6 was a new degraded product reported for the first time (Cox and Cole, 1977; Liu, Jin, Tao, Shan, et al., 2010; Wang et al., 2011; Wang et al., 2012; Stanley et al., 2020). The molecular formula of 5 and 6 was determined to be C₁₇H₁₄O₇ on the basis of HR-ESI-MS, with 18 more daltons than that of AFB₁, implying that 5 and 6 might be transformed from AFB₁ through an addition reaction with H₂O on the double bond (C-8/C-9). The planar and relative configurations of 5 and 6 were established based on 2D-NMR data (Figure 5). Compound 7 was a new degraded product isolated from acetone solvent. The molecular formula of 7 was established to be C₂₀H₁₉O₇ based on HR-ESI-MS. In the ¹H NMR spectrum, an additional methyl (δ _H = 2.12 ppm) and an additional methylene unit (δ _H = 2.72 ppm) were observed compared with that of AFB₁, which indicated that one molecule of acetone might be connected on C-8 or C-9. The ¹H–¹H COSY and HMBC correlations confirmed that the acetonyl group was connected with C-9 (Figure 5). The NOESY correlations from H-9a (δ _H = 3.93 ppm) to H-1' (δ _H = 2.72 ppm) determined the acetonyl group to be α-configuration (Figure 5). Considering that the stereochemistry of C-6a and C-9a were not changed in the photocatalytic reaction, the absolute configurations of (1–7) are shown in Figure 4.

According to the structural features of AFB₁ and degraded products (1–7), the possible photocatalytic reactions were suggested: 1) addition reactions happened between MeOH, H₂O, or acetone with AFB₁ under UV irradiation to produce compounds such as 1–3 and 5–7; 2) compound 4 might be originated from the oxygen free radical attacking the double bond (C-8/C-9) to form an epoxide, which was further attacked by OMe^• or OH^• (Figure 6) (Waiss and Wiley, 1969; Iyer et al., 1994). The photocatalytic mechanism was suggested: MeOH, H₂O, or CH₃COCH₃ formed potential free radicals (H^•, OH^•, OMe^•, or CH₃COCH₂ ^•) under UV irradiation. Then, H^• attacked on the double bond (C-8 or C-9) leading to form carbon-free radicals, which was then coupled with OH^•, OMe^•, or CH₃COCH₂ ^• to shape degraded products 1–3 and 5–7 (Waiss and Wiley, 1969; Iyer et al., 1994; Jamil et al., 2017; Mao et al., 2019). In addition, O₂ in the air under UV irradiation could form O₂ ^•-, which could attack on the double bond C-8/C-9 to produce 8,9-epoxide-AFB₁. Addition reactions then happened fast by the highly unstable intermediate 8,9-epoxide-AFB₁ with OMe^• to form the degraded products of 4 (Figure 6) (Waiss and Wiley, 1969; Iyer et al., 1994; Jamil et al., 2017).

From the structural features of degraded products (1–7), an interesting phenomenon was also observed that the group of C-9 (in 3, 4, 6, and 7) was α-configuration, whereas the group of C-8 in 1 and 2 was α- or β-configuration. This demonstrated that steric hindrance (from right part of AFB₁ structure) might exist and prevent different groups (OH^•, OMe^•, or CH₃COCH₂ ^•) attacking C-9 from the positive face (β-position), whereas C-8 could be attacked from two sides (α- or β-configuration) without steric hindrance. The crystal structure of AFB₁ (Cheung and Sim, 1964; van Soest and Peerdeman, 1970) revealed that the right part of the AFB₁ structure was indeed closer to C-9 than C-8 in space, which might preclude different groups to attack C-9 from the positive face (β-position) due to spatial hindrance. The photocatalytic reactions are depicted in Figure 7.

The cytotoxicity of AFB₁ and seven degraded products (1–7) was evaluated against human normal hepatocytes LO-2 and cancer cell lines Hep-G2 and MCF-7 using the MTS method, with cis-platinum as the positive control; the results are shown in Table 4. AFB₁ displayed stronger cytotoxicity to three cell lines than the degraded products, further supporting that the double bond (C-8/C-9) in the furan ring was the key toxic group, and the toxicity was markedly reduced after the double bond was broken.