Sophora flavescens (S. flavescens) is well-known species of Sophora, and its roots, which are called Kushen in traditional Chinese medicine, are commonly used to treat cancer, skin diseases, and diarrhea [1,2]. To date, a large number of phytochemical studies on S. flavescens have been reported and have demonstrated that flavonoids and alkaloids are the major constituents. Among them, approximately 150 flavonoids have been identified, including flavanones, flavanonols, flavonols, chalcones, isoflavones, and biflavonoids [2,3]. Characteristically, the flavonoids usually contain isoprenyl and lavandulyl groups. Pharmacological studies on these reported flavonoids revealed various biological activities, including antidiabetic, anti-inflammatory, antioxidative, antitumor, and antimicrobial activities [2,4–7]. Recently, several unique skeletons derived from flavonoids, such as sophorapterocarside A (a pterocarpine derivative with a unique pterocarpan-glucose-long chain aliphatic acid structure) [8], sophopterocarpan A (a pterocarpine derivative with a benzotetrahydrofuran-fused bicyclo[3.3.1]nonane) [9], and biflavonoids (formed through rare C_5′-C_β or C₃C_β bonds between dihydroflavones and chalcones) [10] were reported. Based on these results, S. flavescens possesses structurally intriguing flavonoids and further research on its phytochemistry and biological activity is worthwhile.

During our efforts, three high hybrid flavonoids (neosophoflavonoids A–C, 1, 2a, and 2b) were isolated from the roots of S. flavescens (Fig. 1). Compound 1 was characterized by a fantastic heptacyclic 6/6/6/6/6/6/5 skeleton, and compounds 2a/2b were characterized by hexacyclic 6/6/6/6/6/6 skeletons. All compounds were evaluated for their ability to inhibit protein tyrosine phosphatase-1B (PTP1B) and lipid accumulation. Herein, the isolation, structural elucidation, and biological activities of the compounds are described.

Compound 1 was isolated as a pale-yellow powder with ultraviolet spectroscopy (UV) absorption maxima at 204 and 284 nm. Its molecular formula was established as C₃₅H₃₄O₉ based on the [M − H]⁻ ion peak at m/z 597.2134 (calcd. for C₃₅H₃₃O₉, 597.2130) in the high resolution electrospray ionization mass spectrometry (HRESIMS). The infrared spectroscopy (IR) absorptions of 1 suggested the presence of hydroxyl (3207 cm⁻¹), carbonyl (1791 cm⁻¹), and aromatic rings (1607, 1507, and 1465 cm⁻¹). The ¹H NMR spectrum (Table S3 in Supporting information) presented signals for two ABX aromatic ring systems [δ_H 7.26 (1H, d, J = 8.5 Hz, H-6′), 6.48 (1H, dd, J = 8.5, 2.5 Hz, H-5′), and 6.34 (1H, d, J = 2.5 Hz, H-3′); 6.89 (1H, d, J = 8.5 Hz, H-6′′), 6.46 (1H, dd, J = 8.5, 2.5 Hz, H-5′′), and 6.44 (1H, d, J = 2.5 Hz, H-3′′)], one penta-substituted aromatic ring [δ_H 5.93 (1H, s, H-6)], one typical lavandulyl group [δ_H 2.49 (2H, m, H-1a), 2.29 (1H, m, H-2a), 1.88 (2H, m, H-3a), 4.86 (1H, m, H-4a), 1.32 (3H, brs, H-6a), 1.56 (3H, brs, H-7a), 4.49 (2H, m, H-9a), and 1.64 (3H, brs, H-10a)], one methylene [δ_H 3.56 (1H, dd, J = 18.0, 9.0 Hz, H-8′′a) and 2.56 (1H, brd, J = 18.0 Hz, H-8′′b)], two methines [δ_H 5.35 (1H, s, H-2) and 4.06 (1H, d, J = 9.0 Hz, H-7′′)], and one methoxyl [δ_H 3.74 (3H, s, 5-OCH₃)]. According to the ¹³C NMR (Table S3) and heteronuclear single quantum coherence (HSQC) spectra, 35 carbon signals were observed including 18 benzene carbons, 10 lavandulyl carbons, one carbonyl carbon (δ_C 176.6), two quaternary carbons (δ_C 78.9 and 98.4), two methine carbons (δ_C 35.8 and 73.1), one methylene carbon (δ_C 38.5), and one methoxyl (δ_C 56.0). The above information suggested that compound 1 possessed the scaffold mode of 15 (flavone) + 10 (lavandulyl) + 9 (phenylpropanoid). In the heteronuclear multiple-bond correlation spectroscopy (HMBC) spectrum (Fig. 2), the correlations of H-1a to C-7/C-9 confirmed that lavandulyl unit was attached at C-8. Additionally, the correlations from H-2 to C-2′/C-6′/C-4/C-9, H-3′ to C-1′, H-6′ to C-4′, and H-6 to C-10/C-8/C-4 suggested the presence of a functionalized flavone moiety. Furthermore, the phenylpropanoid moiety was deduced by the correlations of H-6′′ to C-2′′/C-4′′, H-7′′ to C-2′′/C-6′′/C-9′′, and H-8′′ to C-1′′. Then, these key correlations from H-7′′ to C-4 and H-8′′ to C-3 revealed that the flavone moiety and phenylpropanoid moiety were linked through C-3/C-7′′. The HRESIMS data indicated that compound 1 contained 19 degrees of unsaturation. However, three degrees of unsaturation remained, not including the 16 degrees of unsaturation of flavone, phenylpropanoid, and lavandulyl unit. Herein, the unusual chemical shifts of C-4 (δ_C 98.4) and C-3 (δ_C 78.9) suggested that these two quaternary carbons linked two oxygen atoms and one oxygen atom, respectively. Thus, the challenge was to determine their linkage. According to the chemical shifts of -C=O (δ_C 176.6), a five membered lactone ring was present (the chemical shift of the carbonyl carbon of the six membered lactone was less than 170 ppm) [11,12]. Therefore, the linkages of C-2′′-O-C-4, C-2′-O-C-4, and -O=C—O-C-3 were confirmed. Accordingly, the planar structure of 1 was determined to be an unprecedented 6/6/6/6/6/6/5 fused skeleton. According to the correlation of H-2/H-6′ in rotating frame Overhauser effect spectroscopy (ROESY) spectrum and the literature, two pyran rings [C-9→O→C-2→C-3→C-4→C-10] and [C-2′→O→C-4→C-3→C-2→C-1′] were presented through cis fusion [11]. Thus, the absolute configuration of 1 possessed the following pairs of potential forms: 2R*, 3R*, 4R*, 7′′S*, 2aR*/2S*, 3S*, 4S*, 7′′R*, 2aS* (Ⅰ), 2R*, 3R*, 4R*, 7′′S*, 2aS*/2S*, 3S*, 4S*, 7′′R*, 2aR* (Ⅱ), 2R*, 3S*, 4R*, 7′′R*, 2aR*/2S*, 3R*, 4S*, 7′′S*, 2aS* (Ⅲ), 2R*, 3S*, 4R*, 7′′R*, 2aS*/2S*, 3R*, 4S*, 7′′S*, 2aR* (Ⅳ), 2R*, 3S*, 4R*, 7′′S*, 2aS*/2S*, 3R*, 4S*, 7′′R*, 2aR* (Ⅴ), 2R*, 3S*, 4R*, 7′′S*, 2aR*/2S*, 3R*, 4S*, 7′′R*, 2aS* (Ⅵ), 2R*, 3R*, 4R*, 7′′R*, 2aS*/2S*, 3S*, 4S*, 7′′S*, 2aR* (Ⅶ), and 2R*, 3R*, 4R*, 7′′R*, 2aR*/2S*, 3S*, 4S*, 7′′S*, 2aS* (Ⅷ). In the ROESY spectrum, the significant correlation of H-2/H-7′′ was observed. However, this correlation cannot be observed in forms Ⅴ-Ⅷ (Fig. S29 in Supporting information). Herein, GIAO ¹³C NMR calculations with DP4+ analyses were utilized to differentiate the forms Ⅰ-Ⅳ [13,14]. In order to increase accuracy of the results, two functions and three basis sets were selected to carry out GIAO ¹³C NMR calculations and DP4+ analyses. As a result, the predicted spectroscopic data for the form Ⅳ was in closer agreement with the experimental values (Table S1 in Supporting information). Then, electronic circular dichroism (ECD) calculations were employed to confirm the absolute configuration of 1. Systematic conformational analyses of the structure of 2R, 3S, 4R, 7′′R, 2aS were carried out using molecular mechanics force field MMFF94 calculations. Optimized conformations were obtained at the B3LYP/6–31G(d) level. At the CAM-B3LYP/6–311G(d, p) level, their ECD spectra were calculated using time-dependent density functional theory (TDDFT). The overall calculated ECD spectra were generated by determining their lowest energy conformers through Boltzmann weighting. Throughout the entire range of wavelengths, the calculated spectrum of 2R, 3S, 4R, 7′′R, 2aS was consistent with the experimental ECD spectrum of 1 (Fig. 3). Finally, the structure of 1 was established and named neosophoflavonoid A.

As diastereoisomers, compounds 2a and 2b were isolated by utilizing a normal-phase chiral column. Herein, only the elucidated process of 2a is presented. The molecular formula of 2a was deduced as C₅₂H₅₈O₁₁, based on HRESIMS analysis (ion peak at m/z 859.4038 [M + H]⁺, calcd. for C₅₂H₅₉O₁₁ 859.4051). The ¹H NMR spectrum of 2a (Table S3) exhibited differences from that of compound 1 as one more penta-substituted aromatic ring [δ_H 6.01 (1H, s, H-5′′′)] and one more lavandulyl unit [δ_H 2.66 (2H, brd, J = 7.5 Hz, H-1′a), 2.29 (1H, m, H-2′a), 2.10 (2H, m, H-3′a), 5.06 (1H, m, H-4′a), 1.58 (3H, brs, H-6′a), 1.64 (3H, brs, H-7′a), 4.62 (1H, m, H-9′aa), 4.55 (1H, d, J = 2.5 Hz, H-9′ab), and 1.72 (3H, brs, H-10′a)] were presented. Compared that of compound 1, the ¹³C NMR spectrum of 2a (Table S3) showed 17 more signals that contained six aromatic carbons, ten lavandulyl carbons, and one methoxyl carbon. HMBC correlations from H-8′′ to C-1′′′, H-5′′′ to carbonyl carbon, H-1′a to C-2′′′, and H-2′a to C-3′′′ suggested that one more penta-substituted aromatic ring was attached at the carbonyl carbon and one more lavandulyl unit was attached at C-3′′′ (Fig. 2). In addition, unlike compound 1, H-3 (δ_H 2.36, 1H, dd, J = 9.5, 2.5 Hz) was presented, indicating that compound 2a did not contain five-membered lactone ring. Thus, the planar structure of 2a was established as shown. Similar to 1, two pyran rings [C-9→O→C-2→C-3→C-4→C-10] and [C-2′→O→C-4→C-3→C-2→C-1′] should be cis fused. Then, quantum chemical calculations were used to determine the absolute configurations of compounds 2a and 2b. First, considering that two flexible lavandulyl units can produce diverse conformations and have an insignificant influence on the chiral centers, simplified planar structures Ⅰ-Ⅳ (representing eight isomers, as shown in Fig. S37 in Supporting information) were used for the GIAO ¹³C NMR (DP4+) calculations and ECD calculations. In GIAO ¹³C NMR calculations, form Ⅱ corresponded better with the experimental values, which was supported by a DP4+ probability of approximately 99.76% at the B3LYP/6–311G(d, p) level (Fig. S38 in Supporting information). In ECD calculations, 2S, 3R, 4S, 7′′S and 2R, 3S, 4R, 7′′R were consistent with the experimental ECD spectra of 2a and 2b, respectively (Fig. 3). The absolute configurations of C-2a and C-2a′ were determined by deep GIAO ¹³C NMR calculations and DP4+ analyses. Isomers 2S, 3R, 4S, 7′′S, 2aR*, 2a′R*, 2S, 3R, 4S, 7′′S, 2aS*, 2a′S*, 2S, 3R, 4S, 7′′S, 2aR*, 2a′S*, and 2S, 3R, 4S, 7′′S, 2aS*, 2a′R* were calculated for their DP4+ probabilities. Similarly, in order to increase accuracy of the calculation results, two functions and three basis sets were used in process. As results, the 2S, 3R, 4S, 7′′S, 2aS*, 2a′R* isomer was the optimal result (Table S2 in Supporting information). Finally, the absolute configurations of 2a and 2b were confirmed as 2S, 3R, 4S, 7′′S, 2aS, 2a′R and 2R, 3S, 4R, 7′′R, 2aS, 2a′R, respectively. In most reported phytochemical researches on Sophora flavescens, the absolute configuration of lavandulyl unit was ambiguous. A few valuable reports assigned the absolute configuration of lavandulyl unit as S or R according to the X-ray crystallographic analysis or quantum chemical calculations [15,16]. Herein, the thorough quantum calculations unambiguous clarified the absolute configurations of lavandulyl units. Compounds 2a and 2b were designated neosophoflavonoid B and neosophoflavonoid C, respectively.

A plausible biosynthetic pathway of 1, 2a, and 2b was presented in Scheme 1. For compound 1, dimer I was generated through a Michael addition reaction between dihydroflavone and chalcone, II was formed through the oxidation of I. Then, two carbonyls were protonated to form the ketal with the corresponding hydroxyl. Finally, compound 1 was generated by lactonization. The biosynthetic pathways of 2a and 2b were similar to that of 1. For these compounds, I was protonated to form the ketal with the corresponding hydroxyl group.

The PTP1B inhibitory effects of all compounds were evaluated in vitro. Compound 1 showed moderate inhibitory activity on PTP1B with half maximal inhibitory concentration (IC₅₀) of 3.94 ± 0.01 µmol/L; 2a and 2b, however, exhibited stronger inhibitory effects on PTP1B with IC₅₀ values of 0.38 ± 0.13 µmol/L and 0.70 ± 0.01 µmol/L (positive control: CCF06240, IC₅₀ = 1.73 ± 0.46 µmol/L), respectively. In addition, the cytotoxic activities of compounds 1, 2a, and 2b were evaluated in HepG2 cells by the cell counting kit-8 (CCK-8) assay. During exposure to compounds 1, 2a, and 2b (0.1–20 µmol/L) for 24 h, the viability of HepG2 cells was not significantly affected, showing that the compounds did not exhibit cytotoxicity (Fig. 4). Furthermore, compounds 1, 2a, and 2b exhibited inhibitory effects on lipid accumulation in the oleic acid (OA)-induced cell model in the concentration range of 1–10 µmol/L (Fig. 4), and the inhibitory effects of 2a and 2b at 5 and 10 µmol/L were comparable or stronger than those of the positive control fenofibrate (Feno) at a concentration of 20 µmol/L. However, compound 1 at 10 µmol/L only showed a 14.9% inhibitory rate, which was lower than that of the positive control fenofibrate (Feno) at a concentration of 20 µmol/L (26.5%). From the perspective of structure-activity relationship, a multiple substituted aromatic ring with lavandulyl unit in compounds 2a and 2b may have an important effect on their inhibitory activities.

In summary, three highly oxidized hybrid flavonoids were isolated from the roots of Sophora flavescens. Their absolute configurations were unambiguous clarified by thorough quantum calculations including GIAO ¹³C NMR (DP4+) calculation protocol and electronic circular dichroism (ECD) calculation method. Additionally, compounds 2a and 2b exhibited significant inhibitory effects both on PTP1B and lipid accumulation. This investigation furnishes novel and promising structures into the development of antihyperlipidemic candidate.

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

This work was supported by the National Natural Science Foundation of China (No. 81973194) and Biomedical High Performance Computing Platform, Chinese Academy of Medical Sciences.

Supplementary material associated with this article can be found, in the online version, at doi:10.1016/j.cclet.2023.109456.