The prepared slices of Chinese crude medicines were purchased from Beijing Tongren Tang Co., Ltd. (Dalian, China). Ginkgetin, isoginkgetin, bilobetin, and amentoflavone were purchased from Herbest Biotechnology Co., Ltd. (Baoji, China). N-ethyl-1,8-naphthalimide (NEN) was synthesized by our lab, and was fully characterized by NMR and MS techniques. Midazolam, NADPNa₂, glucose-6-phosphate dehydrogenase, and D-glucose-6-phosphate (G-6-P) were obtained from Sigma-Aldrich (St. Louis, MO, United States). The pooled human liver microsomes (HLM) were purchased from Bioreclamation IVT (MD, United States).

Medical materials were ground to coarse powder, and then immersed into 10 volumes of 95% ethanol. Subsequently, the powdered materials were extracted twice with ultrasound (600 W) for 30 min. The filtered extracts were enriched by a rotary vacuum evaporator. The extracts were dissolved in DMSO as the stock solution (20 mg/ml) for the subsequent.

The primary incubation mixture contains HLM, potassium phosphate buffer (pH 7.4, 100 mM), G-6-P (10 mM), G-6-P dehydrogenase (1 U/ml), MgCl₂ (4 mM), NEN (50 µM), extracts of herbal medicine (20 µg/ml), or components. The incubation duration and HLM protein concentrations were 30 min and 0.05 mg/ml, respectively. The reaction was initiated by the addition of NADP⁺ and terminated by the addition of 100 µL ice-cold acetonitrile (ACN). NEN 4-hydroxylation were utilized as the probe reaction for CYP3A4, and fluorescence image of 96-well microplates were measured by a fluorescent image analyzer (GE Amersham Typhoon, Sweden) under the excitation wavelength of 488 nm (E _m: 570 ± 15 nm). The formation of NEHN was evaluated by the quantitative analysis of fluorescence signal for individual wells of microplates using the equipment supporting software. The percentage residual activity was calculated from percentage relative value of NEHN formation rate.

The chemical fingerprints of Ginkgo biloba L. (GB) and Selaginella tamariscina (P. Beauv.) Spring (ST) were analyzed based on UPLC with a photodiode array detector.

Fractions were collected with an Acquity UPLC system (Waters, Milford, MA, United States) equipped with a photodiode array detector (2998), an auto sampler, a quaternary gradient pump, and a column compartment. An AQCUITY UPLC HSS C18 (100 mm × 2.1 mm, 1.8 μm, Waters) analytical column was used and kept at 30°C. The mobile phase consisted of 0.3% trifluoroacetic acid in water (A) and Methanol (B). The following gradient conditions was used: 0.0–5.0 min, 85% A, 5.0–38.0 min, 85–20% A, 38.0–41.00 min, 20% A, 41.0–44.0 min, 20–85% A, 44.0–47.0 min, 85% A for analysis of GB extract; 0.0–3.0 min, 65% A, 3.0–10.0 min, 65%–10% A, 10.0–15.00 min, 10% A, 15.0–18.0 min, 10%–65% A, 18.0–20.0 min, 65% A for analysis of ST extract. The precision of the fingerprinting method was previously confirmed by comparison of the HPLC spectrum mainly considering the relative retention time and peak area. Then, the LC fractions of crude extract (20 mg/ml, 10 μL) were collected consecutively at 30 or 60 s intervals, where the flow rate was set at 0.3 ml/min.

All LC fractions were dried in vacuum, and redissolved in DMSO. The fractions were evaluated for their activity moderating effect for CYP3A4 based on the high-throughput screening method. Then, the target fractions underwent further preparation and purification by semi-preparative HPLC.

Varying concentrations of DGK, GK, IGK, and AMF incubated for 30 min with varying concentrations of NEN (15–150 µM). The high-throughput screening assay was performed, and the inhibition curve was fitted with GraphPad Prism (Version 8.0). The inhibition type was confirmed based on the R square of the following four equations and the Dixon and Lineweaver–Burk plots. The inhibition constant (K _i) values were determined using the equations for competitive inhibition (Eq. 1), noncompetitive inhibition (Eq. 2), uncompetitive inhibition (Eq. 3), and mixed inhibition (Eq. 4). v = V m a x S K m ( 1 + I K i ) + S (1) v = V m a x S ( K m + S ) ( 1 + I K i ) (2) v = V m a x S S ( 1 + I K i ) + K m (3) v = V m a x S K m ( 1 + I K i ) + S ( 1 + I / α K i ) (4)

Varying concentrations of DGK, GK, IGK, and AMF incubated for 30 min with gefitinib (10 μM), tamoxifen (20 μM), ticagrelor (30 μM), respectively (Wang et al., 2020). The ExionLC AD HPLC system consisted of an autosampler, a quaternary delivery system and a degasser. The chromatograph was equipped with a Luna Omega PS C18 (2.1 × 100 mm, 3 μm) analytical column, and the mobile phase consisted of 0.1% formic acid aqueous solution (A) and ACN (B), and were conducted with a 0.30 ml/min flow rate as follow for the gradient program: 0–1.5 min (10–10% B), 1.5–3.0 min (10–45% B), 3.0–5.0 min (45–90% B). An AB MDS Sciex QTRAP 5500 Mass Spectrometer equipped with an ESI source was used to analyze the target substances. The positive mode was used with N-demethylated tamoxifen (358.0→58.0) and defluorinated gefitinib (445.0→128.0). The relative parameters were set as follows: ion spray voltage, 5500 V; temperature, 550°C; curtain gas (CUR) flow, 35 L/min; gas1 and gas2 (nitrogen), 45 and 55 psi. The negative mode was used with dehydroxyethoxylated ticagrelor (477.2→361.1). The relative parameters were set as follows: ion spray voltage, -4500 V; temperature, 550°C; curtain gas flow, 35 L/min; gas1 and gas2, 50 and 50 psi. For other parameters, please refer to Supplementary Table S1 in the supplementary data.

The inhibitory effects of 93 herbal medicines toward CYP3A4 were preliminary evaluated using the high-throughput screening system constructed with CYP3A4 selective fluorescent probe NEN (Figure 1). Inhibition of CYP3A4 by herbal medicine extracts was determined based of their effect on the intensity of fluorescence at 570 ± 15 nm that was excited at 488 nm. Visual observation of the fluorescence images of the multi-well plates reveals different degree of modulatory effect of herbal medicines toward CYP3A4 activity. Ginkgo biloba L. (GB, F2) and Selaginella tamariscina (P. Beauv.) Spring (ST, D6) exhibited ultra-strong inhibitory effects toward CYP3A4 resulting in virtual elimination of the fluorescence signal in the corresponding wells of the 96-wells microplate. Quantitative analysis of the inhibitory effects of herbal medicine towards CYP3A4 demonstrate that the residual activities of CYP3A4 after treatment with the extracts of GB and ST were 6.3 and 5.0%, respectively (Supplementary Table S2). The quantitative results were consistent with the visually observed trend of the modulatory effects toward CYP3A4 activity. Additionally, three herbal medicine extracts exhibited relatively strong inhibitory effects towards CYP3A4 with residual activity less than or equal to 50%, namely Tribulus terrestris L. (D2), Areca catechu L. (B10), and Fissistigma glaucescens (Hance) Merr. (F3).

We attempted to reveal the inhibitory components of GB and ST since they exhibited strong inhibitory effect towards CYP3A4. However, the complexity and structural diversity of the two herbal medicines are very high, especially GB. In the present study, a combination of chemical fingerprinting-guided LC fraction collection and screening of the inhibitory potential was used to identify the inhibitory components of herbal medicine. The LC fractions of GB and ST were collected consecutively at appropriate time intervals (30 or 60 s) based on the reliable LC method which was confirmed by reproducible relative retention time of the main components of herbal medicine (RSD < 3%). Subsequently, the inhibitory components of GB and ST toward CYP3A4 were successfully identified by bioassay guided LC fractions studies utilizing high-throughput visualization screening. For GB, three inhibitory components were picked out from a comparison of the UPLC-UV fingerprint with the profile of CYP3A4 inhibition (Figure 2). Then, three key components of GB were prepared by bioassay-guided LC fractionation and identified as bilobetin (7-demethylginkgetin, DGK), ginkgetin (GK), and isoginkgetin (IGK), respectively, according to the comparison of the LC retention times, spectral data, and the mass data of isolated compounds and authentic standards (Wang et al., 2015). Amentoflavone (AMF), meanwhile, was identified as the main component contributing to the strong inhibitory effect of ST (Table 1).

The concentration range of DGK, GK, IGK, and AMF was used to depict the inhibitory plot and characterize the inhibitory effects of four identified biflavones toward CYP3A4. As shown in Figure 3, all of the four biflavones can potently inhibit the formation of N-ethyl-4-hydroxy-1,8-naphthalimide (NEHN) in HLM in a dose-dependent manner. The IC ₅₀ values of DGK, GK, IGK, and AMF against CYP3A4 were evaluated as 0.162 ± 0.022, 0.106 ± 0.004, 0.982 ± 0.006, and 0.186 ± 0.004 μM, respectively. These findings suggested that the four biflavones exhibited a significant inhibition activity towards CYP3A4 and further confirmed the above conclusion that DGK, GK, IGK, and AMF were the main inhibitory components of GB and ST, respectively.

Given that all of the four identified biflavones displayed strong inhibitory effects toward CYP3A4, the inhibition constant (K _i) values of DGK, GK, IGK, and AMF toward CYP3A4 were further determined. As shown in Figure 4A, DGK displayed mixed reversible inhibition against NEN 4-hydroxylation, and the K _i value was determined to be of 0.070 ± 0.018 μM. Correspondingly, the inhibition profiles of GK, IGK, and AMF were all fitted to mixed reversible inhibition equation with the K _i values of 0.034 ± 0.007, 0.278 ± 0.087, and 0.101 ± 0.030 μM (Table 2). These results demonstrated that DGK, GK, IGK, and AMF functioned as strong inhibitors of CYP3A4 with the K _i values ranging from 0.034 to 0.278 µM (Table 2).

The CYP3A4-mediated potential interactions between DGK, GK, IGK, AMF, and clinical drugs were investigated. The selective estrogen receptor modulator tamoxifen, tyrosine kinase inhibitor gefitinib, and P2Y12 platelet inhibitor ticagrelor, which mainly undergo CYP3A4 metabolism (Desta et al., 2004; Li et al., 2007; Zhou et al., 2011), were subjected to our study of their potential drug interaction profiles. As shown in Figure 5, the four biflavones potently inhibit the CYP3A4-mediated oxidation of three clinical drugs in HLM in a dose-dependent manner. The half-inhibition concentration values for the four biflavones are shown in Table 3. The IC ₅₀ values for GK against tamoxifen and gefitinib was 0.478 ± 0.003 and 0.869 ± 0.001 µM, respectively. It was found that GK exhibited the strongest inhibitory effect towards all three clinical drugs. It is followed by DGK, AMF, and IGK; this order is consistent with that determined in the studies with fluorescent probe NEN.