Unveiling the chemical components variation of Sishen formula induced by different prescription ratios by the advanced liquid chromatography/mass spectrometry approaches

2.1 Chemicals and reagents

Forty-one reference compounds (purity ≥ 98 %) (Fig. 2 and Table S1) were purchased from Desite Biotechnology Co., Ltd. (Chengdu, China). Wogonoside, used as the internal standard (I.S.) in the quantitative assay, was obtained from Standard Biotech. Co., Ltd. (Shanghai, China). The Chinese medicines, Psoraleae Fructus (PF; Origin: Yunnan, China), Myristicae Semen (MS; Origin: Guangdong, China), Schisandrae Chinensis Fructus (SCF; Origin: Liaoning, China), and Euodiae Fructus (EF; Origin: Yunnan, China), were purchased from Beijing Tongrentang, Tianjin. Ultra-pure water in-house prepared using a Milli-Q Integral 5 water purification system (Millipore, Bedford, MA, USA), LC-MS-grade acetonitrile, methanol (Fisher, Fair Lawn, NJ, USA), and formic acid (Sigma-Aldrich, MO, Switzerland), were used.

Close

Fig. 2

The chemical structures of 41 reference compounds.

Export to PPT

2.2 Preparation of the reference standard solutions

The reference standard solution for multicomponent characterization and metabolomics analysis was prepared by dissolving the accurately weighed each reference standard (1.0 mg) in 50 % (v/v) methanol. The obtained solutions were mixed and diluted (concentration: 10 μg/mL). After the centrifugation (11,481 × g; 14,000 rotations per minute, rpm) for 10 min controlled at 4℃, the resulting supernatant was taken.

For the quantitative assay, the individual stock solutions for 18 analytes (1, 4, 7, 8, 12–18, 20–23, 36, 38, and 39) and the I.S. were prepared by dissolving each compound in 50 % (v/v) methanol. A stock solution of mixed reference standards was prepared by pooling them with the adequate volumes. A series of calibration solutions were obtained by diluting the mixed reference standards stock solution using 50 % methanol, to construct the calibration curves for the quantitative assays.

2.3 Preparation of the SSF test solutions

The test solutions of SSF were prepared using a water reflux method as recorded in the Chinese Pharmacopoeia (version 2020). The accurately weighed Chinese medicines (PF: MyS: SCF: EF = 4: 2: 2: 1; weight ratio) were soaked in 10-time volume of the ultra-pure water for 30 min, and then decocted for 1 h. Afterward, they were prepared to the solution at a concentration of 100 mg/mL (referring to the sum of all component drugs). By further diluting to 20 mg/mL, the extracted liquid was centrifuged (11,481 × g) for 10 min at 4℃, with the supernatant taken for the multicomponent characterization of SSF. The SSF samples used for the differential analysis, in general, were prepared by the same procedures as described above. In total, 16 different proportions of SSF (Table 1) were considered, with five replicates for each. A Quality control (QC) sample was prepared by pooling the obtained 16 SSF solutions in an equal volume and diluting by the pure water to 20 mg/mL. All the samples of SSF and QC were centrifuged (11,481 × g) for 10 min at 4℃, using the supernatant as the test solution.

For the SSF samples for the quantitative assays, 16 different proportions were prepared in triplicate (100 mg/mL), and further diluted to a concentration of 5 mg/mL. Each SSF sample solution, 50 % methanol, and the I.S. solution, at the volume ratio 1: 3: 1 were pooled, and further centrifuged (11,1481 × g) at 4℃ for 10 min, with the used as the test solution. SSF with the ratio of 4: 2: 2: 1 was used in the methodological establishment and validation experiments.

2.4 Orthogonality, effective peak capacity, and method validation for the off-line 2D-LC/IM-QTOF-MS system

Orthogonality of the developed off-line 2D-LC/IM-QTOF-MS system was evaluated using the asterisk equations proposed by Camenzuli and Schoenmakers, and 65 compounds as the index (Camenzuli and Schoenmakers, 2014). The calculation of orthogonality and peak capacity was consistent with the methods we previously reported, provided in detail as the Supporting Information (Wang et al., 2022a).

2.5 Off-line 2D-LC/IM-QTOF-MS for the multicomponent characterization

The dimension-enhanced off-line 2D-LC/IM-QTOF-MS approach was developed for the in-depth multicomponent characterization of SSF. The ¹D separation by HILIC (hydrophilic interaction chromatography) was performed on an Agilent 1260 HPLC system (Agilent Technologies, Waldbronn, Germany). The water extract of SSF decoction (20 mg/mL) was initially separated on a Waters XBridge Amide column (4.6 × 150 mm, 3.5 μm) maintained at 35 °C. A binary mobile phase, containing 0.1 % formic acid (FA) in H₂O (A) and acetonitrile (B), ran according to an optimal gradient program: 0–3 min, 99–99 % (B); 3–5 min, 99 %–94 % (B); 5–10 min, 94 %–85 % (B); 10–13 min, 85 %–65 % (B); 13–16 min, 65 %–50 % (B); and 16–19 min, 50 % (B). A flow rate of 1.0 mL/min was set, and the injection volume was 50 µL. The ¹D HILIC separation was repeated by four times. The PDA detector monitored the UV signal at 254 nm to guide the fractionation. The eluate for each minute (from 1 to 19 min) was collected, giving 19 sub-fractions (numbered as Fr.1 to Fr.19). They were dried under a steady flow of N₂. Each of the dry residues was dissolved in 500 μL of H₂O and further centrifuged at 11,481 × g for 10 min, with the resultant supernatant taken as the test solution (concentration: 20 mg/mL of drug material). An ACQUITY UPLC I-Class/Vion™ IM-QTOF system (Waters, Milford, MA, USA) was adopted to perform the ²D separation in the RPC (reversed-phase chromatography) mode and acquire the MS data for structural elucidation. An HSS T3 column (2.1 × 100 mm, 1.8 µm) maintained at 30 °C was selected. A binary mobile phase, containing 0.1 % FA in H₂O (A) and acetonitrile (B), was utilized which ran according to an optimal gradient program: 0–7 min, 2–10 % (B); 7–14 min, 10 %–15 % (B); 14–24 min, 15 %–22 % (B); 24–29 min, 22 %–30 % (B); 29–31 min, 30 %–35 % (B); 31–38 min, 35 %–50 % (B); 38–48 min, 50 %–75 % (B); and 48–50 min, 75 %–95 % (B). The flow rate was set at 0.3 mL/min, and the injection volume was 5 µL.

High-accuracy MS data of SSF, in both the positive and negative ESI (electrospray ionization) modes, were acquired by an HDMS^E-HDDDA hybrid scan method available on the Vion^TM IM-QTOF mass spectrometer (Hu et al., 2023; Qian et al., 2022). Parameters for the LockSpray ion source were as follows: capillary voltage, +1.5 kV (ESI+)/−1.0 kV (ESI−); cone voltage: 60 V (ESI+)/−40 V (ESI−); source offset, 80 V; source temperature, 120 °C; desolvation gas temperature (N₂), 500 °C; desolvation gas flow (N₂), 800 L/h; and cone gas flow (N₂), 50 L/h. The mass analyzer scanned over a mass range of m/z 100–1000 at a low collision energy of 6 eV for both two ESI modes by HDMS^E at 0.3 s per scan (MS¹) and HDDDA at 0.15 s per scan (MS¹). The ramp collision energy (RCE) of 10–60 eV (ESI+)/20–60 eV (ESI−) was set in HDMS^E. The MS/MS fragmentation of three most intense precursors was automatedly triggered by HDDDA when the intensity exceeded 200 detector counts under mass-dependent ramp collision energy (MDRCE) of 15–40 eV (ESI+)/20–40 eV (ESI−) in the low mass ramp and 25–50 eV (ESI+)/30–50 eV (ESI−) in the high mass ramp. The MS/MS acquisition stopped until the time exceeded 0.5 s. Default parameters for the traveling wave IM separation were defined, and the calibration of CCS was conducted according to the manufacturer’s guidelines using a mixture of calibrants (Paglia et al., 2015). The MS data calibration was conducted by constantly infusing the leucine enkephalin solution (Sigma-Aldrich, St. Louis, MO, USA; 200 ng/mL) at 10 µL/min.

2.6 UHPLC/QTOF-MS for the untargeted metabolomics analysis

An Agilent 1290 Infinity II UHPLC system coupled with 6550 QTOF mass spectrometer (Agilent Technologies, Waldbronn, Germany) was utilized to acquire the Auto MSMS data of different prescription ratios of SSF samples in the positive ESI mode. In detail, an HSS T3 column (2.1 × 100 mm, 1.8 µm) maintained at 30 °C was chosen for the chromatographic separation. A binary mobile phase consisting of 0.1 % FA in water (A) and acetonitrile (B) was used running according to the following gradient program: 0–5 min, 2 % (B); 5–15 min, 2 %–10 % (B); 15–26 min, 10 %–15 % (B); 26–34 min, 15 %–22 % (B); 34–44 min, 22 %–30 % (B); 44–48 min, 30 %–35 % (B); 48–53 min, 35 %–50 % (B); 53–58 min, 50 %–75 % (B); and 58–64 min, 75 %–95 % (B). A flow rate of 0.4 mL/min and the injection volume of 5 µL were set. For the 6550 QTOF, nitrogen (N₂) was used as both the drying gas (at 12 L/min, 200 °C) and sheath gas (at 11 L/min, 350 °C). Atomizer pressure of 40 psi, threshold of 80 (1 %), capillary voltage of 3.5 kV, nozzle voltage of 1.0 kV, fragmentor of 390 V, and collision energy (CE) =  − 1.586 (m/z) /100 + 54.49, were set. The mass scan range was 100–1000 Da with two spectra recorded per second.

To unveil the chemical composition difference for different SSF samples, 16 different prescription ratios (80 samples, in total; Table 1) were analyzed. The acquired data were imported into the Agilent MassHunter Profinder software to conduct the peak alignment, detection, and extraction. A list of metabolic features including the RT, Mass, and abundance information, were generated. In addition, the “80 % rule” was set to filter the common features (Wang et al., 2023). The remaining were used as the variables imported into the SIMCA-P 14.1 software (Sartorius, Umea, Sweden) for chemometric analysis by PCA (principal component analysis) and OPLS-DA (orthogonal partial least-squares discriminant analysis). Variable importance in projection (VIP), as an indicator, was utilized to filter the potential differential markers for SSF with different prescription ratios.

2.7 UHPLC/QTrap-MRM and method validation for the quantitative assay

A sensitive MRM approach was established on the Waters ACQUITY UPLC I-Class system (Waters) coupled with a QTrap 4500 mass spectrometer (AB Sciex Scientific, Concord, Canada). The data of 48 SSF samples with different ratios and the solutions of PF, SCF, and EF, were acquired. An HSS T3 column (2.1 × 100 mm, 1.7 µm) maintained at 30℃, was eluted by a binary mobile phase consisting of 0.1 % formic acid (FA) in H₂O (A) and acetonitrile (B) at a flow rate of 0.3 mL/min, following a gradient elution program: 0–1 min: 5 % (B), 1–6 min: 5 %–10 % (B), 6–10 min: 10 %–11 % (B), 10–25 min: 11 %–30 % (B), and 25–27 min: 30 %–95 % (B). The injection volume was 4 µL. Key parameters set for MRM were as follows: ESI (+/−); curtain gas (CUR), 35 psi; air injection (CAD), medium; spray gas voltage (IS), ± 4,500 V; ion source temperature (TEM), 550 °C; atomizing gas 1 (GS1) and drying gas 2 (GS2), 45 psi; inlet voltage (EP), ± 10 V; impact room outlet voltage (CXP), ± 13 V. Other parameters, including the parent/product ions, collision energy, declustering potential, and retention time, for the 18 analytes and the I.S. are summarized in Table 2.

The UHPLC/QTrap-MRM quantitative assay approach was validated in terms of selectivity, linearity, sensitivity, precision, stability, repeatability, and recovery, with the detail provided as the Supporting Information.

3.1 Development and validation of an off-line 2D-LC/IM-QTOF-MS approach dedicated to the separation and characterization of complex components from SSF

To comprehensively characterize the complex components occurring to SSF, an off-line 2D-LC/IM-QTOF-MS approach was well established. Key parameters affecting the chromatographic separation in both two dimensions and the MS monitoring were successively optimized by the single-factor experiments.

The separation difference of the SSF components among 26 commercial columns with either the RPC or HILIC mechanism of separation (Table S2) was assessed by the PCA score plot (Wang et al., 2022b). Evidently, 19 RPC columns (including 16 annotated as the ¹D candidates and three as the ²D in Table S2) and seven HILIC columns showed the segregation, which could demonstrate their large selectivity on the SSF components (Fig. 3A). We further compared the separation performance of 16 columns (tightly clustered, annotated with the ¹D candidates in Table S2) on the SSF components, and considered the HSS T3 column as the best enabling the more balanced resolution (Fig. S1). The best performance in the ²D-RPC separation (0.1 % FA in H₂O/acetonitrile as the mobile phase; column temperature, 30 °C) was obtained after the subsequent optimization on the mobile phase, column temperature, and gradient elution program. The remaining ten columns were regarded as the candidates in the ¹D chromatography by preparing the relative retention time scatter plots and calculating the linearity aggression coefficient of determination (R²) of 45 index components (Fig. 3B). By considering both the R² value and the peak shape, the XBridge Amide column was selected. The same mobile phase, as that in the ²D chromatography, was chosen, and the best column temperature was optimized at 35 °C.

Close

Fig. 3

Selection of the stationary phases in establishing the off-line 2D-LC/HRMS system (A: A PCA diagram of 26 chromatographic columns; B: the BPC diagrams of SSF obtained on the candidate ¹D columns and the scatter diagrams for primary orthogonality evaluation between 10 candidate chromatographic columns and HSS T3).

Export to PPT

The key ion source parameters (capillary voltage and cone voltage) of the Vion IM-QTOF mass spectrometer and the collision energy set in the HDMS^E-HDDDA hybrid scan were optimized. Influences by alternating capillary voltage (1.0–3.5 kV) and cone voltage (20–100 V) were tested using six index components in both the positive and the negative modes (Fig. S2). Despite different ion response (viewed by the integrated peak area) changes were observed with the ascending of capillary voltage by ESI + and ESI–, we considered the voltages at 1.5 kV in the positive mode and 1.0 kV in the negative mode as the best choices. In the case of cone voltage, the response of the index components showed inconsistent variation trends by these two different ESI modes (Fig. S2). The ion response, repeatability determined through three parallel injections, and absence of severe ion-source decay, were all taken into account, and ultimately, 60 V of cone voltage in the positive mode and 40 V in the negative mode were selected. For the hybrid scan approach of HDMS^E-HDDDA, RCE of HDMS^E and MDRCE of HDDDA were separately optimized by observing the dissociation degree of the index components. Among three levels of RCE (e.g. 10–60 eV, 20–70 eV, and 30–80 eV) examined in the positive mode, it was found that under 10–60 eV most of the index compounds could generate rich and diversified fragments useful to establish their structures, while 20–60 eV was selected in the negative mode (Fig. S3). MDRCE is accessible for HDDDA, which can apply a customized collision energy ramp on a given m/z of the precursor ion to gain more balanced MS² spectra (Wang et al., 2021). Among the four different levels we had examined, MDRCE of 15–40 eV/25–50 eV in the positive mode and 20–40 eV/30–50 eV in the negative mode were set (Fig. S4).

The performance of the established off-line 2D-LC/IM-QTOF-MS approach in identifying the multicomponents of SSF was evaluated by orthogonality and effective peak capacity (Wang et al., 2022a). The asterisk equations method, proposed by Camenzuli and Schoenmakers (Camenzuli & Schoenmakers, 2014), was utilized to assess the orthogonality. The relative retention time (t_R,norm; Eq. (1)) of 65 components (Table S3) was considered to measure their distribution around four crossing lines (Z_–, Z₊, Z₁, Z₂; Eqs. (2)–(9)). Four Z parameters were calculated to be 0.89 for Z_–, 0.77 for Z₊, 0.99 for Z₁, and 0.92 for Z₂. Accordingly, the orthogonality (A₀, Eq. (10)) of the 2D-LC/HRMS system was 0.79 (Fig. S5). Besides, with the averaged peak width (W_b, Eq. (11)) at baseline in the ¹D (0.24 min) and ²D (0.40 min) chromatography, the peak capacity (n_grd, Eq. (11)) in each dimension was 81 for ¹n_grd and 124 for ²n_grd, and thereby the effective peak capacity (n_2D, Eq. (13)) was 1276.

The intra-/inter-day precision and repeatability for both the ¹D and ²D chromatography were evaluated. Consequently, the intra-/inter-day precision varied in the ranges of 1.25–2.41 % (Table S4) and 4.26–4.84 % (Table S5) for the ¹D chromatography. The RSDs for the intra-/inter-day precision of the ²D chromatography were 1.82–4.35 % (Table S6) and 6.85–11.62 % (Table S7) in the positive mode, and 2.61–9.77 % (Table S8) and 5.99–11.59 % (Table S9) for the negative mode, respectively. Repeatability varied among 2.24–4.37 % (Table S10) for the ¹D chromatography and 6.52–9.11 % (Table S11) for the ²D chromatography. All these data demonstrated high orthogonality, powerful separation ability, and good performance for the off-line 2D-LC/HRMS in the separation and characterization of complex SSF components.

3.2 Comprehensive characterization of the multicomponents from SSF by automatic peak annotation workflows facilitated by UNIFI and searching the in-house database

Multicomponent characterization for SSF was performed by the streamlined automatic data processing workflows established on the UNIFI software. The HDMS^E and HDDDA data (acquired by the hybrid scan approach both in the positive and negative ESI modes) for 19 subsamples prepared from the SSF total extract were interpreted (Fig. S6). As a result, 233 compounds were characterized from SSF, of which 41 ones were confirmed by comparison with the reference standards (Table S12). Fig. 4 displays the typical neutral loss (NL) and diagnostic product ions (DPIs) for rapidly identifying the typical classes of compounds, including the flavonoids, alkaloids, limonoids, lignins, and coumarins.

Close

Fig. 4

The typical neutral loss (NL) and diagnostic product ions (DPIs) used to characterize the typical classes of compounds in SSF (e.g. flavonoids, alkaloids, limonoids, lignins, and coumarins).

Export to PPT

3.2.1

3.2.1 Characterization of flavonoids

Flavonoids are one of the main components of SSF, most of which originate from Psoraleae Fructus and Euodiae Fructus (Koul et al., 2019; Li and Wang, 2020). The flavonoids in SSF can be roughly divided into the flavonones and chalcones. A total of 99 flavonoids were characterized from SSF, of which 14 ones (e.g. 39#, 63#, 64#, 113#, 177#, 184#, 190#, 199#, 202#, 212#, 217#, 221#, 225#, and 226#) were confirmed by comparison with the reference compounds.

Flavonones are prone to undergo the loss of CO, CH₃, and H₂O, together with the common Retro Diels-Alder (RDA) cleavage (the breaking of the 0–2 or 1–3 chemical bond on ring C generating the ^1,3A, ^1,3B, ^0,2A, and ^0,2B fragments). In the positive mode, the DPIs included m/z 149.02 ([C₈H₅O₃]⁺), 147.04 ([C₉H₇O₂]⁺), 137.02 ([C₇H₅O₃]⁺), and 119.04 ([C₈H₇O]⁺). Compound 184# (t_R 37.30 min; CCS 180.82 Å²) gave the [M + H]⁺ ion at m/z 323.1276 (C₂₀H₁₈O₄) in the positive ESI mode, which was identified as neobavaisoflavone by comparison with the reference standard. The fragments at m/z 267.0660, 255.0674, and 239.0697, were ascribed to [M + H − C₄H₈]⁺, [M + H − C₅H₈]⁺, and [M + H − C₄H₈ − CO]⁺, respectively. By the characteristic RDA fragmentation, the ^1,3A fragment at m/z 137.0224 was obtained (Fig. 5). In the case of an unknown compound 143# (t_R 31.78 min; CCS 185.52 Å²), the rich [M + H]⁺ precursor ion was observed at m/z 341.1405 (C₂₀H₂₀O₅, −1.3 ppm). Due to the cracking of RDA^1,3A, the ^1,3A fragment ion at m/z 221.0821 [M + H − C₈H₈O]⁺ was obtained (Fig. 5). With the continuous loss of H₂O and C₃H₆, the fragment ions at m/z 203.0689 and 161.0205 were generated. The characteristic fragment at m/z 149.0225 was assigned as [M + H − C₈H₈O − C₄H₈O]⁺, while the fragments of m/z 147.0441 [M + H − C₇H₆O₂ − C₄H₈O]⁺ and 119.0524 were generated by the continuous loss of C₇H₆O₂, C₄H₈O, and CO. By searching the literature and database, compound 143# was tentatively characterized as brosimacutin D or isomer (Xu et al., 2012).

Close

Fig. 5

Annotation of the positive ion mode MS² spectra and proposed fragmentation pathways for the representative flavonoids and coumarins.

Export to PPT

The fragmentations of the chalcone compounds are featured by the cleavages of I and II chemical bonds (Yang et al., 2012), which can easily produce the DPIs at m/z 149.02, 147.04, and 119.04 in the positive mode. Due to the comparison with the reference standard, compound 202# (t_R 40.17 min; CCS 185.69 Å²) showing the [M + H]⁺ ion at m/z 325.1428 (C₂₀H₂₀O₄), was identified as bavachalcone (Fig. 5). By losing C₄H₈, a fragment ion at m/z 269.0807 was generated. And with the cleavage of ^ⅠB⁺, the DPIs at m/z 147.0429 ([M + H − C₄H₈ − C₇H₆O₂]⁺) and 119.0470 ([M + H − C₄H₈ − C₇H₆O₂ − CO]⁺) were produced. Moreover, the fragments at m/z 205.0850 ([M + H − C₈H₈O]⁺) and 149.0218 ([M + H − C₄H₈ − C₈H₈O]⁺) should result from the cleavage of ^ⅡA⁺. The fragment ion at m/z 121.0272, assigned as [M + H − C₄H₈ − C₈H₈O − CO]⁺, was dissociated from the ion of m/z 149.02. These fragmentation features could assist to characterize the unknown compound 148# (t_R 32.22 min; CCS 184.48 Å²), which exhibited the [M + H]⁺ ion at m/z 341.1361 (C₂₀H₂₀O₅, −1.9 ppm) (Fig. 5). Through the ^ⅡA⁺ cleavage and loss of a molecule of H₂O or C₃H₈O or C₄H₈O, the fragment ions at m/z 203.0693 ([M + H − C₈H₈O − H₂O]⁺), 161.0219 ([M + H − C₈H₈O − C₃H₈O]⁺), and 149.0219 ([M + H − C₈H₈O − C₄H₈O]⁺), could be obtained. Another two fragments at m/z 147.0411 ([M + H − C₁₁H₁₄O₃]⁺) and 119.0490 ([M + H − C₁₁H₁₄O₃ − CO]⁺) were produced owing to the cracking of ^ⅠB⁺. Accordingly, it was tentatively characterized as bakuchalcone or isomer (Xu et al., 2021).

3.3 Holistic comparison among different prescription proportions of SSF by UHPLC/QTOF-MS-based untargeted metabolomics

3.3.2

3.3.2 Multivariate statistical analysis of different proportions of SSF decoctions

The different proportions of SSF decoctions and the QC data were imported into the Profinder software for peak alignment and peak detection. After the filtering using the “80 % rule”, 2948 ions/metabolic features were retained, which were used as the variables for the multivariate statistical analysis by PCA and OPLS-DA. By the unsupervised PCA. the tested 80 SSF samples were grouped into four differentiated clusters: group 1 (PB1, PB5, PB9, and PB13), group 2 (PB2, PB6, PB10, and PB14), group 3 (PB3, PB7, PB11, and PB15), and group 4 (PB4, PB8, PB12, and PB16), of which the dosage of Myristicae Semen (MyS) in each group ranged from 1 to 4 g (Fig. 6A). Clearly, the MyS variation caused very weak effects on the extraction of the other herbal drug components. In contrast, the chemical composition differences among the SSF samples were more pronounced when altering the amounts of Schisandrae Chinensis Fructus (SCF) and Euodiae Fructus (EF). For instance, the dosages of Psoraleae Fructus (PF) and MyS were the same (4: 1) in PB2 and PB3, while the ratios of SCF and EF were 2: 2 in PB2 and 3: 0.5 in PB3, respectively. But PB2 and PB3 showed almost the largest chemical difference, witnessed in the score plot of PCA. OPLS-DA is a supervised classification modeling algorithm for better discovering the potential markers. As a result, the same four groups were largely separated from each other in the score plot (Fig. 6B). The observed difference among the four groups was presumed to be due to the ratio changes of SCF and EF (2: 1, 2: 2, 3: 0.5, and 3: 1 in groups 1–4, respectively). The VIP plot can directly reflect the contributing degree of each variable to the classification, which was thus used to filter the potential markers by setting the cutoff at 1.0 (Fig. 6C). A total of 50 compounds were identified to be essential compounds in the discrimination of SSF with differentiated ratios, and 18 thereof were identified due to the availability of reference compounds comparison (Table S17). By integrating the multicomponent characterization results, 19, one, five, and 25 markers were attributed to PF, MyS, SCF, and EF, respectively. The distribution difference among these 50 identified compounds within four separated groups was illustrated by the heatmap (Fig. 6D), in which the color changes from dark blue to dark red were consistent with in the content ascending for each of the 50 markers among different SSF samples. These data could primarily unveil the overall chemical composition difference of SSF originating from the prescription ratio changes.

Close

Fig. 6

Multivariate statistical analysis of the SSF decoctions with 16 different proportions (A: PCA core plot; B: OPLS-DA score plot; C: VIP diagram; D: heatmap).

Export to PPT

3.4 Quantitative assays of 18 potential markers in SSF with different prescription ratios

Based on the results obtained in the untargeted metabolomics differential analysis, we selected 17 potential differential components (e.g., psoralen, isopsoralen, psoralenoside, isopsoralenoside, daidzein, bavachin, corylifolinin, bavachinin, corylifol A, schisandrin A, dehydroevodiamine, evodiamine, rutaecarpine, rutaevin, evodin, chlorogenic acid, and rutin) and one potentially effective component (hyperoside) (Wang et al., 2022c), and quantitatively determine their contents by a validated UHPLC/Q-Trap-MRM approach. Among them, the coumarins from PF (psoralen/isopsoralen/psoralenoside/isopsoralenoside) and alkaloids from EF (dehydroevolodiamine/evolodiamine/rutacarpine) were reported to have the organ toxicity (Zhang et al., 2021; Zhang et al., 2022).

The MRM method was established by optimizing the key parameters (e.g., CE, DP, Q1 mass, and Q3 mass) of 18 target compounds and the I.S. via the direct-infusion experiment (Table 2). It was achieved by perfusing 18 reference solutions (200 ng/mL) at a constant flow of 7 μL/min using a needle pump in Q1 scan, production ion scan, and MRM modes. Wogonoside was used as the I.S., because it was absent in SSF and gave high response in both the positive and negative ion modes. The MRM chromatograms obtained in the positive and negative modes of a QC sample are shown in Fig. 7. Selectivity, linearity, sensitivity, precision, stability, repeatability, and recovery of the developed UHPLC/Q-Trap-MRM method were validated. Spectral comparison among the blank, methodological sample, and the mixed reference standards, could indicate good specificity of the quantitative assay method (Fig. S10). Linear regression coefficients of determination (R²) for 18 analytes were between 0.9913 and 0.9990 over the tested concentration range. LOD of all analytes varied in the range of 0.0031–6.25 pg, and LOQ between 0.0039 and 25 pg (Table S18). The RSDs of intra-day and inter-day precision at the low, medium, and high levels ranged from 0.38 % to 3.54 % and 2.29 % to 7.40 %, respectively. The stability test of the targeted compounds within 48 h showed a variation of 1.83–8.43 %. The repeatability of the quantitative assay method was in the range of 1.38–10.64 %, and the average recovery of 18 compounds ranged 87.27 %–108.83 % (with RSD 0.96 %–13.84 %). The above data could prove that the UHPLC/Q-Trap-MRM method was suitable for the quantitative analysis of the target marker compounds in SSF.

Close

Fig. 7

The positive and negative MRM spectra for 18 analytes in SSF.

Export to PPT

The quantitative results of 18 compounds in different proportions of SSF decoctions are shown in Table S19 and Fig. 8. Taking PB1 (PF: MyS: SCF: EF = 4: 1: 2: 1), PB5 (4: 2: 2: 1), PB9 (4: 3: 2: 1), and PB13 (4: 4: 2: 1) clustered as group 1 as the example (Fig. 6), the effects on the content of PF (Psoraleae Fructus) components (coumarins and flavonoids) caused by changing the ratio of MyS (Myristicae Semen) were observed (Fig. 8A). The dosage increasing for MyS (from 1 g to 4 g) exerted large influence on the contents of psoralen and isopsoralen (as the free coumarins), but very weak effect on psoralenoside and isopsoralenoside (O-glycosidic coumarins). Most of the flavonoids, such as bavachin, corylifolinin, bavachinin, and corylifol A, were of the high content when the ratio of PF to MS was at 4: 3. In addition, taking PB1 (PF: MyS: SCF: EF = 4: 1: 2: 1) and PB4 (4: 1: 3: 1) located in different groups in Fig. 6 as the example (Fig. 8B), the effects on the content of PF components resulting from SCF changes were investigated. Generally, when the dosage of SCF increased from 2 g to 3 g, the content of PF coumarins showed slight variation. Additionally, the flavonoids including daidzin and bavachin showed decreasing, but the contents of corylifolinin, bavachinin, and corylifol A, significantly increased. Taking the PB 1 (PF: MyS: SCF: EF = 4: 1: 2: 1) and PB 2 (4: 1: 2: 2) as the example (Fig. 8C), the effects of changing EF dosage on the content of PF components were probed. Very weak influence was observed for the four characteristic coumarins, but a very significant increasing trend occurred for some flavonoids, such as bavachinin and corylifol A. However, by comparing PB5 (4: 2: 2: 1; the standard proportion for SSF) and PB6 (4: 2: 2: 2), corylifolinin, bavachinin, and corylifol A otherwise showed a descending trend.

Close

Fig. 8

Box diagrams of various compounds showing their contents in different proportions of SSF decoctions (A: component content variation for PF by changing the dosage of MyS; B: component content variation for PF by changing the dosage of SCF; C: component content variation for PF by changing the dosage of EF; D: component content variation for EF by changing the dosage of SCF; E: component content variation for EF by changing the dosage of MS; F and G: component content variation for SCF by changing the dosage of MyS or EF).

Export to PPT

Taking PB2 (PF: MyS: SCF: EF = 4: 1: 2: 2), PB6 (4: 2: 2: 2), PB10 (4: 3: 2: 2), and PB14 (4: 4: 2: 2) clustered in group 2 in Fig. 6 as the example (Fig. 8D), changing the dosages of MyS on the extraction of EF components showed an overall consistent variation trend, that is, sharply descended and then significantly increased. The contents for three toxic alkaloids (dehydroevodiamine, evodiamine, and rutaecarpine), two limonoids (rutaevin and evodin), and three phenolic components (chlorogenic acid, rutin, and hyperin), were of the lowest level for PB6. By comparing PB9 (4: 3: 2: 1) and PB12 (4: 3: 3: 1), increasing on the dosage of SCF benefited the extraction of three alkaloids and two limonoids, but the variation trend for three phenolic compounds was slight (Fig. 8E).

In addition, by comparing PB3 (4: 1: 3: 0.5), PB7 (4: 2: 3: 0.5), PB11 (4: 3: 3: 0.5), and PB15 (4: 4: 3: 0.5) clustered in group 3 in Fig. 7, the increasing of MyS could significantly affect the content of schisandrol A (a marker compound in SCF): firstly increasing and then decreasing (Fig. 8F). The comparison between PB3 (4: 1: 3: 0.5) and PB4 (4: 1: 3: 1) could indicate the increasing of EF slightly benefited the extraction of schisandrol A.

In fact, the dissolution of a compound is not only related to its physical and chemical properties and the solvent system, but also to the chemical microenvironment in which it is located. The change in the composition ratio of complex solutes can lead to the interaction of “solubilization”. In this work, we could primarily conclude that, changes in the dosages of MyS, SCF, and EF exhibited a relatively slight effects on the contents of four coumarins in PF. Increasing the ratio of MyS, SCF, and EF, could promote the dissolution of some flavonoids in PF, while increasing the SCF ratio promoted the dissolution of alkaloids in EF. By integrating the multicomponent characterization results of SSF, the lignins, as the main components of MyS and SCF showing weak acidity, might contribute to the enhancement of EF alkaloids dissolution.