Validation of Sennae Folium specification grade classification based on UPLC-Q-TOF/MS spectrum-effect relationship

2.1 Plant material and chemicals

A total of 13 bathes of SF (the dried leaves of C. angustifolia Vahl or C. acutifolia Delile) were acquired from various Chinese herbal medicine markets and authenticated by Dr. Dan Zhang, and all the samples were met the Chinese Pharmacopeia 2020 edition (Part I)(Commission, 2020). The specimens were stored at Hebei University of Chinese Medicine. All the leaf samples were classified into three groups according to their color: green leaf (GL, 13 bathes), yellow leaf (YL, 13 bathes), and diseased leaf (DL, 11 bathes). The details of each sample are listed in supporting information (SI) Table S1. The representing pictures of different grades samples are shown in Fig. 1.

Close

Fig. 1

Different grades of Sennae Folium. The green leaf (a), the yellow leaf (b), the diseased leaf (c).

Export to PPT

LC-MS grade methanol, acetonitrile, and formic acid were purchased from Fisher Scientific (Pittsburgh, PA, United States). Sennoside A (PRF9102403), sennoside B (PRF10021348), and rutin (PRF10040801) were purchased from Chengdu Bioprufy Phytochemicals Ltd (Chengdu, China). Vicenin-2 (PS012054), and isorhamnetin 3-O-β-gentiobioside (PS210813-03) were purchased from Chengdu Push Bio-technology Co., Ltd. (Chengdu, China). Ultrapure water was prepared by a Synergy water purification system (Millipore, Billerica, United States). Other chemicals and reagents were of analytical grade. Porcine pancreatic lipase (type II) and 4-methylumbelliferyl oleate (4-MUO) were purchased from Sigma-Aldrich (St Louis, MO, United States).

2.2 Sample preparation

An aliquot of 60 mg of sample powder was immersed in 1.5 mL of 47 % (V/V) methanol, followed ultrasonic extraction at room temperature for 15 min, and the extraction condition was determined by the by response surface methodology (RSM, SI Fig. S2, Table S2, Table S3). The mixture was then centrifuged at 13,000 rpm for 10 min. The extracts were diluted with 47 % methanol to 1 mg/mL, and caffeic acid as internal standard. The supernatant was filtered through a 0.22 µm nylon syringe filter (Agilent Technologies, Shanghai, China) and stored at 4 °C before the UPLC-Q-TOF/MS analysis.

2.3 UPLC-MS analysis

The UPLC-Q-TOF/MS analysis was performed on an Agilent 1290 Infinity II system coupled with an Agilent 6545 quadrupole time-of-flight mass spectrometer system (LC-Q-TOF-MS) (Agilent Technologies, Santa Clara, CA, United States) equipped with an electrospray ionization interface.

Chromatographic separation was performed on an Agilent ZORBAX Eclipse Plus C18 column (2.1 × 50 mm, 1.8 μm, Agilent Technologies, Santa Clara, CA, United States). The binary gradient elution system consisted of acetonitrile (B) and water containing 0.1 % formic acid (A). The separation was achieved at the flow rate of 0.4 mL/min using the following program: 0–5 min, 5 %–13 % B; 5–8 min, 13 %–15 % B; 8–12 min, 15 %–16 % B; 12–13 min, 16 %–18 % B; 13–15 min, 18 %–25 % B; 15–20 min, 25 %–29 % B. The sample injection volume was 0.5 µL, and the column temperature was set to 25 °C.

The MS acquisition parameters were as follows: drying gas (N₂) temperature, 320 °C; sheath gas temperature, 350 °C; drying gas (N₂) flow rate, 10.0 L/min; sheath gas flow (N₂) rate, 11 L/min; nebulizer gas pressure, 35 psi; capillary voltage, 3500 V; fragmentor voltage, 135 V; collision energy, 40 eV. The analysis was operated in a negative mode with the mass range of m/z 100–1,000 Da. Data were analyzed by MassHunter Qualitative Analysis Software Version B.10.00 (Agilent Technologies, Santa Clara, CA, United States). The quality control (QC) samples were prepared by pooling the same amount of SF together from all 37 samples and should be analyzed after each five samples.

2.4 Molecular networking

The molecular networking (MN) was constructed using the UPLC-Q-TOF MS/MS data from GL, YL, and DL. All MS/MS data files were converted into 32-bit mzXML by using ProteoWizard software (https://proteowizard.sourceforge.io/). The converted files were uploaded to the GNPS platform (https://gnps.ucsd.edu) via WinSCP (https://winscp.net) to construct a MN following the online workflow (Wang et al., 2016). The created MN and parameters can be accessed via the link: https://gnps.ucsd.edu/ProteoSAFe/status.jsp?task=3902109bbf36453ca44f4fa5a553c85b. The MN parameters were as follows: minimum cosine score 0.70; minimum matched peaks 6; tolerance 0.02 Da for parent mass and fragments; maximum connected component size 100; minimum cluster size 1, no run MScluster. The results were exported to be visualized on Cytoscape 3.8.2 software (La Jolla, CA, USA).

2.5 In vitro pancreatic lipase inhibitory activity

The PL inhibitory activity of the extracts were determined using a previously reported method with minor modifications (Zhang et al, 2015). The SF extracts (37 bathes), lipase, and 4-MUO were prepared in a Tris-HCl buffer solution (13 mM Tris-HCl, 150 mM NaCl, 1.3 mM CaCl₂, pH 8.0). Taken 1.10 mL of SF extracts and blown to dry with N₂, then added 1.10 mL of Tris-HCl and diluted to concentrations of 20.00, 16.00, 12.00, 8.00, and 4.00 mg/mL with Tris-HCl, respectively. An aliquot of 25 µL at different concentrations (20 mg/mL, 16 mg/mL, 12 mg/mL, 8 mg/mL, and 4 mg/mL) of SF sample solution and 25 µL of PL solution (1 mg/mL) were added into a black bottom 96-well plate. Afterward, 50 µL of 4-MUO was added into the mixture. After incubation at 37 °C for 20 min, the reaction was stopped by adding 100 µL of 0.1 M citrate buffer solution (pH 4.2). The background was prepared by replacing lipase with the same volume of Tris-HCl buffer. The reaction system without the tested SF samples was used as the control and the Tris-HCl buffer was used as the blank. All experiments were repeated in triplicate. The reaction solutions were then analyzed with PerkinElmer VICTOR Nivo Multimode Plate Reader (PerkinElmer Inc., Mountain View, CA, United States) with the excitation and emission wavelengths set at 355 nm and 460 nm. The inhibitory activities (%) of SF samples were calculated using the following formula and the functional strength of the inhibitor was measured by calculating the half maximal inhibitory concentration (IC₅₀) values. $$\mathrm{I}\mathrm{n}\mathrm{h}\mathrm{i}\mathrm{b}\mathrm{i}\mathrm{t}\mathrm{o}\mathrm{r}\mathrm{y}\mathrm{a}\mathrm{c}\mathrm{t}\mathrm{i}\mathrm{v}\mathrm{i}\mathrm{t}\mathrm{y}=(1-\frac{\mathrm{t}\mathrm{e}\mathrm{s}\mathrm{t}\mathrm{s}\mathrm{a}\mathrm{m}\mathrm{p}\mathrm{l}\mathrm{e}-\mathrm{b}\mathrm{a}\mathrm{c}\mathrm{k}\mathrm{g}\mathrm{r}\mathrm{o}\mathrm{u}\mathrm{n}\mathrm{d}\mathrm{s}\mathrm{a}\mathrm{m}\mathrm{p}\mathrm{l}\mathrm{e}}{\mathrm{c}\mathrm{o}\mathrm{n}\mathrm{t}\mathrm{r}\mathrm{o}\mathrm{l}\mathrm{s}\mathrm{a}\mathrm{m}\mathrm{p}\mathrm{l}\mathrm{e}-\mathrm{b}\mathrm{l}\mathrm{a}\mathrm{n}\mathrm{k}\mathrm{s}\mathrm{a}\mathrm{m}\mathrm{p}\mathrm{l}\mathrm{e}})\times 100\%$$

2.6 Mass spectrometry analysis and metabolites annotation

The LC-MS data acquisition was conducted on a MassHunter Workstation (Agilent Technologies, Santa Clara, CA, United States). The MS and MS/MS data files acquired from UPLC-Q-TOF/MS were analyzed using related published literature and metabolite databases. After normalization, the data set was introduced into SIMCA software (version 13.0, Umetrics, Umea, Sweden), for PCA and OPLS-DA and orthogonal projections to latent structures (OPLS). Mean ± standard deviation (SD), ANOVA, and person correlation analysis were performed in SPSS 22.0 statistical software (IBM Corp, Armonk, NY, United States). Heatmap analysis was generated by OriginPro 2019b software (Origin Lab Corporation, Northampton, MA, United States). The significance of differences in data between groups was determined by two-pair test, p < 0.05 was considered to be significantly different.

3.1 Optimal total yields of sennoside A and B from the single-factor experiment and RSM

As shown in Fig. S1-a, the total yield of sennoside A and B from SF increased with the methanol concentration. After the methanol concentration reached 50 %, the yield reached its maximum, but gradually declined as the concentration increased. Ultrasonic time had a parabolic effect on the total yields of sennoside A and B, with the maximum yield observed at 15 min (Fig. S1-b). The influence of the liquid/material ratio on the total yields of sennoside A and B is presented in Fig. S1-c, the maximum yields were obtained when the ratio was 25 mL/g.

RSM experiments were conducted with three parameters, ethanol concentration, ultrasonic time, and liquid/material ratio. The Model F-value was 3.70, and the p-value was 0.0491. These values indicated that the model was significant. As shown in Table S3, when comparing the linear and quadratic coefficients, the factors to consider are as follows: methanol concentration > ultrasonic time > liquid/material ratio. The mutual interaction effects between different factors were investigated using RSM. The combined effects of all the parameters on the extraction yields are shown as three-dimensional response surfaces and two-dimensional elliptical contour plots (Fig. S2). The interaction of parameters had significant effects on the yields of sennoside A and B.

The optimal conditions predicted using RSM by Design Expert software (Version 11.0; Stat-Ease Inc., Minneapolis, MN, USA) were as follows: methanol concentration, 47 %; ultrasonic power, 15 min; liquid/material ratio, 25 mL/g. Under these conditions, the total yields of sennoside A and B from SF was 1.345 %. These values were higher than those obtained under all other conditions. Three experiments were carried out to verify the reliability of the results obtained by RSM: methanol concentration of 47 %, sonication time of 15 min, and liquid/material ratio of 25 mL/g. The mean value of the total extraction of sennoside A and B was 1.334 % with a relative standard deviation of 0.76 %. Thus, the RSM model fitted well with the actual situation and the model was reliable.

3.2 UPLC-Q-TOF/MS profiling of Sennae Folium

Chemical profiling of different grade samples was executed via UPLC-Q-TOF/MS in the negative ionization mode. Elution gradient employed using formic acid in water (0.1 %): acetonitrile allowed for metabolites elution within 20 min. All the QC samples have good sensitivity, specificity, and reproducibility. Overlaid base peak chromatogram of the studied samples is depicted in Fig. 2 (a), and the total ion chromatograms of different grades of SF were displayed in Fig. 2 (b).

Close

Fig. 2

Overlaid base peak chromatograms of Sennae folium methanol extract obtained using UPLC-Q-TOF/MS in negative ionization mode (a); the total ion chromatogram（TIC）of different grades by negative scan (b), 1: Green Leaves; 2: Yellow Leaves; 3: Diseased Leaves.

Export to PPT

3.3 Molecular networking (MN) aided annotation of compounds in GL, YL, and DL

A spectral similarity network provides a means for the visual inspection of tandem mass spectrometry data, aiding in compound annotation alongside the observation of features distributed among the different samples (Wang et al., 2016). In the present study, a comprehensive MN of SF extracts based on their MS/MS spectral similarity was obtained, as shown in Fig. 3. The constructed MN was showed a total of 339 precursor ions visualized as nodes in the molecular map, which included of 26 clusters (nodes > 2) and 181 unique nodes (Fig. S3). As revealed in Fig. 3, four main clusters were generated by MN, including sennoside glycosides (cluster a), anthraquinones (cluster b), benzophenones (cluster c), and flavonoid glycosides (cluster d). All the above clusters were separated from each other by the MS/MS fragments.

Close

Fig. 3

Molecular networking of Sennae folium crude extract obtained using GNPS platform and visualized with Cytoscape 3.8.2 software. Cluster a: sennoside glycosides, cluster b: anthraquinones, cluster c: benzophenones, and cluster d: flavonoid glycosides.

Export to PPT

The chemical constituents from SF were identified by aligning their precursor mass, MS/MS fragments with standard substances (sennoside A, B, and C, Fig. S4), MN compound library, and previously reported literature. In total, 60 compounds were tentatively assigned, including 7 bianthrone glycosides, 4 anthrone glycosides, 7 anthraquinones, 8 benzophenones, 31 flavonoids, and 3 others (Table 1). It is worth mentioning that, due to the same processing method, the area of each pie node in each MN can represent the relative abundance of the SF samples of different grades. The relative abundance of sennosides and anthraquinones in the GL samples were higher than those in the YL and DL samples, while the relative abundance of flavonoids and benzophenones were lower than those in the YL, and the DL was the lowest.

3.3.1

3.3.1 Identification of individual anthraquinones

Anthraquinone compounds (including sennosides) are regarded as the active principles of SF and considered to be the main purgative components (Nesmerak et al., 2020). The MN revealed (Fig. 3) the presence of sennosides and bianthrone aglycones grouped in the cluster a, and distributed mainly in GL. In addition, several mono- and di-glycosides of anthraquinones (cluster b) were identified as derivatives of emodin, aloe-emodin, and rhein based on their MS/MS spectra. The characteristic C₁₀-C₁₀′ dimerization link cleavage was used for rapid identification of sennosides in SF extracts. Sennoside A/B and C/D are the major forms in different grades of blades. In detail, sennoside A and B have a very similar [M−H]⁻ ion at m/z 861 in peak 38, 25, both of them were showed daughter ions at m/z 699 and 537 for the loss of glucosyl moieties (Fig. 4a). It has been reported that the sennoside B (meso form) have the C₁₀-C₁₀′ bond which is easier to be cleaved in contrast to the chiral form, sennoside A (10S, 10′S or 10R, 10′R), where the C₁₀-C₁₀′ bond is more stable (Ye et al., 2007). Therefore, sennoside B was abundant in m/z 386 fragments. Like to sennoside A & B isomers, the ion chromatograms for m/z 847 corresponding to sennoside C, and D in peak 35, 29 showed similar molecular ions at m/z 685 [M−H−162] ⁻ and 523 [M−H−162−162] ⁻, while sennoside D have more abundant ion at m/z 386.

Close

Fig. 4

Proposed MS fragmentation pathway for the [M−H]- ion of sennoside A (a), quercetin 3-O-di-glucoside-6′''-feruliyl (b), cassiaphenone B-2′-glucoside (c).

Export to PPT

Sennidin aglycones (peaks 53, 54, 55, 57) were also most prevalent in GL samples. Whether the enrichment of bianthrone aglycones in these leaf samples were due to limited glucosyltransferase (GT) activity, or degradation of sennosides (primary glycosides) due to the high moisture content in leaves and the simultaneous existence of active glucosidases has yet to be clarified (Farag et al., 2015).

The predominant anthraquinones components were found to be emodin, aloe-emodin isomers, and rhein aglycone. The fragmentations of emodin aglycones (peak 17, 28) shows [M−H]⁻ ion at m/z 269, elimination of CO to produce m/z 241, following by the loss of one hydroxyl group to give m/z 225. The [M−H]⁻ ion of aloe-emodin (peak 14, 32), only produces one fragment at m/z 240 [M−H−CHO]⁻ in agreement with the previous report (Ye et al., 2007).

3.4 PCA and OPLS-DA analyses for the identification of discriminatory compounds

Further, multivariate statistical analysis was performed to verify the accuracy of the MN to identify chemical markers for the grades of GL, YL, and DL of SF. The metabolite differences of individual species were differentiated by employing pattern recognition methods, including an unsupervised model, PCA, and a supervised model, OPLS-DA. The PCA score spots of GL, YL, and DL were shown in Fig. 5a, which had significant differences. The OPLS-DA analysis indicated a more robust clustering discrimination between the three grades of SF (Fig. 5b). Three different clusters were generated, indicating that the components of these three grades are clearly different. These results provided an effective basis for the screening of the unique natural products of SF. Furthermore, the screening of differentially abundant compounds in SF was performed based on the variable importance in projection value (VIP > 1) and p-value (P < 0.05). The list of the chief 22 differential compounds with VIP > 1 is shown in Fig. 5c, Fig. S6, and Table 1. These 22 compounds (including 8 sennosides and anthraquinones, 11 flavonoids, and 3 benzophenones) were then visualized by a heat map and organized into a dendrogram. The results indicated that the relative abundance of sennosides and rhein-O-di-glucoside of the 22 differential compounds were decreasing while the flavonoids were increasing, and occurred when the grades turned from green, yellow to diseased (Fig. 5d).

Close

Fig. 5

PCA, OPLS-DA for different grades of Sennae folium. The PCA score plot (a); and the OPLS-DA model in negative ion mode (b); and S-plot (c); and the heat-map of differential metabolites identified in three grades of SF (d).

Export to PPT

3.5 Changes in pancreatic lipase inhibitory activity of different grades

PL is a major enzyme in lipid absorption, and it hydrolyzes triglyceride into free glycerol and fatty acids, and then promotes fat absorption. Therefore, inhibition of lipase activity can prevent obesity (Nakai et al., 2005; Feng et al., 2020). The lipase inhibitory activity of leaf sample extracts at the concentration from 4 to 20 mg/mL was investigated. The change in PL inhibitory activity is shown in Fig. 6a, and SI Table S5. The GL, YL, and DL extract all showed lipase inhibitory activities with the IC₅₀ values at 6.69 ± 0.15 mg/mL, 13.53 ± 0.15 mg/mL, and 14.60 ± 0.12 mg/mL, respectively. The result showed that the GL had the strongest lipase inhibitory activity, and the content of sennosides and anthraquinones in GL were the highest among the three grades of SF. As a result, it was concluded that the SF pancreatic lipase inhibitory activity was related to the content of some anthraquinones, and it was necessary to screen potential inhibitors of lipase.

Close

Fig. 6

Comparison of inhibitory activities of pancreatic lipase in different grades of Sennae folium (a, **p＜0.01), significance difference was assessed by two-pair test; pearson correlation results between potential quality markers in all samples (b).

Export to PPT

3.6 Correlation analysis among differential compounds and pancreatic lipase inhibitory activity of three SF groups

The correlation analysis between relative abundance and bioactivities was established by heat map and OPLS.

A heat map of correlation coefficients between relative abundance of differential compounds and PL inhibition is shown in Fig. 6b. In vitro, anthraquinones of SF had better PL inhibitory activity than flavonoids, indicating a significantly negative correlation between anthraquinones and IC₅₀. Several studies have shown that traditional Chinese medicine containing anthraquinones has a significant inhibitory effect on PL activity. Chang et al. studied the effects of Polygonum Multifloru in inhibiting pancreatic lipase, and found that anthraquinones have potential lipase inhibitory effects and can be used as a safe alternative to lipase inhibitors in the treatment of obesity (Chang et al., 2016). In addition, Kumar. et al. found that a bianthraquinone of C. siamea exhibited the most inhibitory activity, and some anthraquinones could also be considered as moderate enzyme inhibitors (Kumar et al., 2013). The above results suggest that the anti-obesity potential of SF is achieved through PL inhibition. The lipid-lowering compounds in SF extract are mainly sennosides and anthraquinones, which further substantiates that GL has the most effective lipid-lowering capabilities, and has the highest quality among the three grades of SF.

3.7 OPLS analysis between characteristic compounds and pancreatic lipase inhibitory activity

An OPLS analysis of correlation coefficients between relative abundance of differential compounds and PL inhibition is shown in Fig. 7. The OPLS analysis was performed on three groups of different grades and their in vitro pancreatic lipase inhibitory activity was recorded. In the OPLS score plot (Fig. 7), blue to red indicated that the inhibitory activity was from high to low, and the OPLS model showed no overfitting by 200 iterations of the permutation test (Fig. S7). The parameters of R²X and Q²Y in the model are 0.724, 0.787 (for lipase inhibitory activity), respectively, indicating these established OPLS models have satisfactory explanation ability. Following the application of these OPLS models, the regression coefficient was obtained, which could reflect the positive or negative contribution of each peak to the activity. As is shown in Fig. S8, peak 1, 2, 3, 4, 5, 6, 7, 8, 9, 20, 22, were negatively correlated to the IC₅₀, indicating that these compounds have the best inhibitory activity. Furthermore, the VIP was employed to screen the variables responsible for the bioactivity. Variables above the VIP-value threshold of 1.0 were filtered out as candidate bioactive compounds. Thus, we obtained 6 sennosides (Table 2) as significant differential bioactive metabolites with in vitro pancreatic lipase inhibitory activity by a coefficient and VIP ≥ 1.

Close

Fig. 7

OPLS linear regression of profile-efficacy analysis.

Export to PPT

3.8 In vitro validation of identified active components

In order to verify that the PL inhibitory activity of sennosides from Sennae Folium was superior to that of flavonoids, sennosides A, sennosides B, vincenin-2, rutin and isorhamnetin 3-O-β-gentianoside were selected for in vitro activity verification. The results are shown in Table 3, sennosides A and B were significantly better than vincenin-2, rutin and isorhamnetin 3-O-β-gentianoside in inhibiting PL activity. It further confirms the reasonableness of the specification grade classification of Sennae Folium.