Untargeted metabolomics analysis to unveil the chemical markers for the differentiation among three Gleditsia sinensis-derived herbal medicines by ultra-high performance liquid chromatography/quadrupole time-of-flight mass spectrometry

2.1 Chemicals and reagents

Twenty-six compounds were used as the reference standards in this work. Isoquercitrin (1), quercitrin (2), (-)-catechin hydrate (14), gallic acid (25), and vanillic acid (26), were from Sichuan WeiKeqi Biological Technology Co., Ltd. (Chengdu, China). Cynaroside (3), kaempferol (4), apigenin (6), dihydrokaempferol (7), taxifolin (8), licochalcone B (10), isoliquiritigenin (11), butein (12), epicatechin (13), vitexin (15), orientin (16), betulin (17), betulinic acid (18), echinocystic acid (20), daucosterol (21), and forsythin (22), were purchased from Chengdu Desite Biotechnology Co., Ltd. (Chengdu, China). Meletin (5), eriodictyol (9), caffeic acid (23), and ethyl gallate (24), were from Shanghai Standard Biotech Co., Ltd. (Shanghai, China). Lupenone (19) was from Chengdu Pusi Biotechnology Co., Ltd. (Chengdu, China). Their purity was ≥ 97%. Structures of these reference standards are exhibited in Fig. 1, and the detailed information is given in Table S1. Acetonitrile, methanol, and formic acid (Fisher, Fair lawn, NJ, USA), were all LC-MS grade. Ultra-pure water was in-house prepared using a Milli-Q Integral 5 water purification system (Millipore, Bedford, MA, USA). Information of the GSF/GFA/GS samples (45 batches in total) is available in Table S2. Their authentication was performed according to the Chinese Pharmacopoeia, and the specimens were deposited at the authors′ laboratory in Tianjin University of Traditional Chinese Medicine (Tianjin, China).

Close

Fig. 1

Chemical structures of the 26 reference compounds.

Export to PPT

2.2 Preparation of the test solutions

To each of the sample of G. sinensis, 200 mg of the accurately weighed powder was soaked in 8 mL methanol. Samples were extracted on a water bath at 40℃ with ultrasound assistance for 1 h and centrifuged at 3219 g (4000 revolutions per min, rpm) for 10 min. The supernatant was transferred into a 10-mL volumetric flask and diluted to the constant volume. After mixing, the 1 mL of the liquid was transferred into a 1.5-mL EP tube and further centrifuged at 11,481 g (14000 rpm) for 10 min. the resultant supernatant was used as the test solution for 45 batches of G. sinensis-derived TCM samples (with the final concentration at 20 mg/mL). A QC sample (QC₁), by mixing the equal volume of three representative samples (T-9/F-11/AF-10, Table S2), was used for the establishment of the LC-MS approach. Another QC sample (QC₂), by pooling the equal volume of all test solutions (45 batches), was prepared for monitoring the system stability over the whole analysis batch for chemical markers discovery by untargeted metabolomics.

To the 26 reference compounds, an aliquot of 1.0 mg was accurately weighed and dissolved in methanol yielding a solution at 1 mg/mL as the stock solution for each compound. Equal volume of each stock solution was mixed to prepare four mixed standard samples. After being centrifuged at 11,481 g (14000 rpm) for 10 min, the supernatant obtained was used as the test solutions for the reference compounds.

2.3 Uhplc/Qtof-MS^E

Metabolite profiling of G. sinensis was performed on a Waters ACQUITY UPLC^TM I-Class Plus system coupled to a Xevo G2-XS QTOF high-resolution mass spectrometer (Waters Corporation, Milford, MA, USA). A Waters ACQUITY UPLC® BEH C18 column (2.1 × 100 mm, 1.7 μm) maintained at 40 °C was used. The mobile phase consisted of 0.1% formic acid in H₂O (A) and acetonitrile (B), running in consistency with an optimal gradient program at a flow rate of 0.3 mL/min: 0–7 min: 5–25% (B); 7–8 min: 25–30% (B); 8–15 min: 30–34% (B); 15–16 min: 34–37% (B); 16–20 min: 37–40% (B); 20–20.5 min: 40–62% (B); 20.5–26 min: 62–80% (B); 26–28 min: 80–88% (B); 28–33 min: 88–95% (B); and 33–36 min: 95% (B). The injection volume was 3 μL. The QC₂ sample was analyzed after every six injections of G. sinensis samples for monitoring the system stability. Positive MS^E data were recorded under the ESI source parameters: capillary voltage, 3.0 kV; sampling cone voltage, 20 V; source offset voltage, 60 V; source temperature, 100 °C; desolvation temperature, 500 °C; cone gas (N₂) flow, 50 L/h; and desolvation (N₂) flow, 600 L/h. The QTOF mass analyzer scanned over 120–2400 Da to cover saponins under a low collision energy 6 V for full scan (MS¹) and a high-energy ramp 10–30 V for MS^E. Calibration of the high-resolution MS^E data was performed by reference to leucine enkephalin (Sigma-Aldrich, St. Louis, MO, USA; 1 µg/mL) at m/z 556.2771. Data acquisition was controlled by MassLynx V4.1 (Waters). Random injection was adopted, and a blank control sample and the QC₂ sample were injected after every six samples.

2.4 Multivariate statistical analysis

Raw MS^E data ranging between 0 and 33 min were pretreated using Progenesis QI v2.1 (Waters). The adduct forms, including [M + H]⁺, [M + Na]⁺, [M + NH₄]⁺, [M + K]⁺, and [2 M + H]⁺, were selected or self-edited. Menu-guide and batched processing finally generated a data matrix involving the information of t_R, m/z, and normalized peak area. The obtained metabolic features were filtered by 80% rule and 30% variation limit in turn. The remaining were used as the variates for chemometric analyses by PCA and OPLS-DA facilitated by SIMCA-P v14.1 (Umetrics, Umea, Sweden) after UV scaling (the centered data are divided by the standard deviation of the column variable). Holistic chemical difference among GSF, GFA, and GS was visualized, and the variables showing a VIP value > 8.0 were considered as the major potential markers.

2.5 Elaboration of an in-house library of G. Sinensis

To tackle the lack of specific database in charactering the metabolites from G. sinensis, we sought to create an in-house chemical library, by comprehensively outlining the phytochemistry researches on the species of G. sinensis. The in-house library elaboration was mainly composed by the following steps. 1) Literature searching: focus on the phytochemical research and review of G. sinensis through the public database (e.g. Web of Science, CNKI, and ScienceDirect, etc.). 2) Formulating an Excel table and each structure file: prepare an Excel table with the structure information of each known compound entered and the structure file using the ChemDraw software saved as a single .mol file (name the .mol file consistent with the name of each compound recorded in the Excel table). 3) Import of both the Excel file and all the structure files into the UNIFI™ software.

2.6 Identification of two important marker compounds (saikachinoside A and locustoside A) by LC/MS-based isolation and NMR

To enable more convincing identification for the discovered chemical markers, LC-MS guiding and NMR were utilized to identify representative chemical markers, taking two compounds (M2#: t_R, 2.14 min, m/z 398.1675 for [M + H]⁺; M27#: t_R, 4.56 min, m/z 382.1724 for [M + H]⁺) as the examples. Briefly, the pulverized drug material of Gleditsiae Sinensis Fructus (2.0 kg) was ultrasonically extracted with 75% ethanol twice (v/v, 1 h for each time), and further filtered, yielding the dry extract (582.5 g) after removing the solvent. The crude extract was reconstituted with water and loaded onto a D101 macroporous adsorption resin column, which was sequentially eluted with water, 20%, 40%, 60%, 80%, and 100% ethanol (v/v). Thin layer chromatography (TLC) monitored the combination of elutes, and the combined fractions were further analyzed by LC coupled with an Agilent 6125B quadrupole mass spectrometer configured with an InfinityLab Poroshell 120 EC-C18 column (3.0 × 100 mm, 2.7 µm). The fraction Fr. 3 containing the target compounds was further separated by the MCI column, eluted with gradient ethanol–water (5%, 10%, 15%, 20%, 25%, 30%, and 35% ethanol, v/v). According to TLC examination, 16 MCI subfractions (MFr. 3–1 to 3–16) were obtained. Two samples, MFr. 3–3 (containing M2#) and MFr. 3–7 (containing M27#), were locked by HPLC-MS analysis, which were finally purified by semi-preparative HPLC using a Kinetex® EVO C18 100 Å column (250 × 10.0 mm, 5 µm; mobile phase A: 0.1% ammonia-water, B: acetonitrile, 2.5 mL/min) to obtain compounds M2# (saikachinoside A, 11.0 mg, purity: 98%) and M27# (locustoside A, 13.0 mg, purity: 98%). Chemical structures for these two markers were established by NMR analyses (¹H NMR, ¹³C NMR, and DEPT; M2# in DMSO‑d₆ at 500 MHz; M27# in a mixture of CD₃OD:D₂O = 1:9 at 600 MHz).

3.1 Optimization of a reversed-phase UHPLC/QTOF-MS^E approach dedicated to resolving the multicomponents simultaneously from GS, GSF, and GFA

With the view of developing a UHPLC/QTOF-MS^E method that enables global profiling of the components simultaneously from GS/GSF/GFA, key parameters affecting the UHPLC separation (e.g. stationary phase, mobile phase, column temperature, and gradient elution program) and QTOF-MS detection (capillary voltage and cone voltage) were optimized. And as the first step, the extraction solvent on the components transformation from was evaluated by testing different ratios of methanol and H₂O (pure water, 30% methanol, 50% methanol, 70% methanol, and 100% methanol). The base-peak chromatograms (BPCs) could indicate desirable coverage on both the polar and less-polar components by 50% aqueous methanol (Fig. S1), which was thereby utilized for the sample preparation.

Chromatographic column is a key factor affecting the chromatographic separation due to their differential selectivity and retaining capacity (Feng et al., 2021; Fu et al., 2019; Li et al., 2021). In this section, ten columns, which were purchased from different vendors (Waters, Agilent, and Phenomenex) and involved different silica gel core and functional groups, such as BEH C18, CSH Fluoro-Phenyl, HSS T3, BEH Shield RP18, CSH Phenyl-Hexyl, Cortecs UPLC T3, CSH C18, Zorbax SB-C18, Zorbax Eclipse Plus C18, and Kinetex XB-C18, were examined by observing the overall resolution of the components separated from QC₁. Evidently, the BEH C18 (13377), HSS T3 (13275), and Zorbax Eclipse Plus C18 (12810) columns showed better selectivity than the other seven, and BEH C18 thereof was able to resolve the most peaks (Fig. 2). Moreover, the components separated by BEH C18 showed stronger response than HSS T3 and gave better peak shape than Zorbax Eclipse Plus C18. Notably, the BEH C18 column has ethylene bridge hybrid particles bonded with C18, and is suitable for the retention of medium or weak polar compounds with wide pH tolerance range. HSS T3 is a silica-based C18 column with triple bond bonding to reduce the carbon density, which can enhance the retention of polar molecules and is compatible with 100% aqueous phase. Zorbax Eclipse Plus C18 is spherical fully porous silica particle with double-end capped octadecylsilane chemically bonded phase, suitable for the analysis of acidic and neutral samples, and able to give improved peak shape for the basic compounds. Taking into account of the peak shape and selectivity, we selected the BEH C18 column in the subsequent experiments. Moreover, alternation of column temperature (25–40 °C) on the separation was evaluated, but very minor influence was witnessed (Fig. S2). Comparatively, the setting at 40 °C enabled the separation of the most peaks, which was thus chosen. The gradient elution program was carefully adjusted, which finally achieved the desirable resolution of the major peaks.

Close

Fig. 2

Method development for the reversed-phase UHPLC/QTOF-MS^E approach. A-Selection of the stationary phase for the UHPLC separation by evaluating the base peak chromatograms (in the negative ESI mode) of QC₁ on ten different sub-2 µm particles packed chromatographic columns; B-optimization of key parameters (capillary voltage and cone voltage) of the Waters Xevo G2-XS QTOF mass spectrometer in both the negative and positive ESI modes. In ESI-: 1: 21.31 min, m/z 1931.9424; 2: 28.34 min, m/z 383.1501; 3: 6.97 min, m/z 579.2084; 4: 4.31 min, m/z 459.1297; 5: 8.73 min, m/z 1615.7386; 6: 5.35 min, m/z 741.2612; 7: 6.51 min, 331.0824. In ESI+: 1: 21.31 min, m/z 1933.9568; 5: 8.69 min, m/z 1617.7530; 7: 6.48 min, m/z 333.0968; 8: locustoside A; 9: 8.50 min, m/z 1779.8058; 10: 21.41 min, m/z 1803.8938; 11: 12.81 min, m/z 1981.9415.

Export to PPT

Capillary voltage and cone voltage are two key source parameters for the Xevo G2-XS QTOF mass spectrometer, of which the former determines the ionization efficiency and the latter affects the adduct ions and can cause the in-source fragmentation. The effects of capillary voltage (1.0–3.5 kV) and cone voltage (20–120 V) were assessed by comparing the ion response of the indicator components (negative mode: 1–7; positive mode: 1, 5, 7, and 8–11). In the positive mode, the alternations in the ion intensity showed compound-dependent characteristics while capillary voltage increased, and generally the setting between 2.0 and 3.0 kV could give rise to high response for most indicators. We thus chose to set the capillary voltage at 3.0 kV. Different variation trends were observed for the indicators when increasing cone voltage from 20 to 120 V, for which the response intensity negatively correlated to the ascended cone voltage. Therefore, the cone voltage of 20 V was selected. In the negative mode, the increasing of capillary voltage could favor the obtaining of high response for most of the indicator compounds, which prompted to select 3.0 kV as the negative capillary voltage. Compound-dependent intensity variation was also detected when increasing the cone voltage, and the setting of 60 V could be a good choice to balance all the indicator components. These optimal settings were adopted in the following untargeted metabolomics experiments.

3.2 Application of UHPLC/QTOF-MS^E-based untargeted metabolomics to unveil the potential markers among GS, GSF, and GFA

Metabolomics can measure the global variations and changes of the small-molecule metabolites among different groups, which the untargeted mode more suits the discovery of potential marker compounds (Li et al., 2021; Shi et al., 2018; Wang et al., 2020). Global difference in the chemical compositions (mainly covering the secondary metabolites) among three G. sinensis-derived TCMs (GS/GSF/GFA) was probed following the untargeted metabolomics workflows.

Firstly, we assessed the ionization complexity of G. sinensis components between the negative and positive ESI modes. It was clear that, the ion species for the precursors in the negative mode was rather complex than those in the positive mode (Fig. S3). By ESI+, only the protonated molecules ([M + H]⁺) were generated for the saponin and flavonoid compounds. However, diverse doubly charged deprotonated ions of dimers-FA adducts and the monomers-FA adducts were observed in the negative mode for the saponin, while the deprotonated molecules, dimers, and even trimers, were readily generated for the flavonoid. To obtain simple MS¹ information directly reflecting the contents, we applied UHPLC/QTOF-MS^E in the positive mode to monitor the chemical compositions (120–2400 Da) of 45 batches of G. sinensis samples (Table S2). Secondly, the raw MS^E data were pretreated using the informatics tool, Progenesis QI, to perform the data correction, peak alignment, and peak extraction, which could generate a data matrix consisting of totally 48,011 metabolic features (t_R VS m/z). The metabolic features were further filtered by several criteria to simplify the marker discovery (80% rule and 30% variation limit). Thirdly, the resultant 33,912 metabolic features were used as the variables for chemometric analysis by PCA (unsupervised) and OPLS-DA (supervised), to deduce the potential marker components among GSF/GFA/GS.

PCA score plot displayed very tight clustering of the QC data, indicating good quality for the obtained multi-batch metabolomics data (Fig. S4). In addition, the GS group was remarkably separated from the others, while some samples of GSF and GFA were mixed, which informed the large chemical difference of GS from the other two. Base peak chromatograms of representative GSF (F-11), GFA (AF-10), and GS (T-9) samples, could intuitively embody the metabolomic difference among G. sinensis-derived three TCMs (Fig. 3). An outlier (F-6, from Shandong) was observed. Supervised OPLS-DA (Fig. 4A and 4B) and VIP values (Fig. 4C) were further utilized to exploit the potential marker components. As a result, complete separation among the GS, GSF, and GFA groups was accomplished in the score plot (Fig. 4A), while GFA and GSF that showed less separation could be completely grouped into two clusters (Fig. 4B). A total of 4596 ions were found displaying VIP > 1.0. To screen the major differential components, in the current work, we selected to analyze those with VIP > 8.0, and consequently, 70 differential ions among GSF/GFA/GS were obtained (Fig. 4C). Accordingly, 46 differential components were identified or tentatively characterized based on the integrated analysis of the positive and negative mode MS^E data, and the searching of the in-house library (Table S3) and the literature. Detained information for these 46 potential markers is given in Table S4. High sensitivity and specificity were observed for these discovered differential ions, as the AUC reached 1.0 for the ROC curve (Fig. 4D). The heat map directly reflected the variations of the contents for these 46 significantly different components and the clustering of GSF, GFA, and GS, among the 45 batches of G. sinensis samples (Fig. 4E).

Close

Fig. 3

Base peak chromatograms (BPCs) of three herbal medicines derived from G. sinensis L.

Export to PPT

Close

Fig. 4

Multivariate statistical analysis of the positive MS^E data for 45 batches of G. sinensis samples. A: OPLS-DA score plot of GSF/GFA/GS (GSF: Gleditsiae Sinensis Fructus, GFA: Gleditsiae Fructus Abnormalis, and GS: Gleditsiae Spina); B: OPLS-DA score plot of GSF/GFA; C: VIP plot with the cutoff at 8.0; D: ROC curve; E: Heat map displaying the content differences of 46 potential markers.

Export to PPT

3.3 Structural elucidation of the discovered 46 differential components by automatic peak annotation workflows facilitated by UNIFI and the phytochemical isolation

The bioinformatics software platform UNIFI (Waters), by searching commercial or in-house library, can facilitate automatic peak annotation rendering highly reproducible characterization results independent of the professional skills (Wang et al., 2020; Wang et al., 2021). We applied the well-established UNIFI-based computational workflows to identify the potential marker components discovered by untargeted metabolomics (Zuo et al., 2020). Because of the lack of commercial database specific for G. sinensis, we sought to establish an in-house library of G. sinensis by searching the phytochemistry researches and summarizing the known compounds. Consequently, the chemical library contained the names, molecular formulae, and chemical structures of 203 known components (including 44 triterpenoid saponins, 14 triterpenoids, 10 alkaloids, 36 flavonoids, 15 flavonoid glycosides, 25 lignans, 18 phenols, 12 phenylpropanoids, 5 sterols, 4 amines, 6 coumarins, and 14 others). After processing by UNIFI, the components characterized from three G. sinensis samples were listed in a table, namely “Identified Components”. By comparing the retention time and high-accuracy MS¹ data of these 70 differential ions with those in the “Identified Components” table, we could characterize 46 potential marker components (including 30 saponins, 2 terpenoids, 2 flavonoids, 4 alkaloids, and 8 others; Table S4). Adduct ions filtering and fragmentation pathways analysis based on the MS² data analysis were performed to remove the possible false positive characterization results and confirm the ultimate identities.

3.3.1

3.3.1 Characteristic components (M2#/M7#/M27#) for GSF

The box charts of 46 characterized potential markers among GSF, GFA, and GS, are given in Fig. S5, which were utilized to screen the characteristic components that could be used as the markers for their differentiation. As a result, three key markers, including M2#, M7#, and M27#, were found characteristic for GSF, with the extracted ion chromatograms (EICs) of representative GS/GSF/GFA samples shown in Fig. 5. Evidently, they were relatively rich in GSF, but very rare in the other two species (GFA/GS). To enable the reliable structural elucidation, M2# and M27# were isolated from GSF guided by LC-MS, and their structures were established by both MS and NMR analyses.

Close

Fig. 5

Box plots for the markers of GSF (M2#/M7#/M27#) and GS (M8#/M20#/M21#/M34#), and the corresponding extracted ion chromatograms (EICs) of three typical G. sinensis samples. M46 was used as the internal standard to calculate the peak area ratio.

Export to PPT

M27# (t_R: 4.56 min) exhibited rich molecular ions at m/z 382.1724 for [M + H]⁺ and 380.1570 for [M−H]^–, based on which the molecular formula was established as C₁₆H₂₃N₅O₆. The UV (MeOH; 1.0 mg/mL) spectrum showed strong absorption peaks at 246 and 290 nm, suggesting the presence of a 3,7-disubstituted isoguanine skeleton (Kajimoto et al., 2010a), which could be supported by the ¹H and ¹³C NMR data [δ_H 8.14 (s, 1H, H-8), δ_C 158.1 (C-2), 153.9 (C-4), 104.8 (C-5), 154.9 (C-6), 145.2 (C-8)]. Analysis of the ¹H and ¹³C NMR data (Table S5), together with the DEPT spectrum, revealed the presence of two olefinic methyls [the C-4′ hydrogen signal at δ_H 1.68 (s, 3H) and the C-5′ hydrogen signal at δ_H 1.77 (s, 3H)], two hetero atom bearing methylenes [δ_C 42.8 (C-1′) and 88.1 (C-1″)], five oxygenated methines [δ_C 73.5 (C-2″), 76.7 (C-3″), 69.2 (C-4″), 80.1 (C-5″), 60.1 (C-6″)], an olefinic methine [the C-2′ hydrogen signal at δ_H 5.14 (t, J = 7.2 Hz, 1H)], and a quaternary olefinic carbon [δ_C 139.2 (C-3′)], in addition to the isoguanine unit. The presence of an isopentenyl unit could be testified by the transitions of m/z 382 ([M + H]⁺) → 314.1099 ([M + H–C₅H₈]⁺) and m/z 218 ([M−H−Glc]^–) → 150.0420 ([M−H−Glc−C₅H₈]^–) in the positive and negative MS/MS spectra, respectively (Fig. 6). What’s more, the presence of a hexose unit which was supported by the transition of m/z 314 ([M + H–C₅H₈]⁺) → 152.0570 ([M + H–C₅H₈–C₆H₁₀O₅]⁺) in ESI + and m/z 380 ([M−H]^–) → 218.1043 ([M−H−C₆H₁₀O₅]^–) in ESI–. By comparing the NMR data with the literature (Kajimoto et al., 2010a), M27# was identified as locustoside A (corresponding to compound 167 in the in-house library; Table S3). The similar approaches were utilized to establish the chemical structure of M2# (t_R: 2.14 min). Its molecular formula was established as C₁₆H₂₃N₅O₇, based on the predominant precursor ions detected at m/z 398.1675 ([M + H]⁺) and 396.1517 ([M−H]^–). The presence of an isoguanine unit was supported by the ¹H and ¹³C NMR (DMSO‑d₆) chemical shifts [δ_H 8.09 (s, 1H, H-8), δ_C 154.9 (C-2), 153.0 (C-4), 102.6 (C-5), 153.5 (C-6), 142.9 (C-8)]. The containing of a hexose could be informed by the transition of m/z 398 ([M + H]⁺) → 236.1144 ([M + H–C₆H₁₀O₅]⁺) and m/z 396 ([M−H]^–) → 234.0994 ([M−H−C₆H₁₀O₅]^–) in the positive and negative MS/MS spectra, respectively (Fig. 6), and further confirmed as β-D-glucose because of the signal δ_H 5.42 (d, J = 8.5 Hz, 1H; H-1″), δ_H 3.33–3.66 showing 6 signals of hydrogen on the sugar, and δ_C 86.4 (C-1″) together with the other 5 carbon signals [δ_C 72.5 (C-2″), 76.3 (C-3″), 68.0 (C-4″), 79.3 (C-5″), 58.9 (C-6″)]. Similar to M27#, a rich fragment at m/z 314.1100 was observed, which should be dissociated from the protonated precursors by eliminating 84 Da ([M + H–C₅H₈O]⁺). While the NMR data, [δ_H 4.59 (H-1′), 5.26 (H-2′), 4.11 (H-4′), 1.68 (H-5′); δ_C 121.0 (C-2′), 139.4 (C-3′), 59.9 (C-4′), 21.3 (C-5′)], could testify the presence of a hydroxylated isopentyl, consistent with an additional mass of 84.0575 Da (C₅H₈O). The comparison with the literature could assist to identify M2# as saikachinoside A, consistent with compound 166 in the in-house library (Table S3; Kajimoto et al., 2010b).

Close

Fig. 6

Fragmentation pathways analysis for the characterization of three marker compounds (M2#/M27#/M7#) for GSF.

Export to PPT

M7# (t_R: 14.35 min) was characterized as a furanose analogue of M27#. The protonated precursor ion at m/z 514.2150 ([M + H]⁺) and 512.1997 ([M−H]^–) could suggest the molecular formula C₂₁H₃₁N₅O₁₀, indicative of one pentose more than M27#. By the MS/MS fragmentation in the positive mode, the typical product ions of m/z 446.1522 consistent with the elimination of the isopentane group ([M + H–C₅H₈]⁺), m/z 314.1098 because of the combinational loss of pentose + C₅H₈ ([M + H–C₅H₈–pentose]⁺), m/z 220.1190 by losing all the sugars ([M + H–pentose–Glc]⁺), as well as m/z 152.0570, were observed (Fig. 6). The alkaloid skeleton fragment at m/z 218.1044 was detected by the negative MS/MS fragmentation. By comparison with the literature (Harauchi et al., 2017), M7# was tentatively characterized as locustoside B (in accord with compound 171 in the in-house library, Table S3).

3.3.2

3.3.2 Characteristic components (M8#/M20#/M21#/M34#) for GS

In the similar manner, by analyzing the box charts, four components, including M8#, M20#, M21#, and M34#, were found characteristic for GS. It was clear, they were of certain content in GS, but almost undetectable in GFA and GSF (Fig. 5). By comparison with the reference compounds, M21# (5.97 min, C₁₅H₁₂O₇) was identified as taxifolin (a flavonone compound; Table S4). Structural elucidation of the other three components (M8#/M20#/M34#), characteristic for GS, was conducted based on the MS data analysis.

M20# (t_R: 30.55 min) gave rich precursor ions of m/z 457.3678 ([M + H]⁺)/479.3497 ([M + Na]⁺) by ESI+, and m/z 455.3515 ([M−H]^–)/501.3572 ([M−H + HCOOH]^–) by ESI–, based on which its molecular formula was established as C₃₀H₄₈O₃. The positive-mode MS/MS fragmentation yielded the fragments of m/z 411.3619 by losing formic acid ([M + H–HCOOH]⁺), together with a series of fragments due to ring B cleavage (m/z 301.2163, 243.2109, and 137.1324), while the negative dissociation could generate the fragments at m/z 327.2897, 255.2325, and 237.2199, ascribed to the fragments because of ring cleavages (Fig. 7). In this experiment, we compared the retention time of M20# with those of the reference compounds, betulinic acid (25.78 min), oleanolic acid (26.24 min), and ursolic acid (26.36 min), but matched none of them. Accordingly, M20# was tentatively identified as an isomer of oleanolic acid (Table S4). M8# (t_R: 30.77 min) exhibited the protonated molecular ion at m/z 439.3573, and accordingly, its molecular formula was calculated as C₃₀H₄₆O₂. The MS/MS spectrum gave fragments at m/z 421.3459 ([M + H–H₂O]⁺), 327.2684 ([M + H–C₈H₁₆]⁺), 233.1899 ([M + H–C₁₅H₂₆]⁺), and 191.1794 ([M + H–C₁₈H₃₂]⁺) (Fig. 7). The deduced molecular formula (C₃₀H₄₆O₂) received no hit in the in-house library of G. sinensis, but was consistent with 2-methyl-7-(2-methyl-2-propanyl)-2-[(3E,7E)-4,8,12-trimethyl-3,7,11-tridecatrien-1-yl]-6-chromanol in the ChemSpider database. M8# was thus inferred to be this compound or its isomer. The retention time of M34# was observed at 28.70 min, and its molecular formula was deduced as C₃₀H₄₈O₂, based on the positive protonated precursor ion at m/z 441.3729. By the MS/MS fragmentation, the fragments of m/z 423.3623 ([M + H–H₂O]⁺) and 411.3622 ([M + H–HCOH]⁺) were generated. In addition, the fragments resulting from the cleavage of ring C were observed at m/z 233.1899 and 203.1796. By molecular formula searching, none was matched in the in-house library. Assisted by further searching the ChemSpider database, M34# was tentatively characterized as isomer of betulinaldehyde (Bhattacharyya et al., 1976).

Close

Fig. 7

Fragmentation pathways analysis for the characterization of three marker compounds (M20#/M34#/M8#) for GS.

Export to PPT

3.3.3

3.3.3 Characterization of triterpenoid saponins

Saponins are the most abundant bioactive ingredients occurring to G. sinensis, most of which have an oleanane or lupine skeleton (Zhang et al., 2016). The oleanane-type sapogenins mainly involve oleanolic acid (OA; C₃₀H₄₈O₃) and 16α-hydroxyl oleanolic acid (echinocystic acid; EA; C₃₀H₄₈O₄). The monosaccharide composing the saponins include glucose (Glc, 162.0528 Da), arabinose (Ara, 132.0423 Da), xylose (Xyl, 132.0423 Da), and rhamnose (Rha, 146.0579 Da), while some saponins are esterified with 6(R)-6-hydroxy-2,6-dimethyl-2,7-octadienolic acid (Ter, additional C₁₀H₁₆O₃, 184.1099 Da) or 6(R)-6-hydroxy-2-hydroxymethyl-6-methyl-2,7-octadienolic acid (Ter-OH, additional C₁₀H₁₆O₄, 200.1049 Da). In both the OA- and EA-type saponins, they could be ionized by the positive and the negative ESI modes. What’s different, the precursor ions by ESI– were very complex (e.g. [M−H]^–, [M−H + HCOOH]^–, and [M−2H]^2–, etc), however, in the positive mode, only the [M + H]⁺ ions were readily observed. As previously reported (Wang et al., 2016), The deduced fragmentation rules mainly involve: 1) the characteristic fragmentations of G. sinensis saponins were the B/Y cleavage in the α chain (glycosylated to 28-COOH), easily generating the [Y_0α]^– in ESI– and [B_α]⁺ in ESI+; 2) the product ions ranging m/z 700–900 were useful for identifying the sapogenins. These reported features, elemental composition analysis based on the high-resolution MS¹ data, and searching of an in-house library of G. sinensis, were utilized for structural elucidation of the potential saponin markers. For the convenient expression, Xyl was used to represent all the pentose residues characterized by 132.0423 Da (Table S4).

Structural elucidation of M3# was taken as an example for identifying the EA-type saponin markers. It gave abundant precursor ions at m/z 1965.9487 ([M + H]⁺) by ESI + and m/z 2009.9390 ([M−H + HCOOH]^–) by ESI–, which could inform the molecular formula as C₉₄H₁₄₈O₄₃. The MS/MS fragmentation of the [M + H]⁺ precursors yielded a series of sugar-eliminating product ions at m/z 1701.8638 for [M + H–2Xyl]⁺, m/z 1423.7630 for [M + H–3Xyl–Rha]⁺, and m/z 1291.7215 for [M + H–4Xyl–Rha]⁺ (Fig. 8). The fragment of m/z 1067.4556 was the characteristic fragment of [B_α + H]⁺, accompanied by m/z 935.4136 due to the further elimination of Xyl ([B_α + H–Xyl]⁺). What’s more, the fragment of m/z 477.2491 ([Rha+(Ter-OH) + Ter + H–H₂O]⁺) could be a fragment of [B_α + H]⁺, while m/z 455.3527 could be assigned as [EA + H–H₂O]⁺. By the negative-mode MS/MS, the deprotonated molecular ion of m/z 1964.9353 could continuously eliminate H₂O (18 Da), HCOH (30 Da), C₁₀H₁₄O₃ (182 Da), and C₁₀H₁₆O₄ (200 Da), yielding the fragments at m/z 1946.9255, 1934.9246, 1782.8400, and 1764.8296, respectively, which were representative of the cleavages of the Ter-OH group. The combined loss of the Ter group generated a fragment at m/z 1598.7300. Notably, the medium-intensity ion of m/z 897.4852 was the characteristic fragment of [Y_0α–H]^–, accompanied by m/z 879.4745 after eliminating H₂O. A fragment at m/z 469.1558 was assigned as [EA–3H]^– obtained by eliminating both two saccharide chains. Based on these MS evidences, M3# was tentatively characterized as EA-2Glc-2Rha-4Xyl-(Ter-OH)-Ter, presumably as Caspicaoside J/ Caspicaoside K/Gleditsia saponin C/Gleditsioside E or their isomer (Table S4; Melek et al., 2014; Zhang et al., 1999b). Analogously, the other 17 markers (M4#, M5#, M9#, M10#, M11#, M16#, M17#, M18#, M23#, M26#, M31#, M33#, M36#, M38#, M41#, M44#, and M45#) were identified as the EA-type saponins.

Close

Fig. 8

Fragmentation pathways analysis for the characterization of the representative EA- (M3#) and OA-type (M15#) saponin compounds and a flavonoid compound (M24#).

Export to PPT

The high-accuracy MS^E data of M15# (t_R: 19.79 min) in both the positive and negative modes were analyzed as an example to illustrate the characterization of the OA-type saponin markers. Rich precursor ions at m/z 1621.7996 for [M + H]⁺ in the positive mode and m/z 1619.7795 for [M−H]^–, 1665.7855 for [M−H + HCOOH]^– in the negative mode, could aid to deduce the molecular formula as C₇₈H₁₂₄O₃₅. Upon the positive MS/MS fragmentation, the fragments characteristic for the neutral losses of sugars were observed at m/z 1489.7575 for [M + H–Xyl]⁺ and 1061.6041 for [M + H–3Xyl–H₂O–Rha]⁺. In addition, the diagnostic [B_α + H]⁺ fragment was detected at m/z 739.3019, together with its secondary fragments at m/z 589.2495 and 311.1495. Different from the EA-type, the entire sapogenin ion of OA could be generated at m/z 457.2073 ([OA + H]⁺), as well as its fragment of m/z 439.3575. The similar fragmentation pathways to those EA-type were observed in the negative mode, featured by the neutral loss of sugars and the Ter group (m/z 1453.6858 and 1435.6753). The characteristic [Y_0α–H]^– fragment at m/z 881.4904, together with its fragment of m/z 749.4476, could inform the β-sugar chain composed by Glc-Xyl-Xyl. By searching the in-house library, we tentatively characterized M15# as gleditsioside A or isomer (OA-4Xyl-2Rha-2Glc-Ter) (Fig. 8; Zhang et al., 1999a). In the similar manner, the other 11 markers (M6#, M12#, M13#, M14#, M19#, M22#, M25#, M28#, M30#, M32#, and M39#) were identified as the OA-type saponins (Table S4).