2.1 Chemicals and reagents

A total of 62 ginsenosides (HPLC-UV purity: >95 %; chemical structures shown in Fig. S1, detailed information given in Table S1), purchased from Standard Biotech. Co., Ltd. (Shanghai, China), were used as the reference compounds. Acetonitrile, methanol, and formic acid (FA; Fisher, Fair lawn, NJ, USA) were of the LC-MS grade. Ultra-pure water was purified using a Milli-Q Integral 5 water purification system (Millipore, Bedford, MA, USA). Indium oxide-coated glass slides were purchased from Bruker Daltonics (Bremen, Germany). Crude drug materials of PG, PQ, and PN were purchased from Yucuitang Flagship Store (Tonghua, Jilin), Caowangshenye Flagship Store (Tonghua, Jilin), and Jiaxiangwei Flagship Store (Wenshan, Yunnan), in January 2021. All chemicals, reagents and samples analyzed in this work were deposited in the State Key Laboratory of Component-based Chinese Medicine (Tianjin, China).

2.2 Sample preparation

The samples analyzed in this work include the raw materials of PG, PQ, and PN (named as 0 h), and their processed products prepared at different time points (1 h, 2 h, 4 h, 6 h, 8 h, and 10 h). Detailed information regarding the steaming process is provided in the Supporting Information. The accurately weighed powder of each sample (50 mg) was dispersed in 10 mL of 70 % (v/v) methanol, and further extracted with the aid of ultrasound (400 W, 40 kHz, 40 °C) for 60 min. After being centrifuged at 3,219 × g (4,000 rpm) for 10 min, the supernatant was taken and further diluted to a constant volume with 70 % aqueous methanol in a 10-mL volumetric flask, as the stock solution (5 mg/mL). In addition, all test solutions were mixed in the equal volume (105 batches of samples) to prepare a QC sample, used to optimize the analytical condition and monitor the instrument stability. Additionally, for the multi-component characterization purpose, all the test solutions within the same ginseng species were separately mixed in the equal volumes, which were centrifuged at 11,481 × g (14,000 rpm) for 10 min to give the test solutions for PG, PQ, and PN.

2.3 UHPLC/IM-QTOF-MS

The LC-MS analysis was conducted on an ACQUITY UPLC I-Class/Vion IMS-QTOF system operating under UNIFI control (Waters, Milford, MA, USA). A CSH C18 column (2.1 × 100 mm, 1.8 µm) hyphenated with a VanGuard Pre-column (2.1 × 50 mm, 1.7 µm) maintained at 30 °C was used for the chromatographic separation. A binary mobile phase, containing 0.1 % FA in H₂O (A) and CH₃CN (B), was employed at a flow rate of 0.3 mL/min following an optimal gradient elution program: 0–2 min, 15–22 % (B); 2–4 min, 22 % (B); 4–9 min, 22–30 % (B); 9–10 min, 30–32 % (B); 10–22 min, 32 % (B); 22–25 min, 32–34 % (B); 25–39 min, 34–60 % (B); 39–41 min, 60–95 % (B); and 41–44 min, 95 % (B).

To enable the highly sensitive detection of ginsenosides in the negative ESI mode, the ion source parameters were as the follows: capillary voltage, −1.5 kV; cone voltage, −20 V; source temperature, 120°C; desolvation temperature, 500℃; desolvation gas flow (N₂), 800 L/h; and cone gas flow (N₂), 50 L/h. The hybrid scan approach HDDIDDA was composed by three different scan events (Wang et al., 2022). 1) Ion mobility-enabled full scan (MS¹). The survey scan was performed with the TOF mass analyzer, scanning over a mass range of m/z 100–1500 at 0.3 s per scan under 6 eV of low energy. The default parameters were defined for the travelling wave ion mobility separation (Paglia et al., 2015). 2) Data-independent high-energy HDMS^E scan (MS²). Ramp collision energy (RCE) was set at 40–60 eV and the scan time was 0.3 s. 3) Data-dependent HDDDA (MS²). Three most intense precursor ions (top 3) were set to automatically trigger the MS/MS acquisition when the intensity exceeded 1000 counts, which stopped no longer than 0.7 s (timeout). The mass scan range was m/z 100–1500 and scan rate at 0.2 s per scan. Mass-dependent ramp collision energy (MDRCE) was set at 30–70 eV in the low mass and 40–80 eV in the high mass. Calibration was conducted using an external reference (Lock Spray™) by constantly infusing a solution of leucine enkephalin (200 pg/mL; Sigma-Aldrich, St. Louis, MO, USA) at 10 µL/min. By inputting an in-house library of ginsenosides recording 573 known ginsenosides (Li et al., 2022b), the automatic data processing and annotation workflows separately for the HDDDA and HDMS^E data were established by using the UNIFI™ 1.9.3.0 platform (Waters). Key parameters set in UNIFI, in general, were consistent with our previous report (Yang et al., 2022).

2.4 Multivariate statistical analysis

All the raw HDMS^E data (ESI−) of 105 batches of ginseng samples were corrected by the UNIFI workflows, and then processed by Progenesis QI v2.1 (Waters). The adduct ions, including [M−H]⁻, [M−H+HCOOH]⁻, and [2M−H]⁻, were set to perform the peak correction, peak extraction, and peak annotation to generate a data matrix including the information of t_R, m/z, CCS, and abundance for each metabolic feature. The data matrix was imported into the SIMCA-P v14.1 software (Umetrics, Umea, Sweden) for data modeling after the pre-processing. The unsupervised principal component analysis (PCA) and supervised orthogonal partial least-squares discriminant analysis (OPLS-DA) were used for the chemometrics analysis to depict the ginsenosides variation along with the extension of steaming time. The variable importance in projection (VIP) score, evaluating the importance of each variable in a classifier, was utilized to discover the potential differential components for PG, PQ, and PN induced by the steaming.

2.5 DESI-MSI

All tissue sections were prepared by the frozen section method, with the details given in Supporting Information. The processing parameters of cryo-sectioning were set as the follows: embedding agent, OCT (frozen for 15 min at −80°C); box temperature, −22°C; sample head temperature, −18°C; and section thickness, 40 μm (Jiang et al., 2023).

The MSI data in the negative ion mode were collected using the Waters Xevo-G2 XS QTOF platform equipped with a DESI source (Waters). The relevant parameters were set as follows: capillary voltage, −4.5 kV; cone voltage, −80 V; ion source temperature, 120°C; scanning range, m/z 300–1500; the X and Y pixel sizes, 150 μm; and scanning rate, 400 μm/s. High-definition imaging (HDI) (Version 1.6, Waters Corporation, Manchester, UK) software was applied to process the data of MSI, were processed by using the HDI software (Waters), with the following specific parameters: High-mass precision retention data (number of most intense peaks), 1000; Mass range, m/z 300–1500; m/z window, 0.02 Da; mass lock applied (enable Lock mass): −Ve: 554.2620; Lock mass tolerance, 0.25 amu; Min Signal Intensity, 500 counts (Jiang et al., 2023; Zhao et al., 2023). The identification of the molecules in MSI experiments was performed by comparing the m/z information of the targeted list added during the data processing.

3.1 Optimization of key parameters of the UHPLC/IM-QTOF-MS approach for the multi-component characterization of PG/PQ/PN

Aiming to systematically clarify the chemical composition and explore the metabolome difference of PG, PQ, and PN before and after the steaming, an UHPLC/IM-QTOF-HDMS^E/HDDDA dimension-enhanced multi-component characterization approach was established. For this purpose, key parameters of the chromatography (e.g., stationary phase, column temperature, and gradient elution program) and the Vion IM-QTOF (e.g., capillary voltage, cone voltage, RCE for HDMS^E and MDRCE for HDDDA) were optimized by the single-factor experiments. Considering the high selectivity on ginsenosides, the reversed-phase (RP) columns were selected. Selection of the stationary phase was conducted by comparing the chromatographic columns of 10 reversed-phase matrices, which showed difference in the silica gel core (full porous or core–shell), bonding technology, and bonding groups (Table S2). By observing the general resolution, peak shape, and the number of ions detected in base peak intensity (BPI) chromatograms, the CSH C18 enabled more balanced separation of the peaks with better peak symmetry (Fig. S2). Furthermore, different column temperature settings can have an impact on the chromatographic separation. Here, 30°C (Fig. S3) was selected as the best column temperature considering the advantages in terms of the baseline, peak shape, and resolution.

To enhance the sensitivity in detecting ginsenosides, two key ion source parameters (capillary voltage and cone voltage) and the collision energy (RCE for HDMS^E and MDRCE for HDDDA) were optimized by evaluating the peak areas of ten representative ginsenosides (Table S3). When capillary voltage was set at 1.5 kV, much higher precursor ion response was observed and the variation in the peak areas was stable for all the index components (Fig. S4A). Cone voltage can induce the occurrence of in-source cleavage (Wang et al., 2024). In this work, when the cone voltage was set to 20 V, the response of the index components was high, and the effect of in-source cleavage was not severe (Fig. S4B). For MDRCE in HDDDA, we finally selected MDRCE at 30–70 eV/40–80 eV by comparing the abundance of those secondary fragments produced at different levels. Under this condition, the fragments were intact with high response, which displayed relatively complete fragments information to aid in the structural identification (Fig. S5A). For RCE in HDMS^E, different levels such as 20–40 eV, 30–50 eV, 40–60 eV, 60–80 eV, and 80–100 eV, were examined, and a ramp of 40–60 eV was finally set enabling the generation of rich fragments (Fig. S5B). A simple method validation for the UHPLC/IM-QTOF-MS was also conducted. The range of intra-day precision (RSD, in %) varied among 1.40–11.83 % (Table S4), while the inter-day precision was 4.61–13.36 % (Table S5). Repeatability determined through six copies of the QC samples showed a variation of 3.51–10.73 % (Table S6). These results could demonstrate the established approach was suitable for the qualitative characterization of the ginsenosides in PG, PQ, and PN.

Notably, based on the Waters Vion™ IM-QTOF mass spectrometer platform, the HDDIDDA hybrid scan approach integrates ion mobility separation of all precursors, and the alternating DIA and DDA (Wang et al., 2022). It thus provides one more dimension of structural information, showing the separation potential of isomers and co-eluted components, greatly improving the peak capacity, signal-to-noise ratio, and spectral clarity. The HDDIDDA strategy can not only improve the analysis efficiency, but also combine the advantages of these two scanning modes, that is, high coverage for DIA and high accuracy for DDA. In addition, the determined CCS can assist in the reliable identification due to the comparison with 48 reference compounds (Table S7), thus providing another dimension of evidence supporting the identification of ginsenosides from the samples of PG, PQ, and PN.

3.2 Ginsenosides identification in the raw and steamed products of PG, PQ and PN

As the characteristic and bioactive ingredients of ginseng, the ginsenosides can be roughly divided into the protopanaxadiol (PPD), protopanaxatriol (PPT), oleanolic acid (OA), octillol (OT), malonylated (Mal), C-17 side-chain varied, and other subtypes, according to the sapogenins and attached substituents. The ginsenosides in the raw materials and steamed products of PG, PQ, and PN, were systematically identified by integrating the UNIFI-facilitated automatic peak annotation, reference compounds comparison, and literature searching. And 164, 192, and 182 compounds were ultimately identified or tentatively characterized for PG (Table S8), PQ (Table S9), and PN (Table S10). Here, we illustrated the characterization of the PPD-type, PPT-type, and malonylated ginsenosides, as the typical cases.

Characterization of the PPD-type ginsenosides. We took the reference compound ginsenoside Rb1 (compound 62#, PQ, t_R 13.18 min, C₅₄H₉₂O₂₃) as an example to illustrate its main fragmentation features in the negative ESI mode. The [M−H]⁻ and [M−H+HCOOH]⁻ peaks were observed at m/z 1107.5944 and 1153.6002. Deprotonated precursors could continuously lose Glc (m/z 945.5421 [M−H−Glc]⁻), H₂O (m/z 927.5299 [M−H−Glc−H₂O]⁻), and 3 × Glc, to produce the PPD aglycone ion at m/z 459.3851 ([PPD−H]⁻). Similarly, in the case of an unknown ginsenoside (compound 117#, PQ, t_R 25.17 min, C₅₆H₉₄O₂₄), the deprotonated precursor and the FA-adduct precursor ion were observed at m/z 1149.6029 ([M−H]⁻) and 1195.6091 ([M−H+HCOOH]⁻), respectively. Its MS² spectrum exhibited a series of fragments at m/z 1107.5928, 945.5360, 783.4901, 621.4485, and 459.3842, separately assigned as [M−H−Ace]⁻, [M−H−Ace−Glc]⁻, [M−H−Ace−2Glc]⁻, [M−H−Ace−3Glc]⁻, and [PPD−H]⁻ (Fig. S6A). Accordingly, it was tentatively characterized as PPD-4Glc-Ace (Yang et al., 2020).

Characterization of the PPT-type ginsenosides. The collision-induced dissociation (CID)-MS² behavior for the PPT-ginsenosides was extremely analogous to that of the PPD-type, but the difference was the generation of a sapogenin ion at m/z 475.37. Compound 21# (PQ, t_R 6.71 min, m/z 945.5420 for [M−H]⁻, C₄₈H₈₂O₁₈), consistent with ginsenoside Re, was used to illustrate the fragmentation behavior of the PPT-type neutral ginsenosides (Fig. S6B). For an unknown compound 37# (PQ, t_R 9.55 min, C₅₀H₈₄O₁₉), the [M−H]⁻ peak at m/z 987.5600 and the [M−H+HCOOH]⁻ peak at m/z 1033.5582 were observed. Its CID-MS² spectrum displayed the product ions of m/z 945.5508, 783.4901, 637.4391, and 475.3790, which were assigned as [M−H−Ace]⁻, [M−H−Ace−Glc]⁻, [M−H−Ace−Glc−Rha]⁻ and [PPT−H]⁻, respectively. Accordingly, we could characterize compound 37# as pseudoginsenoside Rs1 or isomer (PPT-2Glc-Rha-Ace, Fig. S6B).

Characterization of the malonylated ginsenosides. The neutral ginsenosides are easy to be acylated, and the polar substitution with the malonyl group generates the malonylginsenosides. Compound 70# (PQ, t_R 14.82 min, C₅₇H₉₄O₂₆, Fig. S6C) was identified as malonylginsenoside Rb1 because of the reference compound comparison. It displayed the [M−H]⁻ precursor ion at m/z 1193.5923. The fragments at m/z 1107.5915 ([M−H−Mal]⁻) and 1089.5858 ([M−H−Mal−H₂O]⁻) should result from the continuous elimination of Mal and H₂O (Shi et al., 2018). Successive cleavages of 3 × Glc and 4 × Glc were observed generating the ions of m/z 621.4346 ([M−H−Mal−3Glc]⁻) and 459.3783 ([PPD−H]⁻). In addition, the secondary fragment of the PPD sapogenin was also detected at m/z 375.2960 ([PPD−H−C₆H₁₂]⁻) (Yang et al., 2020). Similarly, the unknown compound, compound 121# (PQ, t_R 26.71 min, C₅₁H₈₄O₂₁), was deduced to be malonylginsenoside Rd isomer (PPD-3Glc-Mal, Fig. S6C).

3.3 Morphological and ginsenosides variations for PG, PQ, and PN by the steaming

An untargeted metabolomic differential analysis, based on the UHPLC/IM-QTOF-HDMS^E data, was constructed, and the chemical markers related to the steaming of PG, PQ, and PN, were preliminarily discovered and identified.

3.3.1

3.3.1 Preliminary analysis of the appearance and BPI

After being steamed at high temperature for a period of time, the morphology and color of the PG/PQ/PN samples significantly changed (Fig. 2A). As the steaming time prolonged, the moisture of the herb sharply decreased, and the samples shrinked due to the loss of water, rendering the steamed product smaller in size than the fresh sample. Moreover, as the extension of steaming time, the color gradually deepened from the light brown of the raw material to dark brown. According to the BPI chromatograms, remarkable ginsenosides variations were observed. The polar ginsenosides underwent a series of reactions, producing the less polar components in an increasing content (Fig. 2B).

Close

Fig. 2

Preliminary difference analysis of the raw materials and steamed products of PG/PQ/PN on the appearance and the base peak chromatograms (BPCs). A-Changes in the sample morphology and color during the steaming of PG, PQ, and PN; B-BPCs of PG, PQ, PN at different steaming time.

Export to PPT

3.3.2

3.3.2 Application of untargeted metabolomics to unveil the potential differential ginsenosides before and after the steaming for PG/PQ/PN

In the untargeted metabolomics analysis, we initially compared four data pre-treatment methods for screening the robust metabolic features. Method 1: “80 % rule” (remove the features that exceed 80 % of the null values for all sample batches; Bijlsma et al., 2006); Method 2: “30 % rule” (retain the features of peak area showing RSD less than 30 % in the QC sample); Method 3: further normalize the data after “80 % rule” processing; Method 4: 30 % rule processing based on Method 3. Consequently, the PCA score chart (Fig. 3A) demonstrated that the better clustering effect of the QC data gained by the Method 3 treatment. Generally, these samples were divided into three groups: the raw, steamed for 1 and 2 h, and steamed for 4/6/8/10 h (Fig. 3B). When the VIP cutoff was set at 1.5, separately 288, 307, and 325 differential ions were screened for PG, PQ, and PN, respectively. These ions could be assigned as 26 (Table S11), 28 (Table S12), and 18 (Table S13) ginsenosides (Fig. 3C).

Close

Fig. 3

Multivariate statistical analysis of the raw materials and steamed products of PG, PQ, and PN. A-PCA score plot obtained by different data pre-treatment methods: Method 1, 80% rule without normalization; Method 2, 30% rule without normalization; Method 3, 80% rule after normalization; Method 4, 80% rule and 30% rule after normalization; B-OPLS-DA score plots for PG, PQ, and PN; C-VIP plots of PG, PQ, and PN based on the OPLS-DA classification model.

Export to PPT

Characteristic components of PG with different steaming time. In the steaming process of PG samples, the contents of M1, M2 (Re), M4, M6 (m-Rb2), M16, M18, M19, and M21 (m-Rb1), showed the significant decrease trend, while the contents of M3, M10 (20(S)-ginsenoside Rh2), M12 (20(S)-ginsenoside Rh1), and M13 (20(R)-ginsenoside Rh2) clearly increased. It is worth noting that ginsenoside Rh2 has significant pharmacological activity, and the increase in Rh2 content after steaming provides evidence for the superior activity of red ginseng compared to the raw ginseng. Finally, m-Rb1, m-Rb2, 20(S)-Rh2, 20(S)-Rh1, and 20(R)-Rh2 were used as the markers for processing PG, while the unknown markers M16 (pseudoginsenoside Rs1 or isomer, C₅₀H₈₄O₁₉, PPT-2Glc-Rha-Ace) and M19 (malonylginsenoside Rd isomer, C₅₁H₈₄O₂₁, PPT-2Glc-Rha-Mal) deserves the further investigation to fully establish their structures (Fig. 4A and Fig. S7A).

Close

Fig. 4

Box charts illustrating the content variation of some markers for PG (A), PQ (B), and PN (C) steamed at different time.

Export to PPT

Characteristic components of PQ with different steaming time. While steaming PQ, the content of eight characteristic components decreased significantly: M1 (m-Rb1), M3 (Re), M6, M10, M11, M18, M20 (m-Rd), and M22. Meanwhile, and the contents of M2 (Rg5), M26, and M27, showed a remarkable increase trend. In addition, the content of M4 (Rb1) reached the highest when steaming for 2 h, and then showed a downward trend with the prolonging of the steaming time. The effective immunomodulatory and hypoglycemic effect of the steamed PQ may be related to these increased ginsenosides (Sun et al., 2012; Zhang et al., 2023b). By contrast, m-Rb1, Rb1, m-Rd, Rg5, and Re, can be the potential differential markers during the processing of PQ (Fig. 4B and Fig. S7B).

Characteristic components of PN with different steaming time. During the steaming process of PN, the contents of M5, M7 (20(S)-F1), M9 (noto-Fa), M10, and M13 (m-Rb1), decreased significantly, but M14 significantly increased. The overall contents of M1, M2 (Re), M3 (Rb1), M4 (noto-R4), and M6 (noto-Fd), showed a gradual increase followed by a decrease. Compared with the raw material, steamed PN exhibited stronger tonifying effects, which probably could be related to the increase in the content of these ginsenosides (Xiong et al., 2019). During the steaming process, the content of M11 showed a decreasing-increasing–decreasing variation trend. The 20(S)-ginsenoside F1, noto-Fa, m-Rb1, Rb1, Re, and noto-Fd, were finally identified as potential differential markers during the processing of PN (Fig. 4C and Fig. S7C).

3.4 Spatial depicting of the steaming-induced ginsenosides variations for PG, PQ, and PN by DESI-MSI

A DESI-MSI method was developed to globally analyze the ginsenosides of PG/PQ/PN steamed at different periods of time. The key parameters affecting the slice (e.g., freezing time, microtome body temperature, sample head temperature, and slice thickness) and detection (e.g., capillary voltage, cone voltage, and ion-source temperature), were optimized. By comparing different freezing temperature settings (−80℃ and−20℃) and time (1, 2, 5, 8, 10, 15, and 20 min), we could conclude freezing at −80℃ for 15 min was the optimal experimental condition (Tables S14). When setting the temperature of the microtome body and the sample head for the tissue section, the temperature of the sample head should be 2−4℃ higher than the temperature of the box. This temperature difference could help prevent the tissue from rolling up due to the high temperature of the box, ensuring that it can be properly attached to the slide. During the sectioning process, the chamber temperature was set at −22℃, and the sample head temperature was −18℃. We examined slice thicknesses (10–60 μm), a key factor affecting the image resolution and imaging effects. To obtain better integrity of the tissue structure and improve the quality of the spectrum, a slice thickness of 40 μm was considered as the most suitable (Fig. S8). The samples were processed with the optimized slicing condition to acquire the data of the PG/PQ/PN root/rhizome samples. The capillary voltage (−4.5 kV), cone voltage (−80 V) and ion source temperature (120℃) were optimized. The use of 95 % MeOH as the spray solvent gave good imaging effect.

The Fig. 5A shows the tissue diagrams of the PG, PQ, and PN sections. According to the imaging maps of raw materials for these three ginseng varieties, we could obtain the spatial distribution map of m-Rb1/m-Rb2/Rb3/Rg1/Rd/Re/Ro and other components or their isomers (Fig. 5B–D). In PQ, m-Rb1 and Rb1 were mainly distributed in the cork layer, while m-Rb2 in PG and PQ was concentrated in the cork layer, and Ro in the cork layer and xylem. It could thus be concluded that the types and contents of ginsenosides distributed in the cork layer, phloem, and xylem were relatively high, which may be related to the plant self-protection mechanism.

Close

Fig. 5

Slices and mass spectrometry images of the raw materials of PG/PQ/PN. A-Structure of the transverse section of PG, PQ, and PN; B/C/D-DESI-MSI images of the representative ginsenosides of PG/PQ/PN in the negative mode.

Export to PPT

Additionally, the DESI-MSI method was used to detect the possibly transformed ginsenosides in the steamed products of PG, PQ, and PN, and to compare the difference in the spatial distribution of each component. In the section slice of raw PG (marked as 0 h), rich malonylginsenosides were observed (Fig. 6A). Especially, m-Rb1 was mainly distributed in the cork layer and phloem, and only a small amount in xylem. Ginsenoside Rg3 of low content was visualized with the distribution in the cortex, cork layer, phloem, cambium, and xylem. The content of ginsenoside Rk1/Rg5 was very low for PG when steaming for 2 h. Comparatively, the contents of Rb1, Rg3, and Rk1/Rg5 were increased in PG with steaming at 10 h (Fig. 6A). The OA-type ginsenoside Ro and zingibroside R1/chikusetsusaponin Ⅳa (as the isomers) in PQ exhibited a pattern of decreasing, then increasing, and then decreasing again with the prolonged steaming time. The content of calenduloside E was the highest in PQ when steaming for 2 h (Fig. 6B). Noto-R1 (PPT-type) in PN was distributed in the whole area of rhizome in the raw material, particularly in the cork layer. After steaming for 1 h, its content increased significantly. But when steaming for 8 h or even longer, noto-R1 was almost undetectable in the rhizome of PN. The contents of Rh1 and Rh4 were the highest in the steamed product of PN at 8 h, and distributed in the whole area. After steaming for 10 h, their contents then showed the decrease trend (Fig. 6C). It is noted that our conclusions on the ginsenosides variation reflected by MSI refer to all isomers, rather than a single ginsenoside that can be separated and detected by LC-MS.

Close

Fig. 6

The spatiotemporal transformation networks of Mal, OA and PPT-type saponins and the corresponding MSI images of PG, PQ and PN at different steaming time. A-Imaging of malonylginsenoside Rb1 in PG with different steaming time; B-imaging of ginsenoside Ro in PQ with different steaming time; C-imaging of notoginsenoside R1 in PN with different steaming time (undetectable in the sample of steaming 10 h); D-spatiotemporal transformation network of malonylginsenoside Rb1 (malonylated) from PG; E-spatiotemporal transformation network of ginsenoside Ro (OA-type) from PQ; F-spatiotemporal transformation network of notoginsenoside R1 (PPT-type) from PN.

Export to PPT

3.5 Holistic transformation network for ginsenosides occurring to the steaming of PG/PQ/PN

In the steaming process of PG, PQ, and PN, the contents of ginsenosides were affected by both the duration and temperature of steaming. Various patterns of degradation and transformation occurred, leading to the formation of rare ginsenosides (such as Rg3, Rh1, Rk1, Rg5, and Rk3; Fan et al., 2022; Xiong et al., 2019). These saponins have been demonstrated with significant therapeutic effects on the cardiovascular and cerebrovascular diseases, improving the immunity, anti-cancer (Choi et al., 2015), anti-asthmatic (Song et al., 2023), and anti-diabetic properties (Qu et al., 2023). According to the results of the difference analysis, the contents of Rb1 and Rd decreased significantly after the prolonged steaming, whereas the contents of the rare ginsenosides (such as Rg3) was increased. This phenomenon is speculated to be the transformation and formation of components, possibly due to the fact Rb1, Rd, and Rg3 have the same PPD-type parent nucleus. Their glycosidic bonds belong to the carboxyl-formaldehyde structures, which are sensitive to the conditions such as acid, alkali, or enzyme. Under high temperature, hydrolysis reactions occur. The acidic dammarane-type tetracyclic triterpenoid ginsenosides (PPD- and PPT-type) were prone to decarboxylation. Malonyl-substituted ginsenosides, such as m-Rb1, could lose their malonyl groups to generate the neutral PPD-type ginsenosides. The glucose group connected at the C-20 in Rb1 was hydrolyzed to obtain ginsenoside Rd, which was further hydrolyzed to ginsenoside Rg3. Continued dehydration of the sugar groups attached to Rg3 at the C-20 could generate ginsenosides Rg5 and Rk1 (Fig. 6D). The observed ginsenosides transformation in the steaming of PG was generally consistent with the findings reported in previous investigations, and ginsenosides Rg3, Rg5, and Rk1 were the anti-cancer ingredients in black ginseng (Huang et al., 2023b). Similarly, Rg1 (PPT-type) in PQ was hydrolyzed to generate Rh1, which was further dehydrated to generate Rh4/Rk3 (Fig. S9A). p-F11 and p-RT2 (OT-type) underwent the cleavage of the glucosyl moiety to obtain p-RT5 (Fig. S9B). Ginsenoside Ro (OA-type) was hydrolyzed to calenduloside E (Fig. 6E). For noto-R1 in PN, the glycosyl part was degraded by heating. The disaccharide group at the C-3 position was converted into a monosaccharide, and the glucose connected at C-20 could be lost. This resulted in the formation of ginsenoside Rh1, which was further dehydrated to generate Rh4 (Fig. 6F).