Integrating the rapid constituent profiling strategy and multivariate statistical analysis for herb ingredients research, with Chinese official rhubarb and Tibetan rhubarb as an example

2.1 Chemicals, reagents, and plant materials

The dried roots and rhizomes of COR were acquired from AiBaZhou (Sichuan province of China, Rheum tanguticum Maxim), TianShui (GanSu of China, Rheum palmatum L), and BaoJi (ShanXi province of China, Rheum officinale Baill). The dried roots and rhizomes of TR (Rheum likiangense Sam, Rheum australe D. Don, Rheum webbianum Royle) were locally collected in Tibet. The samples were identified by Professor GuoYue Zhong (Jiangxi University of Traditional Chinese Medicine). These herbs, corresponding voucher specimen numbers 2019COR01∼06 and 2019TR01∼06, were preserved at the herbarium of Jiangxi University of Traditional Chinese Medicine. Formic acid and acetonitrile, used in UHPLC-QTOF-MS analysis, were obtained from Fisher Scientific (Fair Lawn, NJ, USA) and Sigma-Aldrich (Sigma-Aldrich, MO, USA), respectively. Pure distilled water was obtained from a Millipore Alpha-Q water purification system (Millipore, USA). All the 23 reference standards (purities > 98% as determined by HPLC-UV), including emodin, aloe-emodin, rhein, physcion, chrysophanol, gallic acid, methyl gallate, ethyl gallate, protocatechuic acid, catechin, caffeic acid, ferulic were purchased from Chengdu Chroma-Biotechnology Co., Ltd and Chengdu Pufei De Biotech Co., Ltd.

2.2 Sample preparation

To prepare the extracts, crude medicinal materials were pulverized and screened through a 100 mesh sieve. Then, 1.0 g powder of COR or TR was suspended in 40.0 mL of methanol, and extraction was performed in an ultrasonic water bath for 1.0 h at room temperature (RT). The extracts were centrifuged at 12,000 rpm for 15.0 min, of which, 2.0 μL was analyzed by UHPLC-QTOF-MS/MS.

To ensure the stability of sequencing analysis, quality control (QC) samples were also prepared by mixing extracts from all the samples. All reference compounds were dissolved in methanol and stored at 4 °C until further use. The stock solutions were diluted with methanol to a final concentration of 10 μg/mL. From this, 2 μL was analyzed by UHPLC-QTOF-MS.

2.3 UHPLC-QTOF-MS/MS conditions

Considering the characteristics of the compounds in the rhubarb plants, the negative ion mode was selected for mass spectrometry data acquisition. Also, mass spectrometry parameters, mobile phase composition, and gradient elution conditions were optimized to achieve a satisfactory ion response intensity of COR and TR. UHPLC was performed on a Shimadzu System (Kyoto, Japan), coupled with an LC-3AD solvent delivery system, a SIL-30ACXR auto-sampler, a CTO-30AC column, a DGU-20A3 degasser, and a CBM-20A controller. The Welch UHPLC (100 mm × 2.1 mm,1.7 μm), maintained at 40 °C, was used as the separation system. The mobile phase consisted of 0.1% formic acid in water (solvent A) and acetonitrile (solvent B). The flow rate was 0.3 mL/min and the injection volume was 2 μL. The pump B gradient program was as follows: 0.01–3 min, 5–10%; 3–12 min, 10–17%; 12–24 min, 17–28%; 24–34 min, 28–43%; 34–42 min, 43–95%; 42–42.1 min, 95–5%; 42.1–45 min, maintain 5% until stop.

UHPLC-QTOF-MS/MS analyses were performed in the negative electrospray ion mode with a Duo Spray source on a Triple TOF^TM 5600^＋ system (AB SCIEX, Foster City, CA, USA). Mass spectrometry was performed in the negative mode using the following parameters: ion spray voltage, −4500 V; ion source temperature, 500 °C; curtain gas, 25 psi; nebulizer gas, (GS1) 50 psi; heater gas (GS2), 50 psi, and declustering potential (DP), −100 V. The mass ranges of TOF-MS and TOF-MS/MS experiments were 50–1250 Da. From each TOF-MS/MS scan, the most intensive eight ions were selected for MS/MS fragmentation. For the UHPLC-QTOF-MS/MS analysis, the collision energy (CE) and the collision energy spread (CES) were 35 eV and (±) 15 eV, respectively.

2.4 Data analysis

The TOF-MS/MS data were analyzed using PeakView 1.2 software (AB SCIEX, Foster City, CA, USA). The components were identified based on chromatographic retention time, formulation composition, MS/MS cleavage pathway, comparison with available standards, references, and using the structural characterization database, such as MassBank ( http://www.massbank.jp/), ChemSpider (http://www.chemspider.com/) and SciFinder scholar.

All raw LC-MS data were imported into MakerView software for data processing and then exported to SIMCA-P 14.1. PCA, an unsupervised multivariate statistical method, was used for analyzing, classifying, and reducing the dimensionality of numerical datasets in the COR and TR data multivariate problems. The differences between COR and TR were grouped with OPLS-DA. Only compounds with VIP >1.0 and P <0.05 were considered significantly different metabolites and further screened against a matching database. Heat map analysis was performed using MetaboAnalyst 4.0 (www.metaboanalyst.ca).

3.1 Identification of the constituents of COR and TR

Firstly, we established the diagnostic ion database based on the reference substance and marked the corresponding retention time. In this study, 23 representative standards were analyzed in the negative ion mode (Fig. 2). Diagnostic ions and standards retention time are listed in Table 1, with their corresponding ions fragmentation behavior as reported in the literature.

Close

Fig. 1

Integrating the constituent profiling strategy and multivariate statistical analysis.

Export to PPT

Close

Fig. 2

The base peak chromatogram (BPC) obtained from UHPLC-QTOF-MS/MS: (a) The mixed reference standards; (b) Anthraquinones; (c) Stilbenes; (d) Tannins; (e) Acylglucosides.

Export to PPT

According to the literature, the major components of rhubarb are anthraquinone, stilbene, acyl glycosides, and catechins (Katsuko et al., 2006; Qin et al., 2011). To efficiently mine the chemical components and systematically identify the rhubarb, we critically studied the fragmentation behavior of the corresponding reference substance(s). The BPC (base peak chromatogram) of the mixed reference standard showed some interference peaks originated from the interfering ions in the solvent. Combining this database and using neutral loss and mass loss filtering strategies, compounds with similar skeletons were mined simultaneously (Table 2). Subsequently, structural identification was performed; the flow process is depicted in Fig. 1. The proposed structure of rhubarb is shown in Fig. 3.

Close

Fig. 3

The proposed structure of constituent components from Chinese official rhubarb and Tibetan rhubarb.

Export to PPT

3.2 Characterization of anthraquinones

Previously, emodin, aloe-emodin, chrysophanol, rhein, physcion, and their derivatives were reported as the major anthraquinone components in rhubarb (Huang et al., 2019; Junko et al., 2007). In this study, we identified a total of 51 anthraquinone components (Table 2). Firstly, Compound 1 (emodin, m/z 269.0454), compound 17 (aloe-emodin, m/z 269.0453), compound 32 (rhein, m/z 283.0252), compound 34 (physcion, m/z 283.0612), and compound 37 (chrysophanol, m/z 253.0507) with the basic skeleton of anthraquinone compounds were successfully identified for their cleavage behavior similar to standard reference substances. Compounds 2–16 are probably the emodin derivatives having diagnostic MS/MS fragment ions at m/z 269.0454 and 225.0552. Compounds 18–31 were inferred as the aloe-emodin derivatives having diagnostic ions at m/z 269.0453 and 240.0432. Peaks 21 and 24, representing the predominant fragment ions m/z 268.0372 and neutral loss 163 Da (C₆H₁₁O₅) through breakage of glycosidic linkages at 3-position, were tentatively identified as aloe-emodin-3-(hydroxymethyl)-glucoside and aloe-emodin-3-(hydroxymethyl)-glucoside-acetyl. Compound 33, based on diagnostic ions at m/z 239.0346 and neutral loss of 162 Da, was identified as rhein-8-glucoside. Compounds 35, 36 were identified as physcion derivatives for m/z 240.0437 as reported previously. Compounds 38–51, with m/z 253.0507 and 225.0559, were identified as chrysophanol derivatives.

3.3 Characterization of stilbenes

Stilbene has the molecular framework of resveratrol (peak 52, m/z 227.0719) (Shang and Yuan, 2002; Yoshiki et al., 1986). Derivatives such as methoxyresveratrol (peak 62, m/z 241.0874), piceatannol (peak 69, m/z 243.0651), and rhapontigenin (peak 83, m/z 257.0809) are among four major types of monomers that exist in rhubarb plants. Using them as framework compounds along with mass loss filter screening, 44 stilbene compounds were successfully identified. Based on the literature, databases, and fragmentation behavior, all stilbenes were classified into two classes, mono-substituted, and di-substituted stilbenes, following the number of substitutions (Sung et al., 1995). Based on the standard reference substance database, the substituents were found to be gallic acid, cinnamic acid, and p-hydroxycinnamic acid. Compounds 52–61 were deduced as resveratrol derivatives. Among these, compounds 52 and 53 were resveratrol isomers and exhibited cleavage behavior similar to resveratrol. However, due to the lack of a reference substance, the hydroxyl group attachment site on the benzene ring could not be determined. Compounds 55 and 56 were identified as resveratrol-glucoside having diagnostic ions m/z 227.0719, 185.0584, and neutral loss m/z 162 Da. For peaks 57 and 58, based on the mass of quasi-molecular ions obtained at m/z 541.1386, molecular formulas were calculated to be C₂₇H₂₆O₁₂ which is more than that of resveratrol-glucoside (152 Da, C₇H₄O₄). These were tentatively identified as resveratrol-glucoside-galloyl. Based on the deprotonated molecular ions at m/z 535.1665, three compounds, 59, 60, and 61 were identified as resveratrol-glucoside-p-coumaroyl. Peaks 62–68, shown in the BPC, presented similar fragmentation behaviors and produced identical diagnostic ions in further auxiliary confirmation, were identified as methoxyresveratrol derivatives. Piceatannol (peak 69, C₁₄H₁₂O₄) with a deprotonated molecular ion [M-H]- at m/z 243.0651 was set as key fragment ions and via neutral loss of 162 Da (glucoside, C₆H₁₀O₅), 152 Da (galloyl, C₇H₄O₄), 146 Da (p-Coumaroyl, C₉H₆O₂) and 168 Da (cinnamoyl, C₉H₈O₂), the compounds 70–82 were identified as piceatannol derivatives, as indicated in the extract chromatogram (Zhu et al., 2005). Peak 83 denotes the deprotonated molecular ion m/z 257.0809. Compounds 84–95 were identified as rhapontigenin derivatives depending on the fragmentation routes and respective retention times.

3.4 Characterization of tannins

Rhubarb tannins exert diverse biological activity (Zeng et al., 2013). In total, 26 tannin derivatives were identified. Catechin (peak 96, precursor ions at m/z 289.0718) was eluted at 5.15 min. It perfectly matched the molecular ion mass, retention time, and secondary fragmentation behavior of the catechin standard. Compounds 97–106, with deprotonated molecular ions at m/z 451.1208, were identified as catechin-glucoside derivatives and produced diagnostic ions m/z 289.0713, 245.0818, 151.0397, 109.0307, and the neutral loss of 162 Da (C₆H₁₀O₅). Compounds 107–113 exhibited predominant ions at m/z 289.0713 and 245.0818. Their deprotonated molecular ion at m/z 613.1834 was 162 Da more than catechin-glucoside, and therefore, these were tentatively characterized as catechin-di-glucoside. Compounds 114 and 115, having 288 Da (C₁₅H₁₃O₄) more than the catechin, were inferred isomeric based on similar retention time and lysis behavior. Compound 114 was unambiguously identified as procyanidin B by comparing it with the standard compound. Peaks, 116, 117, and 118, with quasi-molecular ions at m/z 729.1419, were 152 Da (C₇H₄O₄) more than procyanidin B. Also, their MS/MS fragmentation behavior was consistent with procyanidin B. Therefore, based on the match against the diagnostic ion database of gallic acid (reference substance), these were tentatively identified as procyanidin B-gallate. Compounds 119, 120, and 121, having deprotonated molecular ions at m/z 865.1959 and based on MS/MS fragmentation similarity to previous reports, were tentatively identified as trimers of catechin (Nàdia et al., 2010; Li et al., 2012).

3.5 Characterization of acylglucosides

A total of 52 acylglucoside compounds, having glucose as the skeleton structure, were identified in rhubarb extract (Wei et al., 2017). In the secondary mass spectrometry, 162 Da neutral loss of parent ion produced a robust peak of product ion. In the first step, precise deprotonated molecular ions were produced via mass defect filtering to filter the background-subtracted ion chromatogram and MS² information unambiguously or tentatively identified mono-substituent acylglucosides. Seven types of mono-substituted acylglucosides, including 6-hydroxymusizin-glucoside (peak 122), galloyl-glucoside (peak 131), caffeoyl-glucoside (peak 149), altechromone A-glucoside (peak 161), p-hydroxy benzylacetone-glucoside (peak 158), p-coumaroyl-glucoside (peak 153), and femloyl-glucoside (peak 159) were identified based on diagnostic ion database of previously published standards and mass spectra. In the second step, we looked for di-substituent acylglucosides based on the galloyl-glucoside group. These compounds commenced a similar cleavage pathway generating the predominant ions C₁₃H₁₃O₉ (calculated m/z 313.0556) and C₇H₅O₅ (calculated m/z 169.0133). Subsequently, seven types of di-substituent acylglucosides, including 6-hydroxymusizin-galloyl-glucoside, di-galloyl-glucoside, caffeoyl-galloyl-glucoside, altechromone A-galloyl-glucoside, p-hydroxy benzylacetone-galloyl-glucoside, p-coumaroyl-galloyl-glucoside, and femloyl-galloyl-glucoside were identified (Xu et al., 2019; Zhu et al., 2016). The information is listed in Table 2.

3.6 Multivariate statistical analysis of COR and TR

12 batches of COR and TR mass spectrum data were selected for multivariate statistical analysis. Unsupervised PCA analysis, indicating their difference, divided them into two distinct groups (Fig. 4a). The concentrated QC sample distribution suggested stability during the sampling process. Furthermore, OPLS-DA score plots also identified metabolic differences in COR and TR (Fig. 4b). To verify the reliability of the established OPLS-DA model, 200 rounds of random permutations were performed (Fig. 4c). R²Y (Y matrix interpretation rate) and Q² (prediction rate) were used to evaluate the authenticity of the OPLS-DA model. R²Y 0.998 and Q 0.924 indicate that the model changes the level of individual metabolites in plants have a strong explanatory power (Julien and Douglas, 2013). In general, VIP > 1 means that the variable made a significant contribution to model differentiation and a higher VIP signifies for higher contribution. The compounds, highlighted blue in the S-plot (Fig. 4d), with VIP value > 1.0, contributed high (Hyuk-Hwan et al., 2013). Based on this, the QI software was used for ANOVA analysis, and the variables with P < 0.05 were selected as significant metabolic differences.

Close

Fig. 4

Multivariate statistical analysis of COR and TR: (a) Principal component analysis (PCA); (b) OPLS-DA score plot; (c) cross-validation plot of OPLS-DA model with 200 permutation tests; (d) S-plot of OPLS-DA.

Export to PPT

3.7 Analysis of metabolic differences and biosynthetic pathway

35 chemical markers were identified by PeakView software. The main differences were revealed in anthraquinones (Emodin, rhein, chrysophanic acid, chrysophanol-1-o-glucoside, dihydrochrysophanol, emodin-8-o-glucoside, aloe-emodin-8-o-glucoside, rhein-8-o-glucoside, physcion-8-o-glucoside, acetyl-chrysophanol-glucoside, acetyl-emodin-glucoside, carboxy-emodin-glucoside, aloe-emodin-di-glucoside) and stilbenes (Methoxyresveratrol, piceatannol, resveratroloside, desoxyrhaponticin, piceatannol-o-glucoside, piceatannol-o-glucoside, rhapontin, rhapontigenin-o-glucoside, piceatannol-galloyl-glucoside, piceatannol-galloyl-glucoside, p-hydroxycinnamic-rhapontin, galloyl-desoxyrhaponticin). The other ten metabolites are Catechin glucoside, cinnamoyl-o-galloyl-glucoside, di-galloyl-glucoside, proanthocyanidin, di-glucoside-catechin, galloyl-proanthocyanidin, and sennoside. Furthermore, metabolic differences at the individual level were reflected using the heat map analysis (Fig. 5a). The Red and green colors denote a relatively high and low content respectively (Yousef et al., 2020). The heat map analysis indicate that COR is richer in anthraquinone components while TR has more stilbene. However, an in vivo analysis can better evaluate the distinct therapeutic effects of the two herbs. Similar to PCA findings, hierarchical cluster analysis (HCA) of the heat map also revealed two distinct groups of COR and TR. Furthermore, using the KEGG annotation, we found that the differential metabolite between COR and TR were assigned to stilbenoid, diarylheptanoid, and gingerol biosynthesis (ko 00945), biosynthesis of type II polyketide backbone (ko 01057), and anthocyanin biosynthesis (ko 00942) pathways (Fig. 5b).

Close

Fig. 5

(a) Heat map clustering of 32 metabolic differences; (b) KEGG classification of differential metabolites.

Export to PPT