Integrated quality evaluation strategy for multi-species resourced herb medicine of Qinjiao by metabolomics analysis and genetic comparation

Fifty-eight batches of rhizomes of G. macrophylla, G. straminea, G. crasicaulis and G. daurica were collected from different herbal markets or harvested from various locations of China. All samples were authenticated by Professor Jiuzhi Yuan (Shenyang Pharmaceutical University) or Professor Lihong Wu (Shanghai University of Traditional Chinese Medicine). Voucher specimens were deposited in Department of Pharmacy, The First Affiliated Hospital of Zhengzhou University, Zhengzhou, China or the Institute of Chinese Materia Medica, Shanghai University of Traditional Chinese Medicine, Shanghai, China. Detailed information of collected samples in this study is list in Table 1. The roots were gently washed and dried at 50 °C for 48 h, and then were grounded into powder and stored in glass jars in the dark at room temperature until further analysis.

Methanol-d4 (CD3OD, 99.8%) was obtained from Cambridge Isotope Laboratories (Miami, FL, USA). Methanol of HPLC grade was purchased from Honeywell Inc. (Morristown, NJ, USA). Formic acid of LC–MS grade was bought from ROE Scientific Inc. (Newark, DE, USA). Deionized water was purified using a Milli-Q system (Millipore, Bedford, MA, USA). The chemical reference standards (CRS) of gentiopicroside, loganic acid, swertiamarin, sweroside, 6′-O-β-d-Glucosylgentiopicroside, roburic acid, morroniside, isovitexin, homoorientin, oleanolic acid, ursolic acid, β-sitosterone, citric acid, quercetin, kaempferol, and daucosterol were purchased from Chengdu Must Bio-techenology Co., Ltd. (Chengdu, China), with HPLC purity > 98%. All other chemicals and reagents were of analytical grade and commercially available.

ITS2 was prevalently adopted as a universal barcode for plant, especially herbal medicinal identification [9]. ITS2 barcode has been successfully employed for species identification of Qinjiao [8, 10]. In current research, ITS2 region was compared to verify the authentication of collected Qinjiao samples. To extract total genomic DNA from dried rhizomes (50 mg), the protocol provided by the Plant Genomic DNA Kit (Tiangen Biotech, Co., Ltd., Beijing, China) was used. Extracted DNA sample was stored at − 20 °C until use. The ITS2 sequence was amplified using a pair of primers (ITS2F: 5′-ATGCGATACTTGGTGTGAAT-3′; ITS3R: 5′-GACGCTTCTCCAGACTACAAT-3′) described previously by Chen et al. [12]. The PCR amplification was conducted as described by Luo et al. [13]. Amplification products were examined by electrophoresis in 1% (wt/vol.) agarose gels and visualized under ultraviolet light to detect successfully amplified products and the possible contamination of negative controls. After purifying, the PCR products were directly subjected to sequencing.

Sequences were edited and assembled using DNAMAN software (version 6.0) and refined manually. ITS2 resuquence were identified using DNA barcoding system for identifying herbal medicine (http://​www.​tcmbarcode.​cn). Genetic distances were calculated using the Kimura-2-Parameter (K2P) model. All the newly obtained ITS2 sequences were uploaded to GenBank.

0.5 g power (through No. 3 sieve) of each Qinjiao sample was placed into a separate 50 mL stopper conical flask followed by the addition of 20 mL methanol. The mixtures were vortexed for 1 min, sonicated (40 kHz, 500 W) for 30 min. Cooling down to ambient temperature, the lost weight was made up by methanol. After centrifugation at 10,000 rpm, 4 °C for 10 min, a 2 μL aliquot of the supernatants was injected into a UPLC-ESI-Q-TOF system (AB Sciex, Framinghan, MA, USA) for MS analysis.

Chromatography separation was achieved on an ACQUITY UPLC HSS T3 column (100 × 2.1 mm i.d., 1.8 μm) maintained at 40 °C. The mobile phase consisted of A (0.1% formic acid in water) and B (0.1% formic acid in methanol), using gradient elution: 0–1 min, 5–30% B; 1–5 min, 30–40% B; 5–6 min, 40–90% B; 6–13 min, 90–100% B; 13–21 min, 100% B; 21.01–24 min, 5% B. The flow rate was set at 0.3 mL/min and the injection volume was 2 μL.

The mass spectrometric data were collected on a SCIEX X500R QTOF mass spectrometer (AB Sciex, Framinghan, MA, USA) coupled with an electrospray ionization interface in negative ion modes (ESI-). SCIEX OS software 1.2 (AB, Milford, MA) was employed for data acquisition and procession. The following parameters settings were used: the ion spray voltage of 4000 V; turbo spray temperature (TEM) of 600 °C; declustering potential (DP) of − 80 V; collision energy (CE) of − 45 V; nebulizer gas (gas 1) of 55 psi; heater gas (gas 2) of 55 psi, CAD gas of 7 psi, and curtain gas of 35. Nitrogen was kept as the nebulizer and auxiliary gas. TOF MS and TOF MS/MS were scanned with the mass range of m/z 50–1000. Continuous recalibration was carried out every six samples. In addition, dynamic background subtraction (DBS) trigger information-dependent acquisition (IDA) was used to trigger acquisition of MS/MS information of low-level constituents. The accurate mass and composition for the precursor ions and fragment ions were analyzed using the Markerview™ software (Version 4.1, Waters Co., Milford, MA, USA) integrated with the instrument.

QC samples were prepared by combining equal aliquots from all Qinjiao samples and were injected every six specimens during the whole analysis. QC data obtained was used to assess the stability of the LC/MS platform. For all QCs, 5 characteristic features (list in Additional file 1: Table S1) were picked out to verify the stability. The results proved that variations of retention times were less than 0.2 min, drift values of m/z were less than 10 PPM, and the RSD of peak areas were all below 10% (Additional file 1: Table S1).

Raw data from Q/TOF–MS were analyzed using Markerview for peak deconvolution and peak alignment with the following parameters: initial retention time 0.5 min, final retention time 23 min, mass tolerance 10 PPM, ion intensity threshold (3000 counts) and retention time tolerance 0.1 min. The data were combined into a single matrix by aligning peaks with the same mass-retention time pair together from each data file in the data set. The ion intensities of each peak detected (2806 MS features for ESI- modes) were normalized to the sum of the peak intensities in each sample. After normalization, the data was processed according to the “80% rule”, briefly only variables with values above zero presenting in at least 80% of each group were kept for the following analysis [14].

Output data from TOF–MS or NMR analysis was separately imported into SIMCA (version 14.0, Umetrics, Umeå, Sweden) for multivariate statistical analysis (MS data unit variance scaled, NMR data pareto-scaled). To provide comparative interpretations and visualization of the metabolic differences among the four species of Qinjiao, PCA and OPLS-DA were applied to the TOF–MS or NMR data set. The quality of the models was described by R2X and Q2 values. R2X shows the proportion of variance in the data explained by the models and indicates goodness of fit. The value closer to 1 indicates the goodness of fit. Q2, on the other hand, shows the proportion of variance in the data predictable by the model and indicates predictability. The results were visualized in the form of score plots, where each point represents an individual sample (to show the group clusters), and loading plots or S-plots, where each coordinate represents one mass-retention feature or ¹H-NMR spectral region (to identify the variables contributing to the classification). The variable importance of projection (VIP) is the vector to summarize the total importance of the variable in explaining the model. The corresponding variables with VIP > 1.0 were chosen as potential discriminative metabolites. To justify the OPLS-DS models, analysis of variance testing of Cross-Validated predictive residuals (CV-ANOVA) were conducted. CV-ANOVA is a diagnostic tool for assessing the reliability of PLS and OPLS models. The P value produced by CV-ANOVA indicates the probability level. The common practice is to interpret a p-value lower than 0.05 as pointing to a significant model. Statistical analysis was also performed using one-way analysis of variance (ANOVA) followed by Tukey’s multiple comparison test (SPSS, Chicago, IL, USA). A probability of P < 0.05 was considered to be statistically significant between two groups.

LC–MS Peaks were identified according to actual mass, MS/MS fragments and retention time (RT). First, the m/z value of the molecular ion of interest was searched against a self-build Qinjiao constituent Database, where data was collected from published researches [15, 16]. Then, the putative identifications were verified by comparing the MS2 fragmentations. Part of the constituents were further identified by reference standards. ¹H-NMR signals were assigned by comparing the spectrum of Qinjiao with that of gentiopicroside, loganic acid, sweroside, swertiamarin or sucrose by Chenomx NMR suit (version7.6, Chenomx, Edmonton, Canada) with method described in our previous work [17].

ITS2 sequences from collected Qinjiao samples were submitted to GenBank database (Accession numbers were listed in Table 1), assembled with CodonCode Aligner 3.7.1 (CodonCode Co., Dedham, MA, USA) and aligned using ClustalW. Kimura 2-Parameter (K2P) distances. GC content of base and NJ trees were calculated and constructed using the MEGA X software with the Bootstrap method (500 resampling) and K2P model [18]. The distances within or among species are separately list in Table 2. The average ITS2 region was 497 bp in length in 4 species of Qinjiao, and the G+C content ranged from 55.7% to 56.2%, with an average of 56.0% (Table 2). A neighbor-joining (NJ) tree of ITS2 barcode was formed on the basis of K2P model (Fig. 2).

Representative base peak intensity (BPI) chromatograms of G. crassicaulis is shown in Fig. 3. By comparing actual mass, MS/MS fragments and retention time (RT) of target compounds, 28 constituents were identified, among which 17 were further identified by reference standards. The compound information is list in Table 3, with XIC chromatography shown in Additional file 2: Fig. S1. ANOVA followed by Tukey’s multiple comparison test were conducted for these 28 compounds (partly shown in Fig. 4).

Multivariate analysis of the TOF–MS data was carried out. Initially, unsupervised PCA-X analysis were conducted among groups, showing preferably discriminative distribution (not shown, R2X = 0.661, Q2 = 0.398). Subsequently, to maximize the variation among groups and to determine the variables that contributed to this variation, supervised OPLS-DA model (Fig. 5a) was employed among four species of Qinjiao, with R2X = 0.38, R2Y = 0.666, Q2 = 0.549. The loading plot (Fig. 5b) revealed the correlations between class (species) and variables (MASS feature), where variables clustering close to each class were considered to make great contributions to the classification. Furthermore, it was noticed that G. crassicaulis and G. dahurica clustered furthest from each other in the score plot (Fig. 5a). To explore the difference, OPLS-DA analysis for these two groups was conducted, with R2Y 0.924 and Q2 0.864 (Fig. 5c). Corresponding S-plot (Fig. 5d) was analyzed, which was commonly used and effectively showed the difference between groups. Part of the variable selected from S-plot with VIP value > 1.0 and their abundances were shown in Additional file 3: Table S2. The established OPLS-DA models were further validated by CV-ANOVA test, with P valve less than 0.05 indicating significant models (Additional file 4: Table S3).

Minimum distance within species was observed between G. macrophylla_and G. straminea, with K2P distance value 0.0146. While maximum K2P distance within species was present between G. crassicaulis and G. dahurica, with value 0.0371. The minimum interspecific distance of ITS2 region was higher than the maximum intraspecific distance, indicating that the ITS2 barcode performed well in the discrimination of four species of Qinjiao. The distances among species revealed were in accordance with discrepancy detected by following metabolomics analysis. It was illustrated that species of G. macrophylla, G. straminea, G. dahurica and G. crassicaulis can be clearly distinguished by the NJ tree. ITS2 analysis confirmed the potential discrepancy among species and guaranteed reliability of subsequent metabolomics-based species evaluation.

The score plots (Fig. 5a) shows that samples from species of G. macrophylla, G. straminea, G. dahurica and G. crassicaulis located different areas in score plot, inferring distinctive chemical profiles of four species of Qinjiao. In general, G. crassicaulis and G. straminea were closest in the score plot, while G. macrophylla and G. dahurica were progressively far away from them. Maximum spatial distance was present between G. dahurica and G. crassicaulis. The result was consistent with discrepancy revealed by ITS2 analysis. The loading plot (Fig. 5b) revealed MASS features making great contributions to the classification, which was further confirmed by ANOVA test. Flavonoids (isovitexin, morronside, quercetin) and triterpenoids (oleanolic acid, ursolic acid, roburic acid, β-sitosterone, daucosterol), as well as certain iridoids (macrophylloside A and Qinjiaoside A) were significant higher in G. dahurica. than other 3 species; while other(seco)iridoids (swertiamarin, secologanoside, loganic acid, gentimacroside, loganic acid 11-O-β-glucopyranosylester) and citric aicd were richer in other 3 species of Qinjiao than G. daurica. No obvious distinction was detected in contents of sweroside, sucrose, swertiapunimarin, and swertiajaposide A among four species of Qinjiao.

OPLS-DA score plot (Fig. 5c) and corresponding S-plot (Fig. 5d) of G. crassicaulis and G. dahurica confirmed maximum K2P distance revealed by ITS2 analysis. It was noticed that MASS features of 401.1074_4.19 (gentiopicroside), 375.1276_3.29 (loganic aicd), 419.1193_3.81(swertiamarin), 191.0200_1.09(citric acid), 563.1601_3.50(6′-O-β-d-glucosyl-gentiopicroside), and 389.1087_3.89 (secologanic acid) were richer in G. crassicaulis, while MASS features of 439.3562_15.30 (roburic aicd), 455.3510_9.93 (ursolic Acid), 455.3505_9.99 (oleanolic acid), 457.3658_14.5(β-sitosterone) and 451.1456_3.40 (morroniside) were richer in G. dahurica. The discrepancy was consistent with ANOVA test result (Fig. 4). It seems that (seco)iridoids like loganic acid, gentiopicroside or swertiamarin were richer in specie of G. crassicaulis, while flavonoid (morroniside) and triterpenoids (roburic aicd, ursolic acid, oleanolic acid, β-sitosterone) were richer in specie of G. dahurica. The discrepancy was consistent with previous reports [19, 20], which reported higher contents of gentiopicroside, loganic acid, sertiamarin, and 6′-O-β-d-glucosyl-gentiopicroside in G. crassicaulis comparing with G. dahurica.

ITS2 based gene comparison can present species distance or genetic relationship among species. Currently, ITS2 is the mostly commonly used region for the barcoding and authentication of herbal medicinal materials. In future, species authentication by ITS2 sequence (or other genetic method) may be precondition of quality evaluation of MSRHM. However, environmental implication and chemical discrepancy were not reflected. Information of quality and origin shall ultimately rely on chemical method. On the other side, TOF–MS or NMR based metabolomics analysis providing systemic chemical information among species, and can be powerful tool for quality investigations. TOF–MS bears the advantage of high sensitivity, high specificity, and thus offers maximum amount of information for quality control. However, TOF–MS based platform suffers from the disadvantage of inhomogeneous ionization propensities and time-consuming samples preparation procedures. In contrast, NMR platform possesses disadvantages of lower sensitivity and limited specificity, as well as advantages of homogenous signal response and simple sample preparations [23]. In general, a combination of ITS2 sequence comparations and TOF–MS as well as NMR based metabolomics analysis can validated each other and provided more comprehensive quality investigation for MSRHM [24].