A comparison of the chemical compounds in single- and co-boiled Zhi-Zi-Da-Huang decoctions with UPLC-qTOF-MS/MS and UPLC-DAD

Zhi-Zi-Da-Huang decoction (ZZDHD) is a famous Traditional Chinese Medicine decoction due to its therapeutic effects on clinical hepatobiliary disorders. ZZDHD is composed of Gardeniae Fructus, Rhei Radix et Rhizoma, Fructus Aurantii Immaturus, and Sojae Semen Praeparatum. With the development of current technology, dispensing granules have been widely used for convenience. However, limited research has been conducted to determine differences in the chemical compounds between dispensing granules and traditional decoction. A strategy based on UPLC-qTOF-MS/MS and UPLC-DAD was established to quantitatively and qualitatively analyze the chemical compounds present in single- and co-boiled ZZDHD. First, we utilized UPLC-qTOF-MS/MS to identify the compounds in single- and co-boiled ZZDHD. Then, 15 compounds were quantitatively analyzed in ZZDHD by UPLC-DAD. Finally, ngerprint and chemometric analyses were adopted to evaluate the difference between single- and co-boiled ZZDHD. Zhi-Zi-Da-Huang decoctions. First, chemical compounds were rapidly analyzed by UPLC-qTOF-MS/MS, and the signicantly changed components were identied by comparing their mass spectra with the LC-MS/MS library. Second, 15 compounds in ZZDHD were quantied by UPLC-DAD. Finally, based on the qualitative analysis of UPLC ngerprints for comparison of various samples, several chemometric analyses, including principal component analysis (PCA), partial least-squares discrimination analysis (PLS-DA), and hierarchical cluster analysis (HCA), were achieved to further evaluate the differences in quality of these ZZDHD samples. The results obtained from this study could provide a meaningful basis for explaining the rationality of single- and co-boiled ZZDHD. Determination and chemometric analysis of the ngerprinting were integrated. In 30 common peaks, we found that crocetin, daidzin, physcion, G2, chrysophanol, and emodin with contents changed were included in 15 quantitative The change in content be one of the reasons responsible for the discrimination of CB and SB; therefore, these control markers to differentiate two methods of preparation of ZZDHD. To our knowledge, the quality standard of by the Pharmacopoeia of Republic of China. The quantication, ngerprinting methods, and results of the 15 compounds in study could provide reference for the establishment of quality control of ZZDHD.


Mass spectrometry
High-accuracy mass spectrometric data were recorded on a Waters Xevo G2-XS Q-Tof mass spectrometer (

UPLC-DAD conditions
The UPLC-DAD was carried out on a Shimadzu LC-30A HPLC system, consisting of a Shimadzu LC-30AT dual pump, a Shimadzu SIL-30A autosampler, a CTO-30A column compartment, and an SPD-M30A photodiode array detector. The quantitative samples were detected at 254 nm, 283 nm, and 440 nm. Other chromatographic conditions, such as the chromatographic column, mobile phase composition, gradient program, ow rate, column temperature, and injection volume, were the same as those mentioned in Sect. 2.3.1. Each sample was measured in parallel three times.

Validation of the quanti cation method
The system precision was determined by examining six replicate injections of the same sample solution within a day and performing the procedure on three consecutive days. Six independently prepared solutions from the same sample were determined to check the repeatability. The samples were measured at 0, 2, 4, 8, 12, and 24 h to verify for stability. The recovery test was conducted to evaluate the accuracy of this method. The CB sample (0.1103 g, which equated to 0.5 g crude ZZDHD) was weighed independently six times and spiked with a known amount (approximately equivalent to 0.8, 1.0, and 1.2 times the amount that the actual sample contained) of the corresponding standard compounds. Then, the spiked samples were extracted and quanti ed with the methods mentioned above.

UPLC-DAD conditions
The qualitative samples were detected at 254 nm, 280 nm, and 440 nm as described in section '2.4.1'.

Validation of the ngerprint method
Precision, repeatability, and stability were determined as described in section '2.4.2'.

Similarity analysis of UPLC ngerprints
To establish the representative chromatographic ngerprint, a total of 12 samples (six batches of CB (S1-S6), six batches of SB (S7-S12)) were analyzed following the abovementioned methods. The chromatography ngerprint was established using the 'Similarity Evaluation System of Traditional Chinese Medicine Chromatographic Fingerprint' software (China Committee of Pharmacopeia, 2012).

Statistical analysis
The qTOF-MS/MS data of 12 samples from two methods of ZZDHD preparation were processed using Masslynx 4.1 software. HCA, PCA, and PLS-DA were implemented to characterize the dissimilarities in chemical components among different samples with regard to the processing method. These were operated by SIMCA-P software 14.0 (Umetrics, Umeå, Sweden). Nonparametric tests were used to evaluate the signi cant differences between the different groups by GraphPad Prism software (GraphPad Prism Software Inc., CA, USA). P < 0.05 was considered to be signi cant.

UPLC-qTOF-MS/MS qualitative analysis of chemical constituents in ZZDHD
The MS spectra were obtained in both positive and negative ion modes. Total ion chromatograms of samples were obtained using the optimized UPLC-qTOF-MS/MS system (Fig. 2)

Identi cation of compounds in Gardeniae Fructus
A total of 61 compounds were obtained from ZZ, of which were 30 known compounds. ZZ contained a large number of iridoid glucosides, including gardoside, shanzhiside, genipin-1-β-D-gentiobioside, geniposide, and gardenoside. Crocins, including crocin-I and crocin-II, were deemed the main active compounds in ZZ [30]. Geniposide, the most abundant iridoid glucosides in ZZDHD, and crocin-I, the most abundant crocin, were used to characterize the fragmentation pathways, which are shown in Fig. A1.

Identi cation of compounds in Sojae Semen Praeparatum
The major active constituents of DDC were iso avones [24]. It contained 24 compounds, including eight known compounds, such as daidzin, daidzein, and genistein [31]. Peak 117 (t R =12.79) and 135 (t R =13.46) were recognized as rhein and emodin, respectively, and they were used to characterize the fragmentation pathways ( Fig. A3). In the negative ion mode, peak 117 showed [M-H]-at m/z 283.0239, and then generated a fragment ion at m/z 239.0342, owing to the loss of a CO 2 residue. Subsequently, the fragment m/z 211.0391 was yielded due to the loss of CO from 239.0342. Lastly the ion further lost one molecule of CO to produce an ion at m/z 183.0441. Based upon these data and the literature[28], the compound was rhein. Peak 135 showed [M-H]-at m/z 269.0453, and then generated fragments with m/z 241.0499 and m/z 225.0549, which was due to the loss of CO and CO 2 . Based upon these data and previously published research [29], peak 135 was determined to be emodin. Other constituents of anthraquinones in DH were identi ed using this fragment information (Table 2).

Method validation
Quantitative method validation for the established UPLC-DAD analysis was performed for linearity, precision, stability, repeatability, and recovery, as shown in Table 3. All correlation coe cient values (r > 0.999) demonstrated a good linear relationship between the analyte concentrations and their peak areas within the relatively wide test ranges. The relative standard diviation (RSD) of precision with 15 analytes ranged from 0.29-3.97%. The RSDs for stability were less than 4.22%. As for repeatability, the RSDs were not more than 4.36%. The developed method also had suitable accuracy with spike recovery of 95.35-107.95% for all analytes. These results demonstrated that the proposed quantitative UPLC-DAD method was linear, precise, stable, repeatable, and su ciently accurate for simultaneous determination of fteen compounds in ZZDHD samples.

Analysis of content determination
This study investigated the differences in the chemical compounds of ZZDHD prepared following two different boiling methods. Representative chromatograms of one sample (S1) and reference standards at 254 nm, 283 nm, and 440 nm are presented in Fig. 3. A total of 12 batches of ZZDHD samples from the preparation methods were collected, prepared, and analyzed as described above. Fifteen chemical substances. Speci cally, G1, G2, daidzin, narirutin, naringin, crocin-I, hesperidin, neohesperidin, daidzein, genistein, rhein, crocetin, emodin, chrysophanol, and physcion ( Fig. 1) were chosen as the reference standards for quantitation. The content of each marker compound was determined at its maximum absorption wavelength, using the external standard method. The quantitative data for the marker compounds are presented in Table 4. A slightly variation in the content of other marker compounds (RSD < 30%) was observed, except for the physcion content. A relatively high content of G2, naringin, and neohesperidin were present in both the single-co-boiled samples.
As shown in Fig. 4, the difference was statistically signi cant between the contents of G2, emodin, and chrysophanol in CB and SB (P < 0.05), and the difference was highly statistically signi cantly different between daidzin, physcion (p < 0.01), and crocin-I (p < 0.001). There were no signi cant differences among the other nine compounds.

Method validation
A comprehensive validation of the established UPLC ngerprint method was carried out, including assessments of precision, repeatability, and stability, and the results were summarized in Table 5. The intra-and inter-day variations of all 23 common peaks (RSDs) were within 0.02-0.19% and 0.05-0.32% for the relative retention time (RRT), as well as 1.62-4.73% and 1.83-4.87% for the relative peak area (RPA), respectively.
In the precision, repeatability, and stability tests, RSDs of the RPA from all 30 common peaks were less than 3.80%, 2.74%, and 2.54%, respectively (Table 5).
These results showed that the UPLC ngerprinting method was applicable for the qualitative analysis of ZZDHD.

Fingerprinting analysis
Twelve batches of samples were analyzed under optimized conditions. As shown in Fig. 5, all samples showed similar chromatographic pro les. Peaks existing in all samples were designated as common peaks, and 30 common peaks were found, including 12 common peaks identi ed at 254 nm, six common peaks identi ed at 280 nm, and 12 common peaks identi ed at 440 nm. According to the representative standard peak area of the ngerprint calculated by the Similarity Evaluation System Chromatographic Fingerprint of Traditional Chinese Medicine Chromatographic Fingerprint software, the RPA of the ZZDHD samples was calculated. The peak of the representative standard ngerprint was designated as the reference peak (RPA = peak area of characteristic peak/peak area of reference peak). As shown in Table A1, changes in the contents of single-and co-boiled ZZDHD decoctions were analyzed by RPA. The results indicated that the RPA varied dramatically between samples at 254 nm and 440 nm, and there were two and seven common peaks with an RPA > 15%, respectively, which demonstrated that the content of the single-and the co-boiled was different.

Evaluation of similarity
Similarity evaluation software was utilized to calculate the similarity based on the "vector cosine" method [32] between different samples of ZZDHD and the control ngerprint. The control ngerprint was generated from 12 batches of ZZDHD samples (Table 6). The results showed that the similarity of single-and co-boiled ZZDHD was higher than 0.997 at 254 nm, 280 nm, and 440 nm, indicating that there was no signi cant change in chemical substances in the two different preparation methods.

Chemometric analysis
Further quality assessment of the samples was performed using a PCA on 12 samples with 30 common peaks. The results showed that there was a separation between CB and SB (Fig. 6A). These samples were clearly clustered into two groups: Group 1 including six samples (S1-S6) and Group 2 including six samples (S7-S12), which indicated that the different preparation methods resulted in changes to the ZZDHD. PCA was also employed to map the RPA of 30 common peaks onto the samples in the loading plot (Fig. 6B), which was utilized to nd the compounds accountable for the separation of the groups.
Before subjecting different samples to HCA, 30 common peaks were selected. To visualize the classi cation trends in ZZDHD samples, HCA was conducted based on the RPA of 30 common peaks, which could divide tested samples into different categories. As the dendrogram showed in Fig. 6C, 12 batches of ZZDHD samples were sorted into two clusters. All CB samples (S1-S6) were grouped as group I, and SB samples (S7-S12) were grouped as group II. This corresponded to the PCA plot shown in Fig. 6A, although the similarity values of these samples were much closer. It was concluded that the classi cation of the sample quality might be mostly relevant to the boiling method of ZZDHD.
Based on the results of the HCA, PLS-DA was further established to obtain important variables with CB and SB samples. The VIP plot displays the importance of variables both to explain X correlated to Y. Based on the VIP values (i.e., larger than 1.0), 13 compounds played key roles in the clustering (Fig. 7). These 13 compounds, including C3, A3, crocetin, daidzin, C5, physcion, G2, G1, chrysophanol, emodin, A2, A5, and C6, corresponded to those points far away from zero in the loading plot of 30 common peaks (Fig. 6B).

Discussion
Currently, Traditional Chinese Medicine decoctions and formula granules have been widely studied and applied around the world. The study of chemical compounds is the key factor for the development of traditional Chinese medicine decoctions into formula granules. Unambiguous chemical substances guarantees the quality and therapeutic e cacy of Traditional Chinese Medicine decoctions, so investigations of Traditional Chinese Medicine is essential.
However, the chemical composition of Traditional Chinese Medicines is rather complicated, which poses great challenge for the separation and identi cation of their chemical compositions. UPLC-qTOF-MS/MS has allowed for the perfect combination of ultra-high performance liquid chromatography and highquality mass spectrometry with high resolution, which have improved performances compared with LC-MS. UPLC uses small particle packing (1.7 µ m) as the stationary phase to achieve faster, more sensitive, and higher resolution chromatographic separation of samples, which has advantages over traditional HPLC.
QTOF-MS/MS with high quality resolution and high sensitivity can accurately determine molecular weight and predict the molecular composition [33]. Previous studies showed that iridoids, anthraquinones, avonoids, iso avonoids, coumarins, and crocetin glycosides were the major bioactive compounds in ZZDHD [28]. Based on this, we determined 15 representative compounds by UPLC-DAD in order to further evaluate the quality of samples and comprehensively compare the differences of chemical substances in ZZDHD. ZZDHD has been extensively used for clearing heat and eliminating dampness and jaundice for many years in China, and it is further thought to have liver-protective, choleretic, and anti-in ammatory effects. A previous study has reported that geniposide could prevent hepatocyte injury [34]. Geniposide is hydrolyzed into genipin under the action of beta-D-glucosidases in intestinal bacteria, which is required for its choleretic action [35]. Genipin could also in uence the induction of the in ammatory mediators inducible nitric oxide synthase (iNOS) and tumor necrosis factor (TNF)-α through the inhibition of nuclear transcription factor-κB (NF-κB) activation in hepatocytes; therefore, it might have therapeutic potential for liver injury [36]. Moreover, as one of the major pharmacodynamic substances, G1 generated geniposide and genipin via deglycosylation involving intestinal bacteria, and then distributed in the target organ for hepatinica and cholagogue [37]. As one of the primary constituents of crocin, Crocin-I could reduce oxidative damage in liver tissue in diabetic rats [38]. Crocetin exhibited inhibitory potential against cyclooxygenase-2 (COX-2), iNOS, proin ammatory factors (interleukin [IL]-1β, IL-6, IL-10, and TNF-α) in in vitro assays, and it signi cantly decreased hepatic parameters, such as serum glutamic-oxaloacetic transaminase (SGOT), serum glutamic pyruvic transaminase (SGPT), and alkaline phosphatase (ALP), which suggested protective and therapeutic effects on in ammatory tissue [39]. DH and its anthraquinone constituents, including rhein, emodin, and chrysophanol, could exert a protective effect against liver injury[18, [40][41][42]. Emerging studies suggested that physcion suppressed hepatocellular carcinoma cell metastasis [43]. Narirutin, naringin, hesperidin, and neohesperidin, the active constituents of ZS, could inhibit pro-in ammatory factors, attenuate liver damage, modulate the state of antioxidants, and have hypoglycemic and hypolipidemic effects [44][45][46][47]. Some studies have revealed that the major iso avonoids of DDC, such as daidzein, have a signi cant hepatoprotective effect against oxidative damage induced by tert-butylhydroperoxide (TBHP) in rat hepatocyte BRL-3A cell line, which is closely related to the e cacy of ZZDHD [15,16,18]. Some pharmacokinetics indicated that iso avone glycosides of DDC, including daidzein and genistin, could be hydrolyzed in the intestinal wall by bacterial β-glucosidases after oral administration, and further transformed into corresponding bioactive aglycones, and all of them could be excreted into bile [48]. As a glucoside form of iso avones, daidzin markedly reduced the elevated serum aminotransferase activity and attenuated the apoptosis of hepatocytes [49]. Another study reported that the complex interactions among four herbs may result in a synergistic or antagonistic effect on the solubility of active compounds [50]. Therefore, it was reasonable to select these quantitative compounds, and all of 15 quantitative compounds were representative compounds in four herbs as well.
Then we used ngerprinting to evaluate the similarity of 30 common peaks in 12 batches of samples. The number and retention time of peaks in CB were similar to that of SB, but the peak area of many peaks increased. We therefore conclude that co-boiling was bene cial to the dissolution of effective herbal compounds, providing scienti c support for traditional decoction preparation. Determination and chemometric analysis of the ngerprinting were integrated.
In 30 common peaks, we found that crocetin, daidzin, physcion, G2, chrysophanol, and emodin with contents changed were included in 15 quantitative compounds. The change in content might be one of the reasons responsible for the discrimination of CB and SB; therefore, these compounds may be effective quality control markers to differentiate the two methods of preparation of ZZDHD. To our knowledge, the quality standard of ZZDHD has not yet been described by the Pharmacopoeia of the People's Republic of China. The quanti cation, ngerprinting methods, and results of the 15 compounds presented in this study could provide a reference for the establishment of quality control of ZZDHD.

Conclusion
In the present research, a rapid and e cient method to integrate UPLC with UPLC-qTOF-MS/MS analysis was successfully established to comprehensively discriminate and assess the quality of ZZDHD samples. Simultaneously, combining chemometric analysis with chromatographic ngerprints served as a powerful tool to apply to the quality control of the co-and single-boiled ZZDHD. We identi ed a total of 147 chemical compounds, 15 of which were quanti ed and 12 batches of samples were qualitatively ngerprinted for a systematic analysis of the chemical substances of the two preparation methods of ZZDHD.
This study shows the difference of chemical substances between two preparation methods of ZZDHD, which provided an experimental basis for explaining the rationality of the preparation methods. However, research remains limited on the pharmacodynamics, clinical veri cation, and correlation between changes in chemical composition, pharmacodynamics, and clinical e cacy of ZZDHD prepared via different methods; therefore, further research is required.    UPLC overlap of the co-boiled Zhi-Zi-Da-Huang decoction (S1-S6), the single-boiled Zhi-Zi-Da-Huang decoction (S7-S12) and the representative ngerprint (R).

Figure 7
Variable importance in projection (VIP) plot of the co-boiled Zhi-Zi-Da-Huang decoction (CB) and the single-boiled Zhi-Zi-Da-Huang decoction (SB) with con dence intervals.