Identification and Analysis of Compound Profiles of Sinisan Based on ‘Individual Herb, Herb-Pair, Herbal Formula’ before and after Processing Using UHPLC-Q-TOF/MS Coupled with Multiple Statistical Strategy

Sinisan has been widely used to treat depression. However, its pharmacologically-effective constituents are largely unknown, and the pharmacological effects and clinical efficacies of Sinisan-containing processed medicinal herbs may change. To address these important issues, we developed an ultra-high performance liquid chromatography coupled with electrospray ionization tandem quadrupole-time-of-flight mass spectrometry (UHPLC-Q-TOF/MS) method coupled with multiple statistical strategies to analyze the compound profiles of Sinisan, including individual herb, herb-pair, and complicated Chinese medicinal formula. As a result, 122 different constituents from individual herb, herb-pair, and complicated Chinese medicinal formula were identified totally. Through the comparison of three progressive levels, it suggests that processing herbal medicine and/or altering medicinal formula compatibility could change herbal chemical constituents, resulting in different pharmacological effects. This is also the first report that saikosaponin h/i and saikosaponin g have been identified in Sinisan.


Introduction
Chinese medicine processing and Chinese medicinal formula compatibility are two outstanding characteristics in the clinical applications of Chinese medicine. However, current studies often focus on the compatibility mechanism or processing mechanism alone without combining them together organically, and reports discussing Chinese medicine processing mechanisms in Chinese medicinal Table 1.
Identification of chemical compounds by ultrahigh performance liquid chromatography coupled with electrospray ionization tandem quadrupole-time-of-flight mass spectrometry (UHPLC-Q-TOF-MS/MS).

Multivariate Data Analysis
Using MarkerView TM 1.2.1 data handling software, multivariate data analysis were completed. The principal component analysis (PCA) score plot in negative and positive ion modes were shown in Figure 2. The results showed that all crude and processed samples including individual herb, herb-pair, and complicated Chinese medicinal formula were successfully classified into two categories in both positive and negative ion modes.

Multivariate Data Analysis
Using MarkerView TM 1.2.1 data handling software, multivariate data analysis were completed. The principal component analysis (PCA) score plot in negative and positive ion modes were shown in Figure 2. The results showed that all crude and processed samples including individual herb, herbpair, and complicated Chinese medicinal formula were successfully classified into two categories in both positive and negative ion modes.

Compounds Changed after Processing and Formula Compatibility
The variations of components (p < 0.05) in the individual herb, herb-pair, and complicated Chinese herbal formula before and after processing were shown in Tables 2 and 3 significant differences after processing. Comparing with BR, the intensity of seven peaks increased in VPBR; the other 15 peaks declined in VPBR. Taking compatibility into consideration, it was interesting to find that 14 peaks contributing to differentiate crude and processed individual herbs disappeared in herb-pair, while three new peaks (isorhamnetin-3-rutinoside, HOSSd, 2 -O-AcetylSSd) appeared.
Additionally, prosaikogenin f decreased in individual herb but increased in herb-pair. Compatibility may be responsible for these changes. On the contrary, adonltol, SSh, SSi, SSg, SSb 1 , 3 -O-AcetylSSa, and SSd all showed the same trend after processing of BR in the individual herb and herb-pair. Thus, it was hard to distinguish that the seven components were affected by processing, compatibility, or even their combination. Taking into further account the formula compatibility effect of AFI and GRM, eight peaks showing significant differences in herb-pair vanished in the formula, however seven new peaks (isorhamnetin, buddlejasaponin IV, acetylSSc, 4 -O-AcetylSSa, SSe, 6 -O-AcetylSSa, and 6 -O-AcetylSSd) appeared. Meanwhile SSg, 3 -O-AcetylSSa, and SSd showed the same tendency and this would result in the unidentifiable problem.  For PRA, 20 peaks showed significant differences after processing. Comparing with PRA, the intensity of 13 peaks enhanced in VPPRA, the other seven peaks decreased in VPPRA. Considering compatibility, 10 of these 20 peaks disappeared in the herb-pair, at the same time, 6-O-β-D-glucopyranosyl lactinolide and benzoylpaeoniflorin appeared. Also, cianidanol enhanced in individual herb but decreased in herb-pair. These changes perhaps resulted from compatibility. Moreover, nine peaks had the same trend after processing of PRA in individual herb and herb-pair, and it was also hard to distinguish as BR. Under further influence of formula compatibility with AFI and GRM, five peaks showing significant differences in herb-pair vanished in formula; oppositely, five new peaks (oxypaeoniflora, 6 -O-β-D-glucopyranosylalbiflorin, galloylpaeoniflorin, lactiflorin, benzoylalbiflorin) appeared. Formula compatibility may be responsible for these changes. In addition, seven peaks (6-O-β-D-glucopyranosyl lactinolide, mudanpioside f, albiflorin, isomaltoalbiflorin, paeoniflorigenone, paeonol, and benzoic acid) displayed an identical trend; this still led to the unidentifiable problem. Figure 3 shows the comparison of the contents of the components identified with significant differences. Processing with vinegar and formula compatibility can both regulate the acidity and alkalinity of the solution and promote changes in chemical composition, such as hydrolysis reaction, isomerization reaction, etc., resulting in increased or decreased dissolution of some components. Finally, we found that processing of BR and PRA also had the impact on AFI and GRM, and the results were shown in Table 4.   Compared with SNS, "" represents decrease in contents, "" represents increase in contents, * p < 0.05, ** p < 0.01.  Compared with SNS, "↓" represents decrease in contents, "↑" represents increase in contents, * p < 0.05, ** p < 0.01.
As shown in Table 2, the intensity of paeonol significantly increased after stir-frying with vinegar. According to a previous report [16], adding acid could greatly improve the extraction efficiency of paeonol. Since the boiling point of paeonol is 154 • C, the use of slow fire (130 • C) controlled by infrared radiation thermometer during the processing minimized the loss of paeonol. In addition, acetic acid plays an important role to form intermolecular hydrogen bonds by Van der Waals' force with paeonol, resulting in the increase of dissolution rate. Modern researches indicate that paeonol has analgesic and antiphlogistic pharmacological activities [17,18] and is consistent with TCM theory that processing of medicinal herbs with vinegar can enhance the effects of promoting blood circulation and relieving pain. As an illustration, Figure 4 revealed the course of deducing fragmentation of paeonol.
Molecules 2018, 23, x; doi: FOR PEER REVIEW www.mdpi.com/journal/molecules As shown in Table 2, the intensity of paeonol significantly increased after stir-frying with vinegar. According to a previous report [16], adding acid could greatly improve the extraction efficiency of paeonol. Since the boiling point of paeonol is 154 °C, the use of slow fire (130 °C ) controlled by infrared radiation thermometer during the processing minimized the loss of paeonol. In addition, acetic acid plays an important role to form intermolecular hydrogen bonds by Van der Waals' force with paeonol, resulting in the increase of dissolution rate. Modern researches indicate that paeonol has analgesic and antiphlogistic pharmacological activities [17,18] and is consistent with TCM theory that processing of medicinal herbs with vinegar can enhance the effects of promoting blood circulation and relieving pain. As an illustration, Figure 4 revealed the course of deducing fragmentation of paeonol. As shown in Table 3, we found that the intensity of SSa and SSd declined but the intensity of SSb2 and SSb1 increased in the BR. SSs, a kind of oleanane type triterpenoid saponin, could be divided into seven types according to their different aglycones. SSa, SSd, and SSc are epoxy-ether saikosaponins (type I), while SSb2 and SSb1 with a different aglycone, form a heterocyclic diene saikosaponin (type II) [19]. The glycosidic bond is very easily hydrolyzed in the acidic conditions or being heated [20,21]. Vinegar processing could promote the hydrolyzation from 13 to 28 allyl oxide linkage to its corresponding heteroannular diene structure, resulting in the aglycone accumulation. As shown in Figure 5, peak No. 94 was clearly observed in VPBR, VPBR-VPPRA herb-pair, SNScontaining VPBR and VPPRA, and SNS, and almost undetectable in BR and BR-PRA herb-pair. According to the fragmentations in both positive and negative ion modes and other reports [22][23][24], we suggested that peak No. 94 is SSg. SSg in SNS could be related to the acidic compounds of herbal formula, such as glycyrrhizic acid. Also, peak No. 68 (SSh/i), as the isomer of SSc, had the same change with SSg. Based on these, we hypothesized that SSa and SSd could be transformed to SSb2, SSb1, and SSg, while SSc could be converted to SSh and SSi after processing and formula compatibility. As shown in Table 3, we found that the intensity of SSa and SSd declined but the intensity of SSb 2 and SSb 1 increased in the BR. SSs, a kind of oleanane type triterpenoid saponin, could be divided into seven types according to their different aglycones. SSa, SSd, and SSc are epoxy-ether saikosaponins (type I), while SSb 2 and SSb 1 with a different aglycone, form a heterocyclic diene saikosaponin (type II) [19]. The glycosidic bond is very easily hydrolyzed in the acidic conditions or being heated [20,21]. Vinegar processing could promote the hydrolyzation from 13 to 28 allyl oxide linkage to its corresponding heteroannular diene structure, resulting in the aglycone accumulation. As shown in Figure 5, peak No. 94 was clearly observed in VPBR, VPBR-VPPRA herb-pair, SNS-containing VPBR and VPPRA, and SNS, and almost undetectable in BR and BR-PRA herb-pair. According to the fragmentations in both positive and negative ion modes and other reports [22][23][24], we suggested that peak No. 94 is SSg. SSg in SNS could be related to the acidic compounds of herbal formula, such as glycyrrhizic acid. Also, peak No. 68 (SSh/i), as the isomer of SSc, had the same change with SSg. Based on these, we hypothesized that SSa and SSd could be transformed to SSb 2 , SSb 1 , and SSg, while SSc could be converted to SSh and SSi after processing and formula compatibility.
BR, PRA, AFI, and GRM were obtained from different Chinese pharmacies and pharmaceutical factories, and authenticated by Professor Hao Cai. The quality of all collected samples was strictly evaluated and consistent with the regulations of Chinese Pharmacopoeia (Edition 2015, Part One). VPBR and VPPRA were prepared according to the processing standards described in Chinese Pharmacopoeia (Edition 2015, Part Four). The voucher specimens were deposited in School of Pharmacy, Nanjing University of Chinese Medicine (Nanjing, China).

Sample Preparation
The decoction of BR was prepared as follows. Eight grams of BR were extracted twice in a reflux water heating mantle in 48 mL and 32 mL of deionized water for 1.5 h and 1 h of reflux, respectively. The mixed solution was filtered through a four-layer mesh following the reflux. One milliliter of the
BR, PRA, AFI, and GRM were obtained from different Chinese pharmacies and pharmaceutical factories, and authenticated by Professor Hao Cai. The quality of all collected samples was strictly evaluated and consistent with the regulations of Chinese Pharmacopoeia (Edition 2015, Part One). VPBR and VPPRA were prepared according to the processing standards described in Chinese Pharmacopoeia (Edition 2015, Part Four). The voucher specimens were deposited in School of Pharmacy, Nanjing University of Chinese Medicine (Nanjing, China).

Sample Preparation
The decoction of BR was prepared as follows. Eight grams of BR were extracted twice in a reflux water heating mantle in 48 mL and 32 mL of deionized water for 1.5 h and 1 h of reflux, respectively. The mixed solution was filtered through a four-layer mesh following the reflux. One milliliter of the solution was loaded onto a C 18 RP SPE column and the gradient elution was performed as the following sequence. One milliliter of 20% acetonitrile in water (20:80, v/v), 1 mL of 40% acetonitrile in water (40:60, v/v), 1 mL of 60% acetonitrile in water (60:40, v/v), 1 mL of 80% acetonitrile in water (80:20, v/v), and 1 mL of acetonitrile. After the sequent elution, the collected eluent was eddied for 2 min and centrifuged at 13,000 rpm for 5 min. Finally, the supernatant was collected as the injection solution. The decoctions of VPBR, PRA, and VPPRA were prepared according to the same procedures above.
The decoction of BR-PRA herb-pair consisted of 4 g of BR and 4 g of PRA, and prepared as the same procedures as individual herb described above. The decoction of VPBR-VPPRA herb-pair was prepared using the same procedures as the decoction of BR-PRA herb-pair. The decoction of SNS was consist of 2 g of BR, 2 g of PRA, 2 g of AFI, and 2 g of GRM, and prepared using the same procedures as individual herb. The decoction of SNS containing VPBR and VPPRA was prepared using the same procedures as the decoction of SNS.

MS and MS/MS Data Processing and Analysis
The raw data were obtained by the Analyst TF 1.6 software (AB Sciex, Concord, CA, USA). Before data processing, a database about chemical components of medicinal herbs in SNS, including names, molecular formulas, chemical structures, and accurate molecular weights, was established by searching relevant reported literature and database websites, including PubMed and SciFinder. The data were analyzed by using PeakView TM 1.2 software (AB Sciex, Concord, CA, USA) for a perfect match with the information in the established database, according to fragmentations of the different peaks. The main parameters used were set as follows: retention time range of 0-28 min, mass range of 100 to 2000 Da, and mass tolerance of 10 ppm. By using the method of PCA with MarkerView TM 1.2.1 software (AB Sciex, Concord, CA, USA) to check for outliers and variation trend, the gathered data were more intuitionistic. The Student's t-test was performed to find out a list of peaks that were finally defined as the main contributors to the significant difference between raw and processed medicinal herbs (p < 0.05).

Conclusions
A total of 122 constituents had been identified by creative global analysis in individual herb, herb-pair, and complicated Chinese herbal formula of SNS. Taking BR as an example, 29 kinds of SSs had been identified, including some new discoveries in recent years, such as SSq, SSm, and so forth. Monoterpene glycosides (oxypaeoniflora, mudanpioside f, paeoniflorigenone, etc) showed a marked increase after processing of PRA. This is the first report of SSh/i and SSg being identified in SNS. Through three progressive levels of comparison, it suggests that processing herbal medicine and/or changing medicinal formula compatibility could alter herbal chemical constituents, resulting in different pharmaceutical effects. Herbal formula has always been the predicament of Chinese medicine research, and some scholars only employed SSd and paeoniflorin (the main components of BR and PRA) for research [25], whereas the effects between individual components and herbal formula containing individual components are quite different. We hope that the thoughts of this article would be some helpful for further research of herbal formula.

Conflicts of Interest:
The authors declare no conflict of interest.