Rapid discovery and identification of anti-inflammatory constituents from traditional Chinese medicine formula by activity index, LC-MS, and NMR

The traditional activity-guided approach has the shortcoming of low accuracy and efficiency in discovering active compounds from TCM. In this work, an approach was developed by integrating activity index (AI), liquid chromatography – mass spectrometry (LC-MS), and nuclear magnetic resonance (NMR) to rapidly predict and identify the potential active constituents from TCM. This approach was used to discover and identify the anti-inflammatory constituents from a TCM formula, Gui-Zhi-Jia-Shao-Yao-Tang (GZJSYT). The AI results indicated that, among the 903 constituents detected in GZJSYT by LC-MS, 61 constituents with higher AI values were very likely to have anti-inflammatory activities. And eight potential active constituents of them were isolated and validated to have significant inhibitory effects against NO production on LPS-induced RAW 264.7 cell model. Among them, glycyrrhisoflavone (836), glisoflavanone (893) and isoangustone A (902) were reported to have anti-inflammatory effects for the first time. The proposed approach could be generally applicable for rapid and high efficient discovery of anti-inflammatory constituents from other TCM formulae or natural products.


Results
Characterization of chemical constituents in the extract of GZJSYT by LC-MS. In order to separate as many as possible constituents in one chromatographic separation, chromatographic separation conditions, including the elution gradient and flow rate were optimized to obtain good peak capacity. Finally, the optimum chromatographic conditions were used for LC-MS analysis. The extract of GZJSYT was analyzed by LC-Q-TOF-MS and LC-IT-MS method in negative and positive ion modes to acquire the accurate mass and MS n information that was indispensable for profiling the constituents. The total ions current (TIC) chromatograms obtained by LC-IT-MS are shown in Fig. 1. It was found that most constituents were detected in negative ion mode, except several compounds could only be detected in positive ion mode (Table 1). A total number of 72 constituents, including 34 ones from Glycyrrhizae radix et rhizoma preparata cummelle, 29 ones from Paeoniae radix alba, ten ones from Jujubae fructus, six ones from Cinnamomiramulus, and three ones from Zingiberis rhizoma recens, were detected in the whole extract of GZJSYT. Among them, 58 constituents were identified by comparing with the retention times and MS n data of the reference standards or by referring to the literatures. As shown in   Fig. 2A, the identified constituents mainly included monoterpene glycosides, triterpene saponins, flavonoid and its glycosides, gallic acid derivatives, and organic acids. A total number of 16 monoterpene glycosides in the root of Paeoniae radix alba were observed in negative ion mode. Monoterpene glycosides in the formula mostly possess a "cage-like" pinane skeleton with the hemiketal group or lactone group 9,10 . In the present work, the monoterpene with similar pinane skeleton of paeoniflorin was defined as paeoniflorin type (PT), and that with similar skeleton of albiflorin was defined as albiflorin type (AT). Compounds 7,13,24,27, and 44 were identified as PTs. Their fragmentation pathways were summarized in was observed as a fragment ion in MS 3 spectrum, while for compounds 7, 24, and 27, m/z 165 was generated as base peak ions in MS 4 spectra by the loss of one glucosyl residue (162 Da) (for compound 27) or two glucosyl residues (324 Da) (for compounds 7 and 24) from the precursor ions. For compound 13, the base peak ion in MS 4 spectrum was at m/z 271, which was produced by the loss of pinane skeleton and a glucosyl segment (42 Da) from the precursor ions. Finally, compound 27 was identified as paeoniflorin by comparing with the reference standard. Compounds 7, 13, 24, and 44 were supposed to be isomaltodebenzoyl paeoniflorin, 6′ -O-galloyl desbenzoyl paeoniflorin, isomaltopaeoniflorin, and benzoylpaeoniflorin, respectively, by referring to the literatures 11,12 .     Triterpene saponins were the main constituents of Glycyrrhizae radix et rhizoma, and performed a variety of pharmacological activities [16][17][18] . In this work, 22 oleanane triterpene saponins were identified in the extract of GZJSYT. These oleanane-type triterpene saponins were analyzed in negative ion mode. Their fragment patterns were similar to that reported in our previous works 19,20 . They were grouped into four types according to base peak ions in MS 2 spectra:  were detected. So, compound 33 was characterized as apigenin 6-C-β-xylopyranosyl-8-C-(6′ ″ -O-(3-hydroxy-3-methylglutaroyl)-β-glucopyranoside) by comparing the above MS n data with that in the literature 19 .
Four organic acids (compounds 1, 4, 8, and 14) and two organic acid glycosides (compounds 10 and 11) were identified. The characteristic neutral loss of H 2 O, HCOOH (or CH 3 COOH), and CO 2 occurred in MS 2 or MS 3 spectra of four organic acids. Among them, compounds 4 and 8 were confirmed as citric acid and gallic acid, respectively, by comparing with the reference standards. Prediction of the potential anti-inflammatory constituents by AI method. The whole extract of GZJSYT was separated to 64 fractions by macroporus resin column chromatography and preparative LC. The fractions (except fractions D20, E19, and E20 due to little amount) were analyzed by LC-IT-MS to detect the chemical constituents in these fractions, and to acquire the peak area values of the constituents. As a result, a total number of 903 constituents were detected in the fractions. The anti-inflammatory activities of the whole extract and fractions of GZJSYT were evaluated on LPS-induced RAW 264.7 macrophages at corresponding concentrations without cytotoxicity. As shown in Fig. 4, the whole extract of GZJSYT (A01) exhibited lower inhibition rate against NO production at about 26%. But fractions B03, B04, C19, D04-D08, D11-D19, E01, and E03-E18 performed better inhibition rate than indomethacin (positive control). In particular, fractions B04, D07, D13, D14, E07-E15, E17, and E18 exerted inhibition rate against NO production at more than 80%.
The AIs of the constituents in GZJSYT were calculated according to the equation (1). As shown in Fig. 5, 61 constituents had higher AIs, which existed mainly in fractions E15, E16, E17, and E18. The greater the AI value is, the more likely the constituent is active 1 . Therefore, the 61 constituents were speculated to have anti-inflammatory activities.
Isolation and identification of eight potential active constituents. In order to verify the pre-

Discussion
The chemical profile of GZJSYT was firstly reported in this work. The study of fragmentation rules of MS n spectra for different types of constituents was critical for the rapid identification of their chemical structures by LC-MS. The fragment pathway of monoterpenoids, the main constituents in GZJSYT, was investigated in detail.
Scientific RepoRts | 6:31000 | DOI: 10.1038/srep31000 It was found that paeoniflorin type monoterpenoids mainly eliminated HCHO due to the unstable hemiketal group of pinane skeleton in MS n spectra, and then fragmented the substituent group R 2 , such as BA and Glc, from the mother skeleton under higher collision energy to form typical fragment ion at m/z 165. However, albiflorin type monoterpenoids firstly fragmented the substituent group R 2 , and then lost the lactone group of pinane skeleton under higher collision energy, which was different from the fragmentation behavior of paeoniflorin type monoterpenoids.
Prefraction was not only helpful to decrease the chemical complex of TCM formulae, but also necessary to find the trace constituents that had significant activity. In this work, 903 constituents were detected in GZJSYT fractions while only 72 constituents in the whole extract of GZJSYT. Considering that the identification of chemical structures of those 903 constituents in the fractions was time-consuming even by LC-MS, only the potential active constituents were isolated and identified unambiguously by NMR.
Among the 61 fractions, 40 fractions had anti-inflammatory effects. Many constituents with positive AI values might have contributions to the anti-inflammatory effects. The low inhibition rate against NO production of A01 might be due to the weak activity of constituents with higher contents. However, in fractions, the trace active constituents were enriched, so many fractions had higher inhibition rate against NO production.  These active fractions contained hundreds of constituents. It was not easy to find the active constituents rapidly without the guidance of AI method. It is generally considered that the main constituents in active fraction may be active, but sometime it is not the case. Here we take the active fractions D05 as an example. Fractions D05 exhibited 53% of inhibition rate against NO production. The LC-MS chromatogram of D05 is shown in Fig. 7. It was found that paeoniflorin (27) was the main constituent of fraction D05. It was easy to think that paeoniflorin (27) might be active. As shown in Table 2, the AI value of paeoniflorin (27) was − 0.18. According to our previous work 1 , negative AI value indicated that the constituents might have no anti-inflammatory activity or weak activity. In fact, paeoniflorin (27) was proved experimentally to have no anti-inflammatory activity at 50 μ M, only had weak anti-inflammatory effect at 200 μ M (NO inhibition rate of 38%) 26 . The result indicated the contribution of paeoniflorin (27) to the anti-inflammatory of fraction D05 was weak. Therefore, activity index could predict the probability of the constituents being active to some extent.
The AI results indicated that 61 constituents with higher AI values might have anti-inflammatory activities, and most of them were from Glycyrrhizae radix etrhizoma. The structures of eight active constituents were confirmed by NMR, which included one coumarin and seven flavonoids. Interestingly, most of their structures contained isopentene groups, which might be important for their significant anti-inflammatory activities.

Conclusion
In summary, an approach based on AI, LC-MS, and NMR was developed to rapidly discover and identify the anti-inflammatory constituents from GZJSYT in this study. A total number of 903 constituents were detected in the whole extract and fractions by LC-MS. The AI predicted results showed that 61 constituents with higher AIs were very likely to have anti-inflammatory activities. Eight potential active constituents of them were validated to have significant anti-inflammatory activity in LPS-induced RAW 265.7 macrophages. And more importantly, three of them were first reported to have anti-inflammatory activity in this work.
Calculating activity index of constituents requirs the global analysis of the fractions from TCM formulae, not just the active fractions, which is different from the traditional activity-guided method that only focuses in active fractions. The results indicated that the approach was efficient and feasible. This approach would be generally applicable for the rapid discovery of anti-inflammatory constituents from other TCMs or medical plants.    Dulbecco's modified Eagle's medium (DMEM), Fetal Bovine Serum (FBS), trypsin-EDTA and the penicillin-streptomycin were purchased from Gibico BRL (Grand Island, NY, USA). Lipopolysaccharides (LPS), dimethylsulfoxide (DMSO) and indomethacin (purity ≥ 99%) were acquired from Sigma-Aldrich (St. Louis, MO, USA).

Preparation of sample and standard solutions. According to the recipe of GZJSYT in Shang-Han-Lun,
Cinnamomi ramulus (27.3 g), Paeoniae radix alba (54.6 g), Zingiberis rhizoma recens (27.3 g), Jujubae fructus (72.7 g), and Glycyrrhizae radix et rhizoma preparata cum melle (18.2 g) were immersed in 1.2 L water (6-fold) overnight. Then, they were extracted twice for 1 h of each time and the filtrates were combined. Subsequently, the combined solution was concentrated under reduced pressure using a rotary evaporator, and the residue was dissolved in appropriate volume of water. Finally, the solution was analyzed by LC-MS after it was filtered through 0.22 μ m filter membrane.
The fractions of GZJSYT were prepared as previously reported 1 . Firstly, the extract of GZJSYT was loaded onto a glass column (4.6 cm × 30 cm) packed with the D101 macroporous resin. The column was rinsed with H 2 O first to get fraction B01, and then in turn eluted with 20% ethanol, 40% ethanol, 95% ethanol to give fractions B02, B03, and B04, respectively. Fractions B02, B03 and B04 were subject to preparative HPLC to obtain new subfractions. Subfractions preparation was carried out on a 1200 series LC system (Agilent, Palo Alto, USA) and a Zorbax SB-C18 column (21. LC-Q-TOF-MS conditions. LC-Q-TOF-MS analysis was manipulated on an Acquity TM ultra performance LC system (Waters Corp., Milford, MA, USA) coupled with a high resolution Triple TOF 5600 + (AB SCIEX, Framingham, MA, USA), which was equipped with an ESI source. The chromatographic separation conditions were the same as that in LC-IT-MS. The sample was analyzed in both positive and negative mode, and the parameters in the source were set as follows: curtion gas (CUR), 30 psi; ion source GS1, 50 psi; ion source GS2, 50 psi; source temperature, 550 °C (ESI − ) and 600 °C (ESI + ); ionspray voltage floating (ISVF), − 4500 V(ESI − ) and 5500 V (ESI + ); collision energy (CE), 10 V; declustering potential (DP), ± 100 V; ion release delay (IRD) at 67; ion release width (IRW) at 25. The mass range was m/z 100-1500. The accurate mass acquired were processed by means of the elemental composition calculator incorporated in the PeakView ® software (AB SCIEX).
Isolation and identification of potential active constituents. The Glycyrrhizae radix et rhizoma preparata cum melle (5 kg) was extracted with 95% ethanol (8-fold) for 1 h at the first time. Then the extraction was repeated with 95% ethanol (6-fold) for 1 h. The two extracts were mixed together, and condensed to appropriate concentration for separation experiments. The extract was loaded on a glass column (10 cm × 250 cm) packed with the D101 macroporous resin, and eluted with H 2 O, 40% ethanol and 95% ethanol, respectively. Effluent of 95% ethanol was collected and condensed to appropriate volume. Subsequently, the sample was further separated by medium-pressure column chromatography (10 cm × 50 cm) loaded with ODS. The mobile phases consisted of H 2 O (A) and methanol (B) and the flow speed at 60 mL/min. Isocratic elution with 55% B was carried out, and the eluate was collected and dried for semi-preparation.
Evaluation of the anti-inflammatory activity. The murine macrophage cell line RAW 264.7 (the Type Culture Collection of the Chinese Academy of Sciences, Shanghai, China) was cultured in 90% Dulbecco's Modification Eagle's Medium (DMEM) with 4.5 g/L glucose, L-glutamine and sodium pyrurate supplemented with 10% inactivated Fetal Bovine Serum (FBS) and 1% 10000 U/mL of penicillin, 10000 μ g/mL of streptomycin at 37 °C in a 5% CO 2 humidified atmosphere.
To determine the cytotoxicity of each constituent, RAW 264.7 cells (5 × 10 4 cells/mL, 100 μ L/well) plated in 96 well plates were cultured for 24 h at 37 °C. The samples dissolved in DMSO were diluted with DMEM supplemented with 10% inactivated FBS, penicillin G (100 U/mL) and streptomycin (100 μ g/mL) before co-incubated with the cells for 24 h. Then, MTT solution (100 μ L, 0.5 mg/mL in per well) was added and incubated for 4 h. The solution in per well was drawn and the formazan crystals were adequately dissolved in DMSO (100 μ L/well) by shaking for 10 min at 37 °C. The absorbance at 580 nm was measured in a microplate reader (Bio-Tek ELX800, Winooski, VT, USA). Fresh culture medium was used as a blank in every experiment. The experiments were implemented for three times in parallel.
For the determination of the anti-inflammatory activity, RAW 264.7 cells were seeded into a 96-well plate at a density of 2 × 10 4 cells/well for 24 h. Then, the fresh culture medium containing samples and LPS (200 ng/mL) were added to each well (140 μ L/well) and further incubated for 24 h. Subsequently, 100 μ L of supernatant from each well was mixed with 100 μ L of Griess reagent (0.5% sulfanilamide and 0.05% naphthylene diamide dihydrochloride in 2.5% H 3 PO 4 ) in a separate 96-well plate. After an incubation of 10 min, the absorbance of NO was determined at 535 nm with a microplate reader. Indomethacin (50 μ Μ ) was adopted as a positive control and fresh culture medium was used as the blank in all experiments. The experiments were carried out for three times in parallel.

Calculation of AI values of constituents.
The AI values of constituents were calculated using mathematical equations reported in the previous work 1 .