Comprehensive metabolomics analysis based on UPLC-Q/TOF-MSE and the anti-COPD effect of different parts of Celastrus orbiculatus Thunb.

The root, stem and leaf of Celastrus orbiculatus Thunb. (COT) have all been used as Chinese folk medicine. Aiming at revealing the secondary metabolites and screening the anti-COPD effect of COT, the comprehensive phytochemical and bioassay studies were performed. Based on the ultra-high performance liquid chromatography combined with quadrupole time-of-flight mass spectrometry (UPLC-Q/TOF-MSE), the screening analysis of components in COT was conducted with the UNIFI platform, the metabolomics of the three parts were analyzed with multivariate statistical analysis. Cigarette smoke extract (CSE)-stimulated inflammatory model in A549 cells was used to investigate the biological effect of the three parts. A total of 120 compounds were identified or tentatively characterized from COT. Metabolomics analysis showed that the three parts of COT were differentiated, and there were 13, 8 and 5 potential chemical markers discovered from root, stem and leaf, respectively. Five robust chemical markers with high responses could be used for further quality control in different parts of COT. The root, stem and leaf of COT could evidently reduce the levels of pro-inflammatory factors in a dose-dependent way within a certain concentration range. The stem part had a stronger anti-COPD effect than root and leaf parts. This study clarified the structural diversity of secondary metabolites and the various patterns in different parts of COT, and provided a theoretical basis for further utilization and development of COT.


Introduction
Celastrus orbiculatus Thunb. (COT), belonging to the family Celastraceae and the genus Celastrus L., is widely distributed throughout China. 1 The parts including root, stem and leaf all could been used as Chinese folk medicine to treat rheumatoid arthritis, vomiting, abdominal pain, and snakebites. 2,3 The root of COT was reported to possess anti-tumor, 4 antiviral, 5 bacteriostatic 6 and lipid-lowering 1 activities, and about 50 compounds, including triterpenes, sesquiterpenoids, steroids and organic acids, were isolated from the root or root bark. 3,[7][8][9] The stem was also reported to have anti-inammation, 10 anticancer 11 and fatty liver amelioration 12 effects, and nearly 100 compounds including triterpenes, diterpenoids, steroids, avonoids, phenolics and benzoquinone were isolated. [13][14][15][16] For the leaf part, previous studies have found that the extract of it had insecticidal effect and hypoglycemic effect, 17 while only a few avonoids were isolated. 18 It was revealed that there were signicant variation for the contents of celastrol or total alkaloids in different parts of COT. 19,20 Along with continuous expansion of the folk and clinical application of COT, an indepth study on the chemical constituents in different parts of COT has attracted more and more attention. However, the comprehensive comparative study on the chemical composition between root, stem and leaf parts of COT has not been reported so far.
Recently, the UPLC-Q/TOF-MS method has been innovatively used for screening and identifying chemical components in herbal medicines and traditional Chinese medicine. And the global proling of various metabolites were reported. As part of these research works, we reported this method to detect some natural products including Platycodon grandiorum and Ginseng root. 21,22 Our research results showed that this method is high throughput, comprehensive, simple and efficient. As far as we know, the UPLC-Q/TOF-MS method has not been reported to identify the components in COT. So, the study in this paper comparatively analyzes the phytochemicals of root, stem and leaf parts of COT by using the UPLC-Q/TOF-MS method for the rst time and nds out the similarities and differences between them.
Chronic Obstructive Pulmonary Disease (COPD), predicted to rank as the third leading cause of death in the world, 23 is mainly caused by signicant exposure to harmful gases or particles. 24 Cigarette smoking was the leading environmental risk factor for COPD around the world, and cigarette smokers were more likely to develop respiratory symptoms and had a higher COPD mortality rate. Along with the progressive lung inammation, some pro-inammatory mediators such as IL-1b, IL-6 and TNF-a participated in the occurrence and development of COPD. 25 Although the COT had been used in treating various inammatory diseases, the effect on the cigarette smoke extract (CSE)-induced inammatory reaction has not been reported so far.
In the present study, the main medicinal parts of COT (root, stem and leaf) were chosen as the test sample. On one hand, the similarities and differences of phytochemicals in three parts were analyzed by using UNIFI platform and untargeted metabolomics based on UPLC-Q/TOF-MS E . The components and potential chemical markers to prole diverse classications of metabolites of three parts were investigated. On the other hand, the effects on CSE-induced inammatory reaction of these three parts were explored in A549 cells. The anti-COPD activity of different parts was preliminarily discussed. This comprehensive study could reveal the structural diversity of secondary metabolites and the different patterns of main medicinal parts of COT, and provide the data for further clinical application in anti-COPD. The study on the phytochemistry and the pharmacological activity of various parts were both signicantly valuable to the research and development of COT.

Materials and reagents
A total of 10 batches of fresh COT were collected from different growth areas in China (Table 1). All herbs were authenticated by the authors according to Hunan Province Local Standard for Traditional Chinese Medicine (2009 edition) for "Celastrus orbiculatus Thunb.". The corresponding specimens had been deposited in the Research Center of Natural Drug, Jilin University, China.
Methanol and acetonitrile were of LC/MS grade purchased from Fisher Chemical Company. Formic acid was bought from Sigma-Aldrich Company, St. Louis, MO, USA. Deionized water was puried by Millipore water purication system (Millipore, Billerica, MA, USA). All other chemicals were analytically pure. Cigarettes for bioassay analysis were Xiongshi cigarette (China Tobacco Zhejiang Industrial Co., Ltd, Hangzhou, China), each cigarette contained 11 mg of tar, 0.7 mg of nicotine, and 13 mg of carbon monoxide. Human lung carcinoma A549 cells were obtained from the Department of Pathogen Biology, Basic Medical College, Jilin University. ELISA kits were bought from Nanjing Jiancheng Bio-engineering Institute.

Sample preparation of three parts of COT
Took the whole fresh COT, and separated the root, stem and leaf part respectively to get 30 test samples including root part (R1-R10) samples, stem part (S1-S10) samples and leaf part (L1-L10) samples. The aforementioned parts were air-dried, grinded and sieved with Chinese National Standard Sieve No. 3, R40/3 series to obtain the homogeneous powder respectively. Each powder was weighted (2.0 g) accurately and extracted thrice (3 hours per time) with 100 mL of 80% methanol at 80 C, cooled, ltered, collected and combined the ltrate of each sample, concentrated and evaporated to dryness.
For metabonomics analysis, each residue (all approximately 2.0 mg) was dissolved in 1.0 mL of 80% methanol respectively, aer being ltered with a syringe lter (0.22 mm), 30 test solutions (R M1 -R M10 , S M1 -S M10 and L M1 -L M10 ) were obtained, which was injected into the UPLC system directly. Furthermore, to ensure the suitability consistency and the stability of MS analysis, a sample for quality control (QC) was prepared by pooling 20 mL from every test solution, namely containing all of the constituents in this analysis.
For screening analysis, the test solutions of root part (R S ), stem part (S S ) and leaf part (L S ) were prepared by pooling 100 mL from R M1 -R M10 , S M1 -S M10 and L M1 -L M10 solutions, respectively.
For bioassay analysis, the test samples (R bio , S bio and L bio ) of root part, stem part and leaf part were prepared by combining each residue of R 1 -R 10 , S 1 -S 10 and L 1 -L 10 , respectively. Then, R bio , S bio and L bio were dissolved in water at the concentration of 3.2 mg mL À1 to get the stock solutions stored in 4 C.
2.3. Ultra-high performance liquid chromatography combined with quadrupole time-of-ight tandem mass spectrometry (UPLC-Q/TOF-MS E ) The separation and detection of components were performed on the UPLC system combined with Xevo G2-XS Q/TOF mass    40 min, 10% B) with a ow rate of 0.4 mL min À1 . Set the temperature of column and the sample manager at 30 C and 15 C, respectively. 10% and 90% acetonitrile in aqueous solution were used as weak and strong wash solvents respectively.
The optimized MS parameters were as follows: source temperature (150 C), desolvation temperature (400 C), cone voltage (40 V), capillary voltage at 2.6 kV (ESI + ) and 2.2 kV (ESI À ), cone gas ow (50 L h À1 ) and desolvation gas ow (800 L h À1 ). MS E mode was chosen with low energy of 6 V and high energy of 20-40 V. 26,27 The mass spectrometer was calibrated with sodium formate in the range of 100 to 1200 Da in order to ensure the mass reproducibility and accuracy. Leucine enkephalin (m/z 556.2771 in ESI + and 554.2615 in ESI À ) was used as external reference for Lock Spray™ injected at a constant ow of 10 mL min À1 . The QC sample was injected randomly 4 times throughout the whole work list. All of the volume injection of the samples and QC was 5 mL per run. During data acquisition, the data for screening analysis was performed in MS E continuum mode, the data for metabolomics analysis was performed in MS E centroid mode. Data recording was performed on MassLynx V4.1 workstation (Waters, Manchester, UK).

2.4.
Screening analysis of components in three parts of COT by UNIFI platform UNIFI 1.7.0 soware (Waters, Manchester, UK) was used for data analysis. 28,29 Firstly, in addition to the internal Traditional Medicine Library on UNIFI platform, the chemical constituent investigation was conducted. As the result, a self-built database of chemical compounds isolated from the genus of Celastrus L. was established by searching the online databases including Web of Science, Medline, PubMed, ChemSpider and China National Knowledge Infrastructure (CNKI). The compound name, molecular formula and chemical structure of components were obtained in the database.
Secondly, the raw data obtained from Masslynx workstation were compressed by Waters Compression and Archival Tool v1.10, then were imported into the UNIFI soware.
Thirdly, the compressed data were processed by the streamlined work ow of UNIFI soware in order to quickly identify the chemical compounds which were matched the criteria with Traditional Medicine Library and self-built database. The main parameters of processed method were as follow: 2D peak detection was set to 200 as the minimum peak area. In the 3D peak detection, the peak intensity of high energy and low energy was taken more than 200 and 1000 times as the parameter respectively. Selected +H and +Na as positive adducts and +COOH and -H as negative adducts. Leucine enkephalin was used as reference compound in order to get exact mass accuracy, with [M + H] + 556.2766 for positive ion and [M À H] À 554.2620 for negative ion. As a result, the comprehensive chemical constituents screening list was accomplished.
Finally, a lter was set to rene the results, with the mass error between À5 and 5 ppm and response value over 5000. Each compound was veried by compared with the characteristic MS fragmentation patterns reported in literature or the retention time of the reference substances.

Metabolomics analysis of three parts of COT
The raw data acquired by Masslynx workstation was processed on MakerLynx XS V4.1 soware (Waters, Milford, CT, USA). Firstly, the raw data were processed with alignment, deconvolution, and data reduction, etc. The main parameters of the process method were as follows: retention time (0-28 min), retention time window (0.20), mass (100-1200 Da), mass tolerance (0.10), mass window (0.10), minimum intensity (5%), marker intensity threshold (2000 counts) and noise elimination (level 6). As a result, the list with mass and retention time corresponded to the responses based on all the detected peaks from each data le were shown in Extended Statistics (XS) Viewer. Secondly, multivariate statistical analysis, both principle component analysis (PCA) and orthogonal projections to latent structures discriminant analysis (OPLS-DA), were performed on the MakerLynx soware to analyze the resulting data.
PCA, a classical unsupervised low dimensional pattern recognition model, was used to show pattern recognition and maximum variation, and the overview and classication were obtained. OPLS-DA was used to obtain the maximum separation between two different groups. S-plots, which could provide visualization of the OPLS-DA predictive results, were created to explore the potential chemical markers which contributed to the differences.
Meanwhile, metabolites with VIP value > 4.0 and p-value < 0.001 were considered as potential chemical markers. 30,31 Futhermore, permutation test was also performed to provide a reference distribution with the R 2 /Q 2 values to indicate statistical signicance. Finally, the analysis results were shown in Simca 15.0 soware (Umetrics, Malmö, Sweden).
2.6. Bioassay analysis of three parts of COT 2.6.1 Preparation of cigarette smoke extract. The preparation of the cigarette smoke extract (CSE) was basically the same as previous reports. 32 Put the smoke from one cigarette into 20 mL culture medium (300 s per cigarette). The CSE solution was incubated at 37 C for 30 min aer being ltered with a 0.22 mm sterile lter. The CSE solution was prepared freshly and was used within 30 min. This prepared CSE solution was considered to have the highest concentration (100%).
2.6.2 Cell viability assay. The nal concentrations (20.0, 40.0, 80.0, 160.0, 320.0 mg mL À1 ) of each test samples (R bio , S bio and L bio ) were acquired by diluting the stock solutions with Dulbecco's Modied Eagle Medium (DMEM). A549 cells were cultured in 96-well plates at a density of 5 Â 10 5 cells per well treated with CSE (0%, 5%, 10%, 20%, 30% and 40%) for 18 h, or treated with R bio , S bio and L bio solutions (0.0, 20.0, 40.0, 80.0, 160.0, 320.0 mg mL À1 ) for 24 h. The growth-inhibition effect of CSE and the effect of the drugs on viability of A549 cells were evaluated by MTT assay. Fig. 1 The base peak intensity (BPI) chromatograms in root, stem and leaf of COT in ESI + and ESI À .
This journal is © The Royal Society of Chemistry 2020 RSC Adv., 2020, 10, 8396-8420 | 8407 2.6.3 Drug treatment. For all groups, A549 cells were cultured in 96-well plates at a density of 5 Â 10 5 cells per mL for 18 h. In CSE group, the cells were treated with a certain dose of CSE without drug intervened. In drug groups, the cells were treated with both CSE and drugs (R bio , S bio or L bio ). In control group, A549 cells were cultured normally without the CSE or drugs. In positive group, the cells were treated with both CSE and dexamethasone (5 mg mL À1 ).
2.6.4 Enzyme-linked immunosorbent assay. The contents of IL-1b, IL-6 and TNF-a in the cell culture supernatant were determined with ELISA kits. All procedures were performed according to the manufacturer's instructions.
2.6.5 Statistical analysis. Statistical analysis was performed on Graphpad Prism 6.0 soware (CA, USA). The results were expressed as mean AE SD. Two tailed test or a one-way analysis of variance (ANOVA) was used to calculate statistical signicant difference (p < 0.05).

Screening analysis of components of three parts of COT
A total of 120 compounds, including 91 in ESI + mode and 29 in ESI À mode, were identied or tentatively characterized from three parts of COT ( Table 2). The base peak intensity (BPI) chromatograms were shown in Fig. 1. The chemical structures were shown in Fig. 2, the results showed that COT was rich in natural components with various structural patterns. On one hand, according to the reference, there were nearly 50, 100, 10 compounds were reported from the root, stem, leaf parts of COT, respectively. While in this study, there were 92, 56 and 32 components were identied or tentatively characterized from root part, stem part and leaf part of COT, respectively. And most of the components were identied from COT for the rst time. Various kinds of structures, including triterpeniods, sesquiterpenoids, steroids, avonoids, organic acid and organic acid esters, phenylpropanoids, diterpeniods, monoterpenoids, alkaloid and others, were contained in each part of COT. The numbers (% of the total identied components in each part) and structural types of compounds identied from root, stem and leaf    of COT were shown in Fig. 3. There were 34, 18 and 10 triterpeniods identied from root, stem and leaf, respectively, accounted for 37%, 32% and 31% of the total components in each part. So, it was concluded that triterpeniods were the major constituents in three parts of COT. Moreover, according to each percentage, the root part of COT was also rich in organic acid and organic  This journal is © The Royal Society of Chemistry 2020 RSC Adv., 2020, 10, 8396-8420 | 8413 acid esters, steroids and phenylpropanoids. The stem part was also rich in organic acid and organic acid esters, and avonoids. The leaf part was also rich in steroids, and sesquiterpenoids. The percentage of avonoids in the total identied components in stem was higher than the percentages in root part or in leaf part. The percentages of sesquiterpeniods and steroids in the total identied components in leaf was higher than the percentages in root part or in stem part. On the other hand, the shared components (30 for root and stem, 22 for root and leaf, 23 for stem and leaf, 15 for root, stem and leaf) were also found in our study. As shown in Fig. 3, the structures of shared components were various, while triterpeniods held the majority. Celastrol, one of the triterpeniods, was shown to distribute in root, stem and leaf, which was consistent with the ref. 19. So our research work could provide the scientic data to clarify the chemical composition of COT, particularly for the root and the leaf parts.
Although the study provided evidences to elucidate the chemical composition of COT, there were still some unresolved issues. For example, as shown in BPI chromatograms, there were some unidentied components. Further research should be carried on the identication of these unknown compounds.

Metabolomics analysis of three parts of COT
PCA score 2D plots in both ESI + and ESI À were established as shown in Fig. 4. The QC samples were clustered tightly and were in the middle of the three groups in PCA, which indicated the system had satisfactory stability. The samples from root of COT were clearly gathered together, which indicated there was a good similarity among them, and this phenomenon was also observed in stem and leaf of COT. Meanwhile, the root, stem and leaf groups were easily divided into three clusters, indicating that these three parts of COT could be differentiated in both ESI + and ESI À .
In order to further distinguish one part from the other two parts, OPLS-DA plots, S-plots, permutation tests, and VIP values were obtained to see which variables were responsible for sample separation 97 (Fig. 5-7). In OPLS-DA plots, each spot represented a sample. From the perspective of OPLS-DA, one part was clearly separated from the other two parts. The parameters such as R 2 and Q 2 indicated the model had good   ability of prediction and reliability in both ESI + and ESI À modes. The permutation plots showed the original point on the right was clearly higher than all Q 2 -values (blue) on the le, which indicated the original models were valid. To identify the metabolites contributing to the discrimination, S-plots were generated under OPLS-DA model. Each spot in S-plots represented a variable. The variables with VIP > 4 and p < 0.001 were considered as potential chemical markers. The possible molecular formula of the markers were calculated by highaccuracy quasi-molecular ion with mass error between AE5 ppm. A total of 26 robust known chemical markers (marked in Table 2) enabling the differentiation between one part with the other two parts were identied and marked in S-plots.
The detected result of 88 (celastrol) in our study, with much higher contents in root than in stem and leaf, was consistent with the ref. 19. While there were a few differences between our results and the references. In the present study, compound 39 (pristimerin) was only detected in root, and 11 (orbiculin I) was only detected in leaf. According to the reports, 39 was once isolated from stem 14 though mainly from root, 98,99 and 11 isolated from root. 3 The reason was the concentrations of them were lower than the lowest detection limits. It was worth mentioning that some chemical markers with high responses in UPLC-MS, two triterpenoids (39 and 88) in root, one avonoids (2) in stem and two sesquiterpenoids (11 and 40) in leaf, could be used for further quality control of three parts of COT respectively.
In order to systematically evaluate the chemical markers, a heat-map was generated. The hierarchical clustering heat map, intuitively visualizing the difference level of potential chemical markers in different parts, was shown in Fig. 8. The higher values were indicated by red squares, the lower values were indicated by green squares.

Bioactivity evaluation
3.3.1 Cytotoxicity of CSE and the three parts of COT on the viability of A549 cells. The results of MTT showed that the viability of A549 cells was obviously affected (p < 0.01) by 30% or 40% CSE (Fig. 9). Therefore, 20% CSE was chosen as stimulus in the following experiments. Additionally, as shown in Fig. 10, the viability of A549 cells were not signicantly affected by the R bio , S bio and L bio solutions at 20-160 mg mL À1 . So we evaluated the effects of R bio , S bio and L bio solutions at 20-160 mg mL À1 on CSEstimulated A549 cells.
3.3.2 Effect of root, stem and leaf of COT on CSEstimulated pro-inammatory cytokine levels in A549 cells. The inammatory development was characterized by the release of pro-inammatory mediators such as interleukin-1b (IL-1b), interleukin-6 (IL-6) and tumor necrosis factor-a (TNF-a). Whether the root, stem and leaf of COT could inhibit the release of IL-1b, IL-6 and TNF-a in CSE-stimulated lung epithelial cells was investigated in this paper. As shown in Fig. 11, the production of IL-1b, IL-6 and TNF-a in A549 cells was obviously increased aer treated with CSE (p < 0.01). However, treated with the R bio , S bio and L bio solutions could evidently decrease the levels of pro-inammatory factors in a good dose-dependent way with a certain range of 20-160 mg mL À1 in CSE-stimulated cells. The R bio solution could signicantly decrease the levels of IL-1b, TNF-a (80 mg mL À1 , p < 0.05; 160 mg mL À1 , p < 0.01) and IL-6 (40 and 80 mg mL À1 , p < 0.05; 160 mg mL À1 , p < 0.01). The S bio solution could signicantly decrease the levels of IL-1b, TNF-a (40 mg mL À1 , p < 0.05; 80 and 160 mg mL À1 , p < 0.01) and IL-6 (20 and 40 mg mL À1 , p < 0.05; 80 and 160 mg mL À1 , p < 0.01). The L bio solution could signicantly decrease the levels of IL-6 (80 mg mL À1 , p < 0.05; 160 mg mL À1 , p < 0.01) and TNF-a (80 and 160 mg mL À1 , p < 0.05), but showed no signicantly effect on IL-1b. The above results showed that the S bio solution had a stronger anti-inammation effect than R bio and L bio . It is suggested that to explore the anti-COPD effect of the COT stem in vivo is meaningful in further research. The different activities of root, stem and leaf of COT might be caused by the various phytochemicals in these three parts of COT. The phytochemical study showed that three parts of COT were differentiated. The bioassay study showed that three parts of COT could reduce the levels of pro-inammatory factors to varying degrees. And the stem part had a stronger anti-COPD effect than root and leaf parts. As we all know, the material basis of different pharmacological activities is the different chemical composition. The results of the two parts of our study showed that the different activities of root, stem and leaf of COT might be caused by the various phytochemicals in these three parts of COT.

Conclusions
In conclusion, for screening analysis, a total of 120 compounds (15 shared components), including 92 from root, 56 from stem and 32 from leaf, were identied or tentatively characterized from COT. Each part of COT was rich in various kinds of structures, especially triterpeniods held the majority. For metabolomic analysis, the root, stem and leaf of COT were differentiated in both ESI + and ESI À modes. There were 13, 8 and 5 potential chemical markers identied from root, stem and leaf, respectively. Among the above robust markers, 5 robust chemical markers with high responses in UPLC-MS, 2 triterpenoids (pristimerin and celastrol) in root, 1 avonoids {5,7-dihydroxy-6,8-dimethyl-3(S)-3-(3-methoxy-4 0 -hydroxybenzyl) chroman-4-one} in stem and 2 sesquiterpenoids (orbiculin I and orbiculin A) in leaf, could be used for further quality control in three parts of COT respectively. For bioassay analysis, the root, stem and leaf of COT could evidently reduce the levels of pro-inammatory factors in a dose-dependent way within a certain range of 20-160 mg mL À1 in CSE-induced A549 cells. The results showed that the stem part had a stronger anti-COPD effect than root and leaf parts. The different activities might be caused by the various phytochemicals in these three parts of COT. This comprehensive phytochemical study revealed both the structural diversity of secondary metabolites and the different distributions in different parts of COT. It could provide a theoretical basis for further utilization and development of COT. And the identication of anti-COPD components from COT will be explored deeply in the future based on the current results of this study.

Conflicts of interest
The authors declare no conicts of interests.