Screening for the anti-inflammation quality markers of Xiaojin Pills based on HPLC-MS/MS method, COX-2 inhibition test and protein interaction network

Nowadays, breast disorders seriously affect women’s health in an increasing number. In China, Xiaojin Pills are commonly used in the treatment of breast diseases. Doctors have concluded that the combined use of Xiaojin Pills with conventional therapy can significantly improve the efficacy with fewer side effects. However, the prescription of Xiaojin Pills is complicated and their quality control methods cannot completely ensure the quality of Xiaojin Pills. On the basis of its mechanism, our study combined chemical evaluation and biological evaluation to identify the anti-inflammatory markers of Xiaojin Pills. In this manuscript, 13 compounds in Xiaojin Pills were quantified. At the same time, the cyclooxygenase-2 inhibition rates of different Xiaojin Pills were measured and the possible markers were screened by spectrum-effect relationship. Further, anti-inflammatory activities of markers were verified and protein interaction network was analyzed, identifying the components of Protocatechuate, Beta-Boswellic acid and Levistilide A as the anti-inflammatory quality markers of Xiaojin Pills. We hope our studies can provide a scientific theoretical basis for accurately quality control of Xiaojin Pills and reasonable suggestions for pharmaceutical companies and new ideas for the quality control of other medicines.

Pills showed a good effect on prostatitis 17 , arthralgia, thyroid nodule, mammary cancer, thyroid cancer and other cancers [18][19][20] and the alleviating effect on prostatitis was achieved by inhibiting the expression of COX-2 17,21 . These results provided an important basis for screening the quality markers (Q-markers) of Xiaojin Pills from the anti-inflammatory effect and mechanism.
Xiaojin Pills consist of 10 kinds of Chinese herbal medicine (CHM), including Moschus, Momordicae Semen, Aconiti Kusnezoffii Radix Cocta, Liquidambaris Resina, Olibanum, Myrrha, Faeces Trogopterori, Angelicae Sinensis Radix, Lumbricus rubellus and Fragrant Ink. The sources of these drugs are complex, including plant medicine, resin medicine, animal medicine and mineral medicine. Due to the reliable efficacy, it was first recorded in the 1977 edition of Chinese Pharmacopoeia (CP). However, the latest edition of CP only defined the lower limit of muscone to evaluate the effectiveness, and identified three diester diterpenoid alkaloids by TLC method to control toxicity. Muscone is recognized as a component of effective anti-inflammatory activity and the single indicator is obviously unable to evaluate its quality and clinical effects, thus, it is necessary to study and screen the anti-inflammatory components other than muscone.
In order to accurately control the quality of Chinese medicine and Chinese patent medicine, the concept of Q-markers was proposed by Prof. Liu Chang-xiao (China) in 2016 22 . Q-markers could meet the following basic conditions. They are the chemical substances inherent in the products or formed during processing, and closely related to the functional properties of traditional Chinese medicine, and have definite chemical structures. Meanwhile, they should be the substances can be qualitatively identified and quantitatively measured. Due to the inherent difficulty in characterizing all chemical constituents in botanical drugs, Botanical Drug Development Guidance (BDDG) for Industry by United States FDA's in 2016 clearly suggested the relative content or activity can be characterized by bioassay. Based on this concept and anti-inflammatory mechanism of inhibiting the high expression of COX-2, we put forward a new idea to evaluate the anti-inflammatory activity of Xiaojin pill by COX-2 inhibition test and screen the anti-inflammatory substances.
In the present study, a rapid, sensitive, and accurate HPLC-MS/MS analytical method was developed to detected 13 kinds of candidate anti-inflammatory and analgesic components in Xiaojin Pills, including Aconitine (AC) Mesaconitine (MA), Hypaconitine (HA), Benzoylaconitine (BAC), Benzoylmesaconitine (BMA), Benzoylhypacoitine (BHA), Levistilide A (Lev), Inosine (Ino), Ligustilide (Lig), Acetyl-11-keto-b-boswellic acid (Ace), Beta-Boswellic acid (Bet), Ferulic Acid (Fer) and Protocatechuate (Pro). Validation parameters of the HPLC-MS/MS method were studied systematically 23 , and 10 batches of Xiaojin Pills were evaluated. Then, COX-2 inhibitor screening tests were carried out, and the correlation analysis was further performed to find possible anti-inflammatory ingredients. Finally, the screened markers were verified by COX-2 inhibition tests and protein interaction network analysis. Figure 1 showed the screening process of Xiaojin Pills' quality markers. We hope that our study is benefit for the precise quality control and clinical application of Xiaojin Pills. The approach also provides examples for exploring the bioactive components of Chinese patent medicine and other ethnic drugs.

Results
Clustering analysis (CA) results. The HPLC-MS/MS method was validated in terms of linearity, limit of detection (LOD) and limit of quantification (LOQ), precision, stability, repeatability and recovery based on ICH guidelines 24 . Their results were described in Supplementary Tables S1 and S2. The contents of 13 compounds in 10 samples of Xiaojin Pills were presented in Table S3. Contents of diester diterpenoid alkaloids in all samples were lower than 0.0031%, which showed that Xiaojin Pills were fairly safe. The other ten components differed greatly in different samples. To show the difference intuitively, the concentrations of thirteen components in 10 batches were used to create a 13 × 10 matrix where all the numerical values were displayed in terms of thermography (Fig. 2). Using an appropriate distance level, samples were obviously classified into three clusters. S1, S2 and S6 were categorized into cluster I. S5, S3, S4, S7 and S8 were grouped into cluster II. S9 and S10 were put into cluster III. It was clear that the contents of Fer, Pro, MA, Lig, Bet, HA, Lev, BMA and BHA in cluster III were obviously higher than others.

Results of feasibility.
Results of inhibition curve (S4) were shown in Fig. 3A, and the IC 50 was about 0.5076 mg/ml. At the concentration of 0.5 mg/ml, the actual inhibition rate of S1 and S6 was 34.04% and 37.64%, respectively. The inhibition rate curves of S1 and S6 were shown in Figs 3B and C. It was illustrated that the theoretical inhibition rate of S1 and S6 was 36.04%, and 39.86%, respectively. The error rate of S1 and S6 calculated was 5.55% and 5.57%, which indicated a direct determination method was of good feasibility.
Among them, Y represented COX-2 inhibition rate, and x 1 , x 2 , x 3 represented Pro, Bet and Lev, respectively. The results showed that the F value was 37.974 and the corresponding P value was 0.000 (ie., less than 0.05), demonstrating that the multiple linear regression analysis model was satisfactory. It was found that a strong linear relationship between COX-2 inhibition rate and the contents of Pro, Bet and Lev.
The equation above described the degree which components in Xiaojin Pills contributed to the anti-inflammatory activity. Pro had the greatest important influence on COX-2 inhibition ratio, with correlation coefficient was 0.611. Bet and Lev were also positively correlated to inhibition ratio, and the correlation coefficients were 0.525 and 0.487, respectively. Therefore, Pro, Bet and Lev could be considered as the candidate anti-inflammatory markers.
Results of anti-inflammatory activity of three components. Curve fitting method was used to evaluate the anti-inflammatory activity of three components. The results were shown in Fig. 3D-F. It was clear that the best inhibition rate of Pro was as high as 91.57%, while that of Bet and Lev was about 78.96% and 42.92%. It  could be concluded that the anti-inflammatory inhibition activity of Pro was higher than Bet, and Lev, which was consistent with the correlation coefficient above.

Results of protein interaction network. Compound target information. 73 targets points were obtained
by removing the repeated targets, including 45 of Pro, 33 of Bet, and 6 of Lev. From the number of targets, Pro contributed the most to the network. Among the targets we obtained, PTGS1 (prostaglandin 1/COX-1) and PTGS 2 (prostaglandin 2/COX-2) were the most important node in the process of prostaglandin metabolism. Therefore, Xiaojin Pills might be involved in the prostaglandin metabolic process to play a role in anti-inflammatory effect.

Construction of protein interaction and module analysis.
Protein interaction information of targets from each compound was introduced into Cytoscape 2.8.3 and the protein interaction network diagram of Xiaojin Pills was obtained, including 186 nodes and 834 edges. Detailed information on these hubs is provided in Fig. 4A. Module analysis of protein interaction network had been conducted. As showed in Fig. 4B, 13 modules were identified. Using BinGo, the functions of the protein contained in the modules were annotated. Each module's biological process was shown in Table 1. The module was graded according to the MCODE algorithm. The correlation of proteins in the modules became stronger as the score increased. Modules of 4th and 6th were mainly related to the prostaglandin metabolism process, including PTGS1 (prostaglandin peroxidase 1/COX-1), PTGS2 (prostaglandin peroxidase 2/COX-2). Prostaglandin (PG) was the metabolites produced by PTGS catalytic arachidonic acid. The PG synthesis regulated by PTGS was considered to be a marker of proinflammatory response and played an important role in the initiation of inflammation. The 10th module also played a role in inflammation, and the proteins it contained was related to the leukocyte chemotaxis mediated by chemotactic factors. This module included RB1CC1, PRKCZ, CHUK and IKBKB. Chemokines could cause an inflammatory reaction by inducing leukocytes to express integrins, releasing cytokines, inducing endothelial cells to express adhesion molecules, and collecting more inflammatory cells through the wall of vessels.
During three anti-inflammatory markers screened, Pro and Bet both acted on COX-1 and COX-2, which were directly involved in the arachidonic acid metabolism pathway. However, it was not found a target directly related to this pathway associated with Lev. Through the analysis of protein interaction network, it was predicted that the inhibition effect of Pro and Bet on COX-2 was better than that of Lev.

Discussions and Conclusions
During the study, different chromatography conditions were examined and compared, including various columns, mobile phases, and gradient elution conditions to achieve good resolution, high detection sensitivity, symmetric peak shapes, and short run time. Three kinds of reversed-phase columns, Kromasil C18 column (4.6 mm × 200 mm, 5 μm), Phenomenonex Gemini C18 column (4.6 mm × 150 mm, 5 μm) and Agilent Technologies Zorbax Eclipse XDB-C18 (4.6 mm × 150 mm, 5 μm), were investigated. The Agilent Technologies Zorbax Eclipse XDB-C18 column had good peak separation and sharp peaks. Poor resolution among the peaks was found when methanol was used as the organic solvent of the mobile phase. However, when methanol was replaced by acetonitrile, the resolution was greatly improved. Different concentrations of formic acid (0.05%, 0.1%, and 0.2%) or acetic acid (0.05%, 0.1%, and 0.2%) were added to the aqueous phase, respectively. 0.1% formic acid was selected with the best response and peak shape. The overall chromatographic run time was 30 min. The column temperature was set at 30 °C and the flow rate was 0.3 mL·min −1 to ensure good separation. Mass spectral conditions were optimized in MRM scan type using the reference compounds. LC-MS/MS data of samples were collected by using an Agilent Triple Quadrupole mass spectrometer with an electrospray interface (ESI) operated both in positive and negative modes. The mode of the ESI was depended on the properties of compounds in gaining or losing electrons 25 . In order to get good response, AC, MA, HA, BAC, BMA, BHA, Lev, Ino and Lig were detected by the ESI + mode, while Ace, Bet, Fer and Pro were detected by the ESI − mode. In MRM mode, all compounds could be detected in different channels without interference. In addition, the temperature of 35 °C showed the highest extraction efficiency. For 0.2 g powder, 5 ml methanol was shown to be the optimal volume. The extraction time was studied in the range of 20 to 50 min. Finally, 30 min showed the highest COX, also known as prostaglandin-endoperoxide synthase (PTGS), is an enzyme responsible for the formation of important biological mediators, including prostaglandins, prostacyclin and thromboxane. COX is the central enzyme in the biosynthetic pathway to prostanoids from arachidonic acid. COX-2 is one of the known isoenzymes, which is not expressed on normal conditions in most cells, but elevated levels are found during inflammation. Pharmacological inhibition of COX by non-steroidal anti-inflammatory drugs (NSAID) can provide relief from symptoms of inflammation and pain 10,11 . The experiment indicated that Xiaojin Pills played good anti-inflammatory and analgesic effects of reducing the expression of COX-2 21 .
The quality of Chinese patent medicine is mainly evaluated by the contents of active ingredients. However, it is a huge challenge to select the quality markers accurately for the complex effective material basis of patent medicine. In this study, a method of screening anti-inflammation markers based on spectrum-effect relationship has been practiced and the results provide a scientific basis for the formulation of quality control indicators. We believe our methods could provide a significant way to combine chemical methods and bioactivity evaluation to screen the biological activity markers in Chinese patent medicines or other valuable materials. It will be of interest for pharmaceutical companies and pharmaceutical workers.

Materials and Methods
Materials. Ten batches of Xiaojin Pills samples were collected from four pharmaceutical factories and all the information was listed in Table 2.
HPLC-grade acetonitrile and methanol were purchased from Fisher Chemical (Pittsburg, PA, USA). HPLC-grade formic acid, ammonium acetate, and analytical-grade dimethyl sulfoxide (DMSO) were purchased from Chengdu KeLong Chemical Factory (Chengdu, China). The ultrapure water was obtained by Millipore Milli-Q water purification system (Millipore, Billerica, MA, USA). All solutions were filtered through 0.
The column temperature was 30 °C and 0.1 μL of the sample solution was injected into the system. The mobile phase was composed of (A) 0.1% aqueous formic acid in water (ESI + mode) or 10 mmol/L ammonium acid (ESI − mode) and (B) acetonitrile using a gradient program of 100-80% B for 0-1 min, 80-50% B for 1-2 min, 50-30% B for 2-3 min, 30-80% B for 3-5 min, 80-100% B for 5-30 min, with a mobile flow rate of 0.3 mL/min. Mass spectrometric scan were obtained by ESI in ESI + mode and ESI − mode with a scanning interval 100-1000 m/z. The main parameters for MS were set as follows: gas temperature, 300 °C; gas flow, 11 L/min; nebulizer, 35 psig; capillary voltage, 4000 V; atomizer pressure 15 psi (1 psi = 6.895 Kpa). MS parameters and MRM transitions of each analyte are shown in Table 3. Based on this condition, the MRM chromatogram and chemical structure of each compound are shown in Fig. 5.
Preparation of sample solutions. Break different batch of Xiaojin Pills to powder through a 50-mesh sieve. A total of 0.2 g of powder was accurately weighed and extracted with 5 mL of methanol solution by ultrasonic extraction for 30 min. Extracted solution was cooled, contributed to weight loss during the extraction procedure, and filtered through a 0.22 μm micropore film to yield the subsequent filtrate solution.
Determination of the sample. 10 batches of Xiaojin Pills were determined by the method above.
CA. CA is a multivariate analysis method that is used to sort samples into groups 26 . In the present study, the CA of samples was performed using Metabo Analyst 3.0. Ward's method as the amalgamation rule and squared Euclidean distance as metric were used to establish clusters. To visualize the similarity and dissimilarity of the samples, all the data were expressed as a thermograph.

COX-2 inhibition rate measurement. Sample preparation. An aliquot of 0.2 g Xiaojin Pills powder
(through a 50-mesh sieve) was accurately weighed and transferred into a 15-ml-flask. After being added with 10 mL of DMSO, the flask was weighed and extracted by ultrasonic extraction for 30 min. Extracted solution was cooled and the lost weight was supplied. Then the mixture was centrifuged at 4000 rpm for 10 min, the supernatant was transferred into another 15-ml-flask and the high concentration sample solutions for each batch were obtained.
Determination of suitable sample concentration. It is difficult to complete inhibition curve for each sample and calculate the IC 50 value for the large amounts. Therefore, one sample (S4) was selected to calculate the IC 50 value randomly, and it was used as a reference concentration. Firstly, S4 was prepared into solutions with a concentration of 20 mg/ml, 15 mg/ml, 0.5 mg/ml, 0.3 mg/ml, 0.1 mg/ml, 0.05 mg/ml, 0.01 mg/ml, and 0.005 mg/ml, respectively. Then, the inhibition rate of COX-2 in different concentration samples was measured and the inhibition curve was plotted. IC 50 of S4 was calculated to be 0.5076 mg/ml. Based on it, each sample was diluted to a concentration of 0.5 mg/ml to determine the COX-2 inhibition rate.
Experimental procedure. According to the instructions of COX-2 inhibitor screening kit, the relative fluorescence unit (RFU) was measured after a period of incubation. Excitation wavelength was 560 nm, and emission wavelength was 590 nm. Inhibition rate of each sample was calculated as follows: Among them, RFU 1 , RFU 2 , and RFU S represented the fluorescence value of the blank control group, 100% enzyme activity control group and the sample group, respectively. Feasibility verification. S1 and S6 were randomly selected to compare the results difference between direct determination method and curve fitting method. The former was determined at the fixed concentration of 0.5 mg/ml. It was considered as an actual value. The later was calculated by the best fitted curve equation after the study of relationship between concentrations and inhibition rates. This result was accepted as a theoretical value. Error rate of direct determination method was calculated by formula 2.
Among them, IR a , IR t represented the fluorescence value of the actual inhibition rate and the theoretical inhibition rate, respectively.
Multiple linear regression analysis. Multiple linear regression analysis was used to model the best combination of two or more independent variables (x i ) to predict or estimate the dependent variable (Y) by fitting a linear equation 27 . It showed the contribution of each independent variable to the dependent variable in the following form: where Y was the estimated value and represents the dependent variable and x i were the uncorrelated variables; b 0 represented the estimated constant, and b i was regression coefficients. In this study, multiple linear regression analysis was introduced to combine data from the chemical contents determination and anti-inflammatory activity of Xiaojin Pills. SPSS statistical software (SPSS for Windows 13.0, SPSS Inc., USA) was used to establish the chemical-activity relationships and further explore the anti-inflammatory markers. Verification of anti-inflammation markers. The anti-inflammation activities of the markers we screened were compared by the curve fitting method. After the study of relationship between concentrations and inhibition rates, several best fitted curves of the markers were obtained respectively, and the anti-inflammation effects of several markers could be verified by comparing their best inhibition rates.
Construction and analysis of protein interaction network. The protein interaction information of the target was derived from the String database (http://www.string-db.org/). The protein interaction data we obtained were introduced into Cytoscape 2.8.3 and each of the target protein interaction networks was calculated by the Union method 13,29 . After removing the isolated points, repeated and self-loop edges, we got the interaction network of anti-inflammatory markers.
Pathway enrichment analysis. Because of the great number of compounds of an herb and the multiple targets of each compound, the total number of the targets of all the key herbs is substantial. The great amount of targets makes it difficult to understand the biological meaning, so a pathway enrichment analysis was performed using the Database Visualization and Integrated Discovery software (DAVID, http://david.abcc.ncifcrf.gov/home.jsp, version 6.7.) and based on the pathway data obtained from the Kyoto Encyclopedia of Genes and Genomes database (KEGG, http://www.genome.jp/kegg/, updated on April 18, 2016) 30 .