Chemical profiling and quantification of XueBiJing injection, a systematic quality control strategy using UHPLC-Q Exactive hybrid quadrupole-orbitrap high-resolution mass spectrometry

To clarify and quantify the chemical profiling of XueBiJing injection (XBJ) rapidly, a feasible and accurate strategy was developed by applying ultra high performance liquid chromatography-Q Exactive hybrid quadrupole-orbitrap high resolution accurate mass spectrometry (UHPLC-Q-Orbitrap HRMS). A total of 162 components were characterized, including 19 phenanthrenequinones, 33 lactones, 28 flavonoids and 12 phenolic acids and 51 other compounds. Among them, 38 major compounds were unambiguously quantified by comparing with reference standards. Meanwhile, 38 representative compounds were simultaneously detected in XBJ samples by Q-Orbitrap HRMS. Satisfactory linearity and correlation coefficient were achieved with wide linear range. The precisions, repeatability, stability and recovery were meeting requirements. The validated method was successfully applied for simultaneous determination of 38 bioactive compounds in 10 batches XBJ samples. In addition, the similarity evaluation of fingerprintings was applied to assess the quality of XBJ. And the results were evaluated by multiple statistical strategies and five compounds might be the most important chemical markers for chemical quality control of XBJ. Finally, a rapid and simple UPLC-MS/MS method was developed for determination of five markers in XBJ sample. This research established a high sensitive and efficient strategy for integrating quality control, including identification and quantification of XBJ.

sitive and high-throughput method for rapid identification of the components of XBJ regardless of the macro-and micro-constituents. The total ion chromatograms (TIC) of the XBJsample both in positive and negative ion mode are presented in Fig. 1 38 compounds were unambiguously identified based on comparison of retention time and high-resolution accurate mass with that of available reference standards and their chemical structures were shown in Fig. 2. Moreover, the fragmentation patterns and pathways of the standards were investigated in depth to further confirm the structure of their derivatives. For the compounds without available references, the structures were presumed based on the following steps so as to increase the credibility: (1) the molecular formula was established based on high-accuracy protonated precursors such as [M + H] + , [M + Na] + , [M−H] − , or [M + HCOO] − within a mass error of 10 ppm and the fractional isotope abundance; (2) A class of compounds has the same law of cracking, therefore, the standards were utilized to characterize the fragment pathways and diagnostic fragment ions that could be applied for structural elucidation of their derivatives. In addition, some literatures about the compositions of XBJ and five Chinese herbals could be referred. (3) The fragment ions from mass spectrometry were used to further confirm the chemical structure with the aid of Thermo Scientific TM Mass Frontier 7.0 12 .
As for monoterpene glycosides, the loss of CH 3 , H 2 O and CO was observed clearly in their MS/MS spectra. The mass spectra and proposed major fragmentation of representative compounds Paeoniflorin was shown in Fig. 3A and the proposed fragmentation pathways was presented in Fig. 3B. Other constituents were tentatively deduced by the above steps and paeonisuffrone, phenanthrenequinone, senkyunolide, lactones, flavonoids and phenolic compounds dominated the chemical profiling of XBJ [13][14][15][16][17][18][19][20][21][22][23] . Overall, 162 components, including 19 monoterpene glycosides, 19 phenanthrenequinone, 33 lactones, 28 flavonoids and 63 phenolic acid and other compounds, in XBJ were identified or tentatively characterized with their retention times and MS data, which are summarized in Table 1.
Identification of monoterpene glycosides in XBJ. 19 monoterpene glycosides were identified and listed in Table 1   Identification of phenanthrenequinone in XBJ. 19 phenanthrenequinone were identified and listed in  O 2 ) were observed in the MS/MS 2 experiment of P 5 . This fragmentation information were similar with that of phenanthrenequinone, and their molecular was accordance with tanshinone II B, the major constituent in tanshin, one composition of traditional Chinese medicine in XBJ. Thus, P 3 , P 4 , P 5 , and P 14 were deduced as tanshinone II B and its isomers.       Quantitative analysis of samples. A thorough and complete method validation for assaying 38 bioactive compounds in XBJ was done referring to ICH guidelines 24 . The UHPLC-Q-Orbitrap mass spectrometry was validated with respect to linearity, sensitivity, accuracy and precision, reproducibility and stability.

Method Validation
Linearity, LOD and LOQ. Standard stock solutions containing 38 analytes were prepared and diluted to seven appropriate concentrations for the construction of the calibration curves. Each solution was injected in triplicate, and then the linear regression equation was obtained by plotting the analyte peak area (Y) vs a series of analyte concentrations (X). The regression equation, coefficient of determination (R 2 ) and linear range are given in Supplementary Table S1. All the analytes showed good linearity with R 2 more than 0.9994 in the concentration range. The LOD and LOQ under the optimized chromatographic conditions were evaluated at a signal-to-noise ratio (S/N) of 3 and 10, respectively. The values of LODs and LOQs were in the range of 0.01~35.77 ng·mL −1 and 0.03~119.22 ng·mL −1 , respectively (Supplementary Table S1).
Accuracy and precision. The precision of the established method was evaluated by intra-day and inter-day variability, and the relative standard deviations (RSD) were taken as a measure. The mixed standard solution at   middle concentrations was analyzed in six replicates within one day and on 3 consecutive days. The results are shown in Supplementary Table S1, and the RSD values of the intra-day and inter-day of 38 compounds were all less than 2.97%, which showed good precision of the developed method. The accuracy of the established method was evaluated by recovery test and RE (relative error). The samples were spiked with three concentration levels (80, 100, and 120%) of known amounts of 38 reference compounds. The spiked samples of each concentration were analyzed in triplicate. The accuracy was calculated as the quotient of the measurement and the nominal value of the analyte added to the sample. The detailed accuracy data is presented in Supplementary Table S2. The mean recoveries were ranged from 98.5% to 101.3% with RSDs less than 2.98%.

Reproducibility and Stability.
In order to confirm the reproducibility, six different samples from the same batch sample were analyzed within one day and on three consecutive days. The RSDs were used as a measure and the acceptance criterion should be within 5.0%. The results are shown in Supplementary Table S1 and the RSD values of 38 compounds were all less than 3.0%, which showed good reproducibility of the developed method.
The stability of the sample solution was analyzed at room temperature on three consecutive days. The stability of the standard solutions stored at 4 °C was also examined on three consecutive days. Injections were performed at 0, 12 hour, 1, 2, 3, 5, and 7 days. The stability RSD values of 38 compounds in the sample solution were all less than 2.86% and those in standard solutions were all less than 2.0%, which showed that all analytes in the sample solution (at room temperature) and the standard solutions (at 4 °C) were found to be very stable.

Analysis of chemical profile of XBJ sample
The developed UHPLC-Q-Orbitrap HRMS method was adopted for the routine screening of the 38 bioactive compounds in 10 XBJ samples. 38 bioactive compounds were unambiguously identified by comparing the retention times and high-resolution accurate mass of reference standards. The polarity switching in full scan modes of UHPLC-Q-Orbitrap HRMS was used to achieve the highest response intensities of various types of constituents. In addition, the Q-Orbitrap HRMS as a powerful high resolution mass spectrometry, has the function of qualitative and quantitative simultaneously, namely compounds could be qualitative and quantitative in one analysis. Table 2 showed the obtained quantitative results of each compound calculated according to calibration curves. The results shows that two compounds (Hydroxysafflor yellow A and Paeoniflorin) are the predominant constituents obviously, the contents of which are much higher than other compounds. Hydroxysafflor yellow A and Paeoniflorin are two major marker components in Carthami Flos and Paeoniae Radix Rubra. Moreover, the Q-Orbitrap HRMS has very high sensitivity, so the low-content compounds, such as Levistolide A, Tetramethylpyrazine, Butylidenephthalide and Tanshinone I, were investigated simultaneously. Thus, the constituents with high and low levels contents could be quantified in one analysis.
The RSD of total amounts of investigated 38 compounds in 10 batches XBJ samples was 2.81%, which showed good stability of the total content. However, significant variations were observed as well. The RSD of each compound in 10 batches XBJ samples was in range of 2.48% to 19.43%, which showed instability of the some compounds. However, multiple active components, including macro-and micro-components, are frequently considered to be responsible for the therapeutic effects 25 . So, the present analysis of multiple components is more reasonable for quality control of XBJ injection.

Quality assessment of XBJ with the established strategy
Fingerprinting. Fingerprinting strategies are internationally accepted as an acceptable means of quality control (QC) for TCMs 26 . There are significant advantages of using fingerprinting strategies for sample differentiation, as fingerprinting not only determines the characteristic patterns of each plant type but also reveals the inherent relationships between multiple compounds. The good precision, reproducibility, stability of UHPLC-Q-Orbitrap HRMS analysis were demonstrated. The chromatograms Xcalibur raw files of ten batches sample was imported into the SIEVE software. The batch of 1500181 was selected as reference chromatogram. In order to focus on the most effective information, time windows of 0-60 min was selected to generate chromatographic fingerprinting. The similarity values obtained by SIEVE software was calculated through the overall evaluation of 10 batches total ion current chromatograms. The identical peaks in 10 batches sample chromatograms can be matched in automatic and proceeded peak alignment. The retention time and peak area of all peak in 10 batches sample make a comparison with the reference chromatogram. The correlation coefficients of all introduced chromatograms relative to that of reference chromatogram would be calculated. The similarity values of 10 samples (No.1500181, 1504101, 1504111, 1504121, 1505211, 1505671, 1508171, 1508191, 1509082 and 1509132) in fingerprintings in positive and negative mode were 1, 0.990, 0.988, 0.990, 0.991, 0.989, 0.991, 0.988, 0.977 and 0.993, respectively. The similarity values were all more than 0.9 in positive and negative mode, which indicated that the samples from different batches had strong similarities with high correlation coefficients of similarities. To some degree, this results demonstrate that the fingerprinting chromatograms of these samples might be used to assess the quality of XBJ injection.
Principal component analysis. PCA was used to further classify the 10 samples. PCA is an analytical method that is used to reduce a large set of variable into a smaller set of "artificial"variables known as principal components (PCs), which account f or most of the variance in the original variables. In the present analysis, the data matrix of ten batches samples and 38 bioactive compounds was imported into the multivariate statistical analysis software SIMCA 14.0. The PCA-X model was adopted to match the data and the original 38 variable dimension generated 2 new variables through software automatically, that is the two principal components. After the data fitting, the principal component 1 of variable was accounted for larger percentage of 63.8%, which could reflect the main characteristics of the original data. So PC1 would be suitable for revealing correlations among the different variables. The Score Scatter Plot (Fig. 4A) is used to evaluate the stability of 10 batches XBJ samples. The deviation represented the degree of stability. The deviation represented the degree of stability. The smaller the deviation in the PC 1 axis, the better stability. The Fig. 4A shows the bias of 10 batches was within ± 2 SD, indicating the quality of 10 batches was more stable. In addition, the bias of 8 batches in 10 batches was within ± 1 SD, while the bias of 2 batches in 10 batches was ranged from ± 1 SD to ± 2 SD.
The PCA Loading Plot could reflect the weight size of original variable in the principal component analysis. The greater the absolute value of original variable in the PCA Loading Plot, the more importance role of original variable in the overall distribution. So the PCA Loading Plot can make it possible to discover the variables leading to the difference. In the Loading Column Plot (Fig. 4B) of the scores, the variables of 13 and 15 were the farthest from the origin on the PC 1 and PC 2 . A13 and A15 as the important quality markers, has a relationship with different batch of drugs on the scatter plot distribution location. In the Fig. 4B, A13 and A15 were positive and the absolute value is larger in Loading Plot on PC1, which make the most batch in the positive quadrant portion of the Score Scatter Plot and keep positively correlated with them. Because the level of A13 and A15 is lower than the average level, the batch 1505211 has obvious anomaly in the overall distribution. So the two components made this batch negatively correlated and this batch was spotted in the negative quadrant part of the Score Scatter Plot. That is to say that two markers responsible for the cluster formation were mainly compounds (A 13 and A 15 ) that suggested that the contents of Hydroxysafflor yellow A and Paeoniflorin had a significant relationship with quality of XBJ injection. In addition, the variables of 14, 21 and 30 had a certain statistical significance compared with other variables. The compounds were Albiflorin, Senkyunolide I/H and Benzoylpaeoniflorin, respectively. This  Table 2. Quantitative analytical results for 38 compounds in XBJ from 10 batches (n = 3, μg/mL).
three compounds could provide some reference meaning for quality evaluation of XBJ injection. In 2013 edition Drug Standards of China, Hydroxysafflor yellow A and Paeoniflorin were selected as markers due to the two highest levels chemical composition. However, the therapeutic effects are frequently considered to be connected with multiple active components, including macro-and micro-components. So, the five markers, Hydroxysafflor yellow A, Paeoniflorin, Albiflorin, Senkyunolide I/H and Benzoylpaeoniflorin, were more meaningful for the quality of XBJ injection.
Assay of the five markers in XBJ sample. An UPLC-MS/MS method was developed for the routine determination of five markers in XBJ samples within 5 minutes. And the method was validated according to the above section "Method validation". Satisfactory linearity and correlation coefficient were achieved with linear ranges. The relative standard deviations of precisions, repeatability, stability and recovery were all meeting requirements. The UPLC-MS/MS method could apply for the analysis of five marks in XBJ samples. The typical chromatograms of a standard mixture of five markers (A) and an XBJ sample (B) are shown in Fig. 5. This UPLC-MS/MS method was simpler in operation and higher in data handling efficiency for widely application. Standard solution and samples preparations. The stock standard solutions of 38 reference standards were dissolved in methanol with concentration of 1.0 mg/mL for each compound, respectively. Then, each stock solution was mixed with 50% methanol to prepare a final mixed standard solution. A series of working standard solutions were prepared by the successive dilution of the mixture of standard solutions with 50% methanol. All the solutions were stored at 4 °C before use. Ten batches of commercial preparations of XBJ were directly subjected to UHPLC-MS analysis after being filtered through a 0.22 μm syringe filter.

Mass spectrometric conditions
Statistical data analysis. The fingerprinting was performed on different XBJ samples by SIEVE 2.0 software (Thermo Scientific, San Jose, USA), which was used for evaluating the similarities between different samples. The similarity was evaluated with the correlation coefficients, and the calculation of correlation coefficients was mainly based on the peak area and retention time. The base peak intensity chromatographic data obtained from the positive or negative ion UHPLC-Q-Orbitrap HRMS analyses were imported in the form of Xcalibur raw files into the SIEVE software. With SIEVE software, the chromatogram can be normalized, and the identical peaks in each chromatogram can be matched in automatic or manual mode. All the batches of XBJ samples were used to construct fingerprinting. Subsequently, the correlation coefficients of all introduced chromatograms relative to that of reference chromatogram would be calculated. In a word, the software made the analysis method accurate and rapid.
Principal component analysis (PCA) involves a mathematical procedure that transforms a number of possibly correlated variables into a smaller number of uncorrelated variables called principal components. This transformation is defined in such a way that the first principal component has as high a variance as possible or accounts for as much of the variability in the data as possible 27 . PCA is an unsupervised pattern recognition technique, which is a data visualization method useful for a rapid means of visualizing similarities or differences within multivariate data 28 . PCA makes it possible to represent objects or variables on a graph, with different objectives to study the proximity of objects in order to differentiate them and to detect atypical objects, and also to analyze the position of objects in varied representations. Thus, we could probably speculate the chemical components causing quality differences in different batches. The PCA was performed on different XBJ samples by SIMCA 14.0 software (Umetrics, Sweden).