Chemical Variations among Shengmaisan-Based TCM Patent Drugs by Ultra-High Performance Liquid Chromatography Coupled with Hybrid Quadrupole Orbitrap Mass Spectrometry

Shengmaisan (SMS) is a famous traditional Chinese medicine (TCM) formula to treat coronary heart diseases. It has been developed into several TCM patent drugs to meet the demands of different patients. In this study, a research strategy was proposed to reveal the chemical variations among four SMS-based patent drugs, including Shengmai Oral Solution (Shengmaiyin, SMY), Shengmai Capsule (Shengmai Jiaonang, SMJN), Yiqi Fumai Injection (YQFMI), and Yiqi Fumai Capsule (Yiqi Fumai Jiaonang, YQJN). Firstly, 227 compounds were tentatively identified using an Orbitrap-MS in the full scan/dd-MS2 mode. Secondly, untargeted metabolomics analysis suggested that ginsenosides, steroidal saponins, and lignans were the main types of differential compounds for the four patent drugs. Finally, the contents of 25 compounds were simultaneously determined in 30 batches of samples in the parallel reaction monitoring (PRM) mode. Partial least squares discriminant analysis (PLS-DA) revealed the contents of ginsenosides Re, Rg1, Rb1, Ro, and Rg3, and schisandrin showed the highest intergroup variations. These compounds were chemical markers to differentiate the SMS-based patent drugs.


Introduction
Traditional Chinese medicines (TCMs) are mainly used in clinical practice in the form of formulas [1]. To meet the demands of diverse patients, a lot of popular TCM formulas have been developed into patent drugs [2][3][4][5]. However, the quality of these patent drugs may be affected by crude drug materials and manufacturing technologies. For example, the contents of bioactive compounds in different Gegen Qinlian patent drugs varies significantly [6]. Due to their complex chemical composition, clarification of chemical variation of patent drugs derived from the same TCM formula has been a big challenge.
In this study, we developed a three-step strategy to reveal the chemical variations among the four SMS-based patent drugs. Firstly, the chemical constituents in SMS samples were characterized using a UHPLC/Orbitrap-MS. Secondly, untargeted metabolomics was used to discover potential chemical variations in different SMS samples. Finally, the contents of 17 ginsenosides, 3 steroidal saponins, and 5 lignans in 30 batches of SMS-based samples were determined to confirm the chemical variations.

Chemical Analysis of SMS-Based Patent Drugs
High-resolution mass spectrometry was used to detect and identify the compounds in SMS samples. In total, 227 compounds were tentatively characterized, including 143 ginsenosides, 41 steroidal saponins, and 43 lignans (Table S2). Among them, 36 compounds were identified by comparing with reference standards.

Identification of Lignans
Lignans are characteristic compounds in WWZ ( Figure S1A). The WWZ lignans usually contain methoxyl groups and, thus, could generate a neutral loss (NL) of 15.0238 Da corresponding to a methyl radical ( . CH 3 ) in tandem mass spectrometry [17]. Schisandrin (R21), one of the main lignans in WWZ, contains six methoxyl groups ( Figure 1A). Accordingly, successive NL of 15 Da was observed in its MS/MS spectrum. The normalized collision energy (NCE) of 25% was optimal to yield an abundant product ion at m/z 402.2036 (NL of 15 Da from the parent ion at m/z 417.2274). Thus, using the NL of 15.0238 Da as a filter, 38 lignans were detected in the SMS samples (Table S2). Their structures were characterized by comparing them with reference standards or literature data [17][18][19]. For the other 5 lignans (102, 183, 187, 202, and 208) without NL 15 Da in the MS/MS spectra, they were screened using high-resolution MS data and were tentatively identified by comparing them with reported ones [18,19].

Identification of Steroidal Saponins
Spirostanol-type steroidal saponins are the major compounds in MD [20] ( Figure S1B). Two types of key product ions could be observed in their MS/MS spectra. For type I, the key product ion was m/z 607.3561 (C 33 H 51 O 10 ), corresponding to [ophiopogenin-H+C 6 H 10 O 5 ] − . The optimal NCE was 25% ( Figure 1B) Figure S2A). Using these key product ions, compounds containing the same aglycones could be easily screened. As a result, five type I (9, 55, 68, 131, 137) and seven type II (43-47, 190, 191) compounds were rapidly discovered ( Figure 1B and Figure S2B). The other 29 steroidal saponins were tentatively identified based on previous reports [14,17,20].

Untargeted Metabolomics to Discover Potential Chemical Variations in SMS Samples
The second step of this study was to clarify the potential chemical variations using untargeted metabolomics. Both of the negative and positive ion modes were used ( Figure 2). The data matrices were generated using the Compound Discoverer software (version 3.1, ThermoFisher). There were 681 and 919 variables in the data matrices in negative and positive ion modes, respectively. Partial least squares discriminant analysis (PLS-DA) was performed using the SIMCA-P software (version 13.0) to figure out important chemical markers to discriminate the four patent drugs ( Figure S4). The optimized PLS-DA model for the negative ion mode data described 99.2% of the variations in the response Y (class) (R 2 Y = 0.992), which also predicted 98.8% of the variations (Q 2 = 0.988). The PLS-DA model for the positive ion mode data described 99.4% of the variations in the response Y (class) (R 2 Y = 0.994), which also predicted 99.2% of the variations (Q 2 = 0.992). Finally, 13 important variables showing higher intergroup variance with a variable importance in projection (VIP) value >1.1 [24] were identified, including V113, V214, V337, V338, V441, V472, and V540 in the negative ion mode, and V218, V235, V459, V451, V696, and V762 in the positive ion mode (Table S3). Among them, V113, V214, V338, V441, V472, and V540 are ginsenosides; V337 is a steroidal saponin; and V218, V235, V451, V459, V696, and V762 are lignans. These results indicated that ginsenosides, steroidal saponins, and lignans may be the main chemical differences among the four patent drugs.

Quantitative Analysis of SMS-Based Patent Drugs
To further validate the chemical variations among the four SMS-based patent drug 25 selected compounds, including 17 ginsenosides, 3 steroidal saponins, and 5 lignan were quantitatively determined in 30 batches of samples. The typical PRM chromatograms of the mixed reference standards are displayed in Figure 3. Th quantitative product ion and collision energy for each analyte was optimized using th MS Tune software (Thermo Scientific) and listed in Table 1.

Quantitative Analysis of SMS-Based Patent Drugs
To further validate the chemical variations among the four SMS-based patent drugs, 25 selected compounds, including 17 ginsenosides, 3 steroidal saponins, and 5 lignans, were quantitatively determined in 30 batches of samples. The typical PRM chromatograms of the mixed reference standards are displayed in Figure 3. The quantitative product ion and collision energy for each analyte was optimized using the MS Tune software (Thermo Scientific) and listed in Table 1. The calibration curves of 25 analytes were constructed by plotting the analyte/internal standard peak area ratio (Y) against the concentration (X). The internal standard (IS, astragaloside IV) was used to guarantee precision of the analyses. All the 25 analytes showed good linearity (r 2 = 0.9925 − 0.9993) ( Table S4). The LOQ values ranged from 0.15-  The calibration curves of 25 analytes were constructed by plotting the analyte/internal standard peak area ratio (Y) against the concentration (X). The internal standard (IS, astragaloside IV) was used to guarantee precision of the analyses. All the 25 analytes showed good linearity (r 2 = 0.9925 − 0.9993) ( Table S4). The LOQ values ranged from 0.15-156.41 ng/mL ( Table 1). The RSD values for intraday and interday precisions ranged from 0.28% to 2.80% and 0.10% to 2.40%, respectively, indicating acceptable precision of the method. The RSD values for the stability analysis ranged from 0.35% to 4.06%. The reproducibility test showed a good consistency of the sample preparation process with RSD values ranging from 0.48%-4.70%. Repeatability, precision, and stability variations are listed in Table S5. Recovery of the analytes varied from 98.43% to 108.10% (Table S6), indicating acceptable accuracy of this method.

Sample Analysis
The validated method was used to analyze 30 batches of SMS-based patent drugs, including 12 batches of YQFMI, 6 batches of YQJN, 6 batches of SMY, and 6 batches of SMJN (Table S7). For samples using schisandrin (R21) as a quality control marker, they all met the indicated requirements (0.06 mg/bottle for YQFMI, 0.25 mg/bottle for SMY, 0.15 mg/capsule for YQJN) except for YQJN-1 [5]. On the contrary, the contents of schisandrin in SMJN samples were very low (0.01 mg/capsule), indicating that R21 should not be used as a marker for SMJN. For ginsenoside Re (R3) and ginsenoside Rg1 (R4), all the SMJN samples met the requirements (0.45 mg/capsule) and showed low variations (0.85 ± 0.02 mg/capsule). Similarly, the contents of R3 and R4 in YQMFI samples were also consistent (0.91 ± 0.15 mg/bottle). However, R3 and R4 were hardly detected in SMY and showed great variations in YQJN samples ranging from 0.05 to 1.40 mg/capsule. The total contents of the 25 analytes were consistent in SMJN (3.52 ± 0.06 mg/capsule) and showed moderate variations in SMY and YQFMI (1.30 ± 0.22 and 4.01 ± 0.65 mg/bottle, respectively). For YQJN, significant variations were observed, and the total contents of monitored compounds ranged from 1.45 to 4.30 mg/capsule. The significant intragroup difference among YQJN was mainly due to the variations of the ginsenosides (R3, R4, R9, R12, R13-R16, the total contents ranged from 0.30 to 3.38 mg/capsule) and lignan (R21, ranged from 0.01 to 0.61 mg/capsule).
Due to the different packages and medication regimens, we converted the concentrations (mg/capsule or mg/bottle) into a maximum daily dose (YQFMI, 5.20 g/daily dose; YQJN, 2.22 g/daily dose; SMJN, 2.7 g/daily dose; SMY, 30 mL/daily dose) ( Table S1). As shown in Figure 4A, when SMJN and YQFMI were used by patients, the daily intake of ginsenosides was much higher than the other patent drugs. The total daily intake of ginsenosides in SMJN was 31.55 ± 0.55 mg/day, 30.80 ± 5.11 mg/day for YQFMI, 11.53 ± 8.75 mg/day for YQJN, and 1.08 ± 0.47 mg/day for SMY. For steroid saponins (R19, R20, and R25), their total daily intake also showed a significant difference, ranging from 0.04 to 0.42 mg/day when using different patent drugs. For lignans (R21-R23), the total daily intake via YQJN and SMY (3.06 ± 1.16 mg/day) was much higher than SMJN and YQFMI (0.61 ± 0.37 mg/day). g/daily dose; YQJN, 2.22 g/daily dose; SMJN, 2.7 g/daily dose; SMY, 30 mL/daily dose) ( Table S1). As shown in Figure 4A, when SMJN and YQFMI were used by patients, the daily intake of ginsenosides was much higher than the other patent drugs. The total daily intake of ginsenosides in SMJN was 31.55 ± 0.55 mg/day, 30.80 ± 5.11 mg/day for YQFMI, 11.53 ± 8.75 mg/day for YQJN, and 1.08 ± 0.47 mg/day for SMY. For steroid saponins (R19, R20, and R25), their total daily intake also showed a significant difference, ranging from 0.04 to 0.42 mg/day when using different patent drugs. For lignans (R21-R23), the total daily intake via YQJN and SMY (3.06 ± 1.16 mg/day) was much higher than SMJN and YQFMI (0.61 ± 0.37 mg/day).
The results were re-analyzed via principal component analysis (PCA) using the SIMCA-P software. The first and second principal components accounted for 62.0% and 19.7% of the variation, respectively. In Figure 4B, different patent drugs are grouped into separate clusters. YQJN, SMY, and YQFMI appear closer due to their similar chemical contents. The sample YQJN-1 appears apart from the other YQJN samples, probably due to the high content of ginsenosides (25.70 vs. 8.69 mg, YQJN-1 vs. the other YQJN samples). PLS-DA was used to discover the variables contributing to the grouping of these samples ( Figure S4). As shown in Figure 4C, the contents of R3 (ginsenoside Re), R4 (ginsenoside Rg1), R9 (ginsenoside Rb1), R14 (ginsenoside Ro), R24 (ginsenoside Rg3) from HS, and R21 (schisandrin) from WWZ showed the highest inter-group variance, as suggested by the largest VIP values (>1.0). Their total contents ranged from 2.81-23.89 mg/daily dose for different patent drugs, which could be the main chemical variations.  The results were re-analyzed via principal component analysis (PCA) using the SIMCA-P software. The first and second principal components accounted for 62.0% and 19.7% of the variation, respectively. In Figure 4B, different patent drugs are grouped into separate clusters. YQJN, SMY, and YQFMI appear closer due to their similar chemical contents. The sample YQJN-1 appears apart from the other YQJN samples, probably due to the high content of ginsenosides (25.70 vs. 8.69 mg, YQJN-1 vs. the other YQJN samples). PLS-DA was used to discover the variables contributing to the grouping of these samples ( Figure S4). As shown in Figure 4C, the contents of R3 (ginsenoside Re), R4 (ginsenoside Rg1), R9 (ginsenoside Rb1), R14 (ginsenoside Ro), R24 (ginsenoside Rg3) from HS, and R21 (schisandrin) from WWZ showed the highest inter-group variance, as suggested by the largest VIP values (>1.0). Their total contents ranged from 2.81-23.89 mg/daily dose for different patent drugs, which could be the main chemical variations.

Preparation of Reference Standard Solutions
For qualitative analysis and untargeted metabolomics, an appropriate amount of the 36 reference standards was dissolved in methanol to prepare a mixed standard solution (10 µg/mL for each compound). For quantitative analysis, a mixed stock solution (R1-R25) was prepared by dissolving an appropriate amount of each reference standard in methanol. The mixed standard solution was then serially diluted (dilution factor = 4/3, 2, 4, 8, 32, 128, 512, 2048, 8192, 32768, 131072, and 524288) using methanol. The series of calibration solutions were then diluted by 2-fold using the internal standard solution (IS, astragaloside IV, 500 ng/mL), respectively.

Preparation of Sample Solutions
For the qualitative and untargeted metabolomic analysis, 100 mg of YQFMI, YQJN, and SMJN as well as 1.0 mL of SMY were respectively dissolved in 2 mL of solvent (10% methanol). In addition, 100 mg of HS, MD, and WWZ extracts were individually dissolved in 2 mL of solvent (10% methanol). For quantitative analysis, 100 mg of YQFMI, YQJN, and SMJN as well as 1.0 mL of SMY were respectively diluted with 50 mL of methanol. The samples for LC/MS analysis were then diluted by 2-fold using the IS solution.

Mass Spectrometry
Mass spectrometry analysis was performed on a Q-Exactive hybrid quadrupole Orbitrap mass spectrometer (Thermo Scientific, San Jose, CA, USA) equipped with a heated electrospray ionization source (HESI). It was operated in both negative and positive ion modes. The other parameters were set as follows: spray voltage, ±3.5 kV; sheath gas flow rate, 45 arb; auxiliary gas, 10 arb; capillary temperature, 350 • C; auxiliary temperature, 350 • C; S-lens RF level, 60 V. Full Scan/dd-MS 2 was used to acquire the qualitative and untargeted metabolomics data. The resolutions for MS and MS/MS were set as 70,000 and 17,500, respectively. The scan range was set as m/z 100-1500, and the normalized collision energies (NCE) were 25% and 40%. The ten most abundant ions in each full scan were selected as precursor ions to obtain their MS/MS spectra. Parallel reaction monitoring (PRM) mode was used to acquire the targeted data. Polarity switch was used to detect the 25 selected compounds in a single run. The optimized NCEs and other quantitative information about the 25 reference standards are shown in Table 1. Data were processed using the Xcalibur TM 4.1 software (ThermoFisher, San Jose, CA, USA, 2019).

Method Validation
A mixed standard solution (1250 ng/mL for each analyte) was used for precision and stability tests. Intra-and interday precision was assessed by testing a sample solution six times a day for three consecutive days. The stability was evaluated by analyzing the same solution at 0, 2, 4, 8, 12, and 24 h at 15 • C. The repeatability was described by analyzing six samples (YQFMI-3) prepared using the same method. Accuracy was evaluated by standard addition tests, where a prepared YQFMI-3 sample was mixed with reference standards at the same level (around 100% of the contents, n = 6). Recoveries were calculated using the formula: recovery (%) = (found amount-original amount)/spiked amount × 100%. The limit of quantitation (LOQ) was set at the lowest concentration of the calibration curves according to requirements set by the US Food and Drug Administration [25]. The details are listed in Table 1 and Table S3.

Conclusions
In this study, an integrated strategy was proposed to reveal the chemical variations among four SMS-based patent drugs. Firstly, 227 compounds were identified using a UHPLC/Orbitrap-MS. Secondly, untargeted metabolomics revealed that ginsenosides, steroid saponins, and lignans were significantly different for the SMS-based samples. Finally, the contents of 25 compounds in 30 batches of samples were determined to confirm the chemical variations. Ginsenosides Re, Rg1, Rb1, Ro, and Rg3 as well as schisandrin were the main chemical markers to differentiate the four SMS-based patent drugs.