Simultaneous Determination of 32 Pyrrolizidine Alkaloids in Two Traditional Chinese Medicine Preparations by UPLC-MS/MS

Pyrrolizidine alkaloids (PAs) constitute a class of phytotoxin which demonstrates strong hepatotoxicity. In China, many plants containing PAs are used as traditional medicines or medicinal preparations, which could harm human health and safety. Xiaoyao Tablet (XYT) is an antidepressant drug registered in the European Union (EU), Compound Danshen Dropping Pills (CDDP) is a commonly used drug for coronary heart disease, and phase III clinical study is ongoing in the United States. The purpose of this study is to provide data to support the use of Chinese medicine preparations internationally and to establish analytical methods for 32 PAs in XYT and CDDP. The extraction parameters that were optimized include solid-phase extraction (SPE) cartridge, extraction method, and extraction solvent. Then ultra-performance liquid chromatography coupled with triple-quadrupole linear ion-traptandem mass spectrometry (UPLC-MS/MS) was developed to effectively and efficiently quantify the 32 PAs of the XYT and CDDP. The analytical methods for XYT and CDDP were verified respectively. For XYT, the analytical method for 32 PAs was linear, and the correlation coefficient r was greater than 0.994; the recovery (REC%) at 10–2000 μg/kg was 73.3%–118.5%, and the relative standard deviation (RSD%) was 2.1%–15.4%. The CDDP REC% was 71.8%–112.0%, and the RSD% was 2.0%–17.1%. This study provides technical and data support for the registration of Chinese patented medicines in the EU, controls quality and ensures safety, and is committed to the internationalization and standardization of Chinese patented medicines.


Introduction
PAs and their corresponding N-oxides (PANOs) are a large group of phytotoxins which cause hepatotoxicity [1]. At present, more than 660 pyrrolidine nuclear alkaloids and their corresponding nitrogen oxide forms have been identified. PAs are mainly distributed in the Compositae, the Leguminosae, and the Boraginaceae families and mainly exist in plants in the form of nitrogen oxides. In China, many plants containing PAs are used as traditional medicines or medicinal preparations, which could harm human health and safety, such as Tussilago farfara and Arnebia euchroma [2]. Animal-derived food such as honey [3,4], milk [5], eggs [6], and animal organs is prone to contain PAs. PAs are metabolized by the human body to produce unsaturated dehydropyrrole metabolites, which can be complexed with large molecules such as proteins, causing liver cell necrosis, and liver toxicity [7].
In order to control the intake of PAs, relevant departments have successively proposed regulations and recommendations. In 2016, the European Medicines Agency (EMA) clearly stipulated that at least 28 PAs of Chinese herbal medicine should be controlled, and the limit of PAs should be ≤ 0.35 μg PAs/day, and it was recommended to adopt the article "Determination of Pyrrolizidine Alkaloids in Plant Material by SPE-LC-MS/MS (BfR-PA-Tea-2.0/ 2014)" in the determination of PAs in Chinese herbal medicine by MS [8]. In 2018, the EMA updated the "Guideline on Quality of Herbal Medicinal Products/Traditional Herbal Medicinal Products" and the "Guideline on Specifications: Test Procedures and Acceptance Criteria for Herbal Substances, Herbal Preparations and Herbal Medicinal Products/Traditional Herbal Medicinal Products". e released drafts have added new requirements for the control of PAs in medicinal materials and extracts [9,10] where College ter Beoordeling van Geneesmiddelen (CBG) requires that the maximum daily dose of drugs' PAs and their nitrogen oxides has to be controlled below 0.35 μg/day/ 50 kg. In Germany, Bundesinstitut Für Risikobewertung (BfR) used SPE and liquid chromatography-tandem mass spectrometry (LC-MS/MS) to determine 28 PAs in tea and tea products and 17 PAs in honey [11,12]. According to the Chinese Pharmacopoeia 2020, the content of adonifurin in Senecionis Scandentis Hebra should be not more than 0.004% [13]. In 2021, the EMA adopted new regulations specifying the control of 28 PAs [14]. e most common test method for PAs at present is based on LC-MS/MS. On the one hand, tandem mass spectrometry (MS/MS) has a high sensitivity for limit of detection (LOD), and it can detect PAs which are low in content in some plants [15]; on the other hand, multiple PAs/PANOs with similar structures cannot be simultaneously separated by ultra-performance liquid chromatography (UPLC) separation method, and mass spectrometry is required to confirm their ion fragments for qualitative and quantitative determination of multiple PAs. PAs are alkaloids, which are highly hydrophilic, and they are generally extracted using 0.1% formic acid-methanol [16], 0.1% formic acid, 95% ethanol, and other solvents. Purification of the honey is mostly done using SPE and matrix dispersed solid-phase extraction (MSPD), such as strong cation-exchange solid-phase extraction cartridge (SCX) and strong cation-exchange solid-phase extraction cartridge (PCX) [17], and good recovery (REC%) can be obtained.
Xiaoyao tablet (XYT) [18] is a typical modern Chinese patented medicine, and it is also registered with the EMA. It is derived from the "Taiping Huimin Heji Ju Prescription" in the Song Dynasty, which is a representative prescription for soothing the liver and regulating "qi". It is clinically used to treat the liver "qi" discomfort, the liver "qi" stagnation, and other diseases, including depression, gynecological diseases, etc.  [20,21] and honey. ere is no standard method designed specifically for the identification and quantification of the alkaloids in traditional Chinese medicine preparations.
is study focuses on the quantitative methods for the PAs of two traditional Chinese medicine preparations marketed internationally to ensure that they meet the guidelines for international product safety risks. Based on comprehensive literature and regulatory guidelines [22,23], this paper studied the 32 PAs based on multiple regulatory requirements, comprehensive experimental research, and the availability of reference substances. However, due to the characteristics of PAs, the test involves multiple indicators, low limits, and complex matrices. Ultra-performance liquid chromatography coupled with triple-quadrupole linear iontrap tandem mass spectrometry (UPLC-MS/MS), which is highly specific and sensitive, is used in this study to combine chromatographic separation and mass spectrometry and pass the solid phase at the same time. e purposes are to ensure that the extraction method eliminates the interference of the sample matrix and to analyze the 32 PAs quantitatively.

Chemicals and Materials.
Methanol (LC-MS grade) was purchased from Merck (Darmstadt, Germany); formic acid and ammonium formate (chromatography grade) were obtained from Sigma-Aldrich Corporation (St. Louis, MO, USA); ammonia and sulfuric acid were both of analytical grade; five SPE cartridges including SCX, PCX, X-C, C18, and C18/SCX were all 500 mg/6 mL and from Agela Technologies (Tianjin, China); ultrapure water was obtained from a Milli-Q water purification system (Millipore, Billerica, MA, USA).
Five  Journal of Analytical Methods in Chemistry solution (2) was prepared in the same way as standard solution 1, not including IcNO, ImNO, LyNO, and ReNO. e mixed reference solution for CDDP (standard solution 3) was prepared in the same way as standard solution 1, just including IcNo, ImNo, LyNo, and ReNo. e specific concentration of the reference substance is shown in Table 1. All reference solutions were stored at −4°C and protected from light. XYT was crushed into powder. e capsule of CDDP was removed, and the content was taken 2.0 g of the test product which was accurately weighed; then, 40 mL of 1% formic acid was added and sonication was performed at room temperature for 20 minutes (min). en it was centrifuged at 5000 rpm for 10 min. e supernatant was then transferred to a 50 mL volumetric flask and diluted to volume. en, 25 mL of supernatant was precisely pipetted and applied to the PCX cartridge. en, it was pretreated with 5 mL of methanol and 5 mL of 0.05 M sulfuric acid solution and rinsed with 5 mL of 0.05 M sulfuric acid solution and 10 mL of methanol. e PAs were washed with 10 mL ammonia/ methanol (1 : 3, V/V) solution, evaporated, and finally reconstituted with 50% methanol to a total volume of 5 mL. e test product was taken and a blank matrix solution was prepared in the same way as the preparation method of the test solution.

Chromatographic Conditions.
In this study, the UPLC-MS/MS 8060 (SIL-30AC, LC-30AD, CTO-20AC, Shimadzu, Japan), ermo Hypersil Gold C18 (100 mm × 2.1 mm, 1.9 μm) chromatographic column was used. 0.05% formic acid solution (containing 2.5 mM/L ammonium formate) was used as mobile phase A, and methanol-0.05% formic acid solution (containing 2.5 mM/L ammonium formate) was used as mobile phase B. e elution gradient was 0 min, 95%A, 0 min-0.5 min, and then it was changed with a linear gradient to 90%A, 0.5 min-5 min, 85%A, 5 min-11 min, 80% A, 11 min-13 min, 35%A, 13 min-14 min, 5%A, 14 min-16.5 min, keeping isocratic elution for 2.5 min, 16.5 min-16.6 min, 95%A, 16.6 min-18 min, 95%A. e flow rate was set at 0.4 mL/min, the column temperature was maintained at 40°C, and the injection volume was 2 μL. e mass spectrometry conditions were as follows: electrospray ionization (ESI) was used as the ion source, positive ion scanning mode was used, and the monitoring mode was multiple reaction monitoring (MRM). e atomizing, drying, and heating gas flow were set at 3.0 L/min, 10.0 L/min, and 10.0 L/min, respectively, with an interface, DL, and heating block temperature of 300°C, 250°C, and 400°C respectively and CID gas of 270 kPa. e detection and quantification of the substances were performed using MRM with the MassLynx ™ version 4.1 software (Waters). e retention time, ion pairs, and collision energy (CE) were summarized in Table 1.
In this paper, matrix-matched calibration is used for quantification. e extract of the blank sample is used as the diluent, and solutions of different concentrations are added to prepare the solutions, so that they all have the same ionization conditions, so as to reduce the interference of the matrix on the results. 2 μL of the reference solution and the test solution were accurately drawn respectively for determination, and the contents of PAs were calculated by the standard curve method.

Method Validation.
e method was validated in house by evaluating calibration curves, specificity, LOD, LOQ, precision, stability, and recovery.

Calibration Curves, Specificity, LOD, and LOQ.
Calibration curves were plotted with peak areas versus their corresponding concentrations of the reference substance. An appropriate amount of the mixed reference solution was precisely drawn and added to the blank matrix solution to prepare a linearity reference. en, the linearity reference solution was accurately drawn and sampled for quantitative analysis. Two values were entered for each concentration and the corresponding peak area of each was recorded. e prepared concentration of the solution was taken as the xaxis, while the mean of the two peak areas measured was taken as the y-axis. A six-point calibration was used to plot the standard curve and calculate the r value.
In the specificity test, the interference peak response was less than 5% of the LOQ response of the spiked sample at the retention time of the target peak.
For LOD and LOQ, reference substances were added to blank samples since matrix-matched calibrations were used. LOD and LOQ were determined at a signal-to-noise ratio (S/ N) of about 3 and 10, respectively.

Precision and Stability.
e same batch of samples were obtained and spiked with an intermediate level (100 μg/ kg) of the mixed reference solution to evaluate the precision of the developed method. Based on the preparation method of the test solution, two analysts prepared 6 spiked sample solutions in parallel and calculated the REC% and RSD% of 32 PAs in the 6 samples.
In order to evaluate the stability of the method, the same batch of samples was obtained and added to the mixed standard with an intermediate level (100 μg/kg). Samples were then prepared according to the preparation method of the test sample and stored in the refrigerator (−15-25°C).
e REC% of 32 PAs was analyzed at 0 h, 24 h, and 48 h respectively, and the RSD% of REC% at three points was calculated. ree replicates were prepared for each level, and the REC% and RSD% of 9 spiked samples were calculated. Each standard was spiked with three different levels, 3 samples were prepared for each level, and the REC% and RSD% of the 9 spiked samples were calculated.

Matrix Effect.
Matrix effect was defined as ion suppression or enhancement in the process of substance ionization. e matrix effect was evaluated using the slope of calibration curves of standards in solvent and matrixmatched solutions [24].

Development of the LC-MS/MS Method.
e chromatographic conditions such as the mobile phase and the elution gradient were studied in the preliminary study to achieve a desirable chromatographic profile with optimized retention time and peak profiles. e chromatographic method of reference [17] was optimized, and the separation results of 32 PAs after optimization are shown in Figure 2. 0.05% formic acid in water (containing 2.5 mM/L ammonium formate) and methanol-0.05% formic acid solution (containing 2.5 mM/L ammonium formate) were used as mobile phase A to optimize the gradient elution, and the separation effect was found to be adequate. Compared with the commonly used 0.1% formic acid in water and 0.1% formic acid-acetonitrile solution, the separation effect was found to be better, as isomers Ly, Ic, IcNo, and ImNo which were previously not separated could achieve total separation as LyNo and IcNo (resolution � 1.73).
irty-two types of PAs reference substance mixed solutions were analyzed. Firstly, the flow injection method was used to perform a full scan of the precursor ions in positive ion mode to determine the molecular ion peak, and then the molecular ion of the test compound was identified as the parent ion to perform a full scan of its product ions. Two characteristic product ions were selected where the ion pair with a high S/N, good peak shape, and low interference is selected as the quantitative ion pair. MRM positive ion mode was used to optimize various mass spectrometry parameters. e best mass spectrometry parameters were obtained and the separation effect of 32 PAs was observed to determine the retention time of each substance.

Development of the Sample Extraction and Cleanup.
In order to optimize the PA/PANO separation, the extraction and cleanup procedures were developed. In LC-MS/MS determination, purification methods such as SPE are often used to reduce matrix interference and improve accuracy. In the selection of SPE cartridge, the purification effects of 5 SPE cartridges, PCX, SCX, X-C, C18, and C18/SCX, were investigated. Among all the cartridges studied, the retention effect of the C18 cartridge was found to be extremely poor. e liquid using the X-C cartridge was not easily filtered and had poor retention of Er, ErNo, 7-Im, and 7-Ly, and the REC% was 30.2%-69.2%. Moreover, the SCX cartridge had poor retention effect on Er, ErNo, SnNo, 7-Im, and 7-Ly, and the REC% was 20.0%-68.5%. In addition, the C18/SCX cartridge had poor retention effect on Er, ErNo, and SnNo, and the REC% was 41.9%-53.1%. e PCX, SCX, X-C, C18, C18/SCX cartridges all did not retain ImNO and Ly. After extraction and purification optimization, the PCX cartridge was able to recover the 32 PAs studied. erefore, the PCX cartridge was finally used for enrichment and purification, as shown in Figure 3.
For the selection of extraction solvent, extraction method, and extraction time, the following series of studies were carried out. In order to optimize the extraction solvent selection, five extraction solvents, including methanol, 50% methanol, 2% formic acid solution, 1% formic acid solution, and 2% formic acid-methanol, were selected for study. e results showed that the extraction effects of methanol and 2% formic acid-methanol on the reference substance were both poor. On the other hand, for the solution with 50% methanol, 2% formic acid solution, and 1% formic acid solution, each had no significant difference in the extraction effect of the reference substance respectively. However, since the solution with 50% methanol had poor peak shape and the solution with 2% formic acid-water was more acidic, 1% formic acid-water solution was used as the final extraction solvent. A few selections of extraction methods such as ultrasonic, shake, and vortexing were compared. e REC% of the reference substance under the three extraction conditions (20 min sonication, 20 min shaking, and 20 min vortexing) were studied, respectively. e results showed that there was no significant difference in the REC% of the reference substance under the three extraction conditions. Finally, sonication was selected as it is easy to operate. After determining the extraction method as sonication, the effect of different sonication times on recovery was compared.
ere was no significant difference in the REC% of 32 PA reference substances after sonication for 10 min, 20 min, and 30 min. It was found that sonication for 20 minutes was enough to extract PAs, so an intermediate level of sonication time was used for extraction, as shown in Figure 4.
After the cartridge was identified, the elution cleanup step for the cartridge was optimized. Based on comprehensive studies and literature, three elution solvents have been studied, including 5 mL of 0.05 M sulfuric acid aqueous solution with 10 mL of methanol (solvent 1), 5 mL of water with 10 mL of methanol (solvent 2) and 15 mL of water (solvent 3). It was found that the impurity removal effect of (solvent 3) was poor, and the matrix interfered greatly with the sample recovery. e REC% of (solvent 1) was significantly higher than that of (solvent 2). In addition, 5 mL of 0.05 M sulfuric acid aqueous solution with 10 mL of methanol was in line with the theory of alkaloid acid extraction and alkali precipitation and could fully extract alkaloids in the sample. For satisfactory elution power, 5 mL of 0.05 M sulfuric acid aqueous solution with 10 mL of methanol was selected as the elution solvent. e REC% is shown in Figure 5(a).
Next, further studies were performed on the volume of elution solvent, so various volumes of the elution solvent (0.05 M sulfuric acid aqueous solution with methanol) were studied. ree different volumes (3 mL, 5 mL, and 8 mL) were examined for 0.05 M sulfuric acid aqueous solution, and three different volumes (5 mL, 10 mL, and 15 mL) were examined for methanol. e results showed that the different volumes of the elution solvent had no significant effect on the REC%, so an intermediate volume was selected as the volume of elution solvent, 5 mL of 0.05 M sulfuric acid aqueous solution with 10 mL of methanol, for rinsing to remove impurities. e REC % is shown in Figures 5(b) and 5(c). en, the ratio and volume of the elution solvent were optimized, and the effects of different elution ratios and volumes of ammonia/methanol on the recoveries of PAs were studied. ree different ammonia/methanol ratios (3 : 15, 3 : 12, and 3 : 9) were studied. e results showed that the REC% of ammonia/methanol (3 : 9) was the highest among the three ratios. ree different volumes of ammonia water/ methanol (1 : 3, V/V) (5 mL, 10 mL and 15 mL) were studied. e results showed that 10 mL of ammonia water/methanol (1 : 3, V/V) had the highest elution ability, as shown in Figures 5(d) and 5(e).

Selection of Standard Adding Method.
Preliminarily determined conditions were used to confirm the REC% of the PAs. As a result, it was found that the REC% of IcNO, ImNO, LyNO, and ReNO was significantly higher (close to 150%), and further tests were performed to eliminate the unstable factors of the sample. In the references and review of the operation process, there may be the phenomenon that Ic, Im, Ly, and Re are converted into IcNO, ImNO, LyNO, and ReNO. e results showed that the single standard REC % of IcNO, ImNO, and LyNO test solution and the REC% of IcNO, ImNO, LyNO, and ReNO in the four mixed standard test solutions and the 30 mixed standard test solutions (except IcNO, ImNO, LyNO, and ReNO) were at a normal level. In addition, it was found that 30 mixed standard test products (except IcNO, ImNO, LyNO, and ReNO) showed chromatographic peaks in the IcNO, ImNO, and LyNO channels, while another 30 mixed reference solutions did not show any chromatographic peaks. For CDDP, the REC% of IcNO, ImNO, LyNO, and ReNO in the sample solution was all at a normal level. e results showed that the sample treatment process or the CDDP polyethylene glycol excipient might have caused the conversion of the above substances. erefore, in order to ensure the accuracy of the test results and reduce the interference of the matrix, 28 reference substances and four substances were added to the matrix of the CDDP for study.

Effect of Polyethylene Glycol on Recovery.
In terms of the recovery test, when reference substances were added, polyethylene glycol excipients may have interfered with the conversion of Ic, Im, Ly, and Re into IcNO, ImNO, LyNO, and ReNO. In the study where reference substances were added separately, the REC% of Er and ErNo was below 60%, and it was observed that polyethylene glycol affected the enrichment and elution process of the alkaloids. Polyethylene glycol is a high-molecular polymer with a chemical formula of HO (CH 2 CH 2 O) n H [25]. It has excellent dispersibility and adhesion profile. However, the mechanism of how polyethylene glycol promotes the conversion of PAs to PANOs is unknown and further studies are required.

Method Verification for XYT.
For calibration curves, the correlation coefficient r is greater than 0.994. e standard curve, linear range, and r are shown in Table 2. For specificity study, the interfering peak responses were all less than 5% of     Table 4. In terms of stability, the results showed that the 32 PAs of XYT were stable within 48 hours, of which 28 PAs had RSD% of less than 10% and RSD% of 1.9% (IcNO)-16.2% (JbNO).
For repeatability and accuracy, the mean REC% of the eight different matrices was 73.3%-118.5%, and the RSD% was 2.1%-15.4%. e specific values are shown in Table 5. Both methods had adequate repeatability and accuracy, the REC% was within acceptable range, and the repeatability met the requirements.

Method Verification for CDDP.
e verification method of the CDDP is the same as that of Section 2.5 where reference solutions 2 and 3 were also included when the reference substance was added. e results are shown in Table 6. For accuracy, the REC% of the 32 PAs in the sample was 71.8-112.0%, and the RSD% was 2.0-17.1%.

Analysis of the Necessity for Detection.
e main taxonomic group containing the PAs includes Boraginaceae, Senecio, and Eupatorium of Asteraceae and Crotalaria of Leguminosae [26,27]. e 32 PAs were not detected in the XYT and the CDDP. e XYT only contains Atractylodis Macrocephalae Rhizoma, Paeoniae Radix Alba, and other medicinal materials, while the CDDP contains Salviae Miltiorrhizae Radix Et Rhizoma and other medicinal materials that are not of the above-mentioned principal taxonomic category, which is consistent with other literature [28,29]. Since the plants studied may have the risk of contamination with the other plants containing PAs during the harvesting process, the measures taken to control the PAs in the medicines in this paper are able to reduce the workload and ensure the quality and safety of the medicines. At present, there are few types of PA detection

Conclusion
In summary, by extraction with formic acid solution, purification with PCX cartridge, detection with UPLC-MS/MS, and quantification with external standard method, the content of the 32 PAs in XYT and CDDP was determined, and the method was suitable for the detection of the components of the XYT and the other two intermediate extracts. is study established the relevant detection method for traditional Chinese medicine preparations which have complex matrices. It has high sensitivity but simple operation. It also provides a reference for the detection of samples with complex matrices and is of great significance in ensuring the quality and safety of drugs. In this paper, a detection method for PAs of two international products (XYT and CDDP) was successfully developed. It has major significance as these two drugs are registered internationally. is study demonstrated a breakthrough in the limited detection of the Chinese patented medicines.

Data Availability
Relevant research materials are kept in international Industry Center, Tasly Pharmaceutical Group Co., Ltd.

Conflicts of Interest
e authors declare that they have no conflicts of interest.