Detection of Adulterated Naodesheng Tablet (Naodesheng Pian) via In-Depth Chemical Analysis and Subsequent Reconstruction of Its Pharmacopoeia Q-Markers

Naodesheng Tablet (Naodesheng Pian), a traditional Chinese medicine formula for stroke treatment, is made up of five herbal medicines, i.e., Sanqi, Gegen, Honghua, Shanzha, and Chuanxiong. However, the current Pharmacopoeia quality-marker (Q-marker) system cannot detect possible adulteration. Our study tried to use a new strategy, i.e., standards-library-dependent ultra-high-performance liquid chromatography-quadrupole-Orbitrap mass spectrometry (UHPLC-Q-Orbitrap MS/MS) putative identification, to reconstruct the Q-marker system. Through the strategy, 30 isomers were successfully differentiated (such as 2′-hydroxygenistein, luteolin, and kaempferol; ginsenoside Rg2 and ginsenoside Rg3; ginsenoside Rf and ginsenoside Rg1). In particular, 11 compounds were unexpectedly found in Naodesheng, including 2′-hydroxygenistein, 7,4′-dihydroxyflavone, pectolinarigenin, 7-methoxy-4′-hydroxyisoflavone, scoparone, matrine, 3,3′,4′,5,6,7,8-heptamethoxyflavone, 5-hydroxyflavone, diosgenin, chloesteryl acetate, and (+)-4-cholesten-3-one. In total, 68 compounds were putatively identified and fully elucidated for their MS spectra. Subsequently, relevant compounds were further investigated using UV-vis scanning experiments, semi-quantitative analysis, and quantum chemical calculation. Finally, five adulterated Naodesheng Tablets were used for validation experiments. The experiment successfully detected five adulterated ones via a lower-version LC-MS analysis. On this basis, three new candidates (hydroxy safflor yellow A (HSYA), citric acid, and levistilide A), along with puerarin and notoginsenoside R1, are re-nominated as the Q-markers for LC-MS analysis. The LC-MS analysis of puerarin, notoginsenoside R1, HSYA, citric acid, and levistilide A can clearly detect adulteration regarding all five herbal medicines mentioned above. Therefore, the reconstructed Q-markers are described as a “perfect” quality control system to detect adulteration in Naodesheng and will offer a valuable recommendation for the Pharmacopoeia Commission.


Introduction
Naodesheng Tablet (Naodesheng Pian) is a traditional Chinese medicine (TCM) formula recorded in Chinese Pharmacopoeia (ChP).The Chinese "Naodesheng" means to promote the recovery from cerebral stroke, through activating blood circulation and removing blood stasis as well as clearing the channels.Therefore, it is widely consumed by numerous patients suffering from cerebral stroke [1].Nowadays, there are 60 pharmaceutical factories manufacturing Naodesheng Tablet, according to the National Medical Products Administration of China [2].
The Naodesheng Tablet formula is made up of five herbal medicines, including Gegen, Sanqi, Honghua, Shanzha, and Chuanxiong (Table 1).ChP, however, has already defined its corresponding quality markers (Q-markers) when they were used as individual herbal medicines, that is, puerarin for Gegen, three saponins (ginsenoside Rg1 and Rb1, and notoginsenoside R1) for Sanqi, hydroxy safflor yellow A (HSYA) for Honghua, citric acid for Shanzha, and both levistilide A and ferulic acid for Chuanxiong.However, puerarin, HSYA, levistilide A, and citric acid have been excluded in the current Naodesheng Q-marker system for HPLC analysis (Table 1).This exclusion can cause two limitations.(1) Ferulic acid in Naodesheng cannot specifically characterize the presence of Chuanxiong, although ferulic acid is defined as the Qmarker of individual Chuanxiong by ChP.This is because ferulic acid is also enriched in other herbal medicines, e.g., Honghua [3] and Shanzha [4].(2) ChP has tried to use a TLC tool to analyze puerarin and to characterize the presence of Gegen in Naodesheng [1].However, this characterizing tool is highly tedious and the outcome is so unreliable because it relies on spot comparisons with the Rf value and blue color.The Rf value is well known to be variable and can be affected by external conditions.The blue color is actually a consequence of phenolic -OH interacting with FeCl 3 .Therefore, both Rf value and blue color do not have adequate specificity.
Two limitations further suggest that the current Pharmacopoeia Q-marker system can only specifically characterize Sanqi, because the system uses HPLC to analyze three Sanqi-derived saponins (ginsenoside Rg1, ginsenoside Rb1, and notoginsenoside R1).As a result, the other four herbal medicines lack specific Q-markers in Naodesheng, including Gegen, Honghua, Shanzha, and Chuanxiong.Therefore, the adulteration regarding four herbal medicines will not be detected by the current Pharmacopoeia Q-marker system.For example, if Honghua material is replaced by wood powder, adulterated Naodesheng Tablets will not be detected due to the lack of a Honghua Q-marker.A similar situation may also occur with Chuanxiong, Shanzha, and even Gegen.Now, it has become an inevitable tendency to use some new and high-accuracy technologies, e.g., ultra-high-performance liquid chromatography-quadrupole-Orbitrap mass spectrometry (UHPLC-Q-Orbitrap MS/MS), to reconstruct the Pharmacopoeia Q-marker system.The reconstruction requires a systematical investigation of bioactive compounds in Naodesheng.For this purpose, our study developed a reliable standards-library-dependent UHPLC-Q-Orbitrap MS/MS strategy.
The strategy depends on a set of authentic standards.After these standards were analyzed using UHPLC-Q-Orbitrap MS/MS, numerous and high-accuracy data were obtained and saved in the equipped software.Then, these data were used for matching with those from the Naodesheng sample, which also was subjected to a similar analysis under the same conditions.Through matching tests, the compounds from Naodesheng were putatively identified for their structures and even configurations.
Due to the high efficiency and high accuracy, the strategy is expected to offer reliable outcomes for compound identification.From these identified compounds, appropriate Qmarker candidates will be re-nominated based on experimental and theoretical approaches.Finally, the adulteration detection feasibility of these Q-marker candidates will be further verified by a lower-version LC-MS technology.

UHPLC-Q-Orbitrap MS Identification
Corresponding materials can be found in the following text (Section 3.3).LNT was processed into a sample solution.The LNT sample solution was subsequently assayed by means of the UHPLC-Q-Orbitrap MS/MS method.The total ion current (TIC) diagram is shown in Figure 1.Meanwhile, the main information on chromatographic peaks is detailed in Table 2.The information refers to retention time (R.T.) values, molecular ion peak, main MS/MS fragments, and documental evidence.Through comparison with corresponding authentic standards, 68 compounds were identified (Figure 2).The identification evidence is shown in Supplementary Materials S1-S68.The evidence indicates that the emerging UHPLC-Q-Orbitrap MS analysis was much more effective than previous HPLC-UV analyses [5][6][7][8] because the emerging analysis could simultaneously determine hundreds of compounds.Its high efficiency was further supported by Wu's work which simultaneously identified 189 compounds from Bufei Yishen Formula.However, Wu's work failed to offer a full MS spectrum elucidation of all compounds and also to distinguish isomers [9].Therefore, Wu's work could only be considered as a tentative identification and our work as a putative identification.Our putative identification, however, has been documented to possess evident advantages in MS spectrum elucidation and isomer distinction [10].These advantages could also be found in the present study.As seen in Supplementary Materials S1-S68, all 68 compounds have been elucidated for their MS spectra based on fragmenting principles.The elucidation revealed that there were only 10 −7 RSD values between the calculated and experimental m/z values.For example, a m/z 391 peak in Supplementary Materials S48 was calculated as 2.6 × 10 −7 RSD (391.2848 vs. 391.2850).Such a low RSD value has suggested our identification to be highly reliable.Moreover, the error values (δ) between experimental m/z values and theoretical m/z values of the molecular ions of all identified compounds were also calculated and are listed in Table 2.
In contrast, the previous tentative identification could not offer MS spectrum elucidation and thus had to cite outdated documental data to match their experimental ones [11,12].The identification of calycosin was a typical instance.Its positive model peaks (m/z 285, 270, and 134) were used to match the negative model peak values (m/z 283, 268, and 239).There is obviously no comparability between the two groups of data in m/z values, determination models, and apparatus conditions; correspondingly, the previous study could not offer MS spectrum elucidation and only listed the MS spectrum m/z values [11,12].

UV-Vis Spectrum Scanning and Computational Chemistry Results
To offer further evidence, five Q-marker candidates, citric acid (2), HSYA (7), puerarin (12), notoginsenoside R1 (36), and levistilide A (65), along with two old Q-markers (ginsenoside Rg1 39 and ginsenoside Rb1 50), were scanned for UV-vis spectra.As seen in Figure 3, the five formed a complicated mixture and usually shared the same maximum absorption wavelengths.Even at a range of absorption wavelengths, such as 203, 250, and 325 nm, the detected compounds were limited to several main high-abundance compounds, including puerarin (12), notoginsenoside R1 (36), ginsenoside Rg1 (39), ginsenoside Rb1 (50), and HSYA (7) [6][7][8][43][44][45].This greatly limited the selectivity when monitored by a UV-vis detector.On the other hand, their molecular polarities (i.e., dipole moment values) were close to each other (e.g., HSYA 7, notoginsenoside R1 36, and levistilide A 65, Table 4).As a result, they could not be effectively separated by a polarity-based adsorption chromatographic column (e.g., C 18 ).All these findings from UV-vis spectrum scanning and computational chemistry suggest that conventional HPLC-UV was not applicable for the simultaneous analysis of five Q-markers.Note: The semi-quantification was based on the certified and adulterated Naodesheng Tablet using UHPLC-Q-Orbitrap MS/MS analysis and its results were expressed as mean ± standard deviation (SD) (n = 3).The relevant data are detailed in Supplementary Materials S69.The computational chemistry was conducted using a restricted B3LYP basis set.Dipole moment value, Debye unit; HOMO → LUMO, the energy gap from the highest occupied molecular orbital to the lowest unoccupied molecular orbital, kJ/mol unit.

Adulteration Detection Validation Experiment Based on Five Adulterated Naodesheng Tablets and Low-Version LC-MS
To verify whether the LC-MS technology was applicable for the simultaneous analysis of five Q-marker candidates, this study introduced low-version LC-MS (i.e., UHPLC-ESI-Q-TOF-MS) to analyze CNT 1~CNT 5.As seen in Figure 4A, the UHPLC-ESI-Q-TOF-MS analysis of normal Naodesheng Tablet clearly displayed a puerarin ( 12) peak at R.T. 1.375 min; however, the adulterated Naodesheng Tablet (CNT 1) had no peak at the corresponding site.The comparison suggested the absence of puerarin ( 12) and further indicated the adulteration of Gegen in Naodesheng Tablet.Similarly, the comparison between the two diagrams in Figure 4C evidently illustrates that HSYA (7) was absent in adulterated Naodeshen (CNT 3) and thus, Honghua was adulterated in Naodeshen.Similar successful instances can also be observed in Figure 4B,D,E.Apparently, these successes could be attributed to the high selectivity of the molecular formula extraction technology in LC-MS [46].Meanwhile, these successful experiments also showed that (1) the LC-MS technology was applicable for the analysis of these Q-markers.(2) More importantly, the adulteration regarding all five herbal medicines (Sanqi, Gegen, Honghua, Shanzha, and Chuanxiong) in Naodeshen could be effectively detected.Therefore, the reconstructed adulteration detection Q-marker system was described as a "perfect" one; it would provide valuable consideration for the ChP commission.
Finally, it should be noted that (1) ferulic acid cannot specifically characterize any herbal medicines because it is also distributed in Chuanxiong [28], Honghua [3], and Shanzha [4]; regardless, it has been used as a Q-marker of individual Chuanxiong (Table 1).( 2) The reconstruction of the Q-marker system was based on the analysis of one batch of Naodesheng Tablets in our study.However, these Q-markers were also found in other batches by the previous ones [44,47,48] or Pharmacopoeia itself [1].(3) Although Naodesheng Tablet was reported to be related to the repair of β-amyloid-induced dysfunction [49], the present study does not discuss these bio-pharmacological issues.In fact, the role of β-amyloid is still controversial nowadays [50].

Medicine Materials
Naodesheng Tablet (Lot.Five adulterated Naodesheng Tablets were prepared by our team through the replacement method.Gegen was replaced by wood powder to prepare the first adulterated Naodesheng Tablet, i.e., CNT 1.Similarly, Sanqi was replaced by wood powder to obtain CNT 2. In addition, Honghua, Shanzha, and Chuanxiong were replaced by wood powder to produce CNT 3, CNT 4, and CNT 5, respectively.

Preparation of Lyophilized Aqueous Extract from Naodesheng Tablet and Authentic Standard Solution
To avoid the possible solvent effect [51], Naodesheng Tablet was extracted using distilled water.The extract was lyophilized using a freeze dryer (FDU-1200, Eyela Co., Ltd., Shanghai, China) to prepare a lyophilized powder of Naodesheng Tablet (LNT).The whole process consulted the previous method [52,53] and is summarized in Figure 5.The LNT sample was re-dissolved using methanol under ultrasound treatment and then filtered through a 0.45 µm membrane to prepare the sample solution (~30 mg/mL) [54,55].Similarly, each authentic standard was also dissolved using methanol under ultrasound treatment and then filtered through a 0.45 µm membrane to obtain a standard solution (~10 µg/mL).The sample solution and all standard solutions were kept in a refrigerator (4 • C) for the following analyses.Similar to Naodesheng Tablet, 5 adulterated Tablets (i.e., CNT 1~CNT 5) were, respectively, treated by the above procedure as well.The UHPLC system (Thermo Fisher Scientific, Waltham, MA, USA) was equipped with an Accucore RP-MS LC C 18 column (100 mm × 2.1 mm, 2.6 µm, Thermo Fisher, Waltham, MA, USA) for chromatographic separations.The mobile phase consisted of A (0.1% formic acid in water) and B (methanol) at a flow rate of 0.4 mL min −1 for the negative model.Under the positive model, phase A was replaced by 0.1% formic acid in water containing 5 mmol/L ammonium acetate and phase B was still methanol.The gradient elution was set as follows: 0-5 min, 10% B; 5-14.5 min, 10-100% B; 14.5-16 min, 100% B; 16-16.1 min, 100-10% B; 16.1 min-20 min, 10% B. The column temperature was maintained at 40 • C and the injection volume was 3 µL [56].
The above UHPLC system was coupled with a high-resolution Q-Orbitrap mass spectrometer (Thermo Fisher Scientific, Waltham, MA, USA).The operating parameters were set as follows: auxiliary gas, 10; sheath gas, 40; sweep gas, 0; spray voltage, 4.5 kV.The temperatures of the auxiliary gas heater and capillary were both set at 450 • C. The full MS resolution and data-dependent MS 2 (dd-MS 2 ) were 70,000 and 17500, respectively, while their automatic gain control (AGC) target was 2 × 10 5 .Nitrogen (N 2 ) was applied for spray stabilization and the damping gas in the C-trap.The stepped normalized collision energy was set to 20, 50, and 90 V [57].

Software, Data Acquisition, and Putative Identification
The Xcalibur 4.1 software package and TraceFinder General Quan (Thermo Fisher Scientific Inc., Waltham, MA, USA) were used for data acquisition and analysis.The acquired data included retention time, molecular peak, MS/MS profile, and typical fragments of authentic standards [58].The data were recorded in the software package to build up a database of authentic standards.The data acquisition conditions were set as follows: 100-1500 Da mass range; 5 ppm mass tolerance; 5 S/N threshold; 10 min R.T. window override; 90% isotopic pattern fit threshold.The data of samples were acquired in the software package under the same conditions.Through the comparison, the bioactive compounds from the sample solution were preliminarily identified.After manual elucidation of MS spectrum fragmenting, the bioactive compounds were further confirmed to finish the putative identification.

Semi-Quantification of Re-Nominated Q-Markers
The semi-quantification analyses of 5 re-nominated Q-markers (puerarin 12, ginsenoside Rg1, HSYA 7, citric acid 2, and levistilide A 65) were based on the principle of a previous study with minor modifications [59].Briefly, the linear regression equation was first established through the injection of authentic standard solutions at different volumes into the UHPLC-Q-Exactive-Orbitrap MS system.The equipped Xcalibur 4.1 software offered peak area parameters for these authentic standard solutions.Under the same chromatography and MS spectrum conditions, sample solutions of certified and adulterated Naodesheng Tablets were subsequently injected into the system.According to the linear regression equation and peak area of the Q-markers, their chemical contents were finally quantified and expressed as mean ± SD.

Figure 1 .
Figure 1.The TIC diagrams of Naodesheng Tablet in the UHPLC-Q-Orbitrap MS identification under negative mode (A) and positive mode (B).The positive mode was the supplement for the negative mode.

Figure 2 .
Figure 2. Structures of identified bioactive compounds from Naodesheng Tablet (A) for isomers; (B) for non-isomeric compounds.The chiral atoms in all sugar residue groups have been marked in their absolute configurations to avoid possible misreading.D-glucose is expressed as the Fischer project formula.The wave line in HSYA (7) indicates uncertain stereo configuration.The red tick √ means the old Q-markers.

Figure 5 .
Figure 5.The preparation of the lyophilized aqueous extract of Naodesheng Tablet.

Table 1 .
The information of Naodesheng Tablet and 5 relevant herbal medicines.

Table 2 .
Cont.The peaks with m/z < 50 were also found by the Xcalibur 4.1 Software package, although the scanning mode range was set at m/z 100-1200 in the mass spectra.All identification processes, including MS elucidation, are detailed in Supplementary Materials S1-S68.R.T. values with "*" were detected in positive ion mode, while R.T. values without "*" were detected in negative ion mode.The error values (δ) were calculated using the formula δ = | experimental m/z value − theoretical m/z value | ÷ theoretical m/z value ÷ 10 −6 .

Table 4 .
Semi-quantification results and computational chemistry results (including dipole moment value and HOMO→LUMO energy gap values of Q-marker candidates).