MRM-based strategy for the homolog-focused detection of minor ginsenosides from notoginseng total saponins by ultra-performance liquid chromatography coupled with hybrid triple quadrupole-linear ion trap mass spectrometry

Abstract

Notoginseng total saponins (NGTS) extract has been frequently used for the treatment of cardiovascular disease and its complications in clinical practice. Several major saponins, such as notoginsenoside R₁ and ginsenosides Rg₁, Re, Rb₁, and Rd, are known, however, the minor ginsenosides that account for 25% of all the NGTS have yet to be elucidated. Herein, a step-wise multiple reaction monitoring (MRM)-based strategy employing staggered [M + HCOO]⁻ > [M − H]⁻ transitions was proposed for the homolog-focused detection of the minor ginsenosides in NGTS using liquid chromatography (LC) coupled with hybrid triple quadrupole-linear ion trap mass spectrometry (LC-Q-Trap/MS), and the efficiency and accuracy of the established method were then assessed using LC coupled with hybrid ion trap/time-of-flight mass spectrometry (LC-IT-TOF/MS). The most important parameter for an MRM scan is the collision energy, which was tuned using an online optimization strategy to further advance the detection sensitivity, and the isotopic peaks shown in the MS² spectra assisted determination of the molecular weights of the unknown minor peaks. As a result of this MRM-based strategy, 107 minor ginsenosides were characterized by comparison with reference compounds and matching with mass cracking patterns, which was more efficient than a full scan analysis using LC-IT-TOF/MS. Moreover, an approximately 98% accuracy for the MRM-based profiling was finally confirmed by the target-list-dependent scan on LC-IT-TOF/MS. Taken together, these results suggest that MRM-based strategy could be used as a reliable tool for screening and identifying trace components in ginsenoside-enriched herbal products and other homolog-gathered extracts.

---

1. Introduction

High-resolution tandem mass spectrometry (HR-MS/MS) has been widely recognized as a work horse for profiling the chemical constitution of complex matrices, especially when used in conjunction with the liquid chromatography (LC), because it is capable of supplying not only the accurate quasi-molecular ions but also the abundant fragment ions that are very useful for structural elucidation.^1,2 Customarily, the mass spectral information used for structure characterization on the HR-MS/MS system is collected via an MS¹ full scan mode and automatically triggered mass fragmentation of the most abundant precursor ions, which results in a challenging task for comprehensive data acquisition of trace components due to co-elution with major components and interference from the substrates.

In contrast to the full scan mode in HR-MS/MS, a multiple reaction monitoring (MRM) scan on a triple quadrupole mass spectrometry or a hybrid triple quadrupole-linear ion trap mass spectrometry (Q-Trap) is advantageous for accurately measuring analytes with large concentration spans in complex samples because of the superior sensitivity and the wide dynamic range.³ Furthermore, this unique selectivity can significantly reduce the co-eluting ions and can produce more product ions of the potential targets within a single analytical run. Overall, the MRM scan approach is superior to the full MSⁿ scan in HR-MS/MS for exploring trace components.

Ginsenosides have been widely accepted as the effective ingredients of Panax plants, such as P. ginseng, P. notoginseng, etc., and their activities include inhibiting platelet aggregation,⁴ increasing cerebral blood flow,⁵ and improving neurological behavior,^6,7 among others. Notoginseng total saponins (NGTS), which is an extract purified from P. notoginseng through macro-porous resin column chromatography and is mainly composed of ginsenosides, has been documented in the Chinese Pharmacopoeia for the treatment of cardiovascular disease and its complications.^8,9 In the monograph of NGTS, the total content of ginsenosides Rb₁, Rd, Re, and Rg₁, and notoginsenoside R₁ should account for more than 75% of the whole NGTS.⁹ However, chemical diversity, along with low contents, make the remaining 25% of components in NGTS difficult to purify and characterize. As herbal medicine is a multi-component and multi-target agent, the minor components are just as important as the major ones for clarification of the effective material basis of NGTS and underlying mechanisms for treating cardiovascular disease, which is the main reason for the modernization and wide application of NGTS. Therefore, in-depth characterization of the chemical profile of NGTS, especially regarding the minor components was performed here by proposing an MRM-based strategy.

In previous studies, HR-MS/MS was frequently applied to detect the trace triterpenoid saponins by removing the major constituents.^10,11 However, it is possible that the minor ginsenosides may be removed along with the major constituents when they co-elute. Empirically, [M + HCOO]⁻ ions are usually observed as the primary signals for ginsenosides under the negative ionization mode when formic acid is introduced as the mobile phase addictive, and the MRM-based method adopting formate anion-to-deprotonated ion transitions ([M + HCOO]⁻ > [M − H]⁻) can provide a meaningful choice to screen and identify the saponins in ginsenoside-enriched herbal products.^3,12,13 In addition, the collision energy (CE) value for dissociating the combination between the adduct ion and the neutral molecule is comparable among the ginsenosides,¹² which overcomes the shortcoming of an MRM-based approach that reference compounds are required to optimize the precursor ions, the product ions, and the related mass parameters, especially the CE value. However, the accuracy and efficiency of such a method need to be confirmed. Additionally, we know that the molecular weights (M.W.s) of unknown components determined by the MRM method are not as precise as HR-MS/MS determined. Therefore, in this paper, an MRM-based strategy was first established to detect and analyze the minor ginsenosides in NGTS. Then, a full scan together with a target-list-dependent scan on the hybrid ion trap/time-of-flight mass spectrometry (IT-TOF/MS), was utilized to elucidate the efficiency and accuracy of the MRM-based strategy. We envision the MRM-based method to be efficient and accurate, and to be a reliable choice to screen and identify the minor components of ginsenoside-enriched herbal products and other homolog-gathered extracts.

2. Materials and methods

2.1 Chemicals and materials

A total of 27 reference compounds, including notoginsenoside R₁; ginsenosides Rg₁, F₃, Rb₁, Rb₂, Rb₃, Re, Rc, Rd, F₁, F₂, Rf, Rh₄, Rk₃, Rg₅, and Rg₆; pseudo-ginsenoside F₁₁ (F₁₁); 20(S)-ginsenosides Rg₂, Rh₁, Rh₂, Rg₃, 20(R)-ginsenoside Rh₂; compound K (CK); protopanaxadiol (PPD), panaxadiol (PD), protopanaxatriol (PPT), and panaxatriol (PT), were purchased from the Chengdu Must Bio-Tech Co., Ltd (Chengdu, China). Their purities were determined to be greater than 98% by high performance liquid chromatography (HPLC) analyses. NGTS, containing notoginsenoside R₁ (6.2%) and ginsenosides Rg₁ (26.6%), Re (4.1%), Rb₁ (32.5%), and Rd (6.6%), prepared according to the Monograph of Chinese Pharmacopoeia (version 2015),⁹ was purchased from Yunnan Plant Pharm. Co., Ltd (Yunnan, China), and the quality inspection report of the NGTS is shown in Tables S1 and S2.†

Acetonitrile (ACN), methanol, and formic acid of optima® LC/MS grade were purchased from Thermo-Fisher (Rockford, IL, USA). Deionized water was prepared in-house using a Milli-Q integral water purification system (Millipore, Bedford, MA, USA). The other chemicals were of analytical grade and were obtained commercially from Beijing Chemical Works (Beijing, China).

2.2 Sample and reference solution preparations

The NGTS solution and all of the 27 reference standards were individually prepared with 50% aqueous methanol, and stored at 4 °C until use. Afterwards, the mixed standard stock solution was prepared by pooling all of the stock solutions. All of the abovementioned solutions were centrifuged at 12 000 rpm for 10 min at 4 °C, and each supernatant was filtered through a 0.22 μm membrane.

2.3 Chromatographic separation

An Acquity UPLC® HSS T3 column (100 mm × 2.1 mm i.d., particle size 1.8 μm, Waters, Ltd., USA) protected by a Van Guard™ HSS T3 (5 mm × 2.1 mm i.d., 1.8 μm, Waters, USA) was used in all the chromatographic separations. The mobile phase consisted of 0.01% formic acid ACN (A) and 0.01% formic acid aqueous solution (B). The gradient program was set as follows: 0–8 min, 0–19% A; 8–10 min, 19–20% A; 10–12 min, 20–60% A; 12–12.5 min, 60–100% A; and 12.5–15 min, 100% A. The flow rate was 0.4 mL min⁻¹. The column temperature was maintained at 35 °C. At the end of each run, 100% B was delivered for another 5 min to equilibrate the entire system.

2.4 Constructing the step-wise MRM algorithm for ginsenosides-focused profiling

It is well known that the CE values play the determinant role in the high sensitivity of an MRM scan; therefore the first step is to optimize the CE values by using the representative reference compounds. Herein, an online stepped optimization strategy, an alternative method of manual optimization, was implemented to rapidly obtain the optimal parameters of the formate anion-to-deprotonated ion transitions.¹⁴ It has been reported that the optimal CE value of [M + HCOO]⁻ > [M − H]⁻ ions of the ginsenosides was at −32 eV;¹³ thus, the stepped CE levels from −22 to −47 eV with a step-size of 5 eV were assayed. To circumvent the self-inspection of Analyst software, the pseudo ion transitions were typed into the monitoring list such as m/z 991.500 > 945.500, 991.501 > 945.501, and so on, corresponding to those staggered CE values (−47, −42, etc.). The optimal CE was defined as the CE in which the highest response is obtained for each ion pair. Then, the step-wise MRM algorithm was introduced to globally characterize the chemome of NGTS on an Acquity ultra performance liquid chromatography (UPLC) H-class system (Waters, Corp., Milford, MA, USA), coupled with a Sciex Q-Trap 4500 mass spectrometry (Foster City, CA, USA) via an ESI interface. A list of [M + HCOO]⁻ > [M − H]⁻ ion transitions (that is x > x − 46, x > 565.3, e.g., m/z 991.5 > 945.5) with their corresponding optimal CE values were constructed for the MRM acquisition to detect the ginsenosides in NGTS. To guarantee an adequate dwell time (4 ms) for each ion pair in a total acquisition time (also known as cycle time) of less than 1.5 s, the MRM acquisitions using [M + HCOO]⁻ > [M − H]⁻ ion pairings were performed in four separate runs, i.e., m/z 565.3–789.4, 791.4–999.5, 1001.5–1207.5, and 1209.6–1369.6, respectively. The declustering potential (DP), entrance potential (EP), and cell exit potential (CXP) of all the [M + HCOO]⁻ > [M − H]⁻ ion transitions were fixed at −100 V, −10 V, and −16 V, respectively. Two enhanced product ion (EPI) scans were triggered after each survey scan.

Curtain gas and two source gases (GS1 and GS2) were maintained at 35 psi, 45 psi, and 45 psi, respectively. The source temperature was set to 450 °C. The sprayer voltage was fixed at −4500 V for the negative polarity. Both the Q1 and Q3 cells were operated at unit resolution. Criteria for the information-dependent acquisition (IDA) of EPI was set for the two most intense ions in each dynamic background subtracted survey scan spectrum with an intensity threshold of 500 counts per second (cps). Each ion could be selected for a maximum of two occurrences and then automatically excluded for 20 s. The scan speed for EPI was 10 000 Da s⁻¹. The lower limits of the EPI scan range in all the four separate runs were set as 100 Da, whereas the upper limits were respectively set as 800 Da, 1000 Da, 1250 Da, and 1450 Da. The CE of both dependent experiments was set at −50 eV with a collision energy spread (CES) of 40 eV. The dynamic fill time was used to ensure that the linear ion trap was not overfilled. A 1 μL aliquot of NGTS (8 mg mL⁻¹, sample-loading amount: 8 μg) was injected into the LC-Q-Trap/MS system for analysis.

2.5 LC-IT-TOF/MSⁿ analysis

Following the MRM-IDA-EPI analysis, the ginsenosides were preliminarily detected and characterized, and then a Shimadzu LC-IT-TOF/MS (Kyoto, Japan) was adopted to test its efficiency and accuracy. First, the mass information of the components in the NGTS was collected using the MS¹ full scan coupled with the automatic tandem mass spectrometry which was triggered by using the most abundant precursor ion. Second, the MS¹ full scan along with the excluded ions list (the major pre-known saponins, such as notoginsenoside R₁, and ginsenosides Rg₁, Re, Rb₁, and Rd; Table S3†) was preliminarily applied to detect the minor components in NGTS. Third, all the HR-MS/MS data of the ginsenosides detected by the MRM scan were collected by introducing the quasi-molecular ions of the pre-characterized saponins into the preferred ions list of the full scan (Table S4†). The optimized MS parameters were set as follows: negative ion mode; electrospray voltage, −3.5 kV; detector voltage, 1.7 kV; curved desolvation line (CDL) temperature and heat block temperature, 200 °C; nebulizing gas (N₂), 1.5 L min⁻¹; drying gas (N₂) pressure, 100 kPa; scan ranges, m/z 100–1500 for MS¹, m/z 100–1400 for MS², and m/z 100–1300 for MS³; collision energy, 50% for MS², and MS³ with a region pressure of 1.4 × 10⁻⁴ Pa; ion trap pressure, 1.8 × 10⁻² Pa; and ion accumulation time, 30 ms. The accurate mass determination was calibrated using sodium trifluoroacetate. Ultra-high purity argon was used as the collision gas for the collision-induced dissociation (CID) experiments. A 4 μL aliquot of NGTS (8 mg mL⁻¹, sample-loading amount: 32 μg) was injected into the LC-MS system for analysis. The data acquisition and analysis were achieved using LCMS Solution Version 3 software (Shimadzu).

3. Results and discussion

3.1 Establishment of the MRM-based strategy

It has been shown that the sum content of notoginsenoside R₁, and ginsenosides Rg₁, Re, Rb₁, and Rd account for more than 75% of the total NGTS content (Tables S1 and S2†), but the remaining constituents (nearly 25%) in NGTS are still unclear due to their low abundance. Since the MRM algorithm has advantages for profiling the trace components due to its high sensitivity, a strategy was proposed, as illustrated in Fig. 1. First, the major structure types of ginsenosides and their mass fragmentation patterns were summarized from the accessible databases, such as Pubmed, ACS, China national knowledge infrastructure (CNKI), and Chemspider. The optimal collision energies of the [M + HCOO]⁻ > [M − H]⁻ transitions were obtained by using an online stepped optimization method with the representative reference compounds, and then, the formate anion-to-deprotonated ion transitions (mass range: m/z 565.3–1369.6 for Q1; step-size: 2 Da) with the optimal CE values were introduced into an MRM list to monitor the ginsenosides in NGTS. Two EPI scans were triggered by the IDA method to generate the complementary MS² spectra for the structure characterization. Second, the isotopic signals were utilized to assist in determining the deprotonated molecular ions of the unknown peaks, as well as the formate anions. Consequently, the detected ginsenosides were characterized by comparing with the reference compounds and by analyzing their mass cracking patterns. Finally, the efficiency and accuracy of the MRM-based strategy were investigated by implementing the full scan, the full scan along with an excluded ions list (Table S3†), and the full scan together with a preferred ions list (Table S4†) on a LC-IT-TOF/MS system.

Fig. 1 Workflow of the MRM-based strategy for the targeted detection of ginsenosides in NGTS.

3.2 Mass fragmentation pathways of ginsenosides

Typical ginsenosides are composed of a triterpene sapogenin and one or two sugar chains. 20(S)-Protopanaxadiol (PPD) and 20(S)-protopanaxatriol (PPT) represent the most common sapogenins for ginsenosides in P. notoginseng.^15,16 The minor changes on the C-17 side chain or sugar chains usually yield other subtypes (Fig. 2). Previous studies confirmed that negative ionization afforded a higher response for most ginsenosides than positive ionization,¹⁵ therefore, negative ionization was adopted to achieve a holistic characterization of the ginsenosides in NGTS. In the negative ESI mode, all the investigated reference ginsenosides exhibited predominant formate adduct ions ([M + HCOO]⁻), and then, the deprotonated molecular ions ([M − H]⁻) were produced by the neutral cleavage of the formic acid (HCOOH, 46 Da) unit. Subsequently, the glycosidic bonds cracked and produced the corresponding neutral losses, for instance, glucosyl (162 Da), rhamnosyl (146 Da), xylosyl/arabinosyl (132 Da), etc. Finally, the deprotonated sapogenin ions ([A − H]⁻) were generated, which can be utilized as the preliminary diagnostic signals to determine the sapogenins. As there are various isomeric sapogenins, the other fragment ions generated from the C-17 side chain were adopted to confirm or differentiate the versatile sapogenins (Fig. 2).

Fig. 2 Plausible structures of the sapogenins identified from NGTS.

3.3 Optimization of the CE values and construction of the step-wise MRM scan

The online stepped optimization method is as reliable as the manual optimization process.^17,18 For the 27 reference compounds, the online stepped optimization method can obtain the optimum CE values within 15 min in one LC-MS/MS run. In contrast, at least 135 min (5 min per compound) are required when using the manual optimization method. Therefore, the online stepped optimization method was utilized here to obtain the optimal CE values of [M + HCOO]⁻ > [M − H]⁻ transitions, and the variation in peak area with CE value for the 27 reference ginsenosides is showed in Fig. 3. It is notable that the optimal CE values were related to the number of the glycosyl groups the structure possesses. The ginsenosides containing one glycosyl unit (Fig. 3A) obtained the highest responses corresponding to a CE of −27 eV. For Rg₆, Rf, and F₂, which possess two glycosyl groups, the highest responses were reached at a CE value of −32 eV (Fig. 3B). For the ginsenosides having three glycosyls, both CE values at −32 eV and −37 eV provide considerable responses (Fig. 3C). However, a broad range of CE values, from −32 eV to −42 eV, can provide considerable responses in saponins comprised of four glycosyl residues (Fig. 3D). Unlike saponins, all four sapogenins (PPD, PD, PPT, PT) do not present the predominant formate adduct ions, and are not detected in NGTS. Thus, only the MRM-based strategy was utilized to profile the saponins in NGTS. From the accessible databases, [PPT 1]-O-xyloside (M.W. 524.33) and [PPT 23]-hexa-O-glucoside (M.W. 1320) provided the scan range of m/z 565.3–1369.6 for the MRM analysis. To guarantee an adequate dwell time (4 ms) for each ion pair in a cycle time of less than 1.5 s, the MRM acquisitions using [M + HCOO]⁻ > [M − H]⁻ ion pairings were scanned in four separate runs, m/z 565.3–789.4, 791.4–999.5, 1001.5–1207.5, and 1209.6–1369.6, and the CE values were set as −27 eV, −32 eV, −37 eV, and −42 eV, respectively. Because the charges of pseudo-molecular ions of typical ginsenosides are odd numbers, the step-size of the ion transition list was set as 2 Da.

Fig. 3 Collision energy (CE) levels of the representative ginsenoside references were optimized by an online stepped optimization strategy; (A) the saponins containing monoglycosyl unit, (B) the compounds having di-glycosyl groups, (C) the ginsenosides possessing tri-glycosyl moieties, (D) the ginsenosides occupying tetra-glycosyl residues.

3.4 Detection and characterization of the ginsenosides in NGTS by the step-wise MRM-IDA-EPI analysis

One advantages of the MRM approach is its superior selectivity as it can filter out complicated background ions. Co-eluting ions can be distinguished and identified from specific MRM transitions. Therefore, components present at trace levels in complex matrices can be detected under low background interference using an MRM scan. However, the M.W.s of unknown components determined by the MRM scan were not accurate enough due to the low resolution of the Q-Trap system.¹³ Such a drawback may increase the possibility of false positive assignments.

It is noteworthy that the high sensitivity of the MRM algorithm enables the isotopic peaks to reach an intensity threshold of 500 cps and thus triggers the EPI scans. For instance, peak 88 could be simultaneously extracted out by ion pairs of m/z 1155.4 > 1109.5 and 1153.3 > 1107.5. Both ion pairs exhibited similar fragmentation behaviors (Fig. 4A). Moreover, the 2 Da mass difference and the descending intensities (1153.3: 9.0 × 10⁶ cps; 1155.4: 2.0 × 10⁶ cps, Fig. 4A2 and A3) agree well with the properties of the isotopic ions. As a consequence, m/z 1107.5 with a higher intensity was preliminarily determined as the real pseudo-molecular ion of peak 88. The significant fragment ions at m/z 945.5, 783.4, 621.4, and 459.3 suggested the presence of four hexosyl residues. Additionally, the [A − H]⁻ ion at m/z 459.4 yielded a fragment ion at m/z 375.2 [A − H − C₆H₁₂]⁻, which is highly consistent with that of [PPD 7] (Fig. 2). Thus, peak 88 was tentatively assigned as [PPD 7]-tetra-O-glucoside, and was further confirmed as ginsenoside Rb₁ by the reference compound. Consequently, the isotopic mass profiles were utilized to deduce the real M.W.s of the unknown peaks, along with the formate anions. All of the detected ginsenosides were finally characterized by integrating the usage of the reference compounds and the mass cracking patterns. For instance, when extracting the MRM transitions of 863.3 > 817.4 and 861.4 > 815.5, three sequential peaks (19, 23, and 24) emerged from the substantial background (Fig. S1A1†). Comparing the relative intensities of the fragment ions and their corresponding extracted ion chromatograms (EICs), the quasi-molecular ions of peaks 19 and 23 were determined as m/z 815.5 (Fig. S1A2†). Whereas, the deprotonated molecular ion of peak 24 was concluded to be m/z 817.4 (Fig. S1A3 and S1A4†). Additionally, peaks 19 and 23 exhibited the same deprotonated sapogenin ions at m/z 491.4, indicating their sapogenins could be [PPD 11], [PPT 10], [PPT 11], [PPT 12], [PPT 13], [PPT 14], or [PPT 15] (Fig. 2). The identical fragment ions at m/z 403.4 and 391.2, originated by cracking a C₄H₁₀O₂ unit and a C₆H₁₂O unit from the sapogenin ion, respectively, suggested that their sapogenins are [PPT 12] or [PPT 15]. Similarly, peak 24 was plausibly assigned as di-glucosidated [PPT 16] from the observation of the prominent fragment ions at m/z 565.2 and 403.1.

Fig. 4 The extracted ion current chromatogram (EIC) of peak 90 and its corresponding mass profiles obtained by the MRM on a Q-Trap/MS system (A) and the full scan on a IT-TOF/MS technique (B); (A1) EICs of m/z 1155.4 > 1109.3 and 1153.3 > 1107.5, (A2) MS² spectrum of m/z 1155.4, (A3) MS² spectrum of m/z 1153.3; (B1) EIC of m/z 1153.6006, (B2) MS¹ spectrum of m/z 1153.6047, (B3) MS² spectrum of m/z 1153.6047.

Based on the aforementioned analysis, a total of 109 ginsenosides, including 104 in trace amounts, were detected and plausibly assigned (Tables 1 and S5†). Of these ginsenosides, 12 were affirmed by comparing the mass spectral data and the retention times with those of authentic references.

Table 1 Identification of NGTS components

No.	tR (min)	Identification	[M − H]^− or [M + HCOO]^−		Error (ppm)	Formula	Scan mode^e
			Measured (m/z)	Predicted (m/z)			F	FE	FP	M
a The compound identified by the reference compound.b The compound additionally detected by the full scan along with an excluded and preferred ions list.c The compound revised by the LC-IT-TOF/MS analysis.d The potential new compound.e Scan mode: full scan (F); full scan along with an excluded ions list (FE); full scan together with a preferred ions list (FP) on the LC-IT-TOF/MS^n analysis; and MRM scan (M) on the LC-Q-Trap/MS system.
1	7.62	[PPT 16]-6-Glucosyl-xylosyl-20-glucoside or its isomer	995.5423	995.5427	−0.40	C47H82O19			+	+
2	7.68	Vinaginsenoside R22 or its isomer	879.4974	879.4953	2.39	C42H74O16	+	+	+	+
3	7.80	[PPT 21]-6-Glucosyl-xylosyl-20-glucoside or its isomer	1011.5334	1011.5376	−4.15	C47H82O20		+	+	+
4	7.94	Vinaginsenoside R22 or its isomer	879.4921	879.4953	−3.64	C42H74O16	+	+	+	+
5	8.16	[PPT 21]-6-Glucosyl-xylosyl-20-glucoside or its isomer	1011.5357	1011.5376	−1.88	C47H82O20			+	+
6	8.22	[PPT 21]-6-Rutinosyl-20-glucoside/[PPT 21]-6-rhamnosyl-20-glucosyl-glucoside	1025.5493	1025.5532	−3.80	C48H84O20			+	+
7	8.38	Vinaginsenoside R22 or its isomer	879.4921	879.4953	−3.64	C42H74O16			+	+
8^b	8.39	[PPD 11]-3-Glucosyl-glucoside/[PPT 12]-6,20-di-O-glucoside/[PPT 13]-di-O-glucoside	861.4821	861.4848	−3.13	C42H72O15		+	+
9	8.47	[PPT 21]-6-Rutinosyl-20-glucoside/[PPT 21]-6-rhamnosyl-20-glucosyl-glucoside	1025.5510	1025.5532	−2.15	C48H84O20			+	+
10^b	8.58	[PPD 11]-3-Glucosyl-glucoside/[PPT 12]-6,20-di-O-glucoside/[PPT 13]-di-O-glucoside	861.4869	861.4848	2.44	C42H72O15			+
11	8.73	Vinaginsenoside R13 or its isomer	1025.5497	1025.5532	−3.41	C48H84O20			+	+
12	8.91	Notoginsenoside H or its isomer	993.5226	993.5270	−4.43	C47H80O19		+		+
13^c^,^d	8.91	Dicaffeoyl-[PPD 10]	847.4654	847.4633	2.45	C48H66O10	+	+	+	+
14	8.97	Ginsenoside Re4 or its isomer	977.5329	977.5321	0.82	C47H80O18			+	+
15	9.16	Floaginsenoside B or its isomer	861.4854	861.4848	0.70	C42H72O15	+	+	+	+
16	9.23	[PPT 16]-3(6),20-Di-O-glucoside/[PPD 12]-6(20)-glucosyl-glucoside	863.5006	863.5004	0.23	C42H74O15	+	+		+
17^c^,^d	9.42	Dicaffeoyl-[PPD 10]	847.4641	847.4633	0.94	C48H66O10			+	+
18	9.48	Sanshichisaponin G or its isomer	993.5297	993.5270	2.72	C47H80O19			+	+
19	9.52	Floaginsenoside B or its isomer	861.4833	861.4848	−1.74	C42H72O15		+	+	+
20	9.64	[PPT 16]-20-Glucosyl-6-rutinoside or its isomer	1009.5555	1009.5583	−2.77	C48H84O19		+		+
21	9.65	[PPT 16]-6-Glucosyl-xylosyl-20-glucoside or its isomer	949.5365	949.5378	−1.37	C47H82O19	+	+	+	+
22^d	9.74	[PPT 1]-O-Xylosyl-O-glucoside	731.3854	731.3854	0	C36H60O15		+		+
23	9.76	Floaginsenoside B or its isomer	861.4842	861.4848	−0.70	C42H72O15	+	+	+	+
24	9.86	[PPT 16]-3,20-Di-O-glucoside/[PPT 16]-6,20-di-O-glucoside/[PPT 16]-6(20)-glucosyl-glucoside	863.4997	863.5004	−0.81	C42H74O15			+	+
25	9.89	Majoroside F6 or its isomer	1007.5424	1007.5427	−0.30	C48H82O19			+	+
26	9.99	[PPT 8]-6,20-Di-O-glucoside or its isomer	859.4655	859.4691	−4.19	C42H70O15	+	+		+
27^d	10.01	[PPT 1]-O-Rutinoside	745.4028	745.4016	1.61	C37H62O15		+		+
28	10.02	[PPT 21]-20-Xylosyl-3-glucoside or its isomer	849.4866	849.4848	2.12	C41H72O15	+	+	+	+
29^d	10.04	[PPT 8]-O-Xylosyl-di-O-glucoside	991.5159	991.5114	4.54	C47H78O19			+	+
30	10.24	Ginsenoside Re8 or its isomer	961.5362	961.5372	−1.04	C48H82O19			+	+
31	10.36	[PPT 11]-6-Glucosyl-glucoside/[PPT 11]-3(6),20-di-O-glucoside/[PPT 14]-6,20-di-O-glucoside/[PPD 10]-3-glucosyl-glucoside/[PPT 12]-6,20-di-O-glucoside/[PPT 13]-di-O-glucoside	861.4850	861.4848	0.23	C42H72O15			+	+
32^d	10.37	[PPT 8]/[PPT 9]-O-Xylosyl-di-O-glucoside	945.4997	945.5009	−1.27	C47H78O19			+	+
33	10.37	[PPT 16]-20-Glucosyl-6-rutinoside or its isomer	963.5488	963.5534	−4.77	C48H84O19			+	+
34	10.40	Floralginsenoside C or its isomer	801.4657	801.4642	1.87	C41H70O15			+	+
35	10.49	[PPT 8]-6,20-Di-O-glucoside or its isomer	859.4668	859.4691	−2.68	C42H70O15		+		+
36	10.52	[PPT 21]-20-Xylosyl-3-glucoside or its isomer	849.4835	849.4848	−1.53	C41H72O15			+	+
38^d	10.55	[PPT 19]-O-Rhamnosyl-O-glucoside	861.4829	861.4848	−2.21	C42H72O15	+	+		+
37	10.59	Floralginsenoside I or floralginsenoside J	1023.5367	1023.5376	−0.88	C48H82O20	+		+	+
39	10.65	[PPT 21]-20-Xylosyl-3-glucoside or its isomer	849.4822	849.4848	−3.06	C41H72O15		+		+
40	10.73	Ginsenoside Re8 or its isomer	1007.5373	1007.5427	−5.36	C48H82O19			+	+
41	10.84	[PPD 12]/[PPT 16]-O-Rhamnosyl-O-glucosyl-O-glucuronide	1023.5384	1023.5376	0.78	C48H82O20			+	+
42	10.86	Floralginsenoside C or its isomer	801.4633	801.4642	−1.12	C41H70O15		+	+	+
43	10.91	[PPT 8]-6,20-Di-O-glucoside or its isomer	859.4673	859.4691	−2.09	C42H70O15			+	+
44	11.09	[PPT 21]-20-Xylosyl-3-glucoside or its isomer	849.4837	849.4848	−1.29	C41H72O15			+	+
45^d	11.15	[PPD 10]-O-Glucosyl-O-rutinoside	993.5297	993.5270	2.72	C47H80O19				+
46	11.21	Quinquenoside L9 or its isomer	863.4995	863.5004	−1.04	C42H74O15			+	+
47	11.26	Notoginsenoside SP8 or its isomer	669.4230	669.4219	1.64	C36H62O11		+	+	+
48	11.37	[PPT 9]-3-Rutinosyl-20-glucoside or its isomer	1005.5278	1005.5270	0.80	C48H80O19			+	+
49	11.47	Quinquenoside L9 or its isomer	863.5003	863.5004	−0.12	C42H74O15			+	+
50	11.72	[PPT 8]-6,20-Di-O-glucoside or its isomer	859.4673	859.4691	−2.09	C42H70O15			+	+
51	11.75	Floaginsenoside B or its isomer	861.4808	861.4848	−4.64	C42H72O15			+	+
52^d	11.46	[PPD 10]-O-Glucosyl-O-xyloside	817.4975	817.4949	3.18	C41H72O13			+	+
53	11.78	Notoginsenoside N or its isomer	1007.5402	1007.5427	−2.48	C48H82O19	+	+		+
54	11.85	Ginsenoside B2 or chikusetsusaponin FK1	945.5423	945.5428	−0.53	C48H82O18			+	+
55	11.86	[PPT 21]-20-Xylosyl-3-glucoside or its isomer	803.4769	803.4798	−3.61	C41H72O15			+	+
56^d	11.89	[PPT 8]/[PPT 9]-O-Rutinosyl-O-glucoside	1005.5302	1005.5270	3.18	C48H80O19			+	+
57	11.91	Notoginsenoside A or its isomer	1123.5916	1123.5906	0.89	C54H92O24				+
58^d	11.93	[PPD 12]/[PPT 16]-O-Glucosyl-O-xyloside	833.4912	833.4899	1.56	C42H74O16				+
59	11.96	Quinquenoside L9 or its isomer	863.5000	863.5004	−0.46	C42H74O15			+	+
60	11.97	[PPD 8]-3-Glucosyl-glucosyl-20-glucosyl-arabinoside/[PPD 9]-3-glucosyl-glucosyl-20-glucosyl-xyloside	1093.5794	1093.5800	−0.55	C53H90O23		+	+	+
61	12.09	Yesanchinoside E or its isomer	1107.5947	1107.5957	−0.90	C54H92O23		+		+
62^d	12.14	[PPT 5]-Di-O-glucosyl-O-xyloside	975.5163	975.5165	−0.21	C47H77O18			+	+
63	12.15	Quinquenoside L16 or its isomer	1141.6044	1141.6011	2.89	C54H94O25			+	+
64	12.16	[PPT 6]-6-Glucosyl-xylosyl-20-glucosyl-glucoside/[PPT 6]-3-glucosyl-glucosyl-20-glucosyl-arabinoside (xyloside)	1093.5814	1093.5800	1.28	C53H90O23			+	+
65	12.19	[PPT 4]-6,20-Di-O-glucoside/[PPT 5]-3,20-di-O-glucoside	843.4725	843.4742	−2.02	C42H70O14			+	+
66^a	12.21	Notoginsenoside R1	931.5277	931.5272	0.54	C47H80O18	+	+	+	+
67^a	12.23	Ginsenoside Re	945.5420	945.5428	−0.85	C48H82O18	+	+	+	+
68	12.24	Vinaginsenoside R11 or floraginsenoside D	831.4705	831.4742	−4.45	C41H70O14			+	+
69^a	12.35	Ginsenoside Rg1	845.4898	845.4899	−0.11	C42H72O14				+
70	12.40	Notoginsenoside R2 or its isomer	815.4810	815.4793	2.08	C41H70O13	+	+	+	+
71	12.43	Notoginsenoside N or its isomer	1007.5427	1007.5427	−0.03	C48H82O19			+	+
72	12.45	Notoginsenoside G or its isomer	959.5213	959.5221	−0.83	C48H80O19				+
73	12.50	Vinaginsenoside R11 or floraginsenoside D	785.4693	785.4693	0.00	C41H70O14			+	+
74	12.51	Notoginsenoside R2 or its isomer	815.4773	815.4793	−2.45	C41H70O13			+	+
75	12.53	[PPD 8]-Tri-O-glucoside/[PPD 9]-3-glucosyl-glucosyl-20-glucoside	961.5359	961.5372	−1.35	C47H80O17	+			+
76	12.55	[PPT 6]-6-Acetylglucosyl-20-glucoside/[PPT 6]-6-glucosyl-20-acetylglucoside/[PPT 6]-20-acetyl-6-glucosyl-glucoside	887.4997	887.5004	−0.79	C44H74O15		+	+	+
77	12.56	Gynosaponin V or its isomer	1091.6028	1091.6007	1.92	C54H92O22			+	+
78	12.58	[PPT 12]-6(12,20)-O-Glucoside	699.4329	699.4320	1.29	C36H62O10			+	+
79	12.59	Isomer of notoginsenoside R1	977.5308	977.5321	−1.33	C47H80O18			+	+
80	12.61	[PPD 7]-3-Glucosyl-glucosyl-20-glucosyl-glucosyl-arabinoside (xyloside)/[PPD 7]-3-glucosyl-glucosyl-glucosyl-20-glucosyl-xyloside	1239.6369	1239.6374	−0.40	C59H100O27			+	+
81	12.63	[PPD 7]-3-Glucosyl-glucosyl-20-glucosyl-glucosyl-arabinoside (xyloside)/[PPD 7]-3-glucosyl-glucosyl-glucosyl-20-glucosyl-xyloside	1239.6396	1239.6379	1.37	C59H100O27			+	+
82	12.65	[PPT 4]-6,20-Di-O-glucoside or its isomer	843.4703	843.4742	−4.62	C42H70O14			+	+
83	12.67	[PPD 3]-3-Glucosyl-glucosyl-20-glucosyl-glucoside or its isomer	1105.5802	1105.5800	0.18	C54H90O23		+	+	+
84	12.68	[PPT 6]-6-Acetylglucosyl-20-glucoside/[PPT 6]-6-glucosyl-20-acetylglucoside/[PPT 6]-20-acetyl-6-glucosyl-glucoside	887.5021	887.5004	1.92	C44H74O15		+	+	+
85	12.69	[PPT 6]-3(6,20)-Glucosyl-rhamnoside/[PPT 6]-6-rhamnosyl (glucosyl)-20-glucoside(rhamnoside)	783.4897	783.4900	−0.38	C42H72O13			+	+
86^a	12.76	Ginsenoside F3	815.4772	815.4793	−2.58	C41H70O13	+	+	+	+
87	12.76	[PPT 6]-6(20)-Glucosyl-glucoside/[PPT 6]-3,6-di-O-glucoside	799.4844	799.4849	−0.63	C42H72O14	+	+	+	+
88^a	12.79	Ginsenoside Rb1	1107.5959	1107.5957	0.18	C54H92O23	+	+		+
89^a	12.82	20(S)-Ginsenoside Rg2	829.4928	829.4949	−2.53	C42H72O13	+	+		+
90^a	12.85	Ginsenoside Rc	1077.5804	1077.5851	−4.36	C53H90O22			+	+
91	12.86	Notoginsenoside R2 or its isomer	769.4733	769.4744	−1.43	C41H70O13		+	+	+
92	12.93	[PPT 6]-3(6,20)-Glucosyl-rhamnoside/[PPT 6]-6-rhamnosyl (glucosyl)-20-glucoside(rhamnoside)	783.4910	783.4900	1.28	C42H72O13		+	+	+
93	12.85	Vinaginsenoside R3 or its isomer	929.5481	929.5479	0.22	C48H82O17			+	+
94^a	13.01	Ginsenoside Rh1	683.4348	683.4370	−3.22	C36H62O9		+	+	+
95^a	13.01	Ginsenoside Rd	991.5484	991.5478	0.61	C48H82O18	+	+
96	13.07	Isomer of ginsenoside Rd	945.5434	945.5428	0.63	C48H82O18		+	+	+
97	13.11	[PPT 6]-6,20-Di-O-xyloside or its isomer	739.4637	739.4638	−0.14	C40H68O12				+
98^a	13.15	Ginsenoside F1	683.4352	683.4370	−2.63	C36H62O9	+	+	+	+
99^d	13.21	[PPD 7]-O-Xylosyl-di-O-glucoside	961.5365	961.5372	−0.73	C47H80O17	+	+	+	+
100^d	13.22	[PPD 2]-O-Xylosyl-O-glucoside	795.4539	795.4536	0.38	C41H66O12		+	+	+
101	13.23	Isomer of ginsenoside Rg1	845.4913	845.4899	1.66	C42H72O14			+	+
102	13.24	Ginsenoside Rg9 or its isomer	827.4773	827.4793	−2.42	C42H70O13		+	+	+
103	13.30	[PPD 7]-3-Glucosyl-20-rutinoside/[PPD 7]-3-glucosyl-glucosyl-20-rhamnoside	975.5553	975.5529	2.46	C48H82O17			+	+
104^b^,^d	13.31	[PPD 3]/[PPD 4]/[PPT 2]/[PPT 3]-Decadianoyl-nonenoyl-acetyl-di-O-glucoside	1111.6967	1111.6939	2.5	C63H100O16			+
105^d	13.33	[PPD 4]-O-Xylosyl-O-glucoside	797.4685	797.4693	−1.00	C41H68O12	+	+	+	+
106	13.35	Chikusetsusaponin LT8 or its isomer	763.4633	763.4638	−0.65	C42H68O12			+	+
107	13.38	Ginsenoside Rg9 or its isomer	827.4763	827.4793	−3.63	C42H70O13		+		+
108	13.39	Isomer of ginsenoside Rd	991.5460	991.5478	−1.82	C48H82O18	+	+		+
109^a	13.47	Ginsenoside Rg6	765.4800	765.4795	−0.65	C42H70O12			+	+
110^a	13.50	20(S)-Ginsenoside Rg3	829.4932	829.4949	−2.05	C42H72O13		+	+	+
111^d	13.51	[PPD 7]-O-Xylosyl-di-O-glucoside	961.5334	961.5372	−3.95	C47H80O17		+	+	+
112	13.61	Ginsenoside F2 or its isomer	829.4939	829.4949	−1.21	C42H72O13	+	+		+

---

3.5 Evaluation of the efficiency and the accuracy of the MRM-based strategy by LC-IT-TOF/MSⁿ analysis

LC-IT-TOF/MS has been widely used as a powerful technique for the chemical profiling of complex matrices. Traditionally, the multistage mass data for structure characterization are acquired by triggering the most abundant precursor ion, indicating a great risk for information missing during the data acquisition process, especially for trace ions. As expected, only 27 peaks from NGTS were detected (Fig. 5A and Table 1) by the full scan on the LC-IT-TOF/MS technique with a higher sample-loading amount (32 μg) than the MRM analysis (8 μg). Since the exclusions of the known compound ions could enhance the sensitivity of the trace compounds to some extent,⁹ the HR-MS/MS data of another 23 peaks were further collected (Fig. 5B and Table 1) by the full scan along with an excluded ions list (Table S3†). However, there are still 77 peaks that remain undetected in the LC-IT-TOF/MSⁿ analysis, in comparison with the above results obtained by the MRM-based strategy. It has been reported that the sensitivity and selectivity of the full scan on HR-MS/MS system increased when using a target-list-dependent component detection with the full scan.^20,21 Thus, the predicted deprotonated molecular ions of the saponins detected by the MRM-based method were input into the preferred ions list (Table S4†) of the full scan to establish a target-list-dependent components detection on the LC-IT-TOF/MS system. Therefore, the HR-MS/MS data of the 109 peaks were finally obtained by integrated usage of the full scan, the excluded ions list (Table S3†), and the preferred ions list (Table S4†) on the LC-IT-TOF/MS system. In particular, three components (peaks 8, 10, and 104, Table 1) that did not generate the formate anions were additionally detected. Peaks 8 and 10 shared the same deprotonated molecular ions at m/z 861.48, along with two identical fragment ions at m/z 681.41 and 623.38. The mass difference of m/z 861.48 and 681.41 is 180.07 Da, corresponding to the losses of a glucosyl (162.0528 Da) and an H₂O (18.0106 Da) unit. Finally, they were deduced as [PPD 11] or [PPT 12] or [PPT 13]-di-O-glucoside with an additional cleavage of 58.04 Da (C₃H₆O). The molecular composition of peak 104 was C₆₃H₁₀₀O₁₆, calculated from its pseudo-molecular ion at m/z 1111.6967. The mass difference between the m/z 1111.6967 and 781.4710 is 330.2257 Da, indicating the existence of a nonanoyl (non, 138.1045 Da), a decanoyl (deca, 150.1045 Da), and an acetyl (42.0106 Da) unit (Fig. 2). Finally, it was plausibly assigned as [PPD 3]/[PPD 4]/[PPT 2]/[PPT 3]-deca-non-acetyl-di-O-glucoside due to the observation of another neutral loss of 162.05 Da (glucosyl). Although the MRM scan on the Q-Trap/MS technique missed three components, it is still more efficient than the full scan on the IT-TOF/MS system in consideration of the detected ginsenosides numbers and the lower sample-loading quantity.

Fig. 5 The extracted ion current chromatograms (EICs) of the compounds detected by the full scan (A), the full scan along with an excluded ions list (B), and the full scan together with a preferred ions list (C) on the IT-TOF/MS system.

Furthermore, the accuracy of the ginsenosides detected and identified by the MRM-based strategy was investigated by the obtained HR-MS/MS data. For example, the deprotonated molecular ion of peak 88 was predicted to be m/z 1107.5951 because it had been assigned as [PPD 7]-tetra-O-glucoside based on the MRM-based analysis, which is highly consistent with m/z 1107.5959, measured by IT-TOF-MSⁿ analysis. The other fragment ions at m/z 945.5355, 783.4844, 621.4310, and 459.3793 (Fig. 4B3) indicated the sequential neutral losses of 162.05 Da, confirming the existence of four glucosyl units. In total, the multistage high resolution mass data validated that peak 88 had been correctly characterized by the MRM-based analysis. Similarly, the accuracies of the other assignments (Table 1) were further validated. Consequently, only two components (peaks 13 and 17, Table 1) were falsely characterized because the error values was greater than 5 ppm. Based on the MRM analysis, peaks 13 and 17 were assigned as [PPD 10]-di-O-glucoside by a combination of the sapogenin ion at m/z 477.2, along with the successive neutral losses of 162 Da (Table S5†). However, their molecular composition was calculated as C₄₈H₆₆O₁₀ (M.W., 802.4656) according to the HR-MS/MS analysis, which differed from C₄₁H₇₀O₁₅ (M.W., 802.4715, [PPD 10]-di-O-glucoside) characterized by the MRM analysis. In addition, the mass difference between the deprotonated molecular ion at m/z 801.4487 and the fragment ion at m/z 639.4094 (Table S5†) is 162.0393 Da, agreeing more with the caffeoyl moiety (C₉H₆O₃, 162.0317 Da) than the glucosyl residue (C₆H₁₀O₅, 162.0528 Da). Thus, peaks 13 and 17 were revised as the dicaffeoylated [PPD 10].

At last, by combination of the high sensitivity and the high-resolution analyses, a total of 112 ginsenosides, including 107 minor ginsenosides (total amount < 25%) and 17 potential new ginsenosides were detected and characterized (Tables 1 and S5†). Although the MRM-based strategy falsely characterized two peaks and missed three components, its 98% accuracy and four-fold relevance ratio (MRM scan: 109 peaks; full scan of IT-TOF/MS: 27 peaks) suggested that it could be used as a reliable tool to gain in-depth insights for the ginsenoside-enriched herbal products or other analogue-focused matrices. The retention times, fragment ions, and identities of those compounds are presented in Table S5† and the detailed descriptions of the other components are presented in the electronic ESI.†

3.6 The advantages of the MRM-based strategy

As many trace ginsenosides have been shown to be remarkably effective in treating cardiovascular disease and some other diseases, many strategies have been proposed to screen those trace molecules from P. notoginseng by depleting high-abundance compounds based on the fraction collection^9,19 or two-dimensional LC/MS techniques.^10,22 However, most of these methods are complex, tedious, and time-consuming. Thus, a simpler and more efficient method, i.e., the MRM-based strategy was developed here to comprehensively profile NGTS. With the benefits of the predefined ion transitions and the online stepped optimization strategy, the MRM-based strategy was more sensitive and specific than the aforementioned strategies.^9,10,19,22 In addition, the developed MRM-based strategy increases the detection capability by utilizing the high resolution of UPLC and implementing the four separated scan runs to collect the mass information of NGTS. Moreover, the high specificity of the MRM scan makes the peak retrieval process more concise and rapid than the “five-point screening” approach,²² which demands a reasonable distributed space constructed by a mathematical method.²² Moreover, the remarkable utilization of the MS² spectra of the isotopic ions indicated a new method to determine the M.W.s of unknown peaks. Overall, compared to the pretreatment or two-dimensional approaches, the MRM-based strategy increases the detection sensitivity and specificity and, at the same time, avoids the detector saturation and facilitates the data screen and characterization procedures.

4. Conclusion

Traditional Chinese medicines (TCMs) have been drawing increasing worldwide interest due to the changes of human disease spectrum, particularly, the prevalence of chronic and systematic diseases.^23,24 More and more TCM preparations containing certain categories of phytochemicals have been developed for clinical applications, such as salvia total phenolic acids, centella total glycosides, and so on. However, only a few important components have been disclosed in these homolog-focused extracts. Complex compositions make the compounds difficult to purify and characterize, especially those present at trace levels. Since the low-abundance ions are often undetectable when co-eluted or enveloped by the more intensive ions during the full scan on the HR-MS/MS system,^9,10,19,22 a MRM-based strategy using [M + HCOO]⁻ > [M − H]⁻ transitions was proposed in this paper to rapidly detect and identify those trace phytochemicals, taking NGTS as an example. The corresponding CE values obtained by the online optimization strategy reveal that the optimal CE values are related to the number of glycosyl groups ginsenosides contain. It is worthy noting that the isotopic peaks could reach the threshold (500 cps) and trigger the acquisition of the corresponding MS² spectra. Thus, these isotopic signals could be applied as an alternative method to judge the real M.W.s of unknown components. Subsequently, the identification of those detected ginsenosides was performed on the basis of well-defined mass cracking patterns. Based on the pre-identified phytochemicals, an IT-TOF/MS system was further utilized to test the practicability and efficiency of the MRM-based method for profiling the minor components. Finally, a total of 112 components, including 107 trace components, were detected and characterized in this study. Among the 112 ginsenosides, a total of 17 peaks were found to be potential new ginsenosides compared with literature. The 98% accuracy and four-fold relevance ratio comparable to IT-TOF/MS suggested that the MRM-based profiling method could be used as a reliable tool to gain in-depth insights for ginsenoside-enriched herbal products and other homolog-focused extracts.

Conflict of interest statement

The authors have declared no conflict of interest.

Acknowledgements

This work was financially supported by the National Natural Sciences Foundation of China (Nos. 81573684 and 81222051) and the National Key Technology R & D Program “New Drug Innovation” of China (Nos. 2012ZX09103201-036).

References

1. T. Kind and O. Fiehn, BMC Bioinf., 2007, 8, 1–20.
2. T. Kind and O. Fiehn, Bioanal. Rev., 2010, 2, 23–60.
3. Z. Yan, G. Lin, Y. Ye, Y. Wang and Y. Z. Yan, Anal. Chim. Acta, 2014, 819, 56–64.
4. B. R. Jeon, S. J. Kim, S. B. Hong, H. J. Park, J. Y. Cho and M. H. Rhee, J. Ginseng Res., 2015, 39, 279–285.
5. Z. Guo, G. Cao, H. Yang, H. Zhou, L. Li, Z. Cao, B. Yu and J. Kou, J. Neurosci. Res., 2014, 92, 1295–1306.
6. C. L. Xie, J. H. Li, W. W. Wang, G. Q. Zheng and L. X. Wang, Life Sci., 2015, 121, 145–151.
7. L. C. Wang, Q. Zhang, J. Liu, N. Li, K. Raphael, B. Liu, P. Liu, L. Qi and P. Li, J. Funct. Foods, 2015, 16, 20–27.
8. Y. Fan, Y. Qiao, J. M. Huang and M. K. Tang, J. Evidence-Based Complementary Altern. Med., 2016, 2016, 1–9.
9. Chinese Pharmacopoeia Committee, Pharmacopoeia of the Peoples's Republic of China, Part 1, Medical Science and Technology Press, Beijing, China, 2015, pp. 393–394.
10. Y. Y. Liu, J. B. Li, J. M. He, Z. Abliz, J. Qu, S. S. Yu, S. G. Ma, J. Liu and D. Du, Rapid Commun. Mass Spectrom., 2009, 23, 667–679.
11. Q. Q. Xiang, T. Liang, G. B. Shen, X. L. Wang, Y. Jin and X. M. Liang, Analyst, 2012, 137, 2239–2249.
12. Y. Cai and R. B. Cole, Anal. Chem., 2002, 74, 985–991.
13. Y. L. Song, N. Zhang, S. Shi, J. Li, Y. Zhao, Q. Zhang, Y. Jiang and P. F. Tu, J. Chromatogr. A, 2015, 1406, 136–144.
14. Y. L. Song, Q. Q. Song, J. Li, J. Zheng, C. Li, Y. Zhang, L. Zhang and P. F. Tu, J. Chromatogr. A, 2016, 1454, 58–66.
15. W. Z. Yang, Y. Hu, W. Y. Wu, M. Ye and D. A. Guo, Phytochemistry, 2014, 106, 7–24.
16. M. Peng, T. Zhang, Y. Ding, Y. X. Yi, Y. J. Yang and J. Le, RSC Adv., 2016, 6, 36987–37005.
17. Y. L. Song, Q. Q. Song, J. Li, J. Zheng, C. Li, Y. Zhang, Y. Jiang and P. F. Tu, J. Chromatogr. A, 2016, 1460, 74–83.
18. J. F. Chen, Z. Y. Shi, Y. L. Song, X. Y. Guo, M. B. Zhao, P. F. Tu and Y. Jiang, J. Chromatogr. A, 2016, 1464, 102–114.
19. W. Z. Yang, M. Ye, X. Qiao, C. F. Liu, W. J. Miao, T. Bo, H. Y. Tao and D. A. Guo, Anal. Chim. Acta, 2012, 739, 56–66.
20. H. Zhang, L. Ma, K. He and M. Zhu, J. Mass Spectrom., 2008, 43, 1191–1200.
21. J. Castro-Perez, R. Plumb, L. Liang and E. Yang, Rapid Commun. Mass Spectrom., 2005, 19, 798–804.
22. C. J. S. Lai, T. Tan, S. L. Zeng, L. W. Qi, X. G. Liu, X. Dong, P. Li and E. H. Liu, J. Pharm. Biomed. Anal., 2015, 109, 184–191.
23. A. Huntley and E. Ernst, Thorax, 2000, 55, 925–929.
24. Q. Ruan, Q. C. Ji, M. E. Arnold, W. G. Humphreys and M. Zhu, Anal. Chem., 2011, 83, 8937–8944.

---

Footnote

† Electronic supplementary information (ESI) available. See DOI: 10.1039/c6ra18459f

---