Potential of hyphenated ultra-high performance liquid chromatography-scheduled multiple reaction monitoring algorithm for large-scale quantitative analysis of traditional Chinese medicines

Abstract

It is a great challenge to perform quality control for traditional Chinese medicines (TCMs) that contain a great number of constituents by holistically monitoring hydrophilic and hydrophobic substances. Theoretically, the relatively low scan rate of triple quadrupole (QqQ) equipment makes it quite difficult to meet the demands of reliable quantitation of the narrow peaks generated from ultra-high performance liquid chromatography (UHPLC). Scheduled multiple reaction monitoring (sMRM) algorithm offers the potential to simultaneously monitor numerous analytes without compromising data quality, in particular for co-eluting compounds, by automatically altering the dwell time to maintain the desired cycle time on a QqQ analytical platform. In the current study, UHPLC and sMRM were hyphenated to develop a practical and robust quantitative method for as many as 133 TCM-derived components, including polar and apolar compounds. Efficient separation was achieved on a core–shell-type column (Capcell core ADME column) with adamantylethyl functional groups to generate appropriate surface polarity along with hydrophobicity in comparison with RP-C₁₈ and HILIC columns. To verify the applicability of the developed UHPLC-sMRM method, a formula was simulated by mixing eight TCM raw materials that related to those 133 analytes. Moreover, enhanced product ion scans were triggered by sMRM to acquire MS² spectra to enhance the confidence of peak assignment. Method validation results suggested the developed method to be accurate, precise, and reproducible. In comparison with conventional MRM, sMRM was proved to be advantageous in terms of sensitivity and precision, as well as the dependent MS² spectral quality. Above all, our current study indicated that the integration of UHPLC and sMRM provides the potential to globally and simultaneously quantify the components in TCMs.

---

1. Introduction

Traditional Chinese medicines (TCMs) have been utilized for the prevention and treatment of human diseases in China for centuries, as well as in some East Asian countries.¹ Because of the long historical clinical practices and convincing therapeutic outcomes, TCMs are stimulating scientists’ interests all over the world and an increasing number of famous pharmaceutical companies are employing TCMs as an ideal library for the discovery of lead compounds. However, the features of TCMs efficacy include systematism, multi-target and multi-channel, attributing to their complicated chemical compositions.^2–6 If only a few ingredients are emphasized, the holistic and synergistic ability of TCMs will be neglected, and this thus calls for a comprehensive analytical approach which could reflect the quantitative characteristics of most constituents in TCMs, especially those variations relating to the pharmacological and health benefits, as well as the toxic potential.⁷ Currently, the technical obstacles to draw a complete picture of the quality of TCMs mainly lie in the characterization of the hydrophilic constituents and detection of trace substances. Hydrophilic constituents have been revealed to make primary contributions to some famous crude drugs, e.g., nucleosides and nucleobases for Cordyceps,^8,9 and also, some amino acids are employed as the quality indicators for some raw materials, such as Pheretima and Cervi Cornu Pantotrichum. Moreover, a number of minor and trace constituents sourced from TCMs have been demonstrated to have attractive biological activities, e.g. triterpenoid–diterpenoid heterodimers from Pseudolarix amabilis,¹⁰ and a dimeric sesquiterpene lactone from Inula japonica.¹¹ Therefore, there is an urgent need to develop an analytical method featuring high sensitivity and separation efficiency for globally quantitative analysis of the constituents of TCMs.

The hyphenated liquid chromatography-mass spectrometry (LC-MS) based analytical platform is currently the workhorse of quality control of TCMs. In comparison with time-of-flight (TOF) MS, the multiple reaction monitoring (MRM) mode on triple quadrupole (QqQ) MS equipment exhibits superiority in the linear dynamic range that spans five to six orders of magnitude; however, QqQ-MS is disadvantageous in terms of scan rate (0.5–4 Hz for QqQ versus 20 Hz for TOF).^12–14 Owing to the adoption of sub-2.0 μm particles, the peak width generated by ultra-high performance liquid chromatography (UHPLC) is usually much narrower than that obtained via conventional LC separations, generally in the region of 2–10 s width at the base, thus providing much greater peak capacity. Recently, core–shell-type particles have been introduced to column packing, and they could make analytes spend less time diffusing into and out of the pores of those particles. Hence, core–shell-type columns with approximately 2.5 μm particles could provide comparable peak capacity and width to a column embedded with sub-2.0 μm particles, nonetheless offering lower back-pressure.¹⁵ Therefore, the hyphenation of MRM with UHPLC equipped with a core–shell-type column is expected to be a promising tool for simultaneous determination of dozens of components in TCMs. However, when more than one hundred constituents are desired to be analyzed, acquiring sufficient data points for each narrow peak will be beyond the potency of QqQ equipment due to its slow scan rate. In general, more than ten data points are required for each peak to achieve precise determination.¹⁶ It is feasible to synchronize the UHPLC and QqQ domain by splitting all precursor-to-product ion transitions into several separate runs and/or replacing UHPLC with HPLC to broaden the peaks; however, those two solutions are extremely contrary to the achievement of time- and labor-saving targets. Fortunately, scheduled MRM (sMRM, also known as dynamic MRM) algorithm has shown the potential to simultaneously monitor hundreds of metabolites by monitoring every MRM ion pair in its expected retention time window, consequently decreasing the number of concurrent ion transitions.^17–19 With the application of the sMRM algorithm, both the cycle time and the dwell time are automatically adjusted to be appropriate, leading to the increment of data points for each chromatographic peak.^20–24 In addition, one of the most important advantages of hybrid QqQ-linear ion trap (Q-trap) equipment is that it enables simultaneous quantitative and qualitative analyses without compromising data quality by triggering enhanced product ion (EPI) scans through certain survey experiments, such as MRM and enhanced MS scans.

In order to remove the technical barriers for large-scale quantitative analysis of TCMs, we therefore integrated the merits of UHPLC and Q-trap equipments by integrating a core–shell-type column, sMRM algorithm, and EPI experiment. As many as 133 TCM-derived compounds, including both hydrophilic and hydrophobic ones, were collected to develop and validate an accurate, sensitive, and precise UHPLC-sMRM method, and a simulated TCM formula consisting of eight common raw materials, including Ginseng Radix, Aconiti Lateralis Radix Praeparata, Solani Melongenae Radix, Pheretima, Galli Gigerii Endothelium Corneum, Cistanches Herba, Polygalae Radix, and Draconis Resina, was utilized to confirm the applicability of the developed method. The findings obtained in the current study are expected to propose a robust and flexible solution for the holistic quality control of TCMs.

2. Experimental

2.1 Chemicals and reagents

Seventeen amino acids, namely L-alanine, L-serine, L-valine, L-threonine, L-leucine, L-isoleucine, asparagine, aspartic acid, L-phenylalanine, L-proline, L-tyrosine, L-lysine, glutamine, glutamic acid, γ-aminobutyric acid, L-histidine, L-arginine, and nine nucleosides and nucleobases, namely adenine, uracil, thymine, cytidine, guanosine, uridine, adenosine, thymidine, and inosine, were purchased from Xinjingke Biotechnology Company (Beijing, China). Sixteen ginsenosides, namely ginsenosides Rb1, Rb2, Rh1, Rh2, Rc, Rd, Re, Ro, Rf, Rg1, Rg2, Rg3, F1, F2, pseudo-ginsenoside F11, and compound K, as well as seven diterpenoid alkaloids, namely songorine, neoline, talatisamine, benzoylmesaconine, benzoylaconine, benzoylhypaconine, and hypaconitine, were obtained from Shanghai Standard Biotech Co. Ltd (Shanghai, China). Several organic acids, namely citric acid, fumaric acid, malic acid, tartaric acid, shikimic acid, malonic acid, succinic acid, quinic acid, lactic acid, adipic acid, maleic acid, ascorbic acid, nicotinic acid, and salicylic acid, were provided by Sigma-Aldrich (St Louis, MO, USA). Cinnamic acid was acquired from Sinopharm Chemical Reagent Co. Ltd (Beijing, China). Maltose and rhamnose were acquired from Shanghai Yuanye Biotech Co. Ltd (Shanghai, China). Galactitol, 3,4-dimethoxyphenylethanol, betaine, gallic acid, vanillic acid, nicotinamide, 8-epi-loganic acid, 3,4-dihydroxyphenylethanol, salidroside, 6-deoxycatalpol, gluroside, cistanoside E, sibiricose A₅, sibiricose A₆, mangiferin, geniposide, ferulic acid, alaschanioside A, lancerin, echinacoside, polygalaxanthone VIII, 7-O-methoxyl-mangiferin, polygalaxanthone IX, lariciresinol-4′-O-β-D-glucopyranoside, N-trans-p-coumaroyloctopamine, tenuifoliside B, verbascoside, poliumoside, N-trans-feruloyloctopamine, isoverbascoside, 4-methoxyphenylethanol, pinoresinol-β-D-glucopyranoside, polygalaxanthone VII, cistanoside C, 3,6′-disinapoyl sucrose, 2′-aceylpoliumoside, isocistanoside C, tenuifoliside A, 3,4,5-trimethoxycinnamic acid, tubuloside B, cistanoside D, p-methoxycinnamic acid, N-trans-p-coumaroyltyramine, polygalaxanthone IV, 3-(4-hydroxyphenyl)-N-[2-(4-hydroxyphenyl)-2-methoxyethyl]-acrylamide, loureiriol, N-trans-feruloyltyramine, liquiritigenin, 3-(4-hydroxy-3-methoxyphenyl)-N-[2-(4-hydroxyphenyl)-2-methoxyethyl]-acrylamide, N-trans-feruloyl-3-methoxytyramine, polygalasaponin XXVIII, 5,7,4′-trihydroxyflavanone, cannabisin D, tenuifolin, 6-hydroxy-1,2,3,7-tetramethoxyxanthone, melongenamide B, 3,4′-dihydroxy-5-methoxystilbene, 5,7-dihydroxy-4′-methoxy-8-methylflavane, 2,4′-dihydroxy-4,6-dimethoxydihydro-chalcone, 1,2,3,6,7-pentamethoxyxanthone, 1,7-dimethoxyxanthone, N-trans-feruloyltyramine dimer, cannabisin F, melongenamide D, 4-hydroxy-2,4′-dimethoxydihydrochalcone, 1,2,3,7-tetramethoxyxanthone, pterostilbene, and 4′-hydroxy-5,7-dimethoxy-8-methylflavane were provided by the chemical library of State Key Laboratory of Natural and Biomimetic Drugs, Peking University (Beijing, China). The purity of each reference compound was determined to be more than 95% by normalization of the peak areas detected by UHPLC-DAD-IT-TOF-MS (Shimadzu, Kyoto, Japan). All of the references are also summarized in Table 1.

Table 1 Retention times (t_R), MS¹ and MS² spectral information, compound-dependent mass parameters, limits of detection (LODs) and lower limits of quantification (LLOQs) for the 133 analytes

No.	Compound	tR (min)	MS^1 (m/z)	MS^2^a (m/z)	DP (V)	CE (eV)	LOD (pg mL^−1)	LLOQ (pg mL^−1)
a Product ions in bold are selected for quantitative analysis.
1	Citric acid	0.74	191	129;111;87;85	−30	−13	128	8.00 × 10^3
2	Fumaric acid	0.75	115	71	−35	−15	1.60 × 10^4	2.00 × 10^5
3	D-Malic acid	0.75	133	115;89;71;43	−40	−20	5.12	8.00 × 10^3
4	D-Tartaric acid	0.75	149	103;87;73	−20	−16	128	8.00 × 10^3
5	(−)-Shikimic acid	0.75	173	155;137;129;111;93;73	−70	−15	4.00 × 10^4	2.00 × 10^5
6	Glutamic acid	0.76	148	84	25	23	2.56	12.8
7	Aspartic acid	0.77	134	74	25	21	8.00 × 10^3	4.00 × 10^4
8	L-Proline	0.78	116	70	50	20	1.02	12.8
9	Glutamine	0.79	147	130;84	25	25	1.02	25.60
10	Malonic acid	0.79	103	59	−40	−15	8.00 × 10^3	1.60 × 10^4
11	Succinic acid	0.79	117	99;73	−35	−12	8.00 × 10^3	1.60 × 10^4
12	Quinic acid	0.79	191	173;127;85	−100	−23	3.20 × 10^3	1.60 × 10^4
13	L-Serine	0.80	106	60	40	16	1.02	2.56
14	Asparagine	0.82	133	74	30	23	128	640
15	L-(+)-Lactic acid	0.82	89	43	−40	−14	128	2.00 × 10^5
16	L-Threonine	0.84	120	102	30	10	12.80	1.60 × 10^3
17	L-Alanine	0.85	90	44	25	17	8.00 × 10^3	1.60 × 10^4
18	γ-Aminobutyric acid	0.85	104	87	40	16	1.02	2.56
19	Galactitol	0.86	181	163;113;101;85;71	−100	−16	6.40 × 10^3	8.00 × 10^4
20	3,4-Dimethoxyphenylethanol	0.86	181	89	−40	−16	1.16 × 10^4	7.28 × 10^5
21	Betaine	0.89	118	58	40	41	0.51	5.12
22	L-Arginine	0.90	175	157;130;116;70	25	32	667	1.67 × 10^4
23	Adipic acid	0.90	145	127;101;83	−35	−21	1.60 × 10^3	1.60 × 10^4
24	Gallic acid	0.90	169	151;125;97;81	−60	−21	1.28 × 10^3	3.20 × 10^3
25	L-Histidine	0.91	156	128;110	25	21	1.02	5.12
26	Maleic acid	0.96	115	71	−35	−15	64	128
27	Maltose	0.96	341	179;143;113;89;71	−80	−30	51.20	640
28	Rhamnose	0.97	163	73	−35	−20	2.00 × 10^6	4.00 × 10^6
29	L-Valine	0.98	118	72	25	18	1.02	12.80
30	L-Ascorbic acid	1.25	175	115;87;71;59	−40	−14	64	128
31	Uracil	1.54	113	96	40	27	4.00 × 10^4	1.00 × 10^5
32	L-Isoleucine	1.63	132	115;86	50	18	3.20 × 10^3	1.60 × 10^4
33	L-Tyrosine	1.63	182	165;147;136;123	25	19	1.60 × 10^3	8.00 × 10^3
34	Nicotinic acid	1.76	122	94;78	−50	−20	8.00 × 10^3	1.60 × 10^4
35	L-Leucine	1.81	132	114;86	50	18	3.20 × 10^3	1.60 × 10^4
36	Cytidine	2.11	244	128;112	25	17	1.28 × 10^3	3.20 × 10^3
37	Uridine	2.65	245	227;113;107	40	23	3.20 × 10^3	1.60 × 10^4
38	Vanillic acid	3.22	167	152;123;108	−50	−16	1.60 × 10^4	8.00 × 10^4
39	Thymine	3.92	127	110	40	23	1.28 × 10^3	6.40 × 10^3
40	Inosine	4.88	267	135;92	−80	−30	3.20 × 10^3	6.40 × 10^3
41	L-Phenylalanine	5.00	166	120;103	50	19	3.20 × 10^3	8.00 × 10^3
42	Guanosine	5.51	284	152;135;110	40	25	1.28 × 10^3	6.40 × 10^3
43	Nicotinamide	6.16	123	107;80	30	30	5.12	1.28 × 10^3
44	Adenine	6.51	134	107;92;65	−70	−18	6.40 × 10^2	3.20 × 10^3
45	Salicylic acid	6.77	137	93;65	−50	−21	1.60 × 10^3	8.00 × 10^3
46	8-epi-Loganic acid	7.62	375	213;169;151	−130	−22	4.00 × 10^5	4.00 × 10^6
47	Thymidine	8.20	243	225;131;127	30	16	6.40 × 10^3	1.60 × 10^4
48	3,4-Dihydroxyphenylethanol	8.29	153	123;105;93;77	−40	−20	4.93 × 10^3	4.93 × 10^4
49	Adenosine	9.29	268	136;119	40	30	1.28 × 10^3	6.40 × 10^3
50	Salidroside	11.40	299	119;89	−130	−20	3.20 × 10^3	1.60 × 10^4
51	6-Deoxycatalpol	11.47	345	299;165;101	−50	−12	1.60 × 10^4	3.20 × 10^4
52	Gluroside	12.31	331	161;125;107	−30	−15	3.20 × 10^4	1.60 × 10^5
53	Cistanoside E	12.50	475	329;161;134	−30	−53	2.00 × 10^6	4.00 × 10^6
54	Sibiricose A5	12.94	517	175;160	−190	−32	1.02	25.60
55	Sibiricose A6	13.27	547	529;205;190	−200	−31	1.02	12.80
56	Songorine	13.40	358	340;165;153;115	100	39	<0.10	<0.10
57	Mangiferin	13.56	421	403;385;331;301	−140	−31	4.00 × 10^4	2.00 × 10^6
58	Geniposide	13.62	387	355;225;123;101	−100	−12	3.20 × 10^3	6.40 × 10^3
59	Ferulic acid	13.89	193	178;149;134;117;106	−60	−21	256	1.28 × 10^3
60	Alaschanioside A	14.01	537	375;357;327;312;136	−80	−35	640	6.40 × 10^3
61	Lancerin	14.33	405	369;285;169	−160	−33	1.02	25.60
62	Neoline	14.57	438	420;356;221;152;122	120	40	<0.10	<0.10
63	Echinacoside	14.83	785	623;477;461;161;133	−30	−53	3.20 × 10^3	1.60 × 10^4
64	Polygalaxanthone VIII	15.17	567	447;345;315	−130	−42	320	640
65	7-O-Methoxyl-mangiferin	15.27	435	417;345;315	−140	−30	25.60	320
66	Talatisamine	15.30	422	390;358;181;169;129	120	39	<0.10	<0.10
67	Polygalaxanthone IX	15.84	551	505;431;243;201	−130	−36	25.60	320
68	Lariciresinol-4′-O-β-D-glucopyranoside	15.88	521	359;329;192;121	−60	−30	51.20	640
69	N-trans-p-Coumaroyloctopamine	15.98	298	280;145;133;119	−160	−17	191.36	956.80
70	Tenuifoliside B	16.10	667	461;205;190	−200	−37	25.60	320
71	Verbascoside	16.17	623	461;315;161;133	−50	−41	3.20 × 10^3	6.40 × 10^3
72	Poliumoside	16.21	769	607;461;161;133	−50	−55	3.20 × 10^3	1.60 × 10^4
73	N-trans-Feruloyloctopamine	16.48	328	310;161;133	−120	−18	1.68	42.11
74	Isoverbascoside	16.58	623	461;315;161;133	−50	−41	3.20 × 10^3	6.40 × 10^3
75	4-Methoxyphenylethanol	16.65	151	136;108;92;59	−40	−17	3.11	7.78
76	Pinoresinol-β-D-glucopyranoside	16.70	519	357;342;151;136	−60	−24	51.20	1.28 × 10^3
77	Polygalaxanthone VII	16.76	611	596;576;368;303	−130	−42	320	640
78	Cistanoside C	17.26	637	491;475;161;133	−50	−44	1.28 × 10^3	6.40 × 10^3
79	3,6′-Disinapoyl sucrose	17.29	753	547;367;325;205;190	−200	−39	0.51	1.02
80	2′-Aceylpoliumoside	17.49	811	769;649;607;161;133	−50	−54	3.20 × 10^3	6.40 × 10^3
81	Isocistanoside C	17.63	637	491;473;461;161;133	−50	−44	1.60 × 10^4	6.40 × 10^3
82	Cinnamic acid	17.83	147	103;62	−50	−15	8.00 × 10^3	1.60 × 10^4
83	Tenuifoliside A	18.01	681	443;179;137	−200	−34	0.20	1.02
84	3,4,5-Trimethoxycinnamic acid	18.13	237	178;133;103;89	−50	−17	6.40 × 10^3	3.20 × 10^4
85	Tubuloside B	18.22	665	623;461;443;315;161;133	−50	−45	6.40 × 10^3	1.60 × 10^4
86	Benzoylmesaconine	18.24	590	572;540;166;105	90	48	<0.10	<0.10
87	Ginsenoside Rg1	18.36	845	799;637;475;437;391	−90	−32	2.00 × 10^4	4.00 × 10^4
88	Ginsenoside Re	18.38	991	945;637	−90	−32	6.00 × 10^3	8.00 × 10^3
89	Cistanoside D	18.47	651	615;505;193;175;160	−50	−37	51.20	640
90	p-Methoxycinnamic acid	18.55	177	149;133;118;107	−50	−15	1.60 × 10^4	3.20 × 10^4
91	N-trans-p-Coumaroyltyramine	18.66	282	145;119;117	−120	−34	1.45	7.24
92	Polygalaxanthone IV	18.70	565	521;344;257;242;172	−200	−40	8.00 × 10^3	8.00 × 10^4
93	3-(4-Hydroxyphenyl)-N-[2-(4-hydroxyphenyl)-2-methoxyethyl]-acrylamide	18.81	312	280;145;117	−50	−17	4.01	20.03
94	Loureiriol	19.06	301	195;167;123	−90	−24	2.06	3.10
95	N-trans-Feruloyltyramine	19.07	312	297;178;148;135	−130	−36	40.06	200.32
96	Liquiritigenin	19.14	255	135;119;91	−100	−23	1.05	5.24
97	3-(4-Hydroxy-3-methoxyphenyl)-N-[2-(4-hydroxyphenyl)-2-methoxyethyl]-acrylamide	19.23	342	324;310;160;133	−95	−17	43.90	219.52
98	N-trans-Feruloyl-3-methoxytyramine	19.44	342	327;298;148;135	−120	−35	4.39	43.90
99	Polygalasaponin XXVIII	19.68	1103	1103;745;583;539;469;455;425;	−70	−20	3.20 × 10^3	4.00 × 10^3
100	Benzoylaconine	19.70	604	572;554;522;199;105	100	47	<0.10	<0.10
101	Benzoylhypacoitine	20.23	574	542;510;178;105	103	47	<0.10	<0.10
102	Pseudo-ginsenoside F11	20.63	845	799;653;491	−90	−32	4.00 × 10^3	2.00 × 10^4
103	5,7,4′-Trihydroxyflavanone	20.70	271	177;151;119;93;65	−100	−25	5.58	139.55
104	Ginsenoside Rf	20.71	845	799;637;475;459;391	−90	−32	4.00 × 10^3	2.00 × 10^4
105	Cannabisin D	20.80	623	460;444;350;322;310;158	−190	−38	1.60	3.19
106	Ginsenoside Ro	20.80	955	955;793;569;523	−90	−5	8.00 × 10^3	4.00 × 10^4
107	Tenuifolin	21.20	679	625;455;425;342	−70	−38	128	320
108	6-Hydroxy-1,2,3,7-tetramethoxyxanthone	21.22	331	316;301;157;89	−180	−28	128	320
109	Melongenamide B	21.41	639	621;486;460;415;297	−40	−44	81.92	1.02 × 10^3
110	Ginsenoside Rb1	21.53	1153	1107;945;799;783	−90	−32	1.00 × 10^5	2.50 × 10^6
111	Ginsenoside Rg2	21.66	829	783;637;475;391	−90	−32	2.00 × 10^4	4.00 × 10^4
112	Ginsenoside Rc	21.91	1123	1077;945;915;783;621;459	−90	−32	8.00 × 10^3	1.00 × 10^4
113	3,4′-Dihydroxy-5-methoxystilbene	22.05	241	225;197;181;143	−145	−29	123.55	1.54 × 10^3
114	5,7-Dihydroxy-4′-methoxy-8-methylflavane	22.08	285	191;165;119;79	−130	−28	9.16 × 10^3	4.58 × 10^4
115	Ginsenoside Rh1	22.14	683	637;475;391	−90	−32	4.00 × 10^3	8.00 × 10^3
116	Ginsenoside Rb2	22.18	1123	1077;945;915;783;621;459	−90	−32	4.00 × 10^3	8.00 × 10^3
117	2,4′-Dihydroxy-4,6-dimethoxydihydrochalcone	22.63	301	207;147;135;93	−40	−24	6.19	154.85
118	Ginsenoside Rd	22.87	991	945;917;783;621;459	−90	−32	2.00 × 10^4	4.00 × 10^4
119	Hypaconitine	23.08	616	584;556;524;496;338;197	130	44	<0.10	<0.10
120	Ginsenoside F1	23.11	683	637;475;391;71	−90	−32	4.00 × 10^3	8.00 × 10^3
121	1,2,3,6,7-Pentamethoxyxanthone	23.21	347	332;317;289;218;121	100	27	1.02	5.12
122	1,7-Dimethoxyxanthone	23.40	257	242;213;171;139;115	120	30	2.56	5.12
123	N-trans-Feruloyltyramine dimer	23.43	623	460;445;430;324;297	−200	−30	1.60	7.99
124	Cannabisin F	23.56	623	471;432;402;298	−30	−39	3.19	39.94
125	Melongenamide D	23.93	934	771;739;580;395;319	−100	−50	119.68	598.40
126	4-Hydroxy-2,4′-dimethoxydihydrochalcone	24.43	285	181;149;134;117	−80	−19	1.83 × 10^3	1.83 × 10^4
127	1,2,3,7-Tetramethoxyxanthone	24.82	317	287;259;215;186;132	130	35	1.02	5.12
128	Ginsenoside F2	25.68	829	783;621;459;375;99	−90	−32	2.00 × 10^4	4.00 × 10^4
129	Ginsenoside Rg3	25.91	829	783;621;459	−90	−32	2.00 × 10^4	4.00 × 10^4
130	Pterostilbene	25.92	255	239;224;197;169	−100	−30	5.23	26.17
131	4′-Hydroxy-5,7-dimethoxy-8-methylflavane	27.96	299	179;119	−20	−18	9.60 × 10^3	4.80 × 10^4
132	Ginsenoside Rh2	29.63	667	621;581;459;417	−90	−32	4.00 × 10^3	6.00 × 10^3
133	Compound K	30.16	667	621;459;339;161	−90	−32	2.00 × 10^5	5.00 × 10^5

---

Formic acid, ammonium formate, dimethylsulfoxide (DMSO), methanol, and acetonitrile (ACN) were of HPLC grade and purchased from Merck (Darmstadt, Germany). Ultrapure water was prepared in-house with a Milli-Q system (Millipore, Bedford, MA, USA). The other chemicals were of analytical grade and obtained commercially from Beijing Chemical Works (Beijing, China).

2.2 Raw materials

The raw materials of Ginseng Radix (Chinese name: Renshen), Aconiti Lateralis Radix Praeparata (Chinese name: Fuzi), Solani Melongenae Radix (Chinese name: Qiegen), Pheretima (Chinese name: Dilong), Galli Gigerii Endothelium Corneum (Chinese name: Ji’neijin), Cistanches Herba (Chinese name: Roucongrong), Polygalae Radix (Chinese name: Yuanzhi) and Draconis Resina (Chinese name: Longxuejie) were collected from Beijing Tongrentang Co. Ltd. (Beijing, China) and a local pharmacy (Beijing, China). All crude drugs were authenticated by one of the authors (Prof. Pengfei Tu) and deposited at the Modern Research Center for Traditional Chinese Medicine, Beijing University of Chinese Medicine (Beijing, China).

2.3 Sample preparation

All raw materials were dried using a universal oven with forced convection (FD115, Tuttlingen, Germany) at 40 °C for three days. Then, each crude drug was pulverized into powder using a sample mill (model YF102, RuianYongli Pharmacy Machinery Company, Zhejiang, China) and sieved through a metal drug sieve (0.25 mm, i.d.). Thereafter, the simulated formula was prepared by mixing all accurately weighed raw materials (approximately 0.20 g of each) and was extracted with 20-fold volumes of 50% aqueous methanol for 30 min at 25 °C in an ultrasonicator (230 V, Branson model 5510, Danburry, CT, USA). Following centrifugation at 1500 rpm for 5 min in a centrifuge (Eppendorf, Melbourne, Australia), each supernatant was filtered through a 0.22 μm membrane. An aliquot (50 μL) of the filtrate was 20-fold diluted with 50% aqueous methanol prior to LC-MS/MS measurement. Each raw material was treated in parallel to obtain the extract sample by extracting 0.20 g raw material with 4 mL 50% aqueous methanol. Every experiment was conducted in triplicate.

Stock solutions of all reference standards were prepared individually with methanol, DMSO or water depending on compound solubility, and stored at 4 °C until use. Then, a mixed standard stock solution was prepared by mixing all stock solutions. The working standard solutions were obtained by diluting the mixed standard stock solution with 50% aqueous methanol to desired concentration levels. In addition, each reference solution at appropriate concentration was generated by diluting the corresponding stock solution with methanol or 50% aqueous methanol for manual optimization of those compound-dependent mass spectrometric parameters.

2.4 LC-MS/MS analysis

Liquid chromatography was conducted on a Shimadzu UHPLC system (Kyoto, Japan) that comprised of two LC-20AD_XR solvent delivery units, a SIL-20AC_XR auto-sampler, a CTO-20AC column oven, a DGU-20A_3R degasser, and a CBM-20A controller. Chromatographic separation was achieved on a Capcell core ADME column (2.1 mm × 150 mm, 2.7 μm, Shiseido, Tokyo, Japan) at a flow rate of 0.4 mL min⁻¹, and the column oven was maintained at 40 °C. The mobile phase was composed of 10 mmol L⁻¹ aqueous ammonium formate (A) and acetonitrile containing 0.1% formic acid (B). The gradient elution was programmed as follows: 0–5 min, 0–2% (B); 5–8 min, 2–5% (B); and 8–30 min, 5–65% (B). At the end of each run, the initial composition of mobile phase (0% (B)) was permitted to re-equilibrate the whole system for 5 min. The auto-sampler module was maintained at 10 °C and the injection volume was set at 2.0 μL.

Mass spectrometry was carried out on an ABSciex 5500 Q-trap® mass spectrometer (ABSciex, Foster City, CA, USA) which was equipped with a Turbo V™ electrospray ionization (ESI) interface and operated in sMRM mode. Both positive and negative polarities were adopted according to the results provided by manual parameter optimization. Ion optics were tuned using standard polypropylene glycol (PPG) dilution solvent. Nitrogen was used as the nebulizer (GS1), heater (GS2), curtain (CUR), and collision gas. Ion source parameters were optimized as follows: GS1, GS2, and CUR, 55, 55, and 35 psi, respectively; ionspray needle voltage, 5500 V/−4500 V; heater gas temperature, 550 °C; collisionally activated dissociation (CAD) gas, high level. Entrance potential (EP) and collision cell exit potential (CXP) levels followed the default values, whereas optimized MRM ion transitions (precursor ion-to-the most abundant product ion for each analyte), declustering potential (DP), and collision energy (CE) values for the quantitative ion transitions of all reference compounds are summarized in Table 1. In addition, an accompanying ion transition, which was composed of the precursor ion and the secondary abundant fragment ion, was also utilized for each compound to meet the demands of identity confirmation simultaneously with quantitative analysis.²⁵ The detection time window for each ion transition was set as 60 s (retention time ± 30 s), and the target scan time was maintained at 1.0 s. The information dependent acquisition (IDA) method was employed to trigger two EPI scans with a criterion of 200 cps. The key parameter (CE) of EPI was set as 40 eV and −40 eV for positive and negative polarities, respectively, whereas collision energy spread (CES) was set at 35 eV for both. Analyst software (version 1.6.2, ABSciex) was used for the synchronization of the whole system and for data acquisition and processing.

In addition, in order to compare sMRM and conventional MRM (cMRM) in parallel, cMRM was also performed with the parameters mentioned above, except that the detection time window was replaced with a 10 s dwell time for each ion transition.

2.5 Method validation

For method validation, quantitative terms with respect to linearity, limit of detection (LOD), lower limit of quantification (LLOQ), and recovery were assayed. Among them, LOD and LLOQ assays were performed for all 133 targets (Table 1), whereas the other assays were carried out for 23 selected analytes (Tables S1 and S2, ESI† B) that exhibited abundant distributions in the simulated formula, namely thymine, 1,7-dimethoxyxanthone, 1,2,3,7-tetramethoxyxanthone, 1,2,3,6,7-pentamethoxyxanthone, songorine, benzoylhypaconine, benzoylaconine, L-(+)-lactic acid, nicotinic acid, inosine, salidroside, 4-hydroxy-2,4′-dimethoxydihydrochalcone, 4′-hydroxy-5,7-dimethoxy-8-methylflavane, loureiriol, 2,4′-dihydroxy-4,6-dimethoxydihydrochalcone, 6-deoxycatalpol, alaschanioside A, polygalaxanthone IX, polygalaxanthone VIII, polygalaxanthone VII, tenuifoliside B, polygalasaponin XXVIII, and ginsenoside Rb2. The performance of each validation assay followed the protocols described in the literature.²⁵ For the recovery assay, 23 analytes were added into mixed raw materials at low, medium, and high concentration levels before extraction to prepare desired samples (Table S2, ESI† B). Six replicates of the simulated formula solution were used to evaluate the repeatability, and the sample was maintained in the auto-sampler at 10 °C and then analyzed over three consecutive days to carry out the stability assay. The RSD% (relative standard deviation%) value of the peak area of each analyte was adopted to express the repeatability and stability.

Afterwards, the developed method was applied for the analysis of the simulated formula and all raw materials.

3. Results and discussion

3.1 Development of LC-MS/MS method

3.1.1 Optimization of mass parameters. Aiming to obtain an optimal quantitative response, the MS/MS fragmentation for each compound was investigated. All 133 analyte solutions were diluted to the desired concentrations (50–100 ng mL⁻¹) and directly injected into the ESI interface using a syringe pump (flow rate: 7 μL min⁻¹). Afterwards, optimization of the mass parameters, including precursor-to-product ion transitions, DP, and CE for each analyte, was manually carried out following the procedures described in the literature.^26,27

The mass spectrometric behaviors of ginsenosides, flavonoids, phenylpropanoid amides, phenylethanoid glycosides, xanthones, and aconite alkaloids, including pseudo-molecular ions and fragments, agreed well with some previous descriptions,^25,28–31 while the MS patterns of those hydrophilic components were consistent with the information archived in the literature^32–35 and some accessible databases (e.g. MassBank, METLIN, and HMDB). In addition, the mass cracking rules of those authentic references from Polygalae Radix, namely sibiricose A5, sibiricose A6, mangiferin, polygalaxanthone VIII, 7-O-methoxylmangiferin, polygalaxanthone IX, polygalaxanthone VII, polygalaxanthone VII, polygalasaponin XXVIII, tenuifolin, 1,7-dimethoxyxanthone, 1,2,3,7-tetramethoxyxanthone and 1,2,3,6,7-pentamethoxy-xanthone, were identical to the properties documented in ref. 36. More compounds, 98 in total (corresponding to 196 ion transitions), could afford better responses under negative polarity, while 35 components (corresponding to 70 ion transitions) obtained greater responses with positive ionization mode. All information regarding the MS¹, MS², DP, CE, and quantitative MRM transitions is summarized in Table 1.

3.1.2 Selection of columns. As noted above, we simultaneously targeted both hydrophilic and hydrophobic components in the current study; thus, it is of great importance to select an optimum column that can retain and separate extensive analytes across a great polarity span. In general, a single column is only advantageous at retaining and separating components in a relatively narrow polarity range. However, a few new types of particles, such as pentafluorophenyl (PFP or F5) substituted particles,^24,37 have been developed and proven for universal retention.

Several columns were introduced as candidates to pick out the optimal one for comprehensive retention. After careful comparison in terms of peak capacity, retention performance, peak shape, and low back-pressure, one of the core–shell-type columns, the Capcell core ADME column, was found to be superior to the other columns, including not only the versatile Phenomenex Synergi Polar-RP column³⁸ and the widely recommended PFP and F5 columns, but also some HILIC candidates. Some additives, such as formic acid and ammonium formate, were supplemented into the mobile phase to assess whether they could enhance the peak shapes along with the overall MRM response, and the results suggested the addition of 10 mM ammonium formate and 0.1% formic acid into phases (A) and (B), respectively, as an ideal choice.

The functional group substituted on the silica gel of the ADME particles is the adamantylethyl group. Its surface polarity is 0.65,³⁹ which is considerably higher than that of common RP-C₁₈ columns (approximately 0.4) and makes those particles able to retain hydrophilic components like a HILIC column. Meanwhile, the hydrophobicity of 1.98 indicates that the ADME column could exhibit a retention potency for hydrophobic compounds comparable with a normal C₁₈ column; however, it could tolerate 100% aqueous mobile phase for a long period without stationary phase collapse due to the relatively low hydrophobicity level but big size of the adamantylethyl substitutions. Hence, it is not astonishing to note that the core–shell-type ADME column was advantageous in terms of peak capacity, peak shape, and back-pressure over the other columns for the retention and separation of both polar and non-polar components.

The optimized conditions for LC and MS domains were applied for the analysis of mixed references and the simulated TCM formula, and the representative chromatograms are shown in Fig. 1, while the corresponding chromatogram of each raw material is shown in Fig. S1 (ESI† A). Overall, satisfactory peak shape and separation capacity, but low back-pressure, were obtained. As shown in the chromatograms, most of the hydrophilic components gathered around 0.2–2.0 min, whereas the hydrophobic constituents were widely distributed between 2.0 and 28.0 min. Owing to the adoption of the sMRM algorithm, mutual interferences between co-eluting analytes could be significantly reduced. The signals in the mixed references were subjected to comparison with those existing in the formula for signal assignment in terms of retention times, MS² spectra, and ion transitions, and all 133 analytes could be found in the simulated formula.

Fig. 1 Overlaid extracted ion current (EIC) chromatograms. (A) EIC chromatograms of all 70 ion transitions monitored under positive polarity for mixed references; (B) EIC chromatograms of all 70 ion transitions monitored under positive polarity for the simulated TCM formula; (C) EIC chromatograms of all 196 ion transitions monitored under negative polarity for mixed references; (D) EIC chromatograms of all 196 ion transitions monitored under negative polarity for the simulated TCM formula.

3.2 Method validation

All 133 compounds were subjected to LLOQ and LOD assays, and the results are presented at Table 1. Except for a few analytes, such as fumaric acid, rhamnose, 8-epi-loganic acid, cistanoside E, mangiferin, and compound K, the LLOQs and LODs of all analytes are lower than 50 ng mL⁻¹, suggesting that sensitive quantitative analysis could be achieved using the developed UHPLC-sMRM method. It is worthwhile to mention that the toxic constituents from Aconiti Lateralis Radix Praeparata, including songorine, neoline, talatisamine, benzoylmesaconine, benzoylaconine, benzoylhypaconine, and hypaconitine, could be detected even at extremely low concentrations.

A total of 23 analytes that were observed as the primary ingredients in the simulated formula were employed for linearity, intra- and inter-day, repeatability, stability, and recovery assays. A weight of 1/x was applied for the regression of calibration curves if necessary. All calibration formulae and linear ranges are shown in Table S1 (ESI† B). As described in Table S1 (ESI† B), the correlation coefficients (r) of all calibration curves were higher than 0.999 over their corresponding linear concentration ranges. All RSDs% for repeatability and stability ranged from 0.83% and 12.78%, indicating satisfactory performance in terms of repeatability and stability. Three concentration levels of the mixture of 23 analytes were utilized to assess the intra- and inter-day precisions of the developed method, and all RSD values were observed to be lower than 15% (Table S2, ESI† B), indicating that the method could meet the demands of precise determination. Moreover, known amounts (low, medium and high concentration levels) of mixed 23 standard solutions were added to the mixed raw material powder prior to ultrasound-assisted extraction (Table S2, ESI† B). The recoveries were observed to be between 73.96% and 139.95% for all selected analytes, while most of the related RSDs were calculated as being lower than 15% (Table S2, ESI† B).

Because tandem mass spectrometric detection acted as the additional orthogonal separation dimension and the sMRM algorithm ulteriorly advanced the simultaneous determination, the mutual interferences among the co-eluting substances were expected to be mild. The responses of some selected hydrophilic analytes when they existed in a mixture were almost equivalent to the corresponding response yielded by injecting a single compound individually, suggesting that interferences were negligible during the quantitative characterization. In addition, the impacts of carryover and re-injection were also assessed and the results suggested that their effects could be ignored due to their mild influence.

Above all, the developed UHPLC-sMRM method was demonstrated to be a sensitive, precise, and accurate approach for simultaneous determination of numerous targets. Afterwards, the developed method was subjected to the simultaneous determination of the primary 23 components in the extracted solution of the simulated formula, and the quantitative results are given in Table S1 (ESI† B).

3.3 Comparison of sMRM and cMRM

MRM with fixed dwell time for each ion transition has been widely proved as a promising tool for the simultaneous determination of dozens of compounds; however, acceptable results are difficult to obtain when numerous analytes are targeted. Therefore, in the highly multiplex detection of TCMs, it is essential to employ an sMRM algorithm where the mass spectrometer is scheduled to detect only a limited number of ion transitions in predefined retention time windows.⁴⁰ A significant retention time shift would result in the loss of analyte when it is only programmed to be detected in a narrow retention time window. In the present study, the retention times of all analytes were assessed using the inter-day assays, and only minor migrations (less than 0.1 min) were observed for the retention times of those components. In consideration that most of the peak widths were approximately 10.0 s, the detection window was thereby fixed at 1.0 min for all analytes, while the target scan time was maintained at 1.0 s to satisfy the monitoring of several hydrophilic components that focused at the head of the chromatogram.

The principles of the cMRM and sMRM algorithms are briefly elucidated in Fig. 2, as well as their respective representative chromatograms. In the case of cMRM, all ion transitions are always monitored in every acquisition cycle. In general, it is necessary to assign at least a 10 ms dwell time to each ion pair without seriously compromising the reproducibility of the integrated peak. The cycle time was equal to the total dwell times of all ion transitions plus all pause times (Fig. 2A). In the present study, as many as 196 ion pairs were monitored under negative polarity, and the cycle time was thereby calculated as 2.1 s. For a typical UHPLC peak, the peak width was approximately 10 s; therefore, it is not astonishing that only five points were acquired for a signal peak using cMRM (Fig. 2B). On the other hand, the narrow detection window (1.0 min) of sMRM reduced the number of concurrent ion transitions compared with cMRM, and the dwell time was significantly and automatically maximized without the requirement of a long cycle time (Fig. 2C). The data points of the representative signal corresponding to sMRM were more fifteen, which can meet the demands of reliable quantitation (Fig. 2D). In addition, the intensity of the peak yielded by sMRM is significantly greater than that of cMRM (Fig. 2). Meanwhile, because an adequate dwell time was applied for each ion pair, the noise level of the equipment is thus obviously lower than that of cMRM (Fig. 2).

Fig. 2 Comparisons of algorithm principles and chromatograms acquired in parallel using cMRM and sMRM. (A) Overlaid EIC chromatograms of all 196 ion transitions monitored under negative polarity for mixed references using cMRM algorithm. All ion transitions are always monitored in every acquisition cycle, and the cycle time is equal to the total dwell times of all ion transitions plus all pause times. (B) Representative peak acquired using cMRM. Because the cycle time is too long for the peak width, the data points of this signal are only five, which cannot meet the demands of reliable quantitation. (C) Overlaid EIC chromatograms of all 196 ion transitions monitored under negative polarity for mixed references using sMRM algorithm. Each ion transition is only monitored in its expected retention time window. In the current case, the MRM detection window for each ion transition is fixed as 1.0 min, whereas both of the cycle time and the dwell time are automatically adjusted to be appropriate. (D) Representative peak acquired using sMRM. Because the cycle time is automatically adjusted, and is usually less than the target scan time (1.0 s), the data points of this signal are more than fifteen, which can meet the demands of reliable quantitation.

The quantitative performances of sMRM and cMRM were also elucidated. Overall, all 133 compounds were detected in the simulated formula using the sMRM algorithm, whereas more than 50 analytes could not be observed with the cMRM method. Twelve analytes were picked to compare the sensitivity and precision between sMRM and cMRM. As shown in Table 2, all LODs and LLOQs resulting from sMRM are significantly lower, 5-fold at least, than those of cMRM. In particular, those hydrophilic components that gathered at the head of the chromatogram, e.g. γ-aminobutyric acid, nicotinamide, thymine, adenosine, and malonic acid, could be detected at trace concentrations with sMRM, whereas comparable sensitivity could not be afforded by cMRM (Table 2). In addition, as more data points were distributed in the sMRM peak in comparison with cMRM, the RSDs% of the intra-day assays of sMRM (1.44–7.25%) were significantly lower than those resulting from cMRM (3.37–24.37%).

Table 2 Comparison of the chromatographic performance between sMRM and conventional MRM (cMRM) algorithms in terms of sensitivity and precision

Compound	sMRM			cMRM
	LOD (ng mL^−1)	LLOQ (ng mL^−1)	Intra-day RSD^a (%)	LOD (ng mL^−1)	LLOQ (ng mL^−1)	Intra-day RSD (%)
a Precision data was evaluated from the intra-day relative standard deviations (RSDs) (n = 6).
γ-Aminobutyric acid	0.0010	0.0026	7.25	0.13	1.60	14.12
Nicotinamide	0.0051	0.010	2.62	6.40	16.0	3.37
Thymine	1.28	6.40	4.19	6.40	16.0	14.78
Adenosine	1.28	6.40	5.84	16.00	32.0	7.25
Malonic acid	8.00	16.0	3.59	200.00	400.0	9.21
Cinnamic acid	8.00	16.0	2.66	40.00	200.0	6.13
3,4-Dihydroxyphenylethanol	4.93	49.3	3.31	24.60	123.2	24.37
Inosine	3.20	6.40	4.90	16.00	200.0	10.35
Salidroside	3.20	16.0	2.93	44.00	200.0	7.14
Polygalaxanthone IV	8.00	80.0	4.27	40.00	200.0	5.37
Echinacoside	3.20	16.0	1.44	16.00	80.0	6.83
Ginsenoside Rf	4.00	20.0	3.27	20.00	100.0	13.45

---

The cycle time is of great importance, not only to obtain sufficient data points for a narrow peak, but to avoid the loss of peaks when several analytes are co-eluted.¹⁶ In the present study, EPI scans were triggered by the sMRM experiment with an IDA mode; hence, the loss of signals would result in the absence of MS² spectra. In addition, as previously mentioned, the response of cMRM is usually significantly lower than that of sMRM, and it is thereby difficult to acquire MS² spectra for the minor and trace compounds, because the intensity of cMRM ion transitions might not exceed the IDA threshold. Moreover, even though the intensity of the cMRM ion transition is a bit higher than the threshold, the quality of the MS² spectra should be rough. Taking adenosine for instance, since insufficient precursor ions (m/z 268 [M + H]⁺) were transmitted into the linear ion trap cell (Q3), the intensity of both protonated and fragment ions in the MS² spectrum generated by cMRM (lower part of Fig. 3) were considerably lower than those in the MS² spectrum generated by sMRM (upper part of Fig. 3). Moreover, some noise signals, such as ion species at m/z 251, 195, 156, and 109, are observed in the MS² spectra of cMRM (lower part of Fig. 3), indicating a remarkable obstacle for the confirmation of the peak identity.⁴¹

Fig. 3 Representative MS² spectra (adenosine) were acquired from enhanced product ion experiments which were triggered by sMRM (upper) and cMRM (lower). Obviously, the intensities of most fragments from sMRM are higher than those from cMRM, and also, some noise signals are observed in the MS² spectra of cMRM.

Therefore, sMRM was regarded to be superior to cMRM, being sensitive, reproducible, and giving reliable quantitation by providing higher responses, more data points, and high quality MS² spectra.

4. Conclusions

An algorithm, namely sMRM, was utilized to circumvent the contradiction between the low scan rate of QqQ equipment and the narrow width of the peaks generated from UHPLC, and adequate data points were gained for each peak, although as many as 133 analytes, including hydrophilic and hydrophobic substances, were analyzed in the current study. Efficient separation was achieved using a Capcell core ADME column. A satisfactory quality of MS² spectra was also achieved for all targets using EPI scans on a Q-trap analytical platform, attributed to the introduction of sMRM. Method validation assays indicated the developed UHPLC-sMRM method to be sensitive, accurate, and precise. Collectively, UHPLC-sMRM was suggested as a promising tool to meet the demands of large-scale quantitative analysis of both hydrophilic and hydrophobic compounds in TCMs, which could dramatically advance the quality control of TCMs in comparison with methods where only several hydrophobic components were analyzed. In addition, we can make a prospective view that the developed system provides a feasible analytical platform to simultaneously and universally monitor polar endogenous substances and TCM-derived apolar ingredients following the treatment of TCMs.

Acknowledgements

This work was financially supported by the National Science Fund of China (no. 81403073, to Y.-L. S.), the TCM support project from the Ministry of Industry and Information Technology of China (to J. L.), the Program for New Century Excellent Talents in University (no. NCET-13-0693, to J. L.), and National Science Fund for Excellent Young Scholars (no. 81222051, to Y. J.).

References

1. Y. Jiang, B. David, P. Tu and Y. Barbin, Anal. Chim. Acta, 2010, 657, 9–18.
2. E. Giovannucci, D. M. Harlan, M. C. Archer, R. M. Bergenstal, S. M. Gapstur, L. A. Habel, M. Pollak, J. G. Regensteiner and D. Yee, Ca-Cancer J. Clin., 2010, 60, 207–221.
3. M. M. Gottesman, Annu. Rev. Med., 2002, 53, 615–627.
4. J. J. Monsuez, J. C. Charniot, N. Vignat and J. Y. Artigou, Int. J. Cardiol., 2010, 144, 3–15.
5. M. E. Franks, G. R. Macpherson, E. R. Lepper, W. D. Figg and A. Sparreboom, Drug Resist. Updates, 2003, 6, 301–312.
6. T. Xue and R. Roy, Science, 2003, 300, 740–741.
7. P. S. Xie and A. Y. Leung, J. Chromatogr. A, 2009, 1216, 1933–1940.
8. P. Chen, W. Li, Q. Li, Y. Wang, Z. Li, Y. Ni and K. Koike, Talanta, 2011, 85, 1634–1641.
9. P. X. Chen, S. Wang, S. Nie and M. Marcone, J. Funct. Foods, 2013, 5, 550–569.
10. B. Li, D. Y. Kong, Y. H. Shen, H. Yuan, R. C. Yue, Y. R. He, L. Lu, L. Shan, H. L. Li, J. Ye, X. W. Yang, J. Su, R. H. Liu and W. D. Zhang, Org. Lett., 2012, 14, 5432–5435.
11. J. J. Qin, L. Y. Wang, J. X. Zhu, H. Z. Jin, J. J. Fu, X. F. Liu, H. L. Li and W. D. Zhang, Chem. Commun., 2011, 47, 1222–1224.
12. T. Rousu, J. Herttuainen and A. Tolonen, Rapid Commun. Mass Spectrom., 2010, 24, 939–957.
13. U. Vrhovsek, D. Masuero, M. Gasperotti, P. Franceschi, L. Caputi, R. Viola and F. Mattivi, J. Agric. Food Chem., 2012, 60, 8831–8840.
14. T. Hyotylainen and S. Wiedmer, Chromatographic Methods in Metabolomics, The Royal Society of Chemistry, 2013, p. 44.
15. Y. L. Song, W. H. Jing, G. Du, F. Q. Yang, R. Yan and Y. T. Wang, J. Chromatogr. A, 2014, 1338, 24–37.
16. T. Hyotylainen and S. Wiedmer, Chromatographic Methods in Metabolomics, The Royal Society of Chemistry, 2013, p. 47.
17. J. Liang, W. Y. Wu, G. X. Sun, D. D. Wang, J. J. Hou, W. Z. Yang, B. H. Jiang, X. Liu and D. A. Guo, J. Chromatogr. A, 2013, 1294, 58–69.
18. J. Li, C. X. Hu, X. J. Zhao, W. D. Dai, S. L. Chen, X. Lu and G. W. Xu, J. Chromatogr. A, 2013, 1320, 103–110.
19. Y. L. Song, N. Zhang, S. P. Shi, J. Li, Y. F. Zhao, Q. Zhang, Y. Jiang and P. F. Tu, J. Chromatogr. A, 2015 DOI:10.1016/j.chroma.2015.06.007 , in press.
20. Y. Liu, C. E. Uboh, L. R. Soma, X. Li, F. Guan, Y. You and J. W. Chen, Anal. Chem., 2011, 83, 6834–6841.
21. J. Wang, W. Chow and W. Cheung, J. Agric. Food Chem., 2011, 59, 8589–8608.
22. M. Dziadosz, J. P. Weller, M. Klintschar and J. Teske, J. Chromatogr. B: Anal. Technol. Biomed. Life Sci., 2013, 929, 84–89.
23. F. Song, J. Agric. Food Chem., 2011, 59, 4361–4364.
24. S. Dresen, N. Ferreirós, H. Gnann, R. Zimmermann and W. Weinmann, Anal. Bioanal. Chem., 2010, 396, 2425–2434.
25. Y. L. Song, N. Zhang, Y. Jiang, J. Li, Y. F. Zhao, S. P. Shi and P. F. Tu, RSC Adv., 2015, 5, 6419–6428.
26. Y. L. Song, W. H. Jing, G. Du, F. Q. Yang, R. Yan and Y. T. Wang, J. Chromatogr. A, 2014, 1338, 24–37.
27. Y. Song, W. Jing, F. Yang, Z. Shi, M. Yao, R. Yan and Y. Wang, J. Pharm. Biomed. Anal., 2014, 88, 269–277.
28. X. Q. Su, Y. L. Song, J. Zhang, H. X. Huo, Z. Huang, J. Zheng, Q. Zhang, Y. F. Zhao, W. Xiao, J. Li and P. F. Tu, Fitoterapia, 2014, 99, 64–71.
29. J. Sun, Y. L. Song, J. Zhang, Z. Huang, H. X. Huo, J. Zheng, Q. Zhang, Y. F. Zhao, J. Li and P. F. Tu, J. Agric. Food Chem., 2015, 63, 3426–3436.
30. Y. Jiang, S. P. Li, Y. T. Wang, X. J. Chen and P. F. Tu, J. Chromatogr. A, 2009, 1216, 2156–2162.
31. W. Z. Yang, M. Ye, X. Qiao, Q. Wang, T. Bo and D. A. Guo, Eur. J. Mass Spectrom., 2012, 18, 493–503.
32. J. Qu, W. Chen, G. Luo, Y. Wang, S. Xiao, Z. Ling and G. Chen, Analyst, 2002, 127, 66–69.
33. P. Purwaha, P. L. Lorenzi, L. P. Silva, D. H. Hawke and J. N. Weinstein, Metabolomics, 2014, 10, 909–919.
34. C. Stentoft, M. Vestergaard, P. Lovendahl, N. B. Kristensen, J. M. Moorby and S. K. Jensen, J. Chromatogr. A, 2014, 1356, 197–210.
35. D. Bylund, S. H. Norstrom, S. A. Essen and U. S. Lundstrom, J. Chromatogr. A, 2007, 1176, 89–93.
36. Y. Ling, Z. Li, M. Chen, Z. Sun, M. Fan and C. Huang, J. Pharm. Biomed. Anal., 2013, 85, 1–13.
37. H. Yoshida, T. Mizukoshi, K. Hirayama and H. Miyano, J. Agric. Food Chem., 2007, 55, 551–560.
38. T. J. Whelan, M. J. Gray, P. J. Slonecker, R. A. Shalliker and M. A. Wilson, J. Chromatogr. A, 2005, 1097, 148–156.
39. T. Mochizuki, T. Takayama, K. Todoroki, K. Inoue, J. Z. Min and T. Toyo’oka, Anal. Chim. Acta, 2015, 875, 73–82.
40. L. Guo, Y. Xiao and Y. Wang, Anal. Chem., 2014, 86, 10700–10707.
41. Z. Yan, T. Li, P. Lv, X. Li, C. Zhou and X. Yang, J. Chromatogr. B: Anal. Technol. Biomed. Life Sci., 2013, 928, 22–31.

---

Footnotes

† Electronic supplementary information (ESI) available. See DOI: 10.1039/c5ra09429a

‡ These two authors contributed equally to this article.

---