Selective and comprehensive characterization of the quinochalcone C-glycoside homologs in Carthamus tinctorius L. by offline comprehensive two-dimensional liquid chromatography/LTQ-Orbitrap MS coupled with versatile data mining strategies

Abstract

Quinochalcone C-glycosides (QCGs) are a series of pharmacologically bioactive components chemotaxonomic for Carthamus tinctorius L. The low abundance and ubiquitous interference from flavonoid O-glycosides (FOGs) frequently hinder the systematic exposure and characterization of these QCG homologs. We herein present an offline comprehensive two-dimensional liquid chromatography/linear ion-trap quadrupole/Orbitrap mass spectrometry (2D LC/LTQ-Orbitrap MS) approach coupled with versatile data mining strategies, to systematically characterize the QCGs from C. tinctorius. Initially, an offline 2D LC system, with an orthogonality of 71% and a theoretical peak capacity of 7654, was established by combining an Acchrom XAmide column and a BEH Shield RP-18 column. Subsequently, the water extract of C. tinctorius was separated by first dimensional hydrophilic interaction liquid chromatography (HILIC) yielding twelve fractions, which were further analyzed by reversed-phase ultra-high performance liquid chromatography/LTQ-Orbitrap MS using high energy C-trap dissociation (HCD) and collision-induced dissociation (CID) in the negative ion mode. The characteristic product ion filtering of m/z 119.05 (C₈H₇O⁻) in the HCD spectra, ring double bond equivalent (RDB 10–30), characteristic UV absorption around 405 nm, preferred ^0,2X₀ cleavage for C-glycosides, and diagnostic product ions analysis, were simultaneously employed for the structural elucidation of QCGs. Ultimately, 163 QCQ homologs were putatively characterized, and 149 are potential new ones. Particularly, nine dimers of QCG-FOG have not been previously reported. The obtained results have greatly expanded the knowledge on the structural diversity of QCGs, demonstrating the potency of the offline comprehensive 2D LC/LTQ-Orbitrap MS approach in separation and characterization of minor herbal components.

---

Introduction

Carthamus tinctorius L. (safflower) is a traditional herbal medicine (Hong-Hua) used to treat stroke, coronary heart disease, and angina pectoris, owing to its potency in invigorating blood circulation.^1,2 The water extract of C. tinctorius or its combination with other herbal medicine(s) has been developed into injection agents (such as Honghua Injection and Danhong Injection) to treat a series of stasis resultant cardiovascular and cerebrovascular ailments in China. The chemical composition of C. tinctorius features multiple categories of plant metabolites, such as quinochalcones, flavonoids, alkaloids, polyacetylenes, and lignans, etc.² Amongst them, quinochalcones occur as the C-glycoside form (quinochalcone C-glycosides, QCGs) and are considered to be chemotaxonomic for the identification of this medicinal herb.³ So far as now, a broad range of pharmacological properties, such as the anticoagulant, anti-inflammatory, antioxidant, antihypertensive, anti-tumor properties, and protecting heart/cerebral vessels, have been reported related to the QCG components.¹ New quinochalcone molecules discovery from C. tinctorius is a work of vital significance beneficial to the screening of antithrombotic drug candidates. However, the structural diversity of the QCG analogs in C. tinctorius has not been fully elucidated hitherto, and only 21 QCG compounds were isolated and identified to the end of 2014 (Fig. S1†). It is the low abundance (except for hydroxysafflor yellow A, HSYA), ubiquitous interference from FOGs, and structural instability,^4,5 that render a systematic characterization of QCGs from C. tinctorius practically difficult to implement. We recently reported a UHPLC/QTOF MS based metabolite profiling approach, by characteristic product ion filtering (CPIF) and combined analysis of the MS^E and Fast DDA data, enabling a primary characterization of 27 QCG components from C. tinctorius,⁶ however, we found that there was still a considerable amount of minor and isomeric QCG compounds failing to be identified because of the limited chromatographic separation capacity of one-dimensional liquid chromatography.

Two-dimensional liquid chromatography (2D LC), due to the expanded peak capacity and enhanced selectivity, provides a solution to resolution of complicated herbal components.^7–11 2D LC can be classified into comprehensive 2D LC and heart-cutting 2D LC according to whether a certain part or all of the eluents from the first dimensional separation are transferred and loaded for the second dimensional separation.¹² In contrast, offline comprehensive 2D LC is a promising analytical approach easily accessible among the laboratories (devoid of the relatively complex configuration of pumps and switching waves) by a flexible combination of different separation mechanisms. Meanwhile, it is unnecessary to consider the mobile phase compatibility and separation speed matching in two dimensions by means of the offline 2D LC. There have been several reports that utilized the offline 2D LC to separate and characterize minor herbal components, indicating its potential in discovery of unknown molecules. Given the large polarity of some QCG components, the combination of hydrophilic interaction chromatography (HILIC) and reversed-phase liquid chromatography (RPLC) might be suitable for the resolution of minor QCGs from C. tinctorius. Indeed, the integration of HILIC and RPLC in the offline 2D LC mode exhibited excellent capacity in the separation of polar and medium-polarity herbal components.^13,14 Our laboratory recently reported an offline comprehensive 2D LC system (HILIC × RPLC) coupled with high-resolution mass spectrometric analysis that facilitated the characterization of 646 ginsenosides from the stems and leaves of Panax ginseng.¹¹ We thereby believe that the integration of offline comprehensive 2D LC and high-resolution mass spectrometry has the potential to efficiently expose and characterize new natural bioactive molecules like QCGs from C. tinctorius.

Linear ion-trap quadrupole/Orbitrap mass spectrometry (LTQ-Orbitrap MS), by integrating the multi-stage fragmentation (MSⁿ, n ≥ 2) of the linear ion-trap mass spectrometer and exact mass measurement of Orbitrap, enables high-accuracy MS and MSⁿ acquisition using selectable multiple fragmentation mechanisms and flexible scan methods, thus particularly suitable for the characterization of unknown structures.^15–17 Since the commercial availability of the LTQ-Orbitrap mass spectrometers (including the Discovery, XL, Velos, Velos Pro, and Elite series), three different fragmentation mechanisms, namely collision-induced dissociation (CID), pulsed-Q dissociation (PQD), and high-energy C-trap dissociation (HCD), can be performed on a single MS platform. Simultaneous application of two different fragmentation modes may provide complementary fragments information for a more reliable structural elucidation based on the literatures: (1) quite different MS/MS spectra were obtained between CID and HCD for the analysis of oligosaccharides, and more abundant fragments and more balanced spectrum could be obtained by HCD;^18,19 (2) CID of ginsenosides could give diagnostic information regarding the sugar residues and sapogenins by characteristic neutral eliminations, whilst HCD could offer rich product ions diagnostic for the nature/linkage of the sugars and the composition of oligosaccharide chains.¹¹ On the other hand, versatile integral scan methods can be developed on a hybrid LTQ-Orbitrap mass spectrometer, such as the data-dependent or data-independent acquisition (DDA or DIA) with dynamic exclusion (DE),²⁰ full scan-parent ions list-triggered MSⁿ coupled with DE,²¹ and the so called “MS^M” that achieves the simultaneous DDA and MS^all (or MS^E) acquisition within a single run.²²

To comprehensively expose and discover potentially new QCG molecules from C. tinctorius, in this study, an integrated qualitative analysis approach, by offline comprehensive 2D LC/DAD/LTQ-Orbitrap MS, was developed following a protocol we have established (Fig. 1).¹¹ Both the negative mode HCD and CID were employed to acquire the fragmentation information of all C. tinctorius components. Post-acquisition data mining techniques by CPIF of the HCD spectra, ring double bond equivalent (RDB), characteristic UV absorption, and diagnostic product ions (DPIs) interpretation of the high-accuracy CID-MS³ and HCD-MS/MS data, were simultaneously employed to achieve a selective, comprehensive characterization of the target QCG homologs. As a result, a mass of minor and unknown QCG components were chromatographically separated and structurally characterized by the well-established offline 2D LC system and LTQ-Orbitrap MS analysis, laying the foundation of deeply understanding the therapeutic basis of the traditional herbal medicine.

Fig. 1 A work-flow for the selective characterization of the quinochalcone C-glycoside (QCG) homologs from C. tinctorius by offline comprehensive 2D LC and LTQ-Orbitrap MS coupled with characteristic product ions filtering.

Experimental

Materials and reagents

Four QCG reference standards (purity > 90%), comprising hydroxysafflor yellow A (HSYA), anhydrosafflor yellow B (AnHSYB), isosafflomin C (ISC), and safflomin C (SC), were isolated from the florets of C. tinctorius L. by the authors. Their structures were established by comparing the ¹³C NMR data with those reported in previous literature.⁶ HPLC-grade acetonitrile from Merck KGaA (Darmstadt, Germany), ammonium formate (AF) from Sigma-Aldrich (Sigma-Aldrich, MO, Switzerland), formic acid (FA) from Tedia Company Inc. (OH, USA), and ultra-pure water (18.2 MΩ cm at 25 °C) prepared by a Millipore Alpha-Q water purification system (Millipore, Bedford, USA), were used or tested in the mobile phase.

The florets of Carthamus tinctorius L. were collected in Sinkiang Autonomous Province of China in the summer of 2014. Authentication of the drug materials was confirmed by TLC examination and HPLC assay according to China Pharmacopoeia (2015 version). The voucher specimen was deposited at the author's laboratory in Shanghai Institute of Materia Medica, Chinese Academy of Sciences (Shanghai, China).

Sample preparation

An aliquot of 1 g finely ground powder of C. tinctorius was soaked in 10 mL water for 10 min and extracted on a water bath at 40 °C with ultrasound assistance (1130 W, 37 kHz) for 1 h. The extract was centrifuged at 4000 rpm for 10 min. Subsequently, the supernatant was transferred into a 10 mL volumetric flask and diluted to the constant volume with ultra-pure water as the stock solution. The test solution was prepared from the stock solution by two-fold dilution (at a concentration of 50 mg mL⁻¹) and filtered through a 0.2 μm PTFE filter (Agilent Technologies, Santa Clara, CA, USA).

Offline comprehensive 2D LC separation of the C. tinctorius extract by HILIC × RPLC

The offline comprehensive 2D LC separation of the water extract of C. tinctorius was performed by the combination of HILIC × RPLC. The ¹D HILIC separation was carried out on an Agilent 1100 series HPLC system (Agilent, Waldbronn, Germany) equipped with a quaternary pump, a degasser, an autosampler, a thermostatic column compartment, and a VWD detector. An Acchrom XAmide column maintained at 25 °C was used for the ¹D separation. A binary mobile phase consisting of 2 mM ammonium formate (A) and acetonitrile (B) following an optimized gradient program was used: 0–8 min, 95–86% (B); 8–12 min, 86–85% (B); 12–13 min, 85–72% (B); 13–20 min, 72–70% (B); 20–23 min, 70–50% (B); and 23–30 min, 50% (B). The flow rate was set at 1.0 mL min⁻¹. The VWD detector was set at 280 nm to simultaneously monitor both QCGs and other FOGs. Each 40 μL of the test solution of C. tinctorius was injected onto the column. Peak-dependent collection was employed in this study, that is, the eluents between 1–4 min as Fr. 1, 4–6 min as Fr. 2, 6–9 min as Fr. 3, 9–12 min as Fr. 4, 12–14 min as Fr. 5, 14–15.5 min as Fr. 6, 15.5–17 min as Fr. 7, 17–18 min as Fr. 8, 18–20 min as Fr. 9, 20–23 min as Fr. 10, 23–26 min as Fr. 11, and 26–29 min as Fr. 12. Ten replicate injections were operated in total. The obtained twelve pooled fractions (Fr. 1 to Fr. 12) were concentrated to dryness under a steady flow of dry N₂. Each residue was re-dissolved in 400 μL water and filtered through a 0.2 μm PTFE filter prior to the ²D separation and detection.

The second dimensional (²D) separation was performed on an Ultimate® 3000 UHPLC system (Thermo Fisher Scientific, San Jose, CA, USA) equipped with a binary pump, an autosampler, a DAD detector, and a column compartment. A RP mode UHPLC column, BEH Shield RP-18 (2.1 × 100 mm, 1.7 μm) maintained at 30 °C was used and eluted by a binary mobile phase containing 0.1% formic acid (A, v/v) and acetonitrile (B) in a 20 min gradient program: 0–3 min, 1–13% (B); 3–7 min, 13–16% (B); 7–8 min, 16–19% (B); 8–12 min, 19–26% (B); 12–14 min, 26–50% (B); 14–15 min, 50–95% (B); and 15–20 min, 95% (B). A flow rate of 0.4 mL min⁻¹ was set. The injection volume was 2 μL. The DAD detector recorded the UV signals from 190–600 nm, and at two separate channels of 280 nm and 405 nm.

LTQ-Orbitrap MS

An LTQ-Orbitrap Velos Pro hybrid mass spectrometer (Thermo Fisher Scientific, San Jose, CA, USA), equipped with a new highly sensitive heat electrospray ionization (HESI) probe operating in the negative ion mode, was applied to detect the whole components of C. tinctorius. The source parameters were set as follows: source spray voltage, 3.2 kV; capillary temperature, 320 °C; source heater temperature, 300 °C; sheath gas (N₂), 40 arbitrary units; and auxiliary gas (N₂), 10 arbitrary units. Source fragmentation voltage was not enabled since the deprotonated precursor ions were generated in high abundance. Both data-dependent CID and HCD were employed aimed to obtain the complementary fragment information. In CID mode, the Orbitrap analyzer scanned over a mass range of m/z 200–1500 at a resolution of 30 000 (FWHM at m/z 400) in profile format (p) for MS scan and in centroid format (c) for MSⁿ (n = 2, 3) according to a pre-defined duty circle consisting of four successive scan events (I–IV): Event I performs the full scan; Event II records the MS² fragments of the most intense precursor ions; Events III and IV separately select the 1st and 2nd most abundant MS² product ions recorded by Event II to trigger the MS³ fragmentation. Normalized collision energy (NCE) was set at 50%. In the HCD mode, two scan events were defined to record the MS precursor ions and their MS² fragments. An NCE of 90% was used. DE was enabled for HCD following the parameters: repeat count, 3; repeat duration, 3 s; exclusion list size, 25; and exclusion duration, 10 s. Isolation width was set at 2.0 Da. Data recording and processing were executed using Xcalibur 2.1 software (Thermo Fisher Scientific, San Jose, CA, USA). For prediction of the elemental composition of an observed QCG molecule, the parameters were set as follows: elemental in use, C 1–60, H 1–100, O 1–60, N 0–1; RDB 10–35; mass tolerance, 5 ppm; nitrogen rule, do not use.

Orthogonality evaluation

Orthogonality evaluation of the established offline 2D LC system was performed using an in-house executable program by incorporating a series of asterisk equations using C programming language.²³ A total of 65 major components (contained both QCGs and FOGs) were selected as the index for orthogonality evaluation. First, the t_R (min) of each component was normalized to the relative retention times, ¹t_R,norm(i) (in ¹D separation) and ²t_R,norm(i) (in ²D separation) according to eqn (1), in which t_D and t_G represent the dead time and the total gradient elution time, respectively.

The 2D separation space can be described by four crossing lines named Z₋, Z₊, Z₁, and Z₂. Spreading of all index components around these four lines (S_Z−, S_Z+, S_Z1, and S_Z2) were calculated based on eqn (2)–(5), in which σ stands for the standard deviation of the values of the bracketed equations for all 65 index components.

S_Z− = σ{¹t_R,norm(i) − ²t_R,norm(i)} (2)

S_Z+ = σ{²t_R,norm(i) − (1 − ¹t_R,norm(i))} (3)

S_Z1 = σ{¹t_R,norm(i) − 0.5} (4)

S_Z2 = σ{¹t_R,norm(i) − 0.5} (5)

Then four Z parameters were generated based on the calculated values of S_Z−, S_Z+, S_Z1, and S_Z2, by eqn (6)–(9) to indicate the peaks spreading around four lines. Ultimately, these Z parameters were bundled into eqn (10) to test the orthogonality of the established 2D LC system.

Z₋ = |1 − 2.5|S_Z− − 0.4|| (6)

Z₊ = |1 − 2.5|S_Z+ − 0.4|| (7)

Results and discussion

Construction and parameters optimization for the offline 2D LC system

Aimed to establish a 2D LC system that is potent of exposing the components of C. tinctorius, key parameters for each dimensional separation, including the stationary phase, mobile phase, column temperature, and gradient elution program, were optimized individually.

Given that the ²D chromatography is coupled to the LTQ-Orbitrap MS to acquire the MSⁿ data for the final structural elucidation, we first selected RPLC to optimize the ²D chromatographic conditions mainly based on the following considerations: (1) almost all the previous studies concerning the holistic chemical profiling of C. tinctorius or its metabolites employed RPLC separation, and the results or conclusions (including the retention behavior and MS data) are very important for the characterization of QCGs in this study;^24–28 (2) RPLC could separate a series of herbal components covering a wide polarity range and with good selectivity; and (3) RPLC is potent of separating the QCG isomers. Five different RP columns (including BEH C18, HSS T3, BEH Shield RP-18, Zorbax SB, and SB-Aq) were compared in terms of the selectivity (Fig. S2†). Comparatively, Shield RP-18 exhibited the strongest retention capacity, the best column efficiency, and the best resolution of the major components. Additionally, Shield RP-18 can endure a mobile phase of 100% H₂O, rendering it more suitable for separating polar components. Therefore, the Shield RP-18 column was selected as the column for the ²D separation.

Subsequently, selection of the column for the ¹D separation was based on the separation difference of six candidate columns (Acchrom XAmide, XBridge Amide, BEH-Phenyl, CSH Phenyl-Hexyl, HSS Cyano, and HSS PFP) to the selected Shield RP-18 column. Linearity regression correlation coefficient (R²) of fifteen major components scattered in a 2D plot among different column combinations was compared (Fig. 2a). Apparently, two HILIC columns combined to Shield RP-18 exhibited higher separation capacity (R² < 0.5) than the other candidates. In contrast, the combination of Acchrom XAmide × Shield RP-18 could be regarded as the best choice to configure the 2D LC system for separating the components of C. tinctorius. The satisfactory separation power by integrating HILIC × RPLC has been demonstrated in analysis of the saponins in Panax notoginseng and stems and leaves of Panax ginseng.^11,14

Fig. 2 Comparison of the separation difference of six different stationary phases to Shield RP-18 employed in the second-dimensional separation by plotting 15 index components (a) and orthogonality evaluation of the established offline 2D LC system based on the asterisk equations using 65 index components (b).

After column selection, in the next, the chromatographic conditions for each dimension separation were optimized individually. For the ¹D HILIC separation, the composition of mobile phase is a key parameter that might exert significant influence on the chromatographic separation. Formic acid (FA, 0.1%, v/v) was often used as the additive to obtain symmetric peaks in analysis of the QCG and FOG components of C. tinctorius either in HILIC or RPLC mode.^13,25,27,28 Ammonium formate (AF) or ammonium acetate (AA) (10 mM) was a preferable additive in the HILIC separation mode.^29,30 Here the effects by addition of FA (0.1%) and AF (5 Mm, a medium concentration) in the mobile phase were examined and compared (Fig. S3†). It was inferred that by use of the mobile phase containing 5 mM AF, the most peaks of the water extract of C. tinctorius could be separated. To avoid the additional desalination work, the concentration of AF added was reduced (2 mM and 1 mM). The addition of 2 mM of AF could enable a better resolution of the polar components, compared with 1 mM of AF (Fig. S4†). We thus selected the binary mobile phase of CH₃CN and 2 mM AF for the ¹D HILIC separation. Good precision was obtained under the ¹D HILIC chromatographic condition when the injection volume was increased to 20 μL (Fig. S5†). On the other hand, for the ²D RPLC separation, CH₃CN-0.1%FA was employed as the mobile phase in that good peak shape and selectivity could be obtained as demonstrated in previous studies.^24,25,28 Additionally, reduction of the column temperature (from 35 °C to 25 °C) could slightly enhance the retention of C. tinctorius components on the Shield RP-18 column (Fig. S6†), but increased the back-pressure of the system. We thereby set the column temperature at 30 °C. Finally, desirable chromatographic condition for the ²D RPLC separation was obtained after careful adjustment of the gradient elution program.

Peak capacity and orthogonality evaluation

The established 2D LC system showed distinct improvement on the peak capacity in contrast to that by the conventional single RPLC separation. The theoretical peak capacity for the ¹D and ²D separation was calculated as 89 and 86 according to the total separation time and the average base-line width of the peaks eluted at the beginning, the middle, and the end of the spectrum.³¹ Hence the peak capacity of the 2D LC system in theory reached 7654, indicative of an expansion by tens of folds. The significantly improved peak capacity facilitated the exposure and resolution of a total of 163 QCG compounds from twelve fractionated samples of C. tinctorius, a six-fold expansion compared to our previous report by one dimensional RP-UHPLC separation.⁶

The asterisk equations were used to assess the orthogonality of the established 2D LC system.²³ To obtain a constant orthogonality value, a total of 65 components from C. tinctorius were used as the index for orthogonality evaluation (Fig. 2b). Accordingly, the spreading of these 65 peaks around Z₋, Z₊, Z₁, and Z₂, were calculated to be 98%, 69%, 77%, and 96%, respectively. The orthogonality of the 2D LC system was 71%, suggesting a strong resolution ability of the 2D LC system on C. tinctorius components. The calculated Z values in four different directions indicated that the peaks were equidistributed around Z₋ and Z₂, but slightly gathered around Z₊ and Z₁. Despite several different algorithms used for orthogonality evaluation have been reported, such as linear regression coefficient (R²), Pearson's correlation coefficient,³² bins-counting,¹⁴ nearest-neighbor distance,³³ and effective area distribution methods,³⁴ the asterisk equations approach used in this study not only is convenient to assess the separation difference between two dimensional separations, but also intuitively displays the peaks spreading in four different directions.

It is noted that, an offline RPLC × HILIC 2D LC system has been reported and used to separate the components of C. tinctorius.¹³ In contrast, the 2D LC system established in this study showed higher orthogonality (71% versus 64.81%), higher theoretical peak capacity (7654 versus 2850), and higher analysis efficiency (18 min versus 35 min). Additionally, application of RPLC as the ²D chromatography coupled with MS provided the retention characteristics as an important evidence that could be compared with the results in previous studies, thus rendering the subsequent structural elucidation easier to accomplish.

The CID and HCD features of four QCG reference standards

The CID and HCD features of four QCG reference standards (monomers: HSYA, ISC, and SC; dimers: AnHSYB) were first investigated, aimed to outline the characteristic fragmentation pathways and DPIs that facilitate the characterization of unknown QCGs from twelve fractionated samples of C. tinctorius. As displayed in Fig. 3, four QCG reference standards all exhibited a characteristic UV absorption bond at 400–410 nm and gave a common and abundant product ion at m/z 119.05 (C₈H₇O⁻) in the HCD-MS/MS spectra. Comparatively, this characteristic product ion was very weak in the CID-MS³ spectra of three monomeric QCGs (HSYA: m/z 611.16 > 325.07; ISC/SC: m/z 613.16 > 361.11) and absent for the dimeric AnHSYB (m/z 1043.26 > 923.23) due to the low-mass 1/3 cutoff.¹¹ The UV absorption around 405 nm and the product ion of m/z 119.05 obtained in the HCD spectra thereby can be diagnostic for a primary differentiation of the target QCG homologs from other chemicals present in C. tinctorius.

Fig. 3 The PDA spectra (190–600 nm), HCD and CID spectra of four QCG reference standards, showing the characteristic UV absorption around 405 nm and characteristic product ion of m/z 119.05 in the HCD-MS/MS spectra. The CID spectra refer to the MS³ spectra of m/z 611.16 > 325.07, m/z 613.16 > 361.11, and m/z 1043.26 > 923.23, for HSYA, ISC/SC, and AnHSYB, respectively.

The negative ion mode CID and HCD of the deprotonated QCG precursor ions produced differentiated MSⁿ (n = 2, 3) spectra and thus offered complementary fragments information useful for the characterization of QCG structures. CID of the precursor ion of HSYA (m/z 611.16, RDB 16.5) generated the MS² product ions of m/z 491.12 (^0,2X⁶₀ or ^0,2X⁴₀, a classic fragmentation pathway for C-glycosides),³⁵ 403.10 (–[Glc + H₂O + CO], 208 Da), and 325.07 (^0,2X⁶₀^0,2X⁴₀-CO-H₂O) in high abundance (Fig. S7†), together with the weak aglycone ion observed at m/z 283.07. Further CID-MS³ of the base peak ion of m/z 491.12 gave the MS³ product ions at m/z 473.11, 455.10 due to the successive neutral loss of H₂O, and rich aglycone ion. Moreover, the product ions of m/z 219.03, 205.01, 179.04, and 119.05, were obtained by CID of the 2nd most abundant MS² product ion m/z 325.07. On the other hand, HCD of the HSYA precursors using an NCE of 90% generated the MS² product ions at m/z 353.07, 325.07, 283.06, 205.01, and 119.05. Particularly, the characteristic product ion of m/z 119.05 was in a very high intensity. Based on the product ions information, the fragmentation pathways for HSYA by CID-MS³ were proposed as shown in Fig. S7.† Almost identical fragmentation pathways were observed for CID and HCD of the precursor ions of ISC and SC owing to their similar structures. However, they were quite different from HSYA because of the discriminated 6-substituent. In the case of ISC, CID-MS² of the precursor ions (m/z 613.16) was featured by the neutral elimination of H₂O (m/z 595.15), H₂O + CO₂ (m/z 551.16), and Glc + CO (m/z 361.11) (Fig. S8†). Further CID-MS³ of the ion of m/z 551.16 produced the product ions at m/z 533.15 (–H₂O), 431.10 (–C₈H₈O), 361.11 (–[Glc + CO]), and 241.05 (–[Glc + CO + C₈H₈O]). Similar to HSYA, the HCD of ISC could yield low-mass product ions detected at m/z 361.11, 241.05, and 119.05. Fig. S8† illustrates the possible fragmentation pathways for ISC proposed by analysis of the CID-MS³ fragmentation information. For the dimeric QCG AnHSYB (m/z 1143.27, RDB 23.5), only three MS² product ions, including m/z 1025.26 (–H₂O), 923.23 (−120 Da due to the ^0,2X₀ cleavage of a Glc residue), and 449.11, were produced by CID-MS/MS. The product ion at m/z 449.11 and another weak MS² product ion at m/z 593.15 should be two complementary product ions by the fragmentation of a chemical bond between C-2′ and the C-G′′1 (Fig. S9†), which could offer vital information for characterizing two subsections that compose a QCG dimer. Subsequent CID of the 1st most abundant MS² ion at m/z 1025.26 yielded the product ions owing to the neutral elimination of H₂O (m/z 1007.25), homolytic (−181 Da, m/z 862.20) and heterolytic (−180 Da, m/z 863.20) eliminations of Glc plus H₂O.³⁶ CID-MS³ of the 2nd most abundant MS² product ion of m/z 923.23 could yield the aglycone ion at m/z 287.06. However, the characteristic product ion of m/z 119.05 failed to be detected by the CID-MS³ fragmentation (Fig. 3). This insufficiency can be overcome by means of HCD, by which, as the same as HSYA, low-mass product ions, comprising m/z 461.05, 449.11, 299.06, and 119.05, were obtained. Among them, the product ion of m/z 119.05 was the most intense. Therefore, the HCD of QCG dimers was able to offer rich and complementary fragment information for the structural elucidation of QCG structures. The possible fragmentation pathways, based on both the CID-MS³ and HCD-MS/MS information, for AnHSYB were proposed as shown in Fig. S9.†

Systematic characterization of the QCQ components from C. tinctorius

The ¹D fractionation of the total water extract of C. tinctorius using an Acchrom XAmide HILIC column yielded twelve samples. These samples were further analyzed by reversed-phase UHPLC/LTQ-Orbitrap MS using HCD and CID two fragmentation mechanisms in the negative ion mode. As exhibited in Fig. 4, numerous minor peaks were separated, exposed, and thus detected, because of the excellent fractionation performance of the developed ¹D chromatography. Subsequently, combined data mining strategies, comprising CPIF of m/z 119.05 in the HCD spectrum, typical UV absorption around 405 nm, RDB setting (10–30), and the preferable ^0,2X₀ cleavage of Glc, were employed to distinguish the target QCGs from FOGs. Structural elucidation of these observed QCG homologs was based on the combined analyses of the high-accuracy MS information regarding the precursor ions and the product ions obtained by CID-MS³ and HCD-MS/MS fragmentations. The characterization results were further compared with the known QCG compounds (Fig. S1†) that have been isolated from C. tinctorius. As a result, a total of 163 QCG compounds were identified or tentatively characterized from twelve fractionated samples of C. tinctorius, which comprised 113 monomers and 50 dimers (Table S1†). And 149 are considered as the potential new QCG compounds. It is noted that nine dimers of QCG–FOG were discovered for the first time. The characterization of three representative QCG components (69, 161, and 128) that represent different subclasses was described in detail.

Fig. 4 The UV spectrum (280 nm) of the C. tinctorius extract by the first dimensional HILIC separation and the corresponding base peak chromatograms obtained on an LTQ-Orbitrap Velos Pro hybrid mass spectrometer.

QCG monomers. A total of 113 monomeric QCGs (M.W. < 900; 10 < RDB < 20) were characterized from C. tinctorius, which were featured by the prevalent ^0,2X₀ cleavage of Glc and a weak aglycone ion of 287.06 (C₁₅H₁₁O₆⁻) or 299.06 (C₁₆H₁₁O₆⁻). For convenient description of the QCG structures, the aglycone skeletons with the elemental composition of C₁₅H₁₂O₆ and C₁₆H₁₂O₆ were defined as X and Y, respectively. Additionally, 21 N-bearing QCG monomers were characterized, and some of them shared a common framework of C₁₅H₁₃NO₅, designated as Z. Compound 69 (t_R 10.57 min, Fr. 12) giving the deprotonated precursor ion at m/z 449.1093 was used as an example to illustrate how the QCG monomers were characterized on the basis of the CID-MS³ fragmentation data (Fig. 5a). By CID-MS², the precursor ions were easily fragmented into the product ions of m/z 431.10 ([M − H₂O − H]⁻), 329.05 ([M − C₈H₈O − H]⁻), 299.06 ([M − C₅H₁₀O₅ − H]⁻), 287.06 ([M − C₆H₁₀O₅ − H]⁻), 207.05 ([M − C₁₄H₁₀O₄ − H]⁻), and 153.02 (C₇H₅O₄⁻). Among them, the concomitant ions at m/z 299.06 and 287.06 that had the elemental compositions different in a carbon could help infer the 6-substituent as a fructose–CH₂– of hydrosafflor yellow B (HSYB) or C (HSYC) or a Glc residue of HSYA,³⁷ which could not be clearly differed due to the lack of HSYB or HSYC as the reference standards. We here tentatively characterized the aglycone of 69 as Y (C₁₂H₁₂O₆). CID-MS³ of the dehydrated MS² product ion (m/z 431.10) underwent the similar fragmentation pathways to those of the precursor ions, yielding the major product ions at m/z 413.09 ([M − 2H₂O − H]⁻), 311.06 ([M − H₂O − C₈H₈O − H]⁻), 299.06, and 281.05 ([M − H₂O − C₅H₁₀O₅ − H]⁻). The genuine aglycone ion of m/z 299.06 could be further dissociated by neutral elimination of H₂O (m/z 271.06), CO₂ (m/z 255.07), and p-hydroxystyrene (m/z 179.00). Based on these evidences, compound 69 was tentatively characterized as 4-deglucosyl-HSYB or 4-deglucosyl-HSYC (Y + C₅H₁₀O₅), an unknown QCG compound.

Fig. 5 Illustration of the characterization of compounds 69 and 161 based on the obtained ESI-CID-MS³ fragmentation information.

QCG dimers. A total of 50 dimeric QCGs (720 < M.W. < 1300, 17 < RDB < 27) were characterized from C. tinctorius, which were featured by two complementary residual product ions. Structural characterization of the dimers containing two QCG residues was exemplified by use of compound 161 (t_R 15.16 min, Fr. 9) (Fig. 5b). The CID-MS² of the precursor ion (m/z 955.21) generated two dominant product ions at m/z 505.10 (C₂₃H₂₁O₁₃⁻) and 449.11 (C₂₁H₂₁O₁₁⁻), together with two weak ions of m/z 315.05 and 297.04. To be convenient in elucidation of the fragmentation pathways, the obtained two residues corresponding to the product ions of m/z 505.10 and 449.11 were named as L and R, respectively (Fig. 5b). The elemental composition difference between the residues of L and R was C₂O₂. CID-MS³ of the L fragment yielded the product ions at m/z 487.09 (–H₂O, C₂₃H₁₉O₁₂⁻), 461.11 (–CO₂, C₂₂H₂₁O₁₁⁻), 443.10 (–[CO₂ + H₂O], C₂₂H₁₉O₁₀⁻), 341.05 (–C₈H₈O, C₁₄H₁₃O₁₀⁻), 315.05 (−190 Da, C₁₆H₁₁O₇⁻), 297.04 (−208 Da, C₁₆H₉O₆⁻), 271.06 (−234 Da, C₁₅H₁₁O₅⁻), and 243.07 (−262 Da, C₁₄H₁₁O₄⁻). In addition, the product ions, consistent with those obtained by CID of compound 69, including m/z 431.10 (C₂₁H₁₉O₁₀⁻), 329.05 (C₁₃H₁₃O₁₀⁻), 299.06 (C₁₆H₁₁O₆⁻), 287.06 (C₁₅H₁₁O₆⁻), 261.06 (C₁₀H₁₃O₈⁻), 207.05 (C₇H₁₁O₇⁻), and 153.02 (C₇H₅O₄⁻), were obtained by CID-MS³ of the R residue. We could herein infer that 161 possessed the same or similar R residue as 69. By a survey of the known QCG structures, compound 161 was finally characterized as precarthamin, a QCG dimer ever isolated from the florets of C. tinctorius.³⁸ The observed aforementioned fragmentation pathways can be reasonably assigned to precarthamin as shown in Fig. 5b.

Dimers of QCG–FOG. The most exciting discovery in this study was the putative characterization of nine dimers that encompass both a QCG and a FOG skeletons, which have not been isolated from C. tinctorius hitherto, to the best of our knowledge. Aside from the aforementioned UV and MSⁿ features typical for QCGs, these novel dimers exhibited sequential neutral elimination of entire sugar residues (Glc and Rha) and concomitant product ions of [Y₀ − H]⁻/[Y₀ − 2H]⁻/[Y₀ − 3H]⁻ in the HCD spectra, diagnostic for the characterization of O-glycosyl flavonols.^36,39 Here, compound 128 (t_R 12.13 min, Fr. 11) was used as an example to illustrate the characterization of the dimers of QCG–FOG. Clearly, we could confirm that 128 contained the QCG moiety because of the typical UV absorption at 411 nm (Fig. 6b), rich product ion of m/z 119.05 in the HCD spectrum (Fig. 6c), and a RDB value of 25.5. By CID, the precursor ions of m/z 1131.25 (C₅₀H₅₁O₃₀⁻) could generate the product ions at m/z 1113.24 and 505.10 in high abundance, and m/z 1059.25, 681.13, 637.14, 625.14, and 461.11 in low intensity (Fig. 6d). Amongst them, the product ions at m/z 505.10 ([C₂₃H₂₁O₁₃]⁻) and 625.14 ([C₂₇H₂₉O₁₇]⁻) might represent the QCG and FOG moieties, respectively. CID-MS³ of the ion corresponding to the QCG moiety (m/z 505.10) yielded the product ions at m/z 487.09 (–H₂O), 461.11 (–CO₂), 443.09 (–[CO₂ + H₂O]), 315.05 (−190 Da), 297.04 (−208 Da), and 271.06 (−234 Da). These fragmentation pathways were similar to the CID features of the L residue of 161 as aforementioned. We could thus putatively characterize the QCG moiety as C₁₇H₁₂O₈ + C₆H₁₀O₅. On the other hand, by HCD, the low-mass product ions, comprising m/z 625.14, 463.09, 301.04, 299.02, 271.03, and 151.00, were obtained and considered to be related to the FOG moiety. The fragment ions at m/z 301.04 (C₁₅H₉O₇⁻), 299.02 (C₁₅H₇O₇⁻), 271.03 (C₁₄H₇O₆⁻), and 151.00 (C₇H₃O₄⁻), could help tentatively characterize the aglycone of the FOG moiety as quercetin (a very common FOG aglycone for C. tinctorius),² and these fragments were assigned as [Y₀ − H]⁻/[Y₀ − 3H]⁻, [Y₀ − H − CH₂O]⁻, and ^1,3A, respectively.^40,41 From the transition m/z 625.14 > 463.09 > 301.04 discerned from the HCD spectrum, we could speculate the FOG moiety of 128 contains two Glc residues. Thereby, the FOG moiety was characterized as quercetin-Glc-Glc. Taken together, compound 128 (C₅₀H₅₂O₃₀) was putatively identified as (C₁₇H₁₂O₈-Glc) + (quercetin-Glc-Glc) (Fig. 6a). By searching the on-line SciFinder database by chemical formula, it was an unknown compound. Detailed information with respect to the other eight QCG dimers of this category (25, 95, 129, 139, 141, 143, 146, and 148) was given in Table S1.† As a continuous study, we shall perform a target phytochemical isolation to verify these novel QCG molecules and test their bioactivities.

Fig. 6 Illustration for the characterization of a representative dimer containing both a QCG and a FOG frameworks (128). (a) Schematic diagram for the multi-stage mass spectrometric fragmentation on a speculated structure; (b) the PDA spectrum; (c) the HCD-MS/MS spectra; (d) the CID-MS³ spectra.

Conclusion

Despite QCGs have been demonstrated to be relevant to the blood-invigorating effect of C. tinctorius, a holistic approach that facilitates the systematic exposure and rapid structural elucidation of the contained QCG homologs is not available hitherto. It is the low content and the interference from the other FOGs that renders their systematic characterization rather difficult to accomplish by means of the conventional ¹D RPLC separation. In the present study, we developed an integral approach that integrated the offline comprehensive 2D LC and LTQ-Orbitrap MS coupled with data mining techniques, as a solution to solving the difficulty encountered in resolution and comprehensive characterization of the QCG homologs from C. tinctorius. An orthogonal 2D LC system by combining HILIC × RPLC enabled a desirable separation, whilst the simultaneous use of two different fragmentation mechanisms (HCD and CID) gave rich and complementary fragments information. As established using four reference standards, the CPIF of m/z 119.05 in the HCD spectrum, RDB screening (10–30), UV absorption around 405 nm, and characteristic ^0,2X₀ cleavage of Glc, were diagnostic for discriminating QCGs from FOGs. Further combined analyses of the high-accuracy CID-MS³ and HCD-MS/MS data, ultimately, led to a primary characterization of 163 QCG compounds from C. tinctorius, of which 149 are potential new ones. Notably, due to the potent separation capacity of the established 2D LC system and also the sensitive detection ability by the LTQ-Orbitrap MS, nine dimers containing both a QCG and FOG frameworks were first reported from this plant. In conclusion, the first systematic analysis of the QCG homologs in C. tinctorius was accomplished, and the obtained results would benefit an in-depth understanding of the structural diversity of QCGs in C. tinctorius. Our continuous study will also focus on the contained FOGs, aimed to fully elucidate the therapeutic basis of C. tinctorius as a gynecological herbal medicine.

Conflict of interest

The authors declare no conflict of interest.

Acknowledgements

We gratefully acknowledge the financial support from the National Natural Science Foundation of China (No. 81503240), the Twelfth Five-Year National Science & Technology Support Program (2012BAI29B06), and the National Science and Technology Major Project for Major Drug Development (2013ZX09508104 and 2014ZX09304-307-001-007).

References

1. S. J. Yue, Y. P. Tang, S. J. Li and J. A. Duan, Molecules, 2013, 18, 15220–15254.
2. X. D. Zhou, L. Y. Tang, Y. L. Xu, G. H. Zhou and Z. J. Wang, J. Ethnopharmacol., 2014, 151, 27–43.
3. K. Kazuma, T. Takahashi, K. Sato, H. Takeuchi, T. Matsumoto and T. Okuno, Biosci., Biotechnol., Biochem., 2000, 64, 1588–1599.
4. T. Kanehira and K. Saito, Food Sci. Technol., 2001, 34, 55–59.
5. L. Fan, R. Pu, H. Y. Zhao, X. Liu, C. Ma, B. R. Wang and D. A. Guo, J. Chin. Pharm. Sci., 2011, 20, 47–56.
6. W. Si, W. Z. Yang, D. A. Guo, J. Wu, J. X. Zhang, S. Qiu, C. L. Yao, Y. J. Cui and W. Y. Wu, J. Pharm. Biomed. Anal., 2016, 117, 510–521.
7. S. Y. Wang, L. Z. Qiao, X. Z. Shi, C. X. Hu, H. W. Kong and G. W. Xu, Anal. Bioanal. Chem., 2015, 407, 331–341.
8. X. Qiao, W. Song, S. Ji, Y. J. Li, Y. Wang, R. Li, R. An, D. A. Guo and M. Ye, J. Chromatogr. A, 2014, 1362, 157–167.
9. X. Qiao, W. Song, S. Ji, Q. Wang, D. A. Guo and M. Ye, J. Chromatogr. A, 2015, 1402, 36–45.
10. C. L. Yao, W. Z. Yang, W. Y. Wu, J. Da, J. J. Hou, J. X. Zhang, Y. H. Zhang, Y. Jin, M. Yang, B. H. Jiang, X. Liu and D. A. Guo, J. Chromatogr. A, 2015, 1402, 71–81.
11. S. Qiu, W. Z. Yang, X. J. Shi, C. L. Yao, M. Yang, W. Y. Wu and D. A. Guo, Anal. Chim. Acta, 2015, 893, 65–76.
12. F. Bedani, P. J. Schoenmakers and H. G. Janssen, J. Sep. Sci., 2012, 35, 1697–1711.
13. Y. M. Liu, X. Y. Xue, Z. M. Guo, Q. Xu, F. F. Zhang and X. M. Liang, J. Chromatogr. A, 2008, 1208, 133–140.
14. Q. Q. Xing, T. Liang, G. B. Shen, X. L. Wang, Y. Jin and X. M. Liang, Analyst, 2012, 137, 2239–2249.
15. H. Q. Pan, W. Z. Yang, Y. B. Zhang, M. Yang, R. H. Feng, W. Y. Wu and D. A. Guo, Anal. Bioanal. Chem., 2015, 407, 6057–6070.
16. J. X. Zhang, S. H. Guan, J. H. Sun, T. Liu, P. Chen, R. H. Feng, X. Chen, W. Y. Wu, M. Yang and D. A. Guo, Anal. Bioanal. Chem., 2015, 407, 581–595.
17. X. H. Qiu, J. Zhang, Z. H. Huang, D. Y. Zhu and W. Xu, J. Chromatogr. A, 2013, 1292, 121–131.
18. C. Przybylski and V. Bonnet, Rapid Commun. Mass Spectrom., 2013, 27, 75–87.
19. W. Z. Yang, W. Y. Wu and D. A. Guo, Chin. J. Nat. Med., 2015, 13, 867–872.
20. J. Y. Zhang, F. Wang, H. Zhang, J. Q. Lu and Y. J. Qiao, Phytochem. Anal., 2014, 25, 405–414.
21. J. Y. Zhang, Z. J. Wang, Q. Zhang, F. Wang, Q. Ma, Z. Z. Lin, J. Q. Lu and Y. J. Qiao, Talanta, 2014, 124, 111–122.
22. R. Cho, Y. Y. Huang, J. C. Schwartz, Y. Chen and T. J. Carlson, J. Am. Soc. Mass Spectrom., 2012, 23, 880–888.
23. M. Camenzuli and P. J. Schoenmakers, Anal. Chim. Acta, 2014, 838, 93–101.
24. Y. Jin, X. L. Zhang, H. Shi, Y. S. Xiao, Y. X. Ke, X. Y. Xue, F. W. Zhang and X. M. Liang, Rapid Commun. Mass Spectrom., 2008, 22, 1275–1287.
25. Y. Jin, Y. S. Xiao, F. F. Zhang, X. Y. Xue, Q. Xu and X. M. Liang, J. Pharm. Biomed. Anal., 2008, 46, 418–430.
26. L. Fan, H. Y. Zhao, M. Xu, L. Zhou, H. Guo, J. Han, B. R. Wang and D. A. Guo, J. Chromatogr. A, 2009, 1216, 2063–2070.
27. B. Hong, Z. Wang, T. J. Xu, C. C. Li and W. J. Li, J. Pharm. Biomed. Anal., 2015, 107, 464–472.
28. J. F. Chen, Y. L. Song, X. Y. Guo, P. F. Tu and Y. Jiang, Analyst, 2014, 239, 6474–6485.
29. Y. L. Song, N. Zhang, S. P. Shi, J. Li, Q. Zhang, Y. F. Zhao, Y. Jiang and P. F. Tu, J. Chromatogr. A, 2015, 1407, 106–118.
30. L. Yang, L. L. Chen, S. J. Xu, X. Zeng, Y. Feng and P. S. Xie, Anal. Methods, 2013, 5, 2093–2097.
31. D. X. Li, R. Dück and O. J. Schmitz, J. Chromatogr. A, 2014, 1358, 128–135.
32. E. van Gyseghem, I. Crosiers, S. Gourvenec, D. L. Massart and Y. Vander Heyden, J. Chromatogr. A, 2004, 1026, 117–128.
33. W. Nowik, S. Héron, M. Bonose, M. Nowik and A. Tchapla, Anal. Chem., 2013, 85, 9449–9458.
34. R. Dück, H. Sonderfeld and O. J. Schmitz, J. Chromatogr. A, 2012, 1245, 69–75.
35. H. J. Zhang and X. P. Yang, J. Pharm. Biomed. Anal., 2009, 49, 843–847.
36. W. Z. Yang, X. Qiao, T. Bo, Q. Wang, D. A. Guo and M. Ye, Rapid Commun. Mass Spectrom., 2014, 28, 385–395.
37. S. J. Yue, Y. P. Tang, C. M. Xu, S. J. Li and Y. Zhu, Int. J. Mol. Sci., 2014, 15, 16760–16771.
38. Y. Takahashi, M. Wada and K. Saito, Acta Soc. Bot. Pol., 1984, 53, 187–198.
39. K. Ablajan, Z. Abliz, X. Y. Shang, J. M. He, R. P. Zhang and J. G. Shi, Rapid Commun. Mass Spectrom., 2008, 41, 352–360.
40. W. Z. Yang, M. Ye, X. Qiao, Q. Wang, T. Bo and D. A. Guo, Eur. J. Mass Spectrom., 2012, 18, 493–503.
41. M. Ye, Y. N. Yan and D. A. Guo, Rapid Commun. Mass Spectrom., 2005, 19, 1469–1484.

---

Footnotes

† Electronic supplementary information (ESI) available. See DOI: 10.1039/c5ra23744k

‡ These authors contributed equally to this work.

---