Configurational Isomer Differentiation Between 20(S)‐ and 20(R)‐Ginsenosides in Red Ginseng via Applying [Cu^II(l‐Phe)₂−H]⁺ to Post‐Column Infusion LC–MS/MS

ABSTRACT

Introduction

20(R)‐ginsenosides are formed by Walden inversion during ginseng processing and usually exhibit almost identical MS/MS behaviors, yet display distinct pharmacological activities compared with 20(S)‐configurational isomers. It is thereby crucial for isomer differentiation to facilitate quality control.

Objective

To pursue a program enabling 20(R)‐ and 20(S)‐ginsenosides discrimination in complicated matrices using MS/MS, red ginseng, the processed form of ginseng, was deployed as a representative.

Methods

Five pairs of ginsenoside epimers were collected for method development. MS¹ and MS² spectra of every trimeric complex ion formed by complexation of ginsenoside, l‐Phe, and Cu²⁺ were recorded using direct infusion (DI)–MS/MS, and the abundance ratio between the two dimeric fragment ions was termed R. Moreover, energy‐resolved MS (ER‐MS) was deployed to monitor fragmentation trajectories of precursor and concerned fragment ions and to determine the R feature at optimal collision energy, denoted R ^max. To validate the applicability for differentiating epimers in complicated matrices, e.g., red ginseng extract, post‐column infusion (PCI) was introduced to configure PCI‐LC–ER‐MS.

Results

20(R)‐ vs. 20(S)‐Rg3 as representatives, [Cuᴵᴵ(l‐Phe)₂(20(R)‐/20(S))‐Rg3)−H]⁺ complex ions were observed and R values of the dimeric fragment ions such as [Cuᴵᴵ(l‐Phe)(20(R)‐/20(S))‐Rg3)−H]⁺ and [Cuᴵᴵ(l‐Phe)₂−H]⁺, exhibited differences. R ^max features obtained by DI–ER‐MS showed significant differences when comparing other 20(R)‐ and 20(S)‐configurational isomers, leading to a reliable way for epimer discrimination. The reproducibility of recognition capability on PCI‐LC–ER‐MS platform was justified and applied to epimer differentiation in red ginseng extract. Successes were reached for the differentiation of two pairs of epimers, 20(R)‐ and 20(S)‐Rg3, and 20(R)‐ and 20(S)‐Rh1.

Conclusion

In summary, the proposed strategy combined the complexation driven by [Cu^II(l‐Phe)₂−H]⁺ with PCI‐LC–ER‐MS enables the differentiation of 20(R)‐ and 20(S)‐configurational ginsenoside isomers in complicated matrices.

Short abstract

A novel strategy combining the complexation driven by [Cu^II(l‐Phe)₂−H]⁺ with PCI‐LC–ER‐MS was proposed to enable online differentiation of 20(R)‐ and 20(S)‐configurational ginsenoside in complicated matrices, and it has been successfully applied to epimer differentiation in red ginseng extracts.

1. Introduction

Ginsenosides are a class of triterpenoid glycosides distributed in Panax plants [1], such as P. ginseng , P. notoginseng, and P. quinquefolium. Ginsenosides have been demonstrated as the primary bioactive constituents of these medicinal plants and thereby serve as Q‐markers for the quality assessment. Consequently, structural annotation of ginsenosides has consistently occupied the hot spot in medicinal plant research [2, 3]. From a structural standpoint, ginsenosides comprise a lipophilic scaffold, predominantly of dammarane‐type, and hydrophilic substituents. The widespread occurrence of structural analogs and isomers governs diverse biological activities [4, 5, 6], yet also limits the confidence‐enhanced structural identification. An advanced triple three‐dimensional (3D)‐MS strategy previously proposed by our research group [7] enabled confirmative identification of most ginsenosides, particularly facilitating positional isomeric differentiation through comprehensive acquisition of MS/MS behavior. During processing (e.g., steaming to produce red ginseng), a given 20(S)‐configurational ginsenoside usually undergoes Walden inversion to yield the 20(R)‐configurational isomer, resulting in the coexistence of both epimers in the processed samples [8, 9, 10]. Interestingly, the subtle configurational difference at C‐20 markedly influences the spatial orientation, intermolecular interactions, and binding patterns with biomolecules, thereby leading to distinct pharmacological activities and metabolic fates [11, 12, 13, 14, 15]. However, the aforementioned triple 3D‐MS strategy fell short in distinguishing ginsenoside epimers. Therefore, the development of a method being capable of rapidly and accurately identifying 20(R)‐ and 20(S)‐configurational ginsenosides holds considerable scientific and practical values for elucidating their structure–activity relationships, guiding drug development and quality assessment.

NMR analysis is the most reliable method for identifying unknown compounds. However, due to the low abundance of 20(R)‐configurational rare ginsenosides, traditional isolation and NMR detection are insufficiently efficient [16]. The commonly used approach for epimer discrimination involves physical separation via LC or ion mobility spectroscopy, followed by comparison retention times or drift times with the available authentic compounds [17, 18]. Additionally, machine learning [19, 20, 21] and indirect competitive enzyme‐linked immunosorbent assays [22] have also been deployed to distinguish epimers. Tandem MS (MS/MS) has become the most popular tool for isomer characterization due to its superior selectivity, sensitivity, and high‐throughput advantages [23, 24]. Successful distinction of 20(R)‐ and 20(S)‐ginsenoside Rh1 was achieved using the abundance ratio of fragment ions [M − 2H₂O + H]⁺ and [M − H₂O + H]⁺ in the positive ionization polarity [25]. Metal ion‐driven complexation has emerged as a powerful strategy for MS/MS‐based isomers, even enantiomers identification [26, 27]. Metal ions, i.e., Cu²⁺, Ca²⁺, possess vacant electron orbitals and coordinate selectively with O and N atoms in analytes. The bond strength of coordination bonds depends on the metal ion type, electron donor type, and spatial proximity, and can even rival that of covalent bonds. Consequently, coordination bonds can survive in collision‐induced dissociation (CID), persisting in fragment ions. Crucially, metal chelation magnifies subtle structural differences between isomers, resulting in relative abundance variations between ions [28, 29]. However, before applying to complicated matrices such as herbal extracts, it is essential to properly address two issues, including matrix‐induced ionization suppression and reproducibility of abundance variations.

To achieve efficient formation and MS/MS detection of metal ion‐analyte complexes in complicated matrices, our group previously established a robust post‐column infusion (PCI)‐LC–MS/MS technique [30]. Analytes in complicated matrices were first separated by LC, and then the chiral reference (CR) solution was continuously introduced into the eluate via an external syringe pump. The targeted analytes and CR formed complexes post‐LC separation but prior to MS ionization. Subsequently, various ion transitions such as precursor ion → precursor ion and precursor ion → fragment ions were constructed based on the characteristic fragment ions generated from the complexes, and the intensities were acquired at a specific collision energy (ce) range. Fortunately, a powerful program, namely, online energy‐resolved MS (ER‐MS) has been stably developed to meet the above requirements [31]. By acquiring the sigmoidal breakdown graph of the precursor ion residue and the Gaussian breakdown graph of all fragment ions, the full CE ramp‐MS² (FceR‐MS²) spectrum [32] comprehensively reveals the dissociation behaviors of the precursor ion as a function of ce. In addition to the m/z of the precursor and fragment ions, the FceR‐MS² spectrum also provides key information: RII_max, OCE, and ce ₅₀, defined as the maximum relative ion intensity (RII) of each fragment ion, CE corresponding to RII_max, and CE at which 50% of the precursor ion survives, respectively.

Here, we attempt to pursue a strategy for differentiating ginsenoside epimers by exploiting differences in MS fragmentation behavior of non‐covalent coordination complexes. Five pairs of epimers were chosen for method development. Online complexation was achieved by continuously importing CS through PCI. The dissociation behaviors of resulting complex ions were universally tracked via ER‐MS, with the differential abundance ratio of fragment ions facilitating the epimers' differentiation. The integrated strategy, PCI‐LC–ER‐MS, was thereafter applied to the characterization of 20(R)‐ and 20(S)‐configurational ginsenosides in red ginseng. Following LC separation, 51 compounds were detected in red ginseng, encompassing 42 ginsenosides. Two pairs of epimers were successfully distinguished by PCI‐LC–ER‐MS. The obtained findings are envisioned to provide a robust MS/MS‐based method for distinguishing ginsenoside epimers, and to offer insights for the structural identification in complicated matrices.

2. Experimental

2.1. Chemicals and Materials

Five pairs of ginsenoside epimers (Figure S1), including 20(R)‐protopanaxadiol (No. DST230920‐330), 20(S)‐protopanaxadiol (No. JB313717), 20(R)‐protopanaxatriol (No. J07GB154009), 20(S)‐protopanaxatriol (No. J131B220125), 20(R)‐Rg3 (No. 24121269), 20(S)‐Rg3 (No. 24033111), 20(R)‐notoginsenoside R2 (No. JB241983), 20(S)‐notoginsenoside R2 (No. O11IB228417), 20(R)‐Rh1 (No. 24091453), and 20(S)‐Rh1 (No. M11GB141261), were obtained commercially from DeSiTe Biological Technology Co. Ltd. (Chengdu, China), Bethealth People Biomedical Technology Co. Ltd. (Beijing, China), and Yuanye Biotechnology Co. Ltd. (Shanghai, China). l‐Phe (No. N12HS200901) was supplied by Yuanye, and CuCl₂ was obtained from Acmec Biochemical Technology (Shanghai). Raw materials of steam‐processed P. ginseng , namely, red ginseng, were collected from the crude material market in Anguo (Hebei, China) and authenticated by Prof. Pengfei Tu from School of Pharmaceutical Sciences, Peking University.

LC–MS grade formic acid, methanol (MeOH), as well as acetonitrile (ACN) were supplied by Thermo‐Fisher (Pittsburgh, PA). Deionized water was prepared in‐house using a Millipore Milli‐Q purification apparatus (Bedford, MA).

2.2. Sample Preparation

A series of 10 mM stock solutions were prepared by dissolving accurately weighed authentic ginsenosides in an appropriate volume of MeOH or DMSO, depending on their solubility. CuCl₂ and l‐Phe were co‐dissolved in MeOH to yield the CR solution (200 μM). Two sets of test samples were prepared for DI–MS/MS, differing in whether ginsenosides were fortified with CR solution. Set A samples were obtained by 20‐fold dilution of each stock solution with MeOH. For Set B, stock solutions were first diluted 10‐fold with MeOH and then separately mixed with an equal volume of CR solution. Meanwhile, the five pairs of epimers were thoroughly pooled and diluted with MeOH to obtain a mixed standard solution with a final concentration of 100 μM for LC–MS/MS analysis.

After well pulverization, a 1.0 g aliquot of red ginseng was extracted with 50 mL of 50% aqueous MeOH for 30 min using an ultrasonication‐assisted manner. After centrifugation at 12, 000 rpm for 10 min, the supernatants were filtered through a 0.22 μm Nylon membrane to yield the extract solution.

2.3. DI–MS/MS and DI–ER‐MS/MS Measurements

Sets A and B samples participated in DI–MS/MS experiments on a QTOF‐MS instrument (SCIEX Triple TOF 6600+, Foster City, CA). Each relevant sample was introduced into the electrospray ionization (ESI) source via a syringe pump at a flow rate of 10 μL/min. Notably, Set A samples were analyzed in both positive and negative ion polarities, whereas Set B samples, fortified with CR solution were recorded in positive ion polarity only. The ion source settings were defined as follows: curtain gas (CUR), 20 psi; nebulizer gas (GS1), 20 psi; auxiliary gas (GS2), 15 psi; ion spray voltage, ±3.5 kV; temperature (TEM), 300°C; and declustering potential (DP), ±60 V. The full scan mode was used to record the MS¹ spectrum with a mass range of m/z 100–1200, and ce of ±5 eV was implemented. The product ion scan mode was employed to track the MS² spectrum of user‐defined precursor ion with a mass range of m/z 100–1200, and ce of ±35 eV was applied. SCIEX Analyst TF software was in charge of data acquisition and SCIEX PeakView 1.2 software administered data processing.

Set B samples took part in DI–ER‐MS experiments conducted on the a QTRAP‐MS platform (SCIEX QTRAP 5500, Foster City, CA). The flow rate of syringe pump, ESI parameters, and DP were consistent with those used in QTOF‐MS. ER‐MS was performed with multiple‐reaction monitoring (MRM) mode by tracing the breakdown graph of complex ion at different CEs. Taking the trimeric complex ion m/z 868.4785 [Cu^II(l‐Phe)₂(20(S)‐PPT)−H]⁺ formed between 20(S)‐PPT and CR solution as an example, two major fragment ions m/z 703.3965 [Cu^II(l‐Phe)(20(S)‐PPT)−H]⁺ and 392.0842 [Cu^II(l‐Phe)₂−H]⁺ were generated via CID. Thereafter, three ion transition candidates, including m/z 868.5 → 868.5, 868.5 → 703.4, and 868.5 → 392.1, were constructed. Subsequently, three sets of pseudo‐ion transitions (PITs), including m/z 868.5 → 868.500, 868.5 → 868.501, 868.5 → 868.502, etc., m/z 868.5 → 703.400, 868.5 → 703.401, 868.5 → 703.402, etc., and m/z 868.5 → 392.100, 868.5 → 392.101, 868.5 → 392.102, etc., were derived, corresponding to a series of step‐wise CEs (step‐size as 3 eV) ranging from 5 to 71 eV, e.g., 5, 8, and 11 eV. Analyst 1.6.3 software (SCIEX) was used for spectrum recording and data processing. The peak areas of three sets of PITs for each compound were imported into GraphPad Prism 7.0 (San Diego, CA) for normalization and breakdown graph fitting. Gaussian fitting was applied to generate the breakdown graph for precursor ion → fragment ion, for example, m/z 868.5 → 703.4 or 868.5 → 392.1, whereas precursor ion → precursor ion, such as m/z 868.5 → 868.5, was fitted using Sigmoid‐shaped curve. Each FCER‐MS² spectrum was ultimately constructed via gathering one Sigmoid‐shaped and two Gaussian‐shaped breakdown graphs. Several key structural descriptors, ce ₅₀, OCE, and RII_max, were extracted from FCER‐MS² spectrum to enable structural characterization.

2.4. Post‐Column Infusion‐LC (PCI‐LC)–ER‐MS Measurements

PCI‐LC–ER‐MS measurements were carried out using an Ultimate 3000 dual‐ternary LC equipment (Thermo Fisher Scientific, Waltham, MA) coupled to a QTRAP‐MS device. The mixed standard solution was separated on the CAPCELL CORE ADME column (2.1 × 150 mm, 2.7 μm, Osaka Soda, Japan) under gradient elution conditions: 0.1% aqueous formic acid (A) and ACN (B); 0–5 min, 5%–18% B; 5–5.5 min, 18%–31% B; 5.5–18 min, 31% B; 18–18.1 min, 31%–45% B; 18.1–31 min, 45% B; 31–31.1 min, 45%–95% B; 31.1–33 min, 95% B; 33–33.1 min, 95%–5% B; 33.1–38 min, 5% B; and total flow rate, 0.25 mL/min. The column oven was maintained at 30°C and the injection volume was set as 2 μL. An additional syringe pump was introduced to continuously deliver CR solution at 10 μL/min [29]. After rapid mixing, the combined stream entered the ESI interface. The parameters were set as follows: GS1, 45 psi; GS2, 45 psi; CUR, 30 psi; IS, 4.5 kV; TEM, 300°C; and DP, 60 V. The ion transitions were the same as those in DI–ER‐MS. The data acquisition was managed by Analyst 1.6.3 software. The procedure for breakdown graph fitting was identical to that in DI–ER‐MS. And to further validate the robustness of the PCI‐LC–ER‐MS method, the mixed standard solution was measured in six replicates using identical PCI‐LC and MS conditions.

2.5. LC–QTOF‐MS/MS and PCI‐LC–ER‐MS Measurements of Red Ginseng

The high‐resolution MS¹–MS² dataset of red ginseng was obtained under identically optimized LC and MS conditions. The well‐defined MS fragmentation patterns were employed to preliminarily deduce compound identities. Subsequently, more attention was focused on distinguishing epimers. By employing the same PCI‐LC–ER‐MS approach described in Section 2.4, breakdown graphs and FCER‐MS² spectra of the complexes, formed between the chromatographically separated red ginseng extract and the continuously infused CR solution, were obtained.

3. Results and Discussion

3.1. Differentiation of 20(R)‐ and 20(S)‐Ginsenoside Epimers

It is challenging to reliably distinguish ginsenoside epimers using conventional MS methods because they tend to exhibit similar MS/MS behaviors [29]. Therefore, we introduced a chiral MS approach [30], the core of which involves converting epimer discrimination into complex differentiation by introducing a CS (i.e., l‐Phe) in the presence of Cu²⁺.

MS¹ and MS² spectra of five pairs of epimers were recorded using DI–QTOF‐MS. Taking 20(R)‐ and 20(S)‐Rg3 as representatives, the primary precursor ions in MS¹ spectra appeared at m/z 785.51 [M + H]⁺, accompanied by in‐source CID‐derived fragments, such as m/z 767.50 [M + H − H₂O]⁺ and m/z 749.49 [M + H − 2H₂O]⁺ in the positive ion polarity. Upon CID, similar fragment ions were detected in MS² spectra of 20(R)‐ and 20(S)‐Rg3, including m/z 767.49, 749.48, 443.38, 425.37, and 407.36, assigned as [M + H − H₂O]⁺, [M + H − 2H₂O]⁺, [M + H − 2Glc − H₂O]⁺, [M + H − 2Glc − 2H₂O]⁺, and [M + H − 2Glc − 3H₂O]⁺, respectively (Figure 1A). As shown in Figure 1B, 20(R)‐Rg3 and 20(S)‐Rg3 also exhibit comparable MS/MS behaviors in negative ion polarity. Specifically, in MS² spectra of dominated precursor ions m/z 829.52 [M + HCOO]⁻, except for the fragment ion m/z 375.30 involving the side chain fission of scaffold, those featured fragment ions such as m/z 621.45, 459.39, 221.07, 179.06, and 161.05 were generated by glycosidic bond fission. Overall, DI–MS/MS spectra of ginsenoside epimers were essentially indistinguishable, rendering differentiation based on intrinsic fragmentation patterns unfeasible.

FIGURE 1.

“Head‐to‐tail” MS² spectrum comparison between 20(S)‐Rg3 (upper) and 20(R)‐Rg3 (lower). (A) MS² spectrum of m/z 785.5 [20(R)‐/20(S)‐Rg3 + H]⁺; (B) MS² spectrum of m/z 829.5 [20(R)‐/20(S)‐Rg3 + HCOO]⁻; and (C) MS² spectrum of complex ions m/z 1176.4998 [Cu^II(l‐Phe)₂(20(R)‐/20(S)‐Rg3)−H]⁺.

The chiral MS strategy utilizing CS enabled MS‐based enantiomer differentiation [29], which was previously unattainable. Consequently, we shifted our focus to the MS analysis of complexes formed between ginsenoside epimers and CS, specifically, l‐Phe and Cu²⁺. MS¹ spectra of Set B samples revealed stable trimeric complex ions formed by ginsenosides, l‐Phe, and Cu²⁺. Still using 20(R)‐ and 20(S)‐Rg3 as examples, both Rg3 epimers formed identical complex ions observed at m/z 1176.49 in the MS¹ spectra, corresponding to [Cu^II(l‐Phe)₂(20(R)‐/20(S)‐Rg3)−H]⁺. The crucial differentiation emerged during CID. As depicted in Figure 1C, precursor ions m/z 1176.49 generate the principal fragment ions m/z 1011.47 [Cu^II(l‐Phe)(20(R)‐/20(S)‐Rg3)−H]⁺, 847.40 [Cu^II(20(R)‐/20(S)‐Rg3)−H]⁺, and 392.06 [Cu^II(l‐Phe)₂−H]⁺. Although both complexes produced identical fragment ions, fortunately, strikingly different relative abundances were observed, providing an opportunity to distinguish the epimers. The R feature was introduced to describe the RII ratio between targeted fragment ions. For example, R features of fragment ions m/z 1011.47 vs. 847.40, m/z 1011.47 vs. 392.06, and m/z 847.40 vs. 392.06 for 20(R)‐Rg3 complex were calculated as 1.1876, 13.3333, and 11.2267, respectively. Correspondingly, R for 20(S)‐Rg3 complex were 0.8160, 6.6885, and 8.1967 (Figure 1C). R _epimer (Equation 1), defined as R _R divided by R _S, was introduced to intuitively represent the ability to distinguish epimers. The conceptual meaning of R _epimer is analogous to R _chiral in enantiomer selectivity [33].

$$R_{\text{epimer}}=R_{\mathrm{R}}/R_{\mathrm{S}}$$ (1)

Fragment ions m/z 1011.47 and 392.06 were selected for epimer differentiation, owing to the largest R _epimer, 1.81, which reflects superior distinguish ability between epimer. The same pattern was observed in the other four pairs of ginsenoside epimers, with detailed information presented in Figure S2.

Fragment ions [Cuᴵᴵ(l‐Phe)(20(R)‐/20(S)‐ginsenoside)−H]⁺ and [Cuᴵᴵ(l‐Phe)₂−H]⁺, generated by the neutral loss of one l‐Phe ligand or one ginsenoside from the trimeric complex ions [Cu^II(l‐Phe)₂(20(R)‐/20(S)‐ginsenoside)−H]⁺, were selected as the distinguishing ions. Interestingly, compared with 20(S)‐configurational, 20(R)‐ configurational ginsenosides uniformly exhibited larger R values of fragment ions [Cu^II(l‐Phe)(20(R)‐/20(S)‐ginsenoside)−H]⁺ vs. [Cu^II(l‐Phe)₂−H]⁺. The relative abundance of fragment ions is jointly governed by the generation rate and subsequent dissociation rate, thus by directly reflecting the difficulty of formation and stability of the ions. The preferential fragmentation pathway under competitive CID was correlated to C‐20 configurations.

3.2. Quantitative Differentiation of Epimeric Complexes Using FceR‐MS² Spectra

Because RII is dependent on ce, R value between targeted fragment ions should also vary with ce. As illustrated in Figure 2A, when ce increases from 5 to 71 eV, the R values between m/z 1011.47 [Cuᴵᴵ(l‐Phe)(20(R)‐/20(S)‐Rg3)−H]⁺ and 392.06 [Cuᴵᴵ(l‐Phe)₂−H]⁺ exhibit substantial variability, ranging from 0.16 to 16.95 for 20(R)‐Rg3 complex and from 0.08 to 7.11 for 20(S)‐Rg3 complex. Correspondingly, the R _epimer values also fluctuated within a broad range of 0.32 to 2.97 (Figure 2B), indicating that although the R _epimer value obtained at a single CE could reflect configurational differences, its magnitude was highly sensitive to the chosen energy level. An inappropriate CE might cause R _epimer to approach 1, thereby diminishing or even masking epimeric differentiation and posing risks to the robustness and reproducibility of the method.

FIGURE 2.

Effects of CE on [Cu^II(l‐Phe)₂(20(R)‐/20(S)‐Rg3)−H]⁺ complex ions. Correlations of R (A) and R _epimer (B) values against CE; and (C) “head‐to‐tail” FCER‐MS² spectrum comparison between [Cu^II(l‐Phe)₂(20(S)‐Rg3)−H]⁺ (upper) and [Cu^II(l‐Phe)₂(20(R)‐Rg3)−H]⁺ (lower).

To mitigate this limitation, FCER‐MS² spectrum was introduced to systematically monitor the dissociation of the complex ions across the entire energy range. By recording RII over a series of CEs using ER‐MS, breakdown graphs depicting the generation and degradation trajectory were generated. Following normalization and integration of the breakdown graphs of residual trimeric complex ion and characteristic dimeric fragment ions, a FceR‐MS² spectrum was generated. The FceR‐MS² spectrum encapsulates the comprehensive MS² spectral behaviors of a given compound, including structural descriptors such as OCE and RII_max [32].

FCER‐MS² spectra of m/z 1176.49 [Cu^II(l‐Phe)₂(20(R)‐/20(S)‐Rg3)−H]⁺ were acquired (Figure 2C) and an attempt was made here to calculate the abundance ratio between fragment ions using RII_max instead of RII. For 20(S)‐Rg3 complex, the precursor ion gradually decreased with increasing CE and was completely depleted at 35 eV, with a ce ₅₀ of 21.35 eV. The fragment ions including m/z 1011.47 and 392.08 initially increased and then decreased as CE increased, reaching respective peaks at 24.99 and 32.07 eV. The RII corresponding to the vertexes of breakdown graphs, i.e., RII_max, were 43.50% and 14.78%, yielding a ratio as R ^max of 2.94. For 20(R)‐Rg3 complex, ce ₅₀ was 23.00 eV, and OCEs and RII_max for m/z 1011.47 and 392.08 were 26.18 and 34.04 eV, and 53.94% and 10.15%, respectively. The resultant R ^max value for 20(R)‐Rg3 complex was 5.31. Under such situation, R ^max _epimer value, defined as the ratio of R ^max between 20(R)‐ and 20(S)‐Rg3 complexes, was calculated as 1.81. For all five pairs of epimers, R ^max _epimer values were comparable to or greater than R _epimer (Figure S3 and Table S1). Fortunately, because it was independent of CE setting, the consistent observation that 20(R)‐configurational ginsenoside trimeric complexes exhibited larger R ^max _epimer than 20(S)‐configurational will be more reliable. Furthermore, FCER‐MS² approach encompassed the entire kinetic process of fragment ion generation and decay, reflecting the intrinsic gas‐phase stability differences of the complexes, offering enhanced structural specificity and reproducibility.

Consequently, FCER‐MS² spectrum overcame the inherent limitations of conventional MS/MS measurement performed at a single energy level. The approach significantly enhanced the robustness, reproducibility, and information dimensionality of the epimer differentiation. As core evaluation metrics, R ^max and R ^max _epimer not only quantified the discriminatory capacity of the method but also provided a reliable basis for configurational assignment of unknown samples.

3.3. Comparison of FCER‐MS² Spectra of Complex Ions Obtained by DI–ER‐MS and PCI‐LC–ER‐MS

When CR solution was directly added to complex matrices, such as herbal extracts or biological samples, two key challenges in DI–MS/MS analysis were encountered. First, the characteristic trimeric complex ions formed between ginsenosides and CS were often suppressed due to competitive ionization and matrix effects [34]. This suppression was particularly pronounced when the concentration of target ginsenoside was relatively low. More importantly, signals of epimers were interfered by the existence of other types of isomers. For example, in addition to 20(R)‐Rg3 and 20(S)‐Rg3, positional isomers such as ginsenoside F2 and LXXV [7] also formed complexes with CR solution, leading to signal indistinguishability. Therefore, the established DI–ER‐MS cannot be readily applied to epimer differentiation in complicated matrices.

To address this challenge, the PCI‐LC platform was introduced. An appropriate chromatographic column was responsible for the separation of the complicated sample, and meanwhile, the CR solution was continuously delivered post‐column to ensure thorough contact with the eluted compounds. The resultant stable ternary complexes, [Cu^II(l‐Phe)₂(20(R)‐/20(S)‐ginsenoside)–H]⁺, were then injected into the ESI interface. This design not only effectively circumvents the interference from the complex matrices but also establishes an integrated analytical workflow that seamlessly integrates chromatographic separation, online complexation, and MS detection.

Using PCI‐LC–ER‐MS platform, FCER‐MS² spectra of five pairs of epimers were acquired (Figures 3A and S4). Taking 20(R)‐/20(S)‐Rg3 as examples, complex ions, [Cu^II(l‐Phe)₂(20(R)‐/20(S)‐Rg3)−H]⁺ were detected. It was demonstrated that the formation rate of the noncovalent complex was sufficient to meet the requirements of online analysis. The R ^max values between fragment ions m/z 1011.4744 and 392.0699 for 20(S)‐Rg3 and 20(R)‐Rg3 complex ions were calculated as 3.28 and 5.36, corresponding to 43.56%/13.27% and 49.06%/9.16%, respectively. The resultant R ^max _epimer came out as 1.63, agreeing well with the result obtained via DI–ER‐MS. Systematic comparison revealed that R ^max and R ^max _epimer features obtained by DI–ER‐MS and PCI‐LC–ER‐MS platforms for the five pairs of epimers were almost identical (Figure 3B,C), indicating that the isomer‐discriminating ability of the complex ions was unaffected by the sample introduction manner.

FIGURE 3.

Consistency in dissociation behavior of [Cu^II(l‐Phe)₂(20(R)‐/20(S)‐Rg3)−H]⁺ complex ions analyzed by DI and PCI‐LC. (A) “Head‐to‐tail” FceR‐MS² spectrum comparison of the [Cu^II(l‐Phe)₂(20(S)‐Rg3)−H]⁺ (upper) and [Cu^II(l‐Phe)₂(20(R)‐Rg3)−H]⁺ (lower) obtained via PCI‐LC–ER‐MS. Comparison of R ^max (B) and R ^max _epimer (C) values obtained by DI–ER‐MS and PCI‐LC–ER‐MS.

The experimental results demonstrated identical performance between the data obtained from PCI‐LC–MS/MS and DI–MS/MS, thereby validating the effectiveness and reliability of this strategy for applications in complex matrices.

3.4. Ginsenosides Profiling and Epimers Discrimination of Red Ginseng

Red ginseng was collected as proof‐of‐concept. The representative chromatograms are shown in Figure 4. A total of 51 compounds were detected from red ginseng in both positive and negative ion polarities. Based on the summarized fragmentation patterns [35], series of rules were established to filter out ginsenosides [25, 36, 37, 38]: (1) characteristic fragment ions in negative ion polarity, including m/z 459.4 [PPD − H]⁻, 475.4 [PPT − H]⁻, 621.4 [PPD + Glc − H]⁻, 637.4 [PPT + Glc − H]⁻, 783.5 [PPD + 2Glc − H]⁻/[PPT + Glc + Rha − H]⁻, 799.5 [PPT + 2Glc − H]⁻, 931.5 [PPT + 2Glc + Xyl/Ara − H]⁻, 945.5 [PPD + 3Glc − H]⁻/[PPT + 2Glc + Rha − H]⁻, 955.5 [OA + GlcA + 2Glc −  H]⁻, 1077.6 [PPD + 3Glc + Ara − H]⁻/[PPD + 4Glc − H]⁻, etc.; (2) Neutral loss of 46 Da (HCOOH), 86 Da (C₃H₂O₃), 42 Da (CH₂CO), 132 Da (Xyl‐/Ara‐), 146 Da (Rha‐), and 162 Da (Glc‐) in negative ion polarity; (3) characteristic fragment ions in positive ion polarity, including 443.4 [PPD − H₂O + H]⁺, 459.4 [PPT − H₂O + H]⁺, 623.4 [PPD + Glc + H]⁺, 639.4 [PPT + Glc + H]⁺, 785.5 [PPD + 2Glc + H]⁺/[PPT + Glc + Rha + H]⁺, 801.5 [PPT + 2Glc + H]⁺, 933.5 [PPT + 2Glc + Xyl/Ara + H]⁺, 947.5 [PPD + 3Glc + H]⁺/[PPT + 2Glc + Rha + H]⁺, 957.5 [OA + GlcA + 2Glc + H]⁺, 1079.6 [PPD + 3Glc + Ara + H]⁺/. As a result, 42 ginsenosides were screened out and the results are listed in Table 1.

FIGURE 4.

The base peak chromatogram (BPC) of red ginseng extracts in negative (A) and positive (B) ion polarities.

TABLE 1.

Chromatographic and MS/MS information for the compounds identified from red ginseng extract.

No.	t R (min)	Formular	Negative ion polarity				Positive ion polarity				Identity
			MS1	Adducts	Error (ppm)	MS2	MS1	Adducts	Error (ppm)	MS2
1	1.397	C12H22O11	387.1144	[M + HCOO]−	2.1	341.1138; 221.0701; 179.0590; 161.0478; 89.0257; 59.0147	365.1054	[M + Na]+	−0.1	275.0763; 203.0530; 185.0421	Sucrose
2	1.637	C6H8O7	191.0197	[M‐H]−	3.0	173.0118; 111.0107; 87.0102	193.0347	[M + H]+	2.2	139.0030; 129.0177; 111.0075; 87.0072	Citric acid
3	2.578	C7H10O7	205.0354	[M‐H]−	2.3	173.0109; 143.0635; 111.0099; 87.0095; 67.0194	207.0502	[M + H]+	1.3	143.0331; 139.0028; 129.0177; 111.0074; 101.0230; 83.0125	2‐Methylcitric acid
4	4.214	C13H16O19	315.0764	[M‐H]−	−1.8	225.0453; 152.0101; 109.0289; 108.0207	N/A	N/A	N/A	N/A	Protocatechuic acid 4‐glucoside
5	10.136	C48H82O19	1007.5564	[M + HCOO]−	3.1	961.5550; 799.4862; 781.4889; 637.4479; 619.4322; 475.3811; 179.0578; 161.0474; 119.0365	963.5531	[M + H]+	1.1	783.4873; 621.4372; 603.4236; 441.3720; 423.3622; 405.3509; 325.1127; 203.1785; 163.0592	Notoginsenoside R6/Notoginsenoside R3/isomer
6	10.403	C48H82O19	1007.5568	[M + HCOO]−	3.1	961.5537; 799.5011; 781.4910; 637.4445; 619.4294; 475.3853; 221.0701; 179.0592; 161.0469; 119.0358; 113.0247	963.5538	[M + H]+	1.2	783.4877; 621.4369; 603.4240; 441.3718; 423.3612; 405.3511; 325.1125; 203.1778; 163.0588	Notoginsenoside R6/Notoginsenoside R3/isomer
7	10.528	C54H90O24	1167.5987	[M + HCOO]−	3.7	1121.5988; 1077.6063; 994.5592; 959.5654; 931.5563	N/A	N/A	N/A	N/A	Notoginsenoside B/isomer
8	10.562	C47H80O18	977.5325	[M + HCOO]−	−0.2	931.5281; 799.4922; 637.4187; 179.0508	933.54	[M + H]+	−1.7	621.4343; 603.4207; 585.4100; 441.3716; 423.3601; 405.3518; 325.1125; 295.1019; 259.0819; 203.1792; 163.0589; 145.0492	Notoginsenoside R1
9	10.909	C42H72O14	845.5045	[M + HCOO]−	2.7	845.5063; 799.4999; 637.4436; 475.3870; 179.0586; 161.0477	801.5012	[M + H]+	−0.2	621.4400; 603.4275; 459.3836; 441.3755; 423.3646; 405.3543; 325.1146; 203.1807; 143.0608	Ginsenoside Rg1
10	10.909	C48H82O18	991.5665	[M + HCOO]−	−1.9	991.5708; 945.5634; 799.5020; 783.5096; 765.4955; 637.4459; 619.4328; 475.3882; 179.0586; 161.0480; 119.0362	947.5578	[M + H]+	−0.7	767.4958; 749.4854; 621.4392; 603.4280; 441.3746; 423.3640; 405.3529; 367.3016; 309.1192; 203.1800; 147.0653	Ginsenoside Re
11	11.915	C51H84O21	1031.5405	[M‐H]−	3.3	987.5485; 945.5412; 927.5366; 783.4913; 765.4754; 621.4352; 459.3606; 179.0550; 161.0448	1033.5565	[M + H]+	−1.2	767.4947; 749.4913; 587.4311; 459.3882; 441.3722; 423.3617; 395.1173; 309.1184; 147.0646	Malonyl‐ginsenoside Rd/isomer
12	11.926	C54H94O24	608.3100	[M + 2HCOO]2−	3.6	1125.6310; 963.5676; 801.5143; 783.5033; 639.4580; 221.0706; 179.0582; 161.0472	1127.6201	[M + H]+	−0.6	785.5052; 767.4954; 749.4883; 605.4426; 461.3973; 443.3883; 425.3795; 407.3673; 325.1136; 13.0602; 127.1113	Unknown
13	12.116	C54H90O24	1167.5996	[M + HCOO]−	3.1	1121.5998; 959.5239; 843.5182	N/A	N/A	N/A	N/A	Notoginsenoside B/isomer
14	12.483	C41H70O13	815.4952	[M + HCOO]−	2.8	815.4956; 769.4863; 637.4442; 619.4331; 607.4314; 475.3863; 161.0433; 149.0477	771.4873	[M + H]+	−1.1	591.4244; 459.3838; 441.3718; 423.3577; 405.3512; 217.1962; 149.1295	Pseudoginsenoside RT3/isomer
15	12.856	C44H74O15	887.5163	[M + HCOO]−	3.3	841.5136; 799.5006; 781.4702; 637.4406; 619.4311; 475.3872; 203.0382; 179.0581; 161.0448	843.5102	[M + H]+	0.8	663.4487; 645.4341; 441.3721; 423.3626; 405.3523; 367.3001; 217.1953; 203.1796; 147.1168	Acetyl‐ginsenoside Rg1
16	13.936	C50H84O19	1033.5589	[M + HCOO]−	1.9	987.5543; 945.5523; 927.5515; 799.5012; 637.4397; 619.4340; 571.4121; 475.3886; 179.0560; 161.0469; 119.0374	989.5677	[M + H]+	−0.6	767.4968; 749.4833; 587.4147; 441.3724; 423.3616; 405.3513; 351.1280; 315.1073; 309.1186; 205.0710; 169.0486	Quinquenoside III
17	15.334	C42H72O14	845.5054	[M + HCOO]−	2.6	799.5011; 637.4446; 475.3872; 221.0688; 179.0583; 161.0480; 143.0366; 119.0359; 113.0258	801.5019	[M + H]+	−0.2	783.4862; 765.4813; 621.4367; 603.4299; 441.3750; 423.3637; 405.3525; 325.1133; 203.1795; 163.0597	Ginsenoside Rf
18	16.010	C59H100O27	665.3322	[M + 2HCOO]2−	4.1	553.2966; 1239.6399; 1107.5963; 1077.5738; 945.5456; 783.4975; 765.4821; 353.1085; 323.0975; 221.0694; 191.0546; 179.0564; 161.0444; 149.0448	1241.6515	[M + H]+	−1.2	649.2189; 619.2084; 605.4387; 587.4378; 487.1670; 475.1551; 425.3781; 407.3638; 343.1264; 325.1132; 295.1036; 259.0803; 241.0683; 223.0586; 163.0589; 145.0484	Notoginsenoside R4
19	17.109	C41H70O13	815.4948	[M + HCOO]−	2.2	769.4882; 637.4446; 475.3871; 391.2895; 191.0582; 161.0475; 143.0362; 113.0255	793.4709	[M + Na]+	−2.1	661.4211; 481.3609; 335.0924; 317.0807	Notoginsenoside R2
20	18.262	C41H70O13	815.4935	[M + HCOO]−	1.7	769.4887; 637.4435; 475.3867; 161.0473	771.4886	[M + H]+	−1.5	753.4761; 735.4787; 621.4281; 441.3725; 423.3610; 405.3501; 203.1788	Ginsenoside F3
21	19.072	C42H72O13	829.4994	[M + HCOO]−	1.9	783.4909; 637.4327; 619.4211; 475.3773; 391.2844; 205.0711; 161.0415	785.5046	[M + H]+	−1.9	767.4910; 749.4799; 441.3668; 423.3623; 405.3505; 161.1321	Ginsenoside Rg2
22	19.691	C36H62O9	683.4483	[M + HCOO]−	2.5	683.4511; 637.4443; 475.3872; 161.0469; 113.0251; 101.0253	639.4425	[M + H]+	−1.7	621.4352; 603.4227; 441.3719; 423.3627; 405.3510; 203.1788	20S‐Ginsenoside Rh1
23	20.177	C59H100O27	665.3315	[M + 2HCOO]2−	4.0	553.2967; 1239.6389; 1107.5966; 1077.5739; 945.5458; 783.4978; 765.4828; 353.1080; 323.0991; 221.0692; 179.0568; 161.0451; 149.0448	1241.6505	[M + H]+	−0.8	781.2610; 767.4877; 649.2231; 619.2061; 605.4392; 487.1651; 475.1542; 425.3791; 407.3670; 325.1132; 295.0832; 259.0823; 163.0604; 145.0507	Notoginsenoside Fa
24	20.304	C54H92O23	599.3093	[M + 2HCOO]2−	4.4	1107.5941; 945.5759; 783.4941; 765.4843; 621.4627; 603.4277; 576.2984; 553.2963; 459.3567; 323.0993; 221.0651; 179.0564; 161.0446	1131.5926	[M + Na]+	0.4	789.4773; 365.1037	Ginsenoside Rb1
25	20.979	C36H62O9	683.4490	[M + HCOO]−	2.7	683.4515; 637.4444; 475.3863; 161.0471; 101.0250	639.4427	[M + H]+	−2.0	621.4372; 603.4273; 441.3721; 423.3648; 405.3521; 203.1768	20R‐Ginsenoside Rh1
26	21.355	C48H76O19	955.50779	[M‐H]−	2.0	793.4342; 731.4494; 613.3748; 587.4150; 569.3864; 523.2849; 179.0590; 119.0372	957.5089	[M + H]+	−0.6	795.4587; 777.4496; 759.4336; 633.4024; 615.3891; 587.3946; 439.3583; 393.3524; 339.0930; 321.0822; 203.1798; 191.1799; 163.0603	Ginsenoside Ro
27	22.176	C53H90O22	1123.6097	[M + HCOO]−	2.6	1077.6068; 945.5780; 915.5377; 783.4751; 621.4362; 293.0842; 221.0643; 179.0558; 149.0469	1101.5816	[M + Na]+	−1.8	789.4739; 355.0934	Ginsenoside Rb2
28	22.520	C53H90O22	1123.6112	[M + HCOO]−	2.7	1077.6006; 945.5480; 915.5287; 783.4749; 621.4356; 293.0844; 221.0638; 179.0571; 149.0458	1101.5812	[M + Na]+	−0.4	789.4748; 365.1052; 335.0942	Ginsenoside Rb3
29	22.866	C56H94O24	620.3080	[M + 2HCOO]2−	2.9	574.2965; 553.2962; 1149.6030; 1107.5948; 1089.5884; 987.5502; 945.5404; 927.5274; 825.5038; 783.4923; 621.4338; 459.3807; 323.0993; 221.0640; 179.0561; 161.0425	1151.6198	[M + H]+	−0.9	809.5035; 791.4958; 691.2287; 605.4422; 529.1766; 443.3881; 425.3782; 407.3673; 367.1241; 325.1134; 205.0712; 187.0597; 169.0499; 163.0597; 145.0490	Quinquenoside R1/Acetylginsenoside Rb1/isomer
30	22.906	C36H62O9	683.4469	[M + HCOO]−	2.6	683.4447; 637.4367; 475.3855; 179.0564; 161.0455; 119.0365	661.4277	[M + Na]+	−1.5	481.3644; 203.0533	Ginsenoside Rh1 isomer
31	23.002	C48H82O18	991.5483	[M + HCOO]−	0.3	945.5451; 783.4834; 621.4384; 179.0525; 161.0413	947.5593	[M + H]+	−1.8	767.4969; 749.4852; 605.4423; 587.4317; 443.3892; 425.3788; 407.3681; 325.1142; 217.1961; 189.1632; 163.0604; 145.0496	Ginsenoside Rd
32	23.127	C55H92O23	1165.6207	[M + HCOO]−	2.4	1119.6162; 1077.6047; 1059.5931; 945.5577; 927.5450; 915.5517; 783.5009; 765.4892; 621.4485; 459.3883; 221.0675; 191.0575; 161.0464	1121.6128	[M + H]+	−2.9	923.5434; 809.5068; 791.4962; 605.4431; 529.1766; 499.1656; 443.3881; 425.3794; 407.3680; 367.1247; 295.1025; 217.1958; 205.0717; 187.0603; 169.0492; 127.0388	Pseudoginsenoside F8
33	23.155	C58H98O26	650.3178	[M + 2HCOO]2−	2.8	604.3136; 538.7918; 1209.6433; 1077.5939; 1047.5555; 945.5852; 915.5286; 783.4892; 765.4919; 621.4302; 459.3775; 425.1330; 323.1005; 293.0842; 233.0656; 191.0564; 161.0459; 149.0445; 131.0328	1211.6418	[M + H]+	−2.2	947.5730; 929.5468; 899.5334; 887.5438; 869.5283; 767.4930; 749.4796; 605.4410; 575.4310; 487.1634; 457.1543; 443.3868; 427.3986; 425.3772; 407.3680; 325.1130; 295.1030; 265.0920; 163.0601	Notoginsenoside Fc
34	23.507	C55H92O23	1165.6207	[M + HCOO]−	2.4	1119.6103; 1077.5997; 1059.5896; 945.5560; 927.5415; 915.5517; 783.4982; 765.4872; 621.4437; 459.3889; 221.0675; 191.0568; 161.0458	1121.6094	[M + H]+	−2.1	923.5444; 809.5078; 791.4952; 605.4451; 499.1676; 443.3878; 425.3785; 407.3677; 367.1245; 295.1023; 217.1965; 205.0721; 187.0613; 169.0485; 127.0379	Ginsenoside Rs2
35	23.988	C47H80O17	961.5422	[M + HCOO]−	3.8	915.5411; 783.4903; 765.4884; 753.4773; 621.4426; 603.4251; 459.3795; 161.0442; 149.0418	917.5451	[M + H]+	−1.6	767.4932; 749.4806; 737.4752; 719.4697; 605.4380; 587.4291; 557.4197; 443.3897; 425.3768; 407.3661; 369.2243; 325.1126; 295.1018; 259.0803; 217.1949; 203.1785; 191.1774	Vinaginsenoside R17/Vinaginsenoside R18
36	24.330	C47H80O17	961.5424	[M + HCOO]−	3.6	915.5333; 783.4892; 753.5448; 621.4383; 459.3875; 161.0456	917.5442	[M + H]+	−1.4	767.4941; 749.4855; 737.4741; 719.4699; 605.4378; 587.4286; 557.4191; 443.3889; 425.3767; 407.3671; 369.2234; 325.1123; 295.1011; 259.0811; 217.1951; 203.1775; 191.1769	Vinaginsenoside R17/Vinaginsenoside R18
37	24.388	C50H84O19	1033.5575	[M + HCOO]−	1.5	987.5532; 945.5528; 927.5526; 825.5477; 783.5037; 765.4939; 621.4458; 179.0604; 161.0470; 143.0354; 113.0253	989.5661	[M + H]+	−0.4	767.4977; 749.4831; 587.4139; 441.3734; 423.3612; 405.3521; 351.1277; 315.1071; 309.1169; 205.0711; 169.0459	Pseudoginsenoside RS1
38	24.568	C42H70O12	811.4883	[M + HCOO]−	1.6	765.4812; 619.4275; 601.4141; 553.4204; 503.1555; 457.3736; 205.0658; 163.0585; 161.0436	767.4942	[M + H]+	0.3	749.4843; 605.4421; 587.4302; 569.4197; 443.3896; 425.3784; 407.3675; 639.3128; 351.3054; 325.1142; 217.1956; 203.1795; 163.0601; 145.0498	Ginsenoside Rg5/Ginsenoside Rk1/isomer
39	24.924	C50H84O19	1033.5578	[M + HCOO]−	1.6	987.5750; 945.5527; 927.5513; 825.5471; 783.5082; 765.5014; 621.4558; 221.0661; 179.0593; 161.0471; 143.0364; 113.0265	989.5659	[M + H]+	0.2	767.4978; 749.4829; 587.4126; 441.3744; 423.3611; 405.3523; 351.1271; 315.1069; 309.1171; 205.0712; 169.0445	Pseudoginsenoside RC1
40	24.933	C42H70O12	811.4876	[M + HCOO]−	1.5	765.4809; 603.4280; 221.0686; 179.0547; 161.0440	767.4944	[M + H]+	0.5	749.4841; 605.4431; 587.4311; 569.4177; 443.3886; 425.3779; 407.3669; 639.3131; 351.3051; 325.1138; 217.1951; 203.1783; 163.0600; 145.0493	Ginsenoside Rg5/Ginsenoside Rk1/isomer
41	25.213	C48H82O17	975.5570	[M + HCOO]−	2.7	929.5462; 767.5014; 605.4431; 179.0530; 161.0443	N/A	N/A	N/A	N/A	Vinaginsenoside R3
42	25.561	C36H60O8	665.4318	[M + HCOO]−	1.6	665.4394; 619.4304; 161.0480; 101.0246	621.4371	[M + H]+	1.6	603.4271; 441.3735; 423.3620; 405.3519; 221.1892; 203.1794; 187.1478; 147.1168	Ginsenoside Rh4
43	26.184	C36H60O8	665.4381	[M + HCOO]−	2.2	665.4394; 619.4304; 161.0480; 101.0246	621.4368	[M + H]+	1.4	603.4269; 441.3744; 423.3615; 405.3521; 221.1877; 203.1781; 187.1469; 147.1159	Ginsenoside RK3
44	26.636	C17H26O3	323.1907	[M + HCOO]−	3.8	203.1437; 119.0506	N/A	N/A	N/A	N/A	3‐(3,5‐Di‐tert‐butyl‐4‐hydroxyphenyl)propionic acid
45	27.421	C42H72O13	829.5014	[M + HCOO]−	2.1	783.4998; 637.4319; 619.4207; 475.3775; 391.2848; 205.0709; 161.0419	785.5067	[M + H]+	0.5	767.4954; 749.4839; 605.4418; 587.4292; 569.4190; 443.3883; 425.3376; 407.3678; 325.1132; 217.1952; 203.1796; 163.0604; 145.0494	20S‐Ginsenoside Rg3
46	28.047	C42H72O13	829.5010	[M + HCOO]−	2.0	783.4996; 637.4321; 619.4212; 475.3769; 391.2839; 205.0710; 161.0411	785.5068	[M + H]+	0.5	767.4955; 749.4841; 605.4415; 587.4291; 569.4187; 443.3879; 425.3378; 407.3681; 325.1152; 217.1946; 203.1796; 163.0601; 145.0486	20R‐Ginsenoside Rg3
47	30.348	C33H58O14	723.3933	[M + HCOO]−	2.7	677.3884; 415.1527; 397.1421; 279.2378; 119.0362; 89.0258	679.3909	[M + H]+	1.2	355.2851; 337.2745; 263.2378; 245.2274; 175.1472	Gingerglycolipid B
48	31.348	C33H45NO4	564.3391	[M + HCOO]−	2.1	504.3194; 279.2385; 242.0828; 224.0732; 168.0439	N/A	N/A	N/A	N/A	Unknown
49	34.649	C42H70O12	811.4890	[M + HCOO]−	1.9	765.4806; 603.4271; 221.0673; 179.0553; 161.0464	767.4941	[M + H]+	0.7	749.4836; 605.4428; 587.4321; 569.4181; 443.3891; 425.3772; 407.3663; 639.3111; 351.3044; 325.1129; 217.1961; 203.1777; 163.0598; 145.0475	Ginsenoside Rg5/Ginsenoside Rk1/isomer
50	34.896	C42H70O12	811.4884	[M + HCOO]−	1.6	765.4811; 603.4277; 221.0696; 179.0569; 161.0451	767.4944	[M + H]+	0.5	749.4838; 605.4419; 587.4331; 569.4159; 443.3881; 425.3766; 407.3653; 639.3121; 351.3043; 325.1119; 217.1971; 203.1781; 163.0591; 145.0489	Ginsenoside Rg5/Ginsenoside Rk1/isomer
51	35.473	C28H34O4	433.2416	[M‐H]−	2.3	279.2373; 171.0084; 152.9986	N/A	N/A	N/A	N/A	Unknown

Note: N/A, undetected.

Two pairs of epimers, 22 and 25 along with 45 and 46, were recognized by detailed MS² spectrum comparison. The PCI‐LC–ER‐MS was then employed to plot FCER‐MS² spectra of the corresponding complex ions. 22 and 25 were preliminarily identified as 20(R)‐/20(S)‐ginsenoside Rh1, based on the summarized fragmentation pattern combined with database retrievals. As shown in Figure 5, for the [Cu^II(l‐Phe)₂(compound 22)‐H]⁺ complex, the OCEs for fragment ions m/z 865 and 392 are 20.39 and 25.36 eV, with corresponding RII_max values of 38.05% and 43.14%, yielding an R _max of 0.88. Similarly, R _max of the [Cu^II(l‐Phe)₂(compound 25)‐H]⁺ complex ion was 2.20. According to the ranking styles of R _max for 20(R)‐ and 20(S)‐ginsenosides, 25 and 22 were confirmatively identified as 20(R)‐ and 20(S)‐Rh1, respectively. This recognition workflow was successfully applied to the differentiation of 45 and 46, allowing the unambiguous determination of epimer absolute configurations (Figure S5).

FIGURE 5.

The differentiation of epimers 20(S)‐Rh1 and 20(R)‐Rh1. “Head‐to‐tail” FceR‐MS² spectrum comparison between [Cu^II(l‐Phe)₂(compound 22)−H]⁺ (upper) and [Cu^II(l‐Phe)₂(compound 25)−H]⁺ (lower).

In fact, in addition to [Cu^II(l‐Phe)₂(20(R)‐/20(S)‐ginsenoside)−H]⁺ ion, both [Cu^II(l‐Phe)(20(R)‐/20(S)‐ginsenoside)−H]⁺ and [Cu^II(l‐Phe)(20(R)‐/20(S)‐ginsenoside)₂−H]⁺ ions were observed in the MS¹ spectrum, regardless of whether CR solution was introduced via pre‐mixing or PCI‐LC. The precursor ion [Cu^II(l‐Phe)(20(R)‐/20(S)‐ginsenoside)−H]⁺ with higher intensity was excluded from consideration because R and R ^max features of the fragment ions [Cu^II(l‐Phe)−H]⁺ and [Cu^II(20(R)‐/20(S)‐ginsenoside)−H]⁺ for the complexes were found to be almost identical. Setting aside the risk that [Cu^II(l‐Phe)(20(R)‐/20(S)‐ginsenoside)₂−H]⁺ ion fell outside the MS mass range due to the excessively high m/z, R and R ^max values between fragment ions [Cu^II(l‐Phe)(20(R)‐/20(S)‐ginsenoside)−H]⁺ and [Cu^II(20(R)‐/20(S)‐ginsenoside)−H]⁺ were unsatisfactory for distinguishing 20(R)‐ and 20(S)‐ginsenosides. As a result, [Cu^II(l‐Phe)₂(20(R)‐/20(S)‐ginsenoside)−H]⁺ was ultimately selected in this study due to its superior R and R ^max value differences, consistent with previous research [30]. More fortunately, the R and R ^max of fragment ions [Cu^II(l‐Phe)(20(R)‐/20(S)‐ginsenoside)−H]⁺ vs. [Cu^II(l‐Phe)₂−H]⁺ for 20(R)‐ginsenosides were consistently large that those of 20(S)‐epimers, providing crucial information for epimer differentiation. Furthermore, to evaluate the reliability of the PCI‐LC–ER‐MS platform for analyzing real samples, a systematic stability validation was conducted using five pairs of ginsenoside epimers. Each pair of epimers was analyzed six consecutive times using the PCI‐LC–ER‐MS, and FCER‐MS² spectra were fitted accordingly. R ^max and R ^max _epimer features for each ginsenoside were calculated. As shown in Figure S6, stable differences between the 20(R)‐ and 20(S)‐epimers are consistently observed. Based on the well‐established enantioselective fragmentation patterns of ginsenosides and their correlation with stereochemical configurations, this approach enabled direct configurational identification within complex herbal matrices.

Undoubtedly, ¹³C NMR analysis and single‐crystal X‐ray diffraction are essential techniques for the definitive structural identification [9, 39]; however, they are only suitable for analysis of pure compounds. Moreover, due to the low content of 20(R)‐configurational rare ginsenosides, the traditional preparation is extremely time‐consuming and labor‐intensive [40]. Compared with the extensive research on positional isomers with different substituent positions (C‐3, C‐6, or C‐20) and composition as well as linkage of sugar chain [41, 42, 43], relatively limited studies have been conducted to distinguish 20(R)‐ and 20(S)‐epimers. Thanks to advancements in column packing materials, the 20(R)‐ and 20(S)‐epimers can be successfully separated using reversed‐phase columns, such as T3 and C18, with the retention time of 20(S)‐ginsenosides often being slightly shorter than that of the corresponding 20(R)‐epimer [44]. Nevertheless, the differentiation based on relative retention times still relies on reference standards and fails to address the challenge of detecting only a single peak in the given sample. Yu et al. [26] found that the relative abundance of fragment ion [M + Ag − H₂O]⁺ produced by silver cationized 20(S)‐ginsenoside was significantly higher than that of the 20(R)‐epimer. The silver‐assisted epimeric discrimination strategy was essentially consistent with the introduction of the [Cu^II(l‐Phe)₂−H]⁺ complex in this study. However, Ag⁺ will cause more severe contamination of the MS platform. Ion mobility spectrometry is another viable approach; however, due to the minor structural differences between the epimers, more complex method development is often required to achieve baseline separation [45].

Although Cu²⁺ are widely employed as the core in forming diastereomeric complexes because of the stability of coordination bonds, their use presents significant contamination risks to MS instruments. Upon reduction occurring at the ion source or electrode surfaces, Cu will be deposited on electrodes and quadrupoles. This deposition leads to inhomogeneous electric fields, reduced signal stability, loss of sensitivity, and even permanent instrument damage, especially in high‐precision MS such as Orbitrap. To address this critical limitation, our research group continues to develop a metal ion‐free chiral complexation strategy. Electrostatic attraction between acidic and basic compounds, such as wogonoside and berberine, is considered a viable alternative. This approach enabled the analysis and identification of chiral isomers while effectively avoiding metal‐induced contamination. It provided a green and efficient alternative for chiral analysis and established a methodological foundation for stereoselective studies of chiral drugs in complicated matrices.

In this study, a novel and robust analytical strategy was established for rapid and reliable differentiation of 20(S)‐ and 20(R)‐ginsenoside epimers in complicated matrices, using the [Cu^II(l‐Phe)₂−H]⁺ complex as CS coupled with PCI‐LC–ER‐MS. First, under a single ce, the complex ions formed between the epimers and [Cu^II(l‐Phe)₂−H]⁺ exhibited differences in the fragment ion abundance. Furthermore, by leveraging the differential CID behaviors of diastereomeric ternary complexes, the diagnostic parameter, R ^max _epimer, derived from ER‐MS and FCER‐MS² spectra was introduced to quantitatively and reproducibly discriminate the epimers and overcome the limitations of conventional single‐energy MS/MS approaches. Finally, the integration of PCI‐LC with ER‐MS successfully addressed the challenges of matrix interference and ionization suppression commonly encountered in complex samples such as red ginseng extract. The practical applicability of the method was validated through the successful differentiation of two pairs of ginsenoside epimers Rg3 and Rh1 in red ginseng extract, without the need for prior isolation or purification. The platform enabled online complexation, chromatographic separation, and stereospecific MS detection in a single analytical run, demonstrating excellent consistency with DI experiments. Beyond its application to ginsenosides, this strategy holds significant promise for the stereochemical characterization of other natural product isomers, particularly those with subtle structural differences that are difficult to resolve by conventional MS or chromatographic means. The use of metal–amino acid complexes as chiral selectors, combined with ER‐MS, offers a versatile and extendable platform for isomer differentiation in metabolomics, quality control, and pharmaceutical analysis.

Funding

This work was supported by the National Natural Science Foundation of China (No. 82474202).

Supporting information

Table S1: MS¹ of trimeric diastereomeric ions, relative abundance ratio (R and R ^max), and epimer resolution (R _epimer and R ^max _epimer) for five pairs of ginsenoside epimers employing [Cu^II(l‐Phe)₂‐H]⁺ as CS.

Figure S1: Chemical structures of five pairs of ginsenoside epimers.

Figure S2: “Head‐to‐tail” DI–MS/MS spectrum comparison of the diastereomeric ions [Cu^II(l‐Phe)₂(20(S)‐PPD)‐H]⁺ (upper) and [Cu^II(l‐Phe)₂(20(R)‐PPD)‐H]⁺ (lower) (A), [Cu^II(l‐Phe)₂(20(S)‐PPT)‐H]⁺ (upper) and [Cu^II(l‐Phe)₂(20(R)‐PPT)‐H]⁺ (lower) (B), [Cu^II(l‐Phe)₂(20(S)‐NR2)‐H]⁺ (upper) and [Cu^II(l‐Phe)₂(20(R)‐NR2)‐H]⁺ (lower) (C), [Cu^II(l‐Phe)₂(20(S)‐Rh1)‐H]⁺ (upper) and [Cu^II(l‐Phe)₂(20(R)‐Rh1)‐H]⁺ (lower) (D).

Figure S3: “Head‐to‐tail” FCER‐MS² spectrum comparison of the diastereomeric ions [Cu^II(l‐Phe)₂(20(S)‐PPD)‐H]⁺ (upper) and [Cu^II(l‐Phe)₂(20(R)‐PPD)‐H]⁺ (lower) (A), [Cu^II(l‐Phe)₂(20(S)‐PPT)‐H]⁺ (upper) and [Cu^II(l‐Phe)₂(20(R)‐PPT)‐H]⁺ (lower) (B), [Cu^II(l‐Phe)₂(20(S)‐NR2)‐H]⁺ (upper) and [Cu^II(l‐Phe)₂(20(R)‐NR2)‐H]⁺ (lower) (C), [Cu^II(l‐Phe)₂(20(S)‐Rh1)‐H]⁺ (upper) and [Cu^II(l‐Phe)₂(20(R)‐Rh1)‐H]⁺ (lower) (D) obtained via DI–MS/MS.

Figure S4: “Head‐to‐tail” FCER‐MS² spectrum comparison of the diastereomeric ions [Cu^II(l‐Phe)₂(20(S)‐PPD)‐H]⁺ (upper) and [Cu^II(l‐Phe)₂(20(R)‐PPD)‐H]⁺ (lower) (A), [Cu^II(l‐Phe)₂(20(S)‐PPT)‐H]⁺ (upper) and [Cu^II(l‐Phe)₂(20(R)‐PPT)‐H]⁺ (lower) (B), [Cu^II(l‐Phe)₂(20(S)‐NR2)‐H]⁺ (upper) and [Cu^II(l‐Phe)₂(20(R)‐NR2)‐H]⁺ (lower) (C), [Cu^II(l‐Phe)₂(20(S)‐Rh1)‐H]⁺ (upper) and [Cu^II(l‐Phe)₂(20(R)‐Rh1)‐H]⁺ (lower) (D) obtained via PCI‐LC–ER‐MS.

Figure S5: FCER‐MS² spectrum comparison [Cuᴵᴵ(l‐Phe)₂(20(S)‐Rg3)−H]⁺ (upper) and [Cuᴵᴵ(l‐Phe)₂(20(R)‐Rg3)−H]⁺ (lower) complexes in red ginseng.

Figure S6: Stability evaluation results of five pairs of ginsenoside epimers (n = 6, mean ± SD).

Data Availability Statement

The data that support the findings of this study are available from the corresponding author upon reasonable request.