Development of a Proton Spin Network Fingerprint Library to Support Mass Spectrometry-Based Identification of Pharmacophore-Bearing Constituents in the Botanical Supplement Centella asiatica

Abstract

Centella asiatica (L.) Urban (Apiaceae) has been utilized for centuries in traditional medicine systems in Southeast Asia and Southern Africa, including Madagascar. Previous studies have reported evidence of the therapeutic potential of C. asiatica formulations in models of Alzheimer’s Disease and other dementias. Caffeoylquinic acids (CQAs) have been identified to be among the pharmacologically relevant metabolites contributing to the botanical’s cognitive enhancement and neuroprotective effects. Isomers of CQAs are, however, difficult to differentiate by commonly used LC-MS techniques, making the characterization, standardization, and batch-to-batch consistency of these formulations challenging. Individual CQAs have unique proton Spin Network Fingerprints (pSNFs) that can be used to distinguish between CQA regioisomers within complex extracts. This work describes the development of a CQA-focused pSNF library that can be used to complement LC-MS methods for the accurate metabolite identification and characterization of bioactive C. asiatica fractions and extracts. The isolation of two new (1 and 2) and four known (3 − 6) CQAs and CQA analogues from C. asiatica and their contribution to the pSNF library are also discussed herein.

Graphical Abstract

Centella asiatica (L.) Urban (Apiaceae) is a perennial pennywort that has been used in food and traditional medicine systems for centuries.¹ Many countries have reported its uses in topical treatments for skin conditions and wound healing.² In addition, there is substantial evidence describing the efficacy of oral preparations of C. asiatica to treat a variety of ailments including infectious and gastrointestinal diseases, pain and inflammation, and neurological conditions.² Within the past few decades, the plant has gained interest in the Western world for its applications in skincare and for its medicinal benefits related to neuroprotection and cognition, which has become particularly relevant for Alzheimer’s Disease and other dementias. These effects have been demonstrated in several in vivo models, where oral administration of a C. asiatica water extract (CAW) relieved cognitive deficits^{3, 4, 5} and reduced impairments in dendritic morphology associated with amyloid-β,⁶ improved cognitive performance in aged C57BL/6 mice,⁷ and improved memory and executive function in CB6F1 mice.⁸ These studies have also demonstrated the impacts of CAW on mitochondrial and antioxidant response gene expression, including NRF2, that may be associated with the memory-enhancing properties.^{5, 7–10} Few human trials have been conducted on C. asiatica,¹¹ with inconclusive results in part due to the lack of comprehensive standardization procedures of the botanical samples. In order to achieve consistency across batches and reproduce results in preclinical and clinical settings, accurate identification and quantification of the pharmacophore-bearing chemical constituents present in botanical preparations of C. asiatica is crucial.

Identification of the pharmacophores of C. asiatica

Pentacyclic triterpenoids, chlorogenic acids, and flavonoids have been reported to be the biologically-active constituents found in C. asiatica.¹² Ursane-type triterpenoids, namely asiatic acid (19) and madecassic acid (20), and their glycosylated counterparts, asiaticoside (21) and madecassoside (22), along with caffeoylquinic acids (CQAs, 4 − 12) are abundant in C. asiatica formulations that influence neurological and behavioral effects.¹³ Moreover, Wu and coworkers¹⁴ reported that ursane- and oleanane-type triterpene saponins from C. asiatica displayed neuroprotective effects against 6- hydroxydopamine-induced toxicity and apoptosis in vitro. Asiatic acid,^{15, 16} asiaticoside,^{17, 18} and madecassoside^{19, 20} have attenuated cognitive impairments in various mouse models through different enzymatic and antioxidative mechanisms. Additionally, compounds 19²¹ and 21²² improved memory in adult male Sprague-Dawley rats. CQAs 4 − 6 and 16 have been shown to suppress aggregation and toxicity associated with β-amyloid in MC65 and SH-SY5Y human neuroblastoma cell lines.^{9, 23} Multiple studies have demonstrated the cognitive benefits of CQAs in vivo, including reduced production of reactive oxygen species (ROS) and mitigation of memory loss.^{24, 25} Preliminary data acquired as a result of a collaboration with the Botanicals Enhancing Neurological Function and Resilience in Aging (BENFRA) Center using HepG2-ARE reporter cells showed that the EtOAc fraction of CAW-7 (CAW-E) more strongly activates the antioxidant response transcription factor NRF2 compared to the water fraction of CAW-7 (CAW-W), indicating that the active chemical constituents are concentrated in CAW-E (Figure 1). Liquid chromatography coupled with mass spectrometry (LC-MS) analysis confirmed that CAW-E is rich in non- to mid-polar compounds such as asiatic acid (19), madecassic acid (20), and CQAs (Table S1). These results suggested that CQAs and triterpene aglycones in C. asiatica bear pharmacophores that are responsible for the bioactivity observed. Thus, these compounds have been chosen as markers for standardization and quality control of a C. asiatica botanical sample to be used for preclinical and clinical studies.²⁶ The weak activity observed for CAW-W can be attributed to the presence of minor amount of CQAs in the fraction as well as polar components (e.g., asiaticoside (21) and madecassoside (22)).

Figure 1.

NRF2 activation by CAW and derived fractions in HepG2-ARE reporter cells. Significance is relative to control unless otherwise indicated. *p<0.05, **p<0.01, ***p<0.001, n = 8–22 wells across 3 experiments.

With the current refinement of ionization techniques and expansion of mass libraries, LC-MS is the most utilized method for the detection, characterization, and quantification of complex mixtures including botanical dietary supplements. Currently, standardization of botanical supplements concentrates on the presence of abundant metabolites or metabolites characteristic to the specific plant species, regardless of whether those compounds are related to the plant’s pharmacology.²⁷ When developing a botanical sample for clinical studies, however, more comprehensive analysis focused on standardizing the pharmacologically-active constituents is recommended. For C. asiatica, the similarity of fragmentation patterns of many isomeric compounds can make their individual characterization using mass spectrometry a challenging task. Bioactive mono-, di-, or tri-substituted CQAs (4−16) can be difficult to discern using MS analysis as this class of compounds consists of many constitutional isomers. MSⁿ methods have been previously used to differentiate quinic acid regioisomers,²⁸ but because fragmentation patterns can vary between instruments and fragmentation methods, results can be complicated to interpret. Yang et. al. established an LC-MS method that disinguishes mono- and di-CQAs by retention time,²⁹ and standardization efforts of complex botanical samples would benefit from comprehensive structural data. Additional spectroscopic tools, namely nuclear magnetic resonance (NMR), can provide valuable information for accurate identification of metabolites using coupling constants and conformation-induced shift differences. For instance, the resonances and coupling constants of the quinic acid protons in CQAs can determine the location of substituents and stereochemical configuration, information that many mass spectrometry techniques are unable to specify. The goal of the current project is to complement the existing LC-MS library of C. asiatica metabolites with NMR data that provides additional spectroscopic information to contribute to overall standardization efforts.

Total Correlation Spectroscopy (TOCSY) is an NMR experiment utilized to identify and assign all coupled protons belonging to the same spin system. A series of selective one-dimensional TOCSY (S1DT) experiments were performed on isolated compounds (1−6 and 18) to build a library of proton Spin Network Fingerprints (pSNF) that are characteristic to each compound. In this study, we focus on the identification of CQAs, which belong to a class of secondary metabolites of the pharmacophore-bearing constituents of C. asiatica.

Structurally, CQAs are mono, di-, tri-, or tetra-acylated quinic acids ((1S,3R,4S,5R)-1,3,4,5-tetrahydroxycyclohexane-1-carboxylic acid). Mono- and di-CQAs are more commonly found in nature than tri- and tetra-CQAs, which are not abundant in the plant kingdom.³⁰ Acyl substituents in CQAs range from acetoxyl groups to caffeoyl, feruloyl, and other related cinnamoyl moieties. The unique spin network of coupled quinic acid protons (H-2ab, H-3, H-4, H-5, and H-6ab) present in each CQA with their defined splittings and J-coupling values has been explored for fingerprinting and accurate identification within sample mixtures. Proton signals of an acylated carbinol of quinic acid (e.g., H-3 of 10) always shift downfield when compared to those of non-acylated sites (H-3 of 18). Thus, the presence of downfielded H-3, H-4, and/or H-5 in the spin system can be used to indicate the positions of acylation. We thus hypothesize that the structures of pharmacophore-bearing CQA isomers identified in a C. asiatica extract by LC-MS can be confirmed by their pSNFs. To test this hypothesis, we acquired pSNFs of CQAs using S1DT and began constructing a pSNF library of CQAs from compounds isolated from C. asiatica plant material (compounds 1–7) and data from published literature (compounds 8–16). Compounds 1–7 were isolated in-house from plant material provided by the BENFRA Center and compound 17 was synthesized in-house by reacting quinic acid (18) with equal volume of acetic anhydride and pyridine.

To avoid confusion regarding the nomenclature of CQAs, the numbering and stereochemical assignment at carbons of the quinic acid moiety will be as depicted in structure 18 throughout the paper.

RESULTS AND DISCUSSION

Isolation and structure elucidation of secondary metabolites of C. asiatica

Successive reversed-phase column chromatography procedures using Combiflash and high-pressure liquid chromatography (HPLC) on CAW-E (the EtOAc fraction of the water extract of C. asiatica) led to the isolation of two new CQAs identified as 1 and 2, together with 3-caffeoyl-5-feruloylquinic acid (3), 3,5-DCQA (4), 1,5-DCQA (5), and 4,5-DCQA (6).

Compound 1 was isolated as a yellow amorphous solid with a molecular formula C₂₇H₂₆O₁₃ determined by the observation of the deprotonated ion peak at m/z 557.13028 in negative-High Resolution Electrospray Ionization Mass Spectrometry (Neg-HRESIMS, 557.13006, calculated for C₂₇H₂₅O₁₃⁻). The IR spectrum showed stretching and absorption bands characteristics of hydroxy, carbonyl, and benzyl functionalities at 3384, 1709, and 1598 cm⁻¹, respectively. The ¹H NMR spectrum displayed signals of two caffeoyl groups (Table 1), one methyl singlet of an acetyl group (δ_Η 2.01, br s, 3H), and a spin system arising from the methylene protons at H-2 [δ_Η 2.14 (dd, J= 14.3, 5.1 Hz) and δ_Η 2.41 (dd, J= 14.3, 2.6 Hz), 2H] and H-6 (δ_Η 2.27, m, 2H). The downfielded signals of three oxygen-bearing methines [δ_Η 5.21 (dd, J= 8.6, 3.3 Hz, 1H, H-4), δ_Η 5.58, (br m, 1H, H-5), and δ_Η 5.61 (dt, J= 4.3, 3.6 Hz, 1H, H-3)] were ascribable to a quinic acid esterified at C-3, C-4, and C-5. The ¹³C NMR data (Table 1) confirmed the presence of resonances due to two caffeoyl groups, an acetyl group, and an acyl-substituted quinic acid moiety. The location of the acetoxyl group at C-4 and the two caffeoyl groups at C-3 and C-5 were deduced by the interpretation of the Heteronuclear Multiple Bond Correlation (HMBC) experiment. Long-range cross-peaks were observed from the characteristic H-4 doublet-doublet signal at δ_Η 5.21 to the acetyl carbonyl (δ_C 171.9) that was confirmed to show ²J coupling with the acetyl methyl at δ_Η 2.01. In addition, the downfield shifts to the signals of H-3 and H-5 observed in the ¹H-NMR spectrum allowed us to assign the two caffeoyl groups to be attached at C-3 and C-5. These assignments were confirmed by the HMBC correlations from H-3 and H-5 to the carbonyl carbons of the two caffeoyl groups (δ_C 168.64, C-9’ and δ_C 168.14, C-9”, Figure 2). Moreover, the similarity of the ¹H NMR chemical shifts of H-3, H-4, and H-5 of 1 with those of 3,4,5-TCQA (16) indicated that attachment of caffeoyl or acetoxyl group does not differentiate the acylation effect.^{31, 32} Noteworthy, the ¹³C NMR chemical shifts of 1 and 16 did not show significant difference.³³ The absolute configuration of 1 was deduced to be as depicted in Figure 2 since the circular dichroism (CD) spectra of 1 and 4 (Figure S1.3 and Figure S3.1), both with caffeoyl moieties attached at C-3 and C-5, were superimposable. Collectively, the structure of 1 was deduced to be 4-acetyl-3,5-dicaffeoylquinic acid.

Table 1.

¹H (700 MHz) and ¹³C (175 MHz) NMR Spectral Data (CD₃OD) for 1 and 2.

	1a		2a
position	δC, type	δH (J in Hz)	δC, type	δH (J in Hz)
1	74.95, C		Not observed C
2ax	36.78, CH2	2.14 dd (5.1, 14.3)	32.85, CH2	2.41
2eq		2.41 dd (2.6, 14.3)		2.85
3	69.93, CH	5.61 dt (3.6, 4.3)	73.28, CH	5.39 dt (3.3, 3.1)
4	73.17, CH	5.21 dd (3.3, 8.6)	72.24, CH	3.92 dd (9.6, 3.6)
5	69.01, CH	5.58 br m	71.43, CH	5.50 td (10.1, 4.0)
6ax	38.87, CH2	2.27	38.86, CH2	1.94
6eq		2.27		2.60
7	177.55, C		Not observed, C
1’, 1”	127.78, 127.94, C		127.93, 127.96, C
2’, 2”	115.34, 115.38, CH	7.05 d (1.87) 7.06 d (1.87)	115.33, 115.28, CH	7.06 d (2.0) 7.08 d (2.0)
3’, 3”	149.82, 149.93, C		149.83, 149.97, C
4’, 4”	146.98, C		147.01, 147.07, C
5’, 5”	116.64, 116.65, CH	6.77 d (8.0) 6.78 d (8.0)	116.73, 116.65, CH	6.78 d (8.0) 6.79 d (8.0)
6’, 6”	123.24, 123.30, CH	6.96 dd (1.95, 8.5) 6.97 dd (1.95, 8.5)	123.28, 123.16, CH	6.97 dd (2.0, 8.2) 6.99 dd (2.0, 8.2)
7’, 7”	147.67, 147.83, CH	7.55 d (15.8) 7.61 d (15.8)	147.71, 147.44, CH	7.62 d (15.8)
8’, 8”	114.72, 115.20, CH	6.23 d (15.8) 6.31 d (15.8)	115.67, 115.24, CH	6.34 d (15.8) 6.31 d (15.8)
9’, 9”	168.14, 168.64, C		167.97, 168.88, C
OAc	171.99, C		172.6, C
OAc	20.90, CH3	2.01 br s	21.44, CH3	1.95 s

^aMeasured in CD₃OD

Figure 2.

Structures of compound 1 (with key HMBC correlations) and compound 2.

Compound 2 was isolated as a yellow powder, with neg-HRESIMS analysis showing a [M − H]^– ion peak at m/z 557.13093, consistent with the molecular formula C₂₇H₂₆O₁₃ (557.13006, calculated for C₂₇H₂₅O₁₃⁻). The IR spectrum exhibited stretching of conjugated carbonyl (1677 cm⁻¹) and absorption bands of benzyl (1597 cm⁻¹) and hydroxy (3380 cm⁻¹) functionalities. The ¹H NMR spectrum of 2 displayed resonances due to four trans alkene protons, two trisubstituted aromatic rings of two caffeoyl groups, one downfield shifted methyl singlet of an acetyl group (δ_Η 1.95, s, 3H), and four methylene protons at δ_Η 1.94 and δ_Η 2.60, and δ_Η 2.41 and δ_Η 2.85 (H-6ab and H-2ab, respectively (Table 1)). Signals at δ_Η 5.39 (dt, J= 3.3, 3.1 Hz, H-3) and δ_Η 5.50 (td, J= 10.1, 4.0 Hz, H-5) are assignable to those of esterified quinic acid. Irradiation of the resonance for H-4 at δ_Η 3.92 (dd, J= 9.6, 3.6 Hz) using the S1DT experiment generated the proton spin network containing resonances of H-3, H-5, and the methylene protons H-2 and H-6 of the quinic acid moiety. The mass spectrometric data together with the resonances observed in the ¹³C NMR data (Table 1) suggested the presence of quinic acid esterified at three different positions, although the chemical shifts of the quaternary carbon at C-1 and the carboxylic acid at C-7 were not observed. The downfield signals of protons H-3 and H-5 were determined to be a result of acylation shifts and confirmed acylation on positions C-3 and C-5. The proton signal at H-4 remained upfield and non-acylated, thus the three acyl groups (two caffeoyl and one acetyl groups) present in 2 must be attached at C-1, C-3, and C-5. Interpretation of the above data together with compounds having similar resonance profiles published in the literature³⁴ allowed the determination of compound 2 to be 1,5-dicaffeoylquinic acid with an acetyl group substituted at position 3. The location of the two caffeoyl groups at C-1 and C-5 with the acetyl group at C-3 were deduced as follows.

Inspection of the ¹H NMR data of C-1 acylated CQAs showed that the chemical shift difference between the most downfielded and most upfielded methylene signals (Δ δ_Η H-2ab, H-6ab) was ≥ 0.49 ppm compared to CQA derivatives without acylation at C-1.^{31–32, 41–46} Compound 2 has a Δδ_Η of 0.91 ppm between its methylene protons, nearly identical to the Δδ_Η value found from 1,3,5-TCQA (13) of 0.92 ppm.³⁴ Additionally, the CD spectrum for compound 2, which displayed a positive Cotton effect at 341 nm and negative Cotton effects at 285 and 212 nm, was superimposable to that of 1,5-dicaffeoylquinic acid (5) and the inverse to that of 3,5-dicaffeoylquinic acid (4) (Figures S3.3, S6.1, and S5.1), demonstrating that 2 and 5 possess the same substitution of caffeoyl chromophores that induce excitonic effect.^{16, 35, 36} This allowed us to conclude the acetoxyl group to be attached at C-3 as no acylation shift has been observed for the resonance due to H-4. Therefore, compound 2 was determined to be 3-acetyl-1,5-dicaffeoylquinic acid (Figure 2).

Since EtOAc was used during the liquid-liquid partition of the water extract to yield compounds 1 and 2, we conducted additional experimentation to understand if the new acetylated CQAs are artifacts. A second extraction and isolation procedure, free from acetylated solvents such as EtOAc and acetone, was performed. C. asiatica plant material (CA6) was suspended in water, filtered, and partitioned with hexanes to afford a defatted water fraction (CA6-W1) that was further partitioned with CHCl₃ to afford CHCl₃- (CA6-C) and water- (CA6-W2) soluble fractions. CA6-W2 was dried, suspended in MeOH, centrifuged, and the supernatant collected to yield a MeOH-soluble fraction, CA6-W2M. The precipitate from this fraction was then suspended in water, centrifuged, and the supernatant collected to afford fraction CA6-W2W. Targeted LC-HRMS experiments of the obtained fractions detected compounds 1 and 2 in the CHCl₃ (CA6-C), MeOH (CA6-W2M), and water (CA6-W2W) fractions of C. asiatica (CA6), indicating that these compounds occur naturally (Figure S7.1 – S7.3).

Compound 3 was identified as the known 3-caffeoyl-5-feruloylquinic acid (C₂₆H₂₆O₁₂) by the observation of the [M − H]⁻ ion peak at m/z 529.13556 in the neg-HRESIMS (529.13515, calculated for C₂₆H₂₅O₁₂⁻) and interpretation of its one- and two-dimensional NMR spectroscopic data. The ¹H NMR spectrum exhibited signals characteristics of C-3 and C-5 di-acylated quinic acids (δ_Η 5.41, dt, J=3.7, 3.5 Hz, 1H, H-3, and δ_Η 5.49 m, 1H, H-5) with a caffeoyl and feruloyl moiety. Irradiation of the signal of H-3 using S1DT identified that H-3 (δ_Η 5.41), H-4 (δ_Η 3.93), H-5 (δ_Η 5.49), and H-2,6 (δ_Η 2.05–2.35) belong to the proton spin network of 3,5-di-acylated quinic acid. The methoxy group at δ_Η 3.90 showed long-range correlation with the carbon singlet at δ_C 149.6 confirming the presence of a feruloyl group.³⁷ The assignment of the caffeoyl and feruloyl group to be attached at C-3 and C-5, respectively, was carried out by interpretation of its MS² fragmentation patterns since the NMR data were inconclusive. Hetero-disubstituted quinic acids have been characterized by their MSⁿ fragmentation patterns due to the differences in lability of the 3-position and 5-position esters, as reported by Clifford and coworkers.^{28, 38} Previous work shows that 3-feruloyl-5-caffeoyl quinic acid generated a base peak at m/z 367 while 3-caffeoyl-5-feruloyl quinic acid exhibited a base peak at 353.^{28, 39, 40} While fragmentation patterns are expected to be similar for different instruments using related fragmentation methods, drawing structural conclusions based on potentially minor differences in relative intensities must be done cautiously when not compared to authentic standards. By comparison to the fragment intensity data reported in literature, our MS-MS data speculated the identity of compound 3 to be 3-caffeoyl-5-feruloylquinic acid (Figure S3.3). However, additional experiments with positive controls are being conducted for confirmation before making structural assignments on the basis of MS² data. Compound 3 is a known natural product compound, though this is the first report of its isolation from C. asiatica. Additionally, Eze et. al. tentatively identified a feruloyl-CQA (1F-5CQA) found in a C. asiatica phenolic extract using UHPLC-ESI-QTOF-MS analysis⁴¹ but did not conduct further analysis to confirm its substitution pattern.

Library of Proton Spin Network Fingerprints (pSNF) of Centella asiatica for accurate identification and standardization of secondary metabolites

The Selective 1D-TOCSY (S1DT) experiment has been utilized to identify coupled proton spin networks in a molecule. Two-dimensional TOCSY has been coupled with HSQC for metabolite identification in natural products research and metabolomics.^{42, 43} Interpretation of two-dimensional NMR such as HSQC/TOCSY and 2D-TOCSY of complex mixtures however is challenging because of signal overlap. The ¹H NMR spectrum of a sample mixture represents a series of ¹H spin networks belonging to proton-bearing compounds present. Neighboring protons within a compound that are connected through J-couplings can be easily identified by exciting a defined proton resonance using the S1DT experiment to generate resonances of spin network-coupled protons of up to five or six bonds away.⁴⁴ The spin networks of coupled protons in each compound can thus be individually identified in a mixture by scanning characteristic resonances using S1DT. Moreover, splitting patterns of proton signals are conserved in TOCSY-generated spin networks, providing information about the geometrical orientation of protons and relative stereochemistry at the connected carbon. Thus, S1DT is capable of differentiating isomeric structures that could not be readily identified by many mass spectrometric methods. The molecular fingerprints of proton spin networks offer supplementary information that can be used to assist LC-MS for isomeric differentiation, dereplication, and accurate metabolomic identification.

Although the ¹H NMR of mixtures contain many signals, not all proton resonances in a defined spin system are always overlapped. Targeted irradiation of isolated signals using S1DT is an optimal method to clearly capture coupled proton resonances present in a sample mixture. More importantly, when the ¹H spin networks of molecules are well characterized and recorded in a library as fingerprints, S1DT can be used on an extract or mixture and can define compounds by their proton spin network fingerprints (pSNFs) when matched with the library-recorded data. We hypothesized that performing S1DT scanning on a range of proton resonances characteristic to those of known metabolites will generate a series of pSNFs that correspond with previously recorded pSNF library data, allowing for the accurate identification of isomers. Non-related signals that are generated by accidental irradiation of overlapping or neighboring signals may appear in the pSNF, but they can be readily identified and removed by focusing only on the data that match with the pSNF library of reference compounds.

We have probed this method using the botanical dietary supplement C. asiatica to identify pharmacophore-bearing constituents such as caffeoylquinic acids (CQAs) within the plant extract. CQAs have a characteristic pSNF arising from the quinic acid core that displays acylation shifts allowing the accurate assignment of positional isomers due to the various attachments of acyl groups. We first recorded the ¹H NMR of D-(–)-quinic acid ((1S,3R,4S,5R)-1,3,4,5-tetrahydroxycyclohexane-1-carboxylic acid), to determine the stereochemistry-dictated splitting patterns of each proton signal in the molecule. Methylenes H-2ab (ddd, J= 14.5, 3.9, 2.2 Hz and dd, J= 14.6, 3.3, 1H each) and H-6ab (dd, J= 13.2 and 10.8 Hz and ddd, J= 13.2, 4.6, 1.8 Hz, 1H each) of quinic acid resonated at δ_Η 2.02, 2.07, 1.86, and 2.13, respectively, while the resonances due to H-3, H-4 and H-5 were at δ_Η 4.09 (td, J= 3.4, 3.4 Hz), δ_Η 3.39 (dd, J= 9.1, 3.2 Hz), and δ_Η 3.99 (ddd, J= 10.8, 9.1, 4.6 Hz), respectively (Figure 3). The methylene signals of H-6 can be easily differentiated from H-2 since they display a larger J coupling value due to its axial-axial coupling with H-5.

Figure 3.

¹H NMR spectrum of 18 (in CD₃OD, 400 MHz).

The resonances of H-3, H,4, and H-5 of quinic acid (18) shift downfield without changes of their splitting pattern when their oxygen-bearing carbons are acylated. The location of acyl groups at C-3, C-4, and/or C-5 can be thus determined by the downfield shift of their corresponding signals. For instance, the ¹H NMR chemical shift of H-5 of 5-caffeoylquinic acid (12) resonated at 5.32 ppm, which was shifted about 1.3 ppm downfield compared to that of H-5 of free quinic acid (δ_Η 3.99, H-5), indicating C-5 acylation.⁴⁴ Since C-1 of CQAs are not protonated and acylation shifts are not obvious in ¹³C resonances, the chemical shifts of H-2 and H-6 were used for assignments as they are always affected by the presence of an acyl group at C-1. At least one of the methylene proton signals (H-2ab and/or H-6ab) shifted downfield when a caffeoyl group or an acyl derivative is attached at the axially-oriented hydroxy group at C-1. Therefore, the difference of the most downfielded and most upfielded signals of H-2 and H-6 of each compound were calculated to detect the presence of C-1 acylation. Results showed that compounds with acylation at C-1 displayed chemical shift differences (Δ δ_Η) of ≥ 0.49 ppm between H-2 and H-6 resonances. This can be observed in the ¹H spin network spectrum of compound 5 in Figure 5c. From these observations, we conclude that pSNFs give information about the stereochemistry and substitution patterns of important secondary metabolites in Centella asiatica and can assist mass spectrometry in identifying isomers present in complex mixtures such as botanical dietary supplements. To develop the identification method, we focused on CQAs and have collected pSNFs of acylated quinic acids isolated during the present study and those published in the literature in order to observe patterns that can differentiate CQAs present in C. asiatica sample material.

Figure 5.

a) ¹H NMR of the EtOAc fraction of a C. asiatica water extract (CAW-E) along with pSNFs of CQAs b) 4, c) 5, d) 6, e) 7 generated from irradiation using S1DT.

Here we describe efforts on the generation of a library of pSNFs that can characterize CQAs present in Centella asiatica. pSNFs for the resonances of H-2, H-3, H-4, H-5, and H-6 were generated using the S1DT NMR experiment on isolated compounds (1−7) and acylated quinic acids published in the literature. pSNFs of quinic acid (18) and seventeen acylated quinic acids (1−17) were recorded and results showed that the chemical shifts of protons H-3, H-4, and H-5 of the quinic acid resonate below δ_Η 4.60 when C-3 or C-4, or C-5 were non-acylated, while chemical shifts above δ_Η 4.80 indicate attachment of acyl groups (Figure 4). We concluded that quinic acids acylated at C-3 displayed ¹H resonances ranging from δ_Η 5.3 to δ_Η 5.7, while acylation of C-4 and C-5 exhibited H-4 and H-5 signals ranging from δ_Η 4.8 to δ_Η 5.3 and δ_Η 5.3 to δ_Η 5.8, respectively.

Figure 4.

Illustration of proton NMR chemical shift values of H-3, H-4, and H-5 of quinic acid and acylated derivatives. The upper half of the chart (turquoise) distinguishes positions which are acylated whereas the lower half (pale yellow) shows non-acylated positions. All ¹H NMR data were recorded in CD₃OD.

*Synthesized in-house

For references: ^aZhu et. al. (2009)⁴⁵, ^bSefkow et. al. (2001)⁴⁶, ^cLin et. al. (2020)⁴⁷, ^d,iIndy Tamayose et. al. (2019)³³, ^eChen et. al. (2014)⁴⁸, ^fAgata et. al. (1993)³⁴, ^gMerfort (1992)⁴⁹, ^hKim et. al. (2012)⁵⁰

To test the newly developed pSNF library in the identification of CQAs in C. asiatica, we recorded the ¹H NMR of the EtOAc fraction of the bioactive water extract (CAW-E) that has been shown to be rich in CQAs by mass spectrometric data (Table S1). Selective 1D-TOCSY scanning at the expected frequency ranges of protons H-3, H-4, and H-5 of CQAs (δ_Η 5.00 to δ_Η 5.45, Figure 5a) generated pSNFs corresponding to CQA spin networks previously recorded in the library, allowing for their identification. For instance, S1DT irradiation scanning at δ_Η 5.451 (H-5), δ_Η 5.406 (H-5), δ_Η 5.136 (H-4), and δ_Η 5.017 (H-4) generated pSNFs readily identified to be those of 4, 5, 6, and 7, respectively (Figure 5) after comparison with the library data (Figure 4). Additionally, the Δ δ_Η values of H-2 and H-6 can be used to determine if C-1 is acylated. As observed in Figure 5c, the chemical shifts of the observed H-2 and H-6 protons of compound 5 showed Δδ_Η of ≥ 0.49 ppm, indicating that the C-1 position must be acylated. In comparison, the compounds with no acylation at C-1 (4, 6, and 7) have Δδ_Η2/Η6 values less than 0.49 ppm (Figure 5b, 5d, 5e).

This research analyzes the TOCSY-generated pSNFs of CQAs and provides an initial proof-of-concept for metabolite characterization in C. asiatica samples using pSNFs. The methodology describes the downfield shifts of ¹H NMR signals in the quinic acid spin system containing protons of acylated carbinol substituents. This work has led to the development of a C. asiatica-specific pSNF library capable of differentiating 17 substituted QAs and ongoing work utilizes this methodology to include the botanical’s bioactive triterpenoid isomers and their glycosides. Further standardization efforts and quantitation studies are to be conducted using ultra-highfield NMR instrumentation such as the 1.2 GHz (28.2 T) NMR spectrometer available at The Ohio State University. With high resolution spectra, fingerprints generated from overlapped signals can be analyzed and readily assigned by comparing with pSNFs already recorded in the library. Outside of botanical specific libraries, the long-term goal is to expand the pSNF database for broader applications in metabolomics and dereplication.

To conclude, the use of S1DT to generate pSNFs is a novel technique that can complement LC-MS to accurately identify CQA isomers. This concept is most effectively implemented in instances where LC-MS alone is not sufficient and additional structural information is needed for metabolite authentication purposes. Accurate identification of the pharmacophore-bearing secondary metabolites in C. asiatica using LC-MS/pSNFs will contribute to batch-to-batch consistency and standardization of a botanical formulation that is suitable for clinical studies.

EXPERIMENTAL SECTION

General Experimental Procedures.

Optical rotation ([α]_D) of samples were measured at 20 °C using an Anton Paar MCP 150 polarimeter (Anton Paar OptoTec GmbH, Seelze-Letter, Germany). Ultraviolet (UV) absorption spectra were measured in a 1 cm quartz cuvette using a Hitachi U-2910 UV/Vis double-beam spectrophotometer (Hitachi High-Technologies America, Schaumburg, IL, USA). Circular dichroism (CD) spectra were obtained on a Jasco J-810 spectropolarimeter (Jasco Products Company, Oklahoma City, OK, USA) and measured in a 1 mm CD cell. Infrared (IR) spectra were obtained on a Nicolet 6700 FT-IR spectrometer (Thermo Scientific, Waltham, MA, USA). 1D (¹H, ¹³C, selective gradient TOCSY and NOESY) and 2D (COSY, HSQC, and HMBC) NMR spectra were recorded at 300 K on Bruker Avance III HD 400 MHz (9.4 T), Bruker Avance III HD Ascend 700 MHz (16.4 T), and Bruker Avance III HD 800 MHz (18.8 T) instruments (Bruker, Billerica, MA, USA) using standard Bruker pulse sequences and referenced with residual solvent signals (CD₃OD δ_H/C 3.31/49.15 ppm). NMR data processing and analysis were completed on TopSpin software version 4.0.7 (Bruker, Billerica, MA, USA). High-Resolution ElectroSpray Ionization Mass Spectra (HRESIMS) and MS² data were acquired on a Q-Exactive Orbitrap (Thermo Electron) using a HESI II probe for negative mode ESI ionization. Details on the acquisition of mass data of each compound are described in the Supporting Information. Data were processed using Thermo Scientific FreeStyle software. Column chromatography procedures were conducted on Teledyne ISCO CombiFlash Rf+ (Teledyne ISCO, Lincoln, NE, USA) with columns and flow rates as described below. Analytical thin-layer chromatography (TLC) was performed on aluminum-backed silica gel plates (200 μm thickness) purchased from Silicycle Inc. High-performance liquid chromatography (HPLC) was performed on a Hitachi Primaide HPLC (Hitachi High-Technologies Corporation, Tokyo, Japan) equipped with a Primaide 1430 diode array detector, a Primaide 1210 autosampler, and a Primaide 1110 pump with degasser. Separation and collection were carried out using a reversed-phase semi-preparative column (Phenomenex Luna C₁₈(2), 100Å, 10 × 250 mm x 5 μm; Phenomenex Inc., Torrance, CA, USA), equipped with a guard column. Elution was conducted with a mobile phases and flow rates as indicated in the Extraction and Isolation section. All chemical solvents were acquired from Fisher Scientific. Compound 18 was purchased from Sigma Aldrich.

Plant material.

Centella asiatica plant material (CA6) and hot water extract (CAW-7) were provided for this project by the BENFRA Botanical Dietary Supplements Research Center, Oregon Health & Science University, Portland, OR, USA. A commercial sample of Centella asiatica herb (batch number X20090016) consisting of dried aerial parts was obtained by the BENFRA Center from Oregon’s Wild Harvest (OWH, Redmond, OR, USA). The material was authenticated as Centella asiatica by thin layer chromatographic comparison with botanical reference material from the American Herbal Pharmacopoeia, and genetic fingerprinting, both performed at the BENFRA Center. Voucher samples of the plant material are deposited in the BENFRA Center laboratories (code number BEN-CA-6) and at the Herbarium at Oregon State University (code number OSC-V-265416). BEN-CA-6 was extracted by boiling the herbal material in hot water, followed by filtration and lyophilization of the filtrate as described previously,²⁹ to produce a dried water extract (CAW-7; yield 20.5% w/w of plant material).

Extraction and Isolation.

Preliminary biological and quantitative LC-MS data (Figure 1 and Table S1) were conducted on CAW-7 and its EtOAc and water fractions (CAW-E and CAW-W, respectively). Liquid-liquid partition of 1g of CAW-7 dissolved in distilled water (250 mL) using EtOAc (~250 mL, twice) afforded CAW-E (395 mg) after evaporation while the remaining water fraction was evaporated to give CAW-W (960 mg). The compounds reported in this paper were isolated from C. asiatica plant material (CA6) provided by BENFRA. Most in vitro and in vivo studies that affirm the biological activity of C. asiatica focus on the water-soluble fraction of the plant (CAW), prepared by refluxing the raw plant material for up to two hours. For this study, dried plant material (CA6) was suspended in warm water and partitioned with EtOAC to yield two fractions (CA6-W and CA6-E). This extraction procedure was used to 1) facilitate the isolation of metabolites by concentrating based on their solubilities and 2) discover any constituents that may be absent in CAW samples due to degradation from reflux conditions. Approximately 110 g of dried plant material was suspended in water at 50°C for 30 minutes before partitioning to exhaustion with EtOAc. The partition procedure yielded 39.6 g and 4.1 g of the water partition (CA6-W) and the EtOAc partition (CA6-E), respectively. CA6-E (1 g) was prepared with solid sample loading on celite and subjected to reversed-phase Combiflash Rf (Biotage SNAP Cartridge KP-C18-HS, 60g column) using a MeOH/H₂O mobile phase gradient method (20% MeOH isocratic over 3 min, 20% to 100% MeOH over 22 min, 100% MeOH isocratic over 5 min, 48 mL min-1 flow rate) monitored at 254 nm and 365 nm. The initial run was followed by a 100% MeOH isocratic wash (7 min), 100% CHCl₃ isocratic wash (5 min), and 100% MeOH isocratic wash (5 min). After TLC analysis, the Combiflash collections were combined to yield a total of 17 fractions (CA6-E1 – E17).

Compounds 2 and 6 as well as fraction CA6-E3–1 were separated from fraction CA6-E3 of the EtOAc partition (31 mg/mL) on reversed-phase semi-preparative HPLC using MeCN/H₂O gradient with 0.1% formic acid (FA) at a 3 mL min⁻¹ flow rate. Chromatographic separation was carried out as follows: 15% MeCN+FA/H₂O+FA isocratic over 7 min, 15% to 20% MeCN+FA over 1 min, 20% MeCN+FA isocratic over 21 min, 20% to 25% MeCN+FA over 1 min, 25% to 30% MeCN+FA over 20 min, 30% to 100% MeCN+FA for 1 min, 100% MeCN+FA isocratic over 8 min, with monitoring at 330 nm.

Compounds 4 and 5 were purified from fraction CA6-E3–1 (28 mg/mL) on a reversed-phase column using H₂O with 0.1% trifluoroacetic acid (TFA) and MeOH (30% to 40% MeOH/H₂O+TFA over 10 min, 40% MeOH isocratic over 10 min, 40% to 50% MeOH over 15 min, 50% to 60% MeOH over 6 min, 60% to 100% MeOH over 1 min, 100% MeOH isocratic for 5 min, 100% to 30% MeOH over 1 min, 30% MeOH/H₂O+TFA isocratic for 5 min, 2 mL min⁻¹ flow rate) with monitoring at 330 nm.

Compounds 1, 3, 4, 5, and 6 were isolated from fraction CA6-E4 of the EtOAc partition (31 mg/mL) that was subjected to reversed-phase semi-preparative HPLC using H₂O with 0.1% TFA and MeOH (30% to 40% MeOH/H₂O+TFA over 10 min, 40% MeOH isocratic over 10 min, 40% to 50% MeOH over 31 min, 50% MeOH isocratic over 5 min, 50% to 100% MeOH over 1 min, 100% MeOH isocratic for 5 min, 100% to 30% MeOH over 1 min, 30% MeOH/H₂O+TFA isocratic for 5 min, 2 mL min⁻¹ flow rate) with monitoring at 330 nm.

4-acetyl-3,5-dicaffeoylquinic acid (1). pale yellow powder; [α]²⁰_D −77 (c 0.1, EtOH); UV (MeOH) λmax (log ε) 201 (4.11), 218 (3.99), 246.5 (3.77), 298.5 (3.82), 332 (3.93) nm; CD (MeOH) λmax (Δε) 343 (−3.74), 287 (+1.31), 226 (−0.35), 208 (+1.06) nm; IR (diamond ATR cell) ν_max 3384, 2924, 2852, 1709, 1598, 1370, 1180, 1019 cm−1; ¹H and ¹³C NMR see Table 1; HR-ESIMS m/z 557.13028 [M - H]⁻ (calcd for C₂₇H₂₅O₁₃, 557.13006).

3-acetyl-1,5-dicaffeoylquinic acid (2). pale yellow powder; [α]²⁰_D 31 (c 0.1, EtOH); UV (MeOH) λmax (log ε) 201.5 (4.25), 218 (4.20), 245.5 (4.01), 297.5 (4.07), 331.5 (4.19) nm; CD (MeOH) λmax (Δε) 341 (+4.26), 285 (−0.83), 212 (−9.09) nm; IR (diamond ATR cell) ν_max 3380, 2924, 2853, 1677, 1597, 1378, 1182, 1050 cm−1; ¹H and ¹³C NMR see Table 1; HR-ESIMS m/z 557.13093 [M - H]⁻ (calcd for C₂₇H₂₅O₁₃, 557.13006).

Data Availability Statement

The NMR data for compounds 1 and 2 have been deposited in the Natural Product Magnetic Resonance Database (NP-MRD; www.np-mrd.org) and can be found at NP0350766 (Compound 1) and NP0350755 (Compound 2).