Investigation of Red Clover (Trifolium pratense) Isoflavonoid Residual Complexity by Off-line CCS-qHNMR

Abstract

The importance of Trifolium pratense L. as a dietary supplement and its use in traditional medicine prompted the preparation of a thorough metabolite profile. This included the identification and quantitation of principal constituents as well as low abundant metabolites that constitute the residual complexity (RC) of T. pratense bioactives. The purity and RC of isoflavonoid fractions from standardized red clover extract (RCE) was determined using an off-line combination of countercurrent separation (CCS) and two orthogonal analytical methodologies: quantitative ¹H NMR spectroscopy with external calibration (EC-qHNMR) and LC-MS. A single-step hydrostatic CCS methodology (Centrifugal Partition Chromatography [CPC]) was developed that fractionated the isoflavonoids with a hexanes-ethyl acetate-methanol-water (HEMWat) 5.5/4.5/5/5, v/v solvent system (SS) into 75 fractions containing 3 flavonolignans, 2 isoflavonoid glycosides, as well as 17 isoflavonoids and related compounds. All metabolites were identified and quantified by qHNMR spectroscopy. The data led to the creation of a complete isoflavonoid profile to complement the biological evaluation. For example, fraction 69 afforded 90.5% w/w biochanin A (17), with 0.33% w/w of prunetin (16), and 0.76% w/w of maackiain (15) as residual components. Fraction 27 with 89.4% w/w formononetin (13) as the major component had, in addition, a residual complexity consisting of 3.37%, 0.73%, 0.68% w/w of pseudobaptigenin (11), kaempferol (10) and pratensein (8), respectively. Despite the relatively high resolving power of CPC, and not unexpectedly, the chromatographic fractions retained varying degrees of the original metabolomic diversity. Collectively, the extent of metabolomic diversity should be recognized and used to guide the development of isolation strategies, especially when generating samples for bioactivity evaluation. The simultaneous structural and quantitative characterization enabled by qNMR, supported by LC-MS measurements, enables the evaluation of a relatively large number of individual fractions and, thereby, advances both the chemical and biological evaluation of active principles in complex natural products.

Graphical Abstract

1. Introduction

Virtually all natural product (NP) extracts contain complex mixtures of metabolites created by the intricate maze of the biosynthetic pathways [1] and embedded into “matrix compounds” and artifacts [2] that tend to receive less attention. The network of biosynthetic pathways generates a brew of various NP classes, each including many congeners, whose identities and quantities are seldom fully known, whereas they typically differ in their pharmacological activities [3]. A cursory examination of crude extracts will only reveal a handful of major components, which tend to mask the underlying proliferation of low abundant metabolites [4]. This approach may be characterized by turning the idiom of “not seeing the forest for the trees” into “not seeing the trees for the forest”. Taking this idiomatic perspective further: once one sees the trees, the shrubbery under the trees may still be hidden, and underneath it, further plants and also other organisms reside. In other words, (too) little attention is generally paid to low abundance metabolites, which can be referred to as “residual complexity” within the forest of complex metabolomic mixtures [5].

The concept of residual complexity (RC) was established to clarify the relationship between impurities in materials from natural sources and their corresponding biological profile (https://go.uic.edu/residualcomplexity). RC is divided into two forms: static and dynamic. Static RC is essentially a chemical record of the provenance and subsequent manipulations of a sample as revealed by its components [6–8]. Even highly purified single chemical entities retain minor components as nearly-impossible-to-separate metabolites that occur as “minor” or “underlying” peaks in chromatograms or spectra. In contrast, dynamic RC is generated by degradation, rearrangement(s), inter/intramolecular reaction chemistry, isomerization or epimerization of any species contained in a given sample [6–8]. Importantly, careful identification and quantitation of the chemical composition and variation of a sample creates an RC profile that is essential for understanding the bioactivity associated with the sample [5–9].

The knockout extracts (KOEs) methodology target known metabolites which have been previously identified as biological active compounds in crude extracts. By selectively depleting/enrichment of these metabolites, KOEs become possible to evaluate the role of minor constituents and/or matrix components that are often considered “inert” materials [10,11,14]. On the other hand, depleting the concentration of major compounds in complex mixtures, through separation technologies, will reveal other metabolites whose analytical or biological signals were overlapping with the major compounds and/or occupying the baseline noise. Separation technology must be summoned to develop methodology able to distribute crude natural product extracts in such a way as to allow the subtraction of major metabolites. A benign, non-destructive separation technology is imperative for this process in the event that the resulting fractions will be submitted for biological activity testing. For this purpose, countercurrent separation (CCS) is a method with preparative loading capacity, good resolution, reproducible yield, and practically complete recovery of all sample components (i.e., it lacks irreversible absorption properties) [12–14]. Thus, CCS also has a higher sample recovery than conventional liquid-solid chromatography, being essentially 100% [13,14]. The main hydrodynamic variant of CCS, high-speed countercurrent chromatography (HSCCC) shares these key properties with a different mixing and settling mechanism [15].

The Generally Useful Estimate of Solvent Systems (GUESS) methodology is an empirical and laboratory practice-oriented approach to the prediction of CCS optimal liquid-liquid partition coefficients (K) [16]. GUESS utilizes thin layer chromatography (TLC), which is widely practiced including for preparative fraction monitoring, to accelerate matching of the partition-based polarity of known and unknown analytes with biphasic CCS solvents systems. The separation resolution of an analyte in CCS depends on its liquid-liquid partition coefficient (K) relative to the K values of the other components in the mixture. An ideal biphasic solvent system (SS) will deliver an analyte into a distinctive range of K values, referred to as the “sweet spot” of optimal performance. This range is between K = 0.4 and K = 2.5 [16] because large K values tend to produce excessive sample band broadening [17]. However, the sweet spot range can be extended to K = 0.25 to K = 16 with elution extrusion methodology [18].

HSCCC has previously been applied to the isolation of 13 and 17 from the crude ethanolic extract of chickpea sprouts. In that study, the final purity of 13 (92.3%) and 17 (95.9%) was determined by HPLC [19]. Various SSs have been developed in order to isolate important bioactive isoflavones, such as 6 and 9. For instance, Shi described the purification of 6 and 9 using HSCCC, with petroleum ether-ethyl acetate-methanol-water (2/3/5/4, v/v) and (2/3/2/3, v/v) SSs. These two SSs provided an excellent purification yield with K values for target compounds between 0.5 and 2.0 [20]. Another example applied HSCCC with n-hexane-ethyl acetate-ethanol-water (5/8/8/10, v/v) for the purification of 6, 9, 14, 15, and 17; all have been reported as active compounds in red clover extract [21]. In a further case, a preparative-scale approach was developed with two SSs methyl tert-butylether-acetonitrile-water (2/2/3, v/v) and (6/3/8, v/v), leading to the isolation of 6 and 9 from soy flour [22]. An earlier proposed method used ultrasound-assisted dynamic extraction (UADE) coupled with two HSCCC systems in parallel. The n-hexane-ethyl acetate-ethanol-water (0.62/1.00/0.99/1.25, v/v) SS yielded 6, 9, 14, 15, and 16 from red clover extract [23]. On the basis of these prior examples, it is possible to conclude that CCS combined with orthogonal analytical techniques, such as quantitative NMR, UHPLC-UV, and LC-MS, represent powerful tools for metabolite profiling of red clover extracts [9,24].

Red clover (Trifolium pratense L.) is an herb that belongs to the Fabaceae (legume) family. In herbal medicine, red clover is used to treat asthma, bronchitis, eczema, arthritis and women’s health problems such as menopausal symptoms [25,26]. T. pratense documentary and reference standards are included in the United States Pharmacopoeia [27]. Bioactive isoflavones from red clover, particularly daidzein (6), genistein (9), formononetin (13), and biochanin A (17) have shown promise as candidates as bioactive markers that rationalize the use of red clover for the amelioration of menopausal symptoms in recent clinical trials [28,29], while a standardized red clover extract did not defeat the null hypothesis in a trial conducted by our Botanical Center [30]. However, it remains a general consensus in the literature that isoflavonoids may contribute desirable therapeutic properties for (pre-)menopausal women, because of the affinity of isoflavonoids for estrogen receptors [25,26]. Still, with disclosures of the side effects of estrogen supplementation, careful attention should be paid to the dose limitations of these (dietary) interventions [31].

One broader goal of this work was evaluate the pros and cons of a well-established paradigm in the analysis of bioactive NPs: the combination of fractions that exhibit compounds with similar TLC R_f values prior to biological evaluation. Thus, by analyzing all individual fractions both qualitatively and quantitatively using (q)NMR, the study aimed at gaining a better understanding of the RC and structural diversity of the eluents as well as obtain reliable measures of the extent of chromatographic peak overlap that is unavoidable during the preparative fractionation schemes.

The present study focused on the development of a preparative CPC method for the purification and characterization of major and minor isoflavonoids and their “adjacent” metabolites with similar structures and/or chromatographic characteristics. The purity and RC of the CPC-derived materials (fractions, compounds) was determined using the 100% qHNMR method, as well as an absolute qHNMR method with DMSO₂ and caffeine as external calibrants (ECs). This study also evaluated the potential of combining LC-MS and qHNMR as orthogonal analytical methods for analyzing RC in every fraction separately. A total of 75 CPC fractions were processed and expressed purity/content in %w/w using both methods. In summary, two isoflavonoid glycosides as well as 17 isoflavonoid aglycones and related compounds were identified and quantified. Furthermore, three tricin derivatives (B-ring methoxylated flavonoid aglycones) were identified from red clover for the first time.

2. Experimental

2.1. Materials

Chemical solvents and reagents were purchased from Fisher Scientific (Hanover Park, IL, USA); HPLC grade solvents from Sigma-Aldrich (St. Louis, MO, USA); and DMSO-d₆ (99.9 atom % D) from Cambridge Isotope Laboratories Inc. (Andover, MA, USA). Isoflavone standards were purchased from Indofine Chemical Co. (Hillsborough, NJ, USA). Standardized red clover extract (RCE) was a T. pratense auto-hydrolyzed, enriched extract manufactured by PureWorld Botanicals Inc. (South Hackensack, NJ, USA, acquired by Naturex in 2005). As described earlier, it contained 61.7% total isoflavonoids, with high concentrations of daidzein (6), genistein (9), formononetin (13), irilone (14), prunetin (16), and biochanin A (17) [25,30,32].

The external calibrants, dimethylsulfone (Code N. 048-33271; 100%, CRM), and caffeine (CAS-N. 5808-2; 99.2%, by qHNMR/100% method) were purchased from Wako Pure Chemical Ind., Ltd. (Osaka, Japan) and Sigma-Aldrich, respectively. The internal calibrant, 3,5-dinitrobenzoic acid (CAS-N. 99-34-3; 99.3% by qHNMR/100% method) was purchased from Fluka Analytical (Buchs, Switzerland). A gastight (1.0 mL) Pressure-Lok syringe from VICI Precision Sampling (Baton Rouge, LA, USA) was used for volumetric qNMR sample preparation.

Analytical TLC was performed on Alugram precoated 0.2 mm thick silica gel G/UV254, 10×20 cm aluminum plates (Macherey-Nagel GmbH & Co, Düren, Germany). Syringe filters (CHROMAFIL Xtra PTFE-20/13, pore size: 0.20 μm, 13 mm diameter, Macherey-Nagel) were used for HPLC sample filtration.

2.2. Instrumentation

The qHNMR spectroscopy experiments were performed on a JEOL RESONANCE Inc. (Akishima, Tokyo, Japan) JNM-ECZ400S NMR spectrometer, with the sample temperature regulated at 25°C (298 K) equipped with a SuperCOOL liquid N₂ cryoprobe. The NMR data were processed and analyzed with JEOL Delta v5.0.4.4 NMR processing software and/or MestReNova 14.1.0 software from Mestrelab Research S. L. (Santiago de Compostela, Spain).

The SCPC-250-B centrifugal partition chromatography (CPC) extractor from Gilson Inc. (Madison, WI, USA) had a 264 mL total volume, 220 mL of which is attributed to cell volume (12 disks; twin cell design; 20 cells per disk). The instrumentation was operated at a flow rate of 50 mL/min and a rotation speed of 2,500 rpm, with a pressure maximum of 300 bar. The Spot Prep peripheral solvent delivery and operating system was equipped with a 50 mL sample loop, binary pump, a four-channel UV/VIS detector, a sample collector, and data collection software. A Shimadzu (Kyoto, Japan) Nexera UHPLC-UV system was equipped with a DAD and employed a Phenomenex (Torrance, CA USA) Kinetex 1.7 μm XB-C₁₈ 100 Å column (2.1 × 50 mm). Data analysis was performed using the Shimadzu LabSolutions software package.

LC-MS-MS analysis was carried out using a Waters 2695 (Milford, MA USA) solvent delivery system connected to a Waters SYNAPT quadrupole/time-of-flight (Q/ToF) mass spectrometer operating in the negative ion electrospray mode.

2.3. Selection of the CPC solvent system (SS)

The GUESS method of SS selection was performed as described earlier [12,16]. This method has been classified as a semi-empirical strategy as it involves a TLC-based method for estimating the K values of target compounds [12]. Target compounds 13 and 17 were developed on a normal phase TLC plate (5 × 4 cm) with the organic phase of various biphasic SSs. Developed TLC plates were analyzed under UV 254 nm, then in a second step, sprayed with vanillin reagent, and heated at 120°C until the spots were revealed.

2.4. Liquid-liquid partitioning experiments

A biphasic SS was formulated and equilibrated in a separatory funnel at room temperature. Suitable quantities of 13 (1.3 mg) and 17 (1.4 mg) were added to 20.0 mL vials. Then, 8.0 mL of each phase were delivered to the vials. The contents were mixed thoroughly and separated into two layers. Equal volumes of upper phase and the lower phase were dried overnight in a miVac speed vacuum system (SpeedVac, Model: QUC-12060-C00, Genevac LTD. Ipswich, England). Each dried phase was redissolved in 1.0 mL of HPLC-grade methanol for UHPLC-UV analysis.

A UHPLC method was developed for a total of 11 min analysis time. Solvents A and B were water and acetonitrile respectively, both containing 0.1% formic acid. The mobile phase gradient consisted of 20–26% B in 0.5 min, 26–31.5% B in 6.5 min, held isocratic at 31.5% B to 6.7 min, re-equilibrated from 31.5–20% B at 8 min, and reconditioning at 20% B to 11 min. The flow rate was 0.6 mL/min, and the injection volume was 1.0 μL. The column oven, detector cell, and autosampler temperature were maintained at 40°C, 40°C and 4°C, respectively, throughout the analysis.

The K (partition coefficient) value was expressed as the peak area ratio of the upper and lower phases (K = A_UP/A_LP).

2.5. LC-MS-MS methodology

Separations were carried out using a Waters YMC C₁₈ column (2 × 100 mm, 3 μm particle size). The mobile phase consisted of 0.1% formic acid on water (solvent A) and acetonitrile (solvent B) with a linear gradient from 30% to 95% B in 25 min. The flow rate was 0.2 mL/min, and the column temperature was set at 30°C. High-resolution accurate mass measurements were performed on Waters SYNAPT hybrid quadrupole/time-of-flight mass spectrometer operating in the negative ion electrospray mode. The measurements were carried out at a resolving power of 10, 000 FWHM using leu-enkephalin as the mass lock. Product ion spectra were recorded at 15 or 25 eV using Ar as the collision gas. For compound identification, molecular compositions and tandem mass spectra were compared with the standard spectra from the public and in-house generated databases as well as with spectra published in the primary literature [25,32].

2.6. Centrifugal partition chromatography (CPC) fractionation

A 234.2 mg portion of RCE was dissolved, with sonication, in equal amounts of upper (UP) and lower phase (LP) to a total volume of 46 mL (Put the 234.2 mg of RCE in 23 mL of UP overnight next day adding the 23 mL of LP slowly under sonication). The sample was then filtered and loaded into the 50 mL sample loop. The residue remaining after filtration was weighed in order to determine the effective solubility of RCE in the SS. The CPC column was filled with the (upper) stationary phase at 50 mL/min and a rotation speed of 500 rpm. Pre-equilibration of the column with the mobile phase occurred over 10 min with the same flow rate and a rotation speed of 2500 rpm. The stationary phase volume retention ratio (S_f) was 80% as determined by collecting the displaced stationary phase in a graduated cylinder. After introducing the RCE sample into the column, elution (15 min) was performed in descending mode. The CPC experiment was performed at room temperature without additional temperature control. Sweep elution and extrusion (15 min) were initiated by pumping the upper phase into the column. Fractions (10.0 mL each) were collected in 150 previously weighed test tubes.

All fractions were dried in a centrifugal vacuum evaporator. TLC and NMR were performed on the contents of each test tube. Fractions 22–40 were found to be poorly soluble in DMSO-d₆, therefore each dry sample was redissolved in 1.0 mL of methanol, then a 5.0 mL aliquot of solution was taken for drying and qHNMR spectroscopic analysis (Table S1).

2.7. qHNMR spectroscopic analysis

Prior to NMR analysis, each sample was dried under vacuum (<1 mbar) in a desiccator overnight to remove residual solvents. Samples were dissolved in 250 μL DMSO-d₆ measured with a 1000 μL analytical syringe. A 200 μL aliquot was delivered into a 3-mm tube (NORELL, Landisville, NJ, USA) with a calibrated glass pipette.

The qHNMR spectra of the RCE fractions were acquired using standard qHNMR spectroscopy conditions [33–35], which included a 15 ppm spectral window (SW), 5 ppm offset (O1P), a 60 s acquisition delay (D1), 46 as receiver gain (RG), 32 scans (NS), a 90° (6.4 μs) flip angle (P1), and automatic gradient shimming. Data processing utilized a Lorentzian-Gaussian window function (exponential factor −0.3, Gaussian factor 0.05 in GF mode) and fifth order polynomial baseline correction. The residual DMSO-d₅ signal at 2.500 ppm was used for chemical shift referencing. The preparation of external and internal calibrants (EC and IC, respectively) as well as tests for establishing linearity have been described previously [14,36] (Fig. S22, Supporting Information, describes the IC vs EC calculation of biochanin A and formononetin in RCE). Limit of quantitation (LOQ) and detection (LOD) parameters were taken from previous articles [14,36]. The S/N ratio calculators in MestReNova and JEOL Delta Processing Software were used to determine the S/N of individual signals.

3. Results and Discussion

3.1. Solvent system (SS) selection

The GUESS method of matching a compound’s behavior in TLC with its CCS properties was employed to select the SS for the CPC separation with 13 and 17 as target compounds (Table 1) (for further information of GUESS, see [16,37–39]). The SS hexanes-ethyl acetate-methanol-water (HEMWat) (6/4/5/5, v/v) gave TLC R_f values below 0.5 and HEMWat (5/5/5/5, v/v) gave values above 0.5. Therefore, an optimized HEMWat (5.5/4.5/5/5, v/v) SS was formulated that gave R_f values closer to 0.5. Previous work has indicated that the polarity match of the TLC-based method is more important than the difference between two compared R_f values, interpreted as apparent resolution [37].

Table 1.

The SS for the countercurrent separation (CCS) method developed in this study was selected by the GUESS method, using the organic phase of the biphasic SS as the TLC mobile phase. GUESS methodology links the standard TLC with the liquid-liquid partition coefficient (K) values. R_f. values around 0.5 are the best choice for a success SS developing.

HEMWat	HEMWat, v/v	Biochanin A 17	Formononetin 13
SS#		Rf
−1	6/4/5/5	0.47	0.33
−0.5	5.5/4.5/5/5	0.51	0.42
0	5/5/5/5	0.63	0.55
+1	4/6/5/5	0.88	0.69

Subsequent liquid-liquid partitioning experiments with HEMWat (5.5/4.5/5/5, v/v) showed that the K values of 13 and 17, were 1.18 and 1.25, respectively. The experimental values were found to be K = 0.5–1.12 and 2.03–3.68 for 13 and 17, respectively. These K values were taken from the data presented in Fig. 2A (fractions 20 to 32 for 13 and 52 to 83 for 17), and calculated using known equations [16]. Actually, experimental K values are influenced by the degree of mixing in the CPC chambers, interactions between components of a mixture, extra column volumes, and the possible loss of stationary phase during elution. Initial SS determination must be followed up by additional experimentation to take into account other parameters such as stationary phase retention, loading capacity, injection volume, the ratio of UP/LP that the botanical extract may be dissolved in, and separation time.

Fig. 2.

The UV-CPC profiles of RCE in the HEMWat solvent systems (SSs) HEMWat −0.5 and HEMWat 0. The flow rate was 50 mL/min and the separation performed in descending mode. The rotation speed was 2500 rpm, the stationary phase volume retention ratio (S_f) 80%.

3.2. CPC fractionation

Fig. 2 shows that the retention volumes of target compounds were sensitive to the HEMWat SS composition. The HEMWat (5.5/4.5/5/5, v/v, 234.2 mg) SS demonstrated sufficient resolution compared with the previous experiments [14,40]. HEMWat (5/5/5/5, v/v, 250 mg) was tried in order to see if resolution could be improved. Gaussian deconvolution of these two experiments indicated that separation of 13, 14, 16 and 17 was not improved, when the retention time of compounds was shifted towards longer retention times. A 46 mL sample of a 1:1 mixture of UP and LP phase HEMWat (5.5/4.5/5/5, v/v) SS dissolved 213.2 mg of a 234.2 mg RCE sample. qHNMR analysis of the combined fractions 34 – 52 (8.32 mg) yielded 81% of 14 and combined fractions 42 – 55 (8.88 mg) yielding 50.5% of 14 from SS HEMWat (5.5/4.5/5/5 v/v) and (5/5/5/5 v/v) respectively (Figure. S28, Supporting Information). Additional CPC experiments were performed with other HEMWat SS (4/6/5/5, v/v). The solubility of the RCE (310 mg) in HEMWat (4/6/5/5, v/v) was more than that in HEMWat (5.5/4.5/5/5, v/v). In HEMWat (4/6/5/5, v/v), the RCE solubility was improved to 232.5 mg. However, fractions containing 13, 14, and 16 were overlapped when using HEMWat (4/6/5/5, v/v), the combined fractions 40 – 60 (11.32 mg) yielded 64.9% of 17, 15.7% of 16, and 19% of 14.

The CPC separation of 213.2 mg of RCE resulted in the suitable distribution of 15 target compounds in the collected fractions (Fig. 3, Table 2 and Fig. 4). The K value ranges for 13, 14, 16, and 17 were 0.50–1.12, 1.12–1.85, 2.03–3.07, and 2.03–3.68 respectively. In a single CPC run, highly enriched fractions of 13 (89.4%; fraction 27), 14 (74.2%; fraction 37), 16 (25.2%; Fraction 45) and 17 (94.8%; fraction 79) were collected. Extrusion of the column was performed by switching the phase being pumped into the column from the LP (mobile) to the UP with continuing rotation. This ensured the recovery of all lipophilic metabolites with a strong affinity for the upper phase. Extrusion fractions 84 to 150 were found to be enriched in fatty acids/materials. Fractions 1 to 7 contained the void volume, therefore, 75 fractions were investigated for major and minor components by LC-MS and NMR.

Fig. 3.

TLC profile (hexanes-ethyl acetate 2/1) and reconstructed CPC chromatogram of 213.2 mg of RCE, separated using the HEMWat (hexanes-ethyl acetate-methanol-water) (5.5/4.5/5/5, v/v) SS. The flow rate was 50 mL/min in descending mode, the rotation speed 2500 rpm, and the stationary phase volume retention ratio (S_f) 80%. The distribution of four isoflavonoids, 13, 14, 16, and 17 are shown as deconvoluted Gaussian peaks.

Table 2.

Compounds identified in fractions 8–83. The table follows the order of elution starting with the more polar isoflavonoid glycosides and ending with aglycone 17 as identified by LC-MS-MS. Listed are the (H-2) chemical shifts of the isoflavonoids and their yields in mg.

Metabolite	MW	Frs.#	δH(ppm) H-2	Yield (mg) CPC
Onion 1	430.40	08–11	8.453	>LOQ
Sissotrin 2	446.41	08–11	8.497	>LOQ
Salcolin p-coumarate	672.18	08–11	*	-
Salcolin A and B	526.15	08–11	*	-
Coumestrol 3	268.22	14–17	-	0.50
Tricin 4	330.07	15–17	-	0.20
Calycosin 5	268.26	12–18	8.275	0.53
Daidzein 6	254.24	12–18	8.283	0.89
Quercetin 7	302.23	13–19	-	1.85
Pratensein 8	300.26	13–27	8.318	2.50
Genistein 9	270.24	17–26	8.318	0.98
Kaempferol 10	286.24	18–29	-	2.30
Pseudobaptigenin 11	282.24	22–33	8.345	1.12
Medicarpin 12	270.28	41–67	-	0.72
Formononetin 13	268.26	21–45	8.336	33.26
Irilone 14	298.25	31–45	8.437	6.32
Maackiain 15	284.27	56–83	-	0.20
Prunetin 16	284.27	47–73	8.408	1.41
Biochanin A 17	284.26	44–83	8.362	36.43

^*Compounds tentative identification by LC-MS-MS

Fig. 4.

The CPC separation profile of the major and minor metabolites in RCE. Conditions are the same as reported in Fig. 3.

3.3. ¹H NMR signal assignment in 75 fractions

A thorough analysis of the ¹H NMR spectrum of each fraction provided identification and quantitation of major and minor components (Fig. 4 and Fig. 5). The ¹H NMR spectra were divided into four regions of interest as shown in Fig. 5. The first (upfield) region, between 3.6 to 4.0 ppm, revealed methoxy signals attached to the B-ring C-4’ or, in the case of 16, at A-ring C-7. Region 2, between 4.2 and 6.7 ppm, contained signals of the A-ring hydrogens H-5, H-6, and H-8. The two hydrogens on the benzopyrone of 7, at H-6 and H-8, appear as doublets at 6.183 ppm and 6.403 ppm, respectively; their multiplicity arises from long-range ⁴J_HH W-coupling (2.02 Hz) across the aromatic π-system. The methylenedioxy moieties of 14 (fractions 31–45), 11 (fractions 22–33), and 15 (fractions 56–83) are also found in this region at 6.184 (2H, s), 6.046 (2H, s) and 5.923 (2H, dd) ppm, respectively (for a detailed analysis of the methylenedioxy region of 14 and 11 [41] in the ¹H NMR spectra exhibiting singlets, see Fig. 5). In the case of 15 [21], the chemical shift was affected by the stereocenters, which led the methylenedioxy signal to appear as an AX pattern (Δδ/J_AX=14.21/0.85=16.72 Hz) [42]. The pterocarpins, 12 and 15, could be distinguished via the characteristic methylenedioxy moiety linked to C-8 and C-9 for 15, compared with 12 containing a methoxy group (3.869 ppm) at C-9. Compounds 12 and 15 were also reported from Spatholobus suberectus, taking into account minor differences in their reported ¹H and ¹³C NMR data in CD₃OD [41].

Fig. 5.

Stacked plot of the ¹H NMR (400 MHz) spectra of 21 fractions compared with that of the RCE. The methylenedioxy moiety related with 11, 14, and 15 is marked with red asterisks, and key compounds 9, 10, 13, 16, and 17 are highlighted in blue.

Region 3, between 6.7 and 8.2 ppm, showed B-ring signals as AA’XX’ and AMX patterns. The signal assignments in this region were the most challenging due to spectral overlap. However, the signals provided evidence of the presence of phenolic rings from major and minor components throughout the 75 fractions. Fig. 5 shows evidence for 3, 4, 7, 10, 13, 16, and 17 in the highlighted boxes. This region also includes H-2’ (7.536 ppm) and H-6’ (7.675 ppm) hydrogens on the catechol substituent that appeared as doublet and doublet of doublets, respectively, for 7 and 4 (fractions 15–17). Tricin (4) was also identified by H-3 at 7.327 ppm that appears as a singlet. Compound 3 was identified as coumestrol in fractions 14–17, matching nearly all reported ¹H NMR signals [43]. However, only the H-5 signal at 7.046 ppm, a doublet (1H, 1.64 Hz), was used for quantification, despite the fact that the signal for the same hydrogen was previously reported to appear at 7.16 ppm (d, J=2.0 Hz, 1H) [43].

Region 4, located downfield between 8.1 and 8.5 ppm, included the C-ring H-2 singlets indicative of the 1,4-benzopyrone moiety of isoflavonoids. These assignments are in accordance with prior reports [32]. The 3-phenyl-4H-1-benzopyran-4-one group possesses a number of interesting structural, chemical, and spectroscopic properties thought to be associated with the interaction of its ketone, ring ether, and ethylene components also leading to aromatic ring stabilization [44,45]. Interestingly, the resonance of H-2 located in the 1,4-benzopyrone moieties was found between 8.348 – 8.334 ppm and 8.372 – 8.355 ppm for 13 and 17, respectively. The chemical shifts of these hydrogens exhibit variation with analyte concentration, especially in the presence of intermolecular hydrogen bonding [46]. Figs. S1–S9 (Supporting Information) illustrates how H-2, corresponding to 5, 6, 8, 9, 11, 13, 14, 16 and 17, is shifted in different fractions (Δδ_H = 3 ppb for 5, Δδ_H = 2 ppb for 6, Δδ_H = 8 ppb for 8, Δδ_H = 10 ppb for 9, Δδ_H = 1 ppb for 11, Δδ_H = 14 ppb for 13, Δδ_H = 4 ppb for 14, Δδ_H = 6 ppb for 16, and Δδ_H = 19 ppb for 17). The boxes for 1 and 2 were observed on fractions 11 to 12 representing the glycosides present in unprocessed RCE identified by the typical H-2 singlet peak. The boxes for 5 and 6 show their co-elution due to similar polarities in fractions 12–18 and were identified by their H-2 singlet. The H-2 signals for 8 and 9, contained in fractions 18–22, were assigned by comparing chemical shifts with commercially available reference compounds.

The signals associated with 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, and 17 were assigned and quantified by comparing ¹H NMR spectra with literature data and consulting relevant LC-MS-MS databases (Figs. S13–S19, Supporting Information). Methoxy hydrogens assigned to compounds 5, 6, 8, 13, 16, and 17 were quantitated by comparing integrals following deconvolution of their singlet resonances. In cases where ¹H NMR signals were overlapped, integrals derived from Global Spectral Deconvolution (GSD) using MNova software were employed.

In fractions 8 to 83, upfield signals for residual solvents were commonly noted, i.e., acetic acid (δ_H 1.904 ppm), methanol (δ_H 3.164 ppm), water (δ_H 3.334 ppm), 1,2-dichloroethane (δ_H 3.903 ppm), 1,4-dioxane (δ_H 3.574 ppm), and ethylene glycol (δ_H 3.341 ppm). These solvents present may be residual solvents used in the preparation of the RCE or during chromatography and less likely from NMR sample preparation. It is important to note that even organic solvents purified by repeated distillation may contain minor or trace impurities that can be identified and quantified by qNMR [47].

3.4. Purity/content determination

The purity/content and residual complexity (RC) of enriched fractions were first investigated with qHNMR spectroscopy in order to identify and quantitate RC components, using the 100% method. The contents (i.e., purity of the major components) of individual compounds in each test tube were determined by EC qHNMR. As hydrogen is near ubiquitous in organic molecules, ¹H NMR offers almost universal detection of organic compounds and bears a specific advantage of simultaneous access to both qualitative and quantitative information. The 100% method [48] reflects the relative concentration of a targeted compound (“isolate”) compared to its co-eluted impurities. The most abundant metabolite, 17, exhibited mass percentages in the range of 28.6 to 94.8% in fractions 46–81. The other major isoflavonoids were present in fractions 22–45 such as 13 (25.6–89.4%), 14 (38.6–74.4%) and 16 (23.2–25.2%), representing highly enriched fractions of these compounds (Table S1, Supporting Information). The ¹H NMR and MS data (relative abundance) were in agreement with published data [25].

Fractions 8–11:

These fractions contained a complex mixture of hydrophilic compounds and led to the identification of the two compounds, ononin 1 and sissotrin 2, as well as tentative identification of salcolins A and B as well as salcolin p-coumarate. The isoflavone glycosides, 1 and 2, were identified by ¹H NMR (Fig. 5, region 4), at levels below qHNMR LOQ (Fig. S12, Supporting Information,), as well as by LC-MS-MS operating in negative mode and, thus, were corroborated based on their fragmentation patterns (Fig. S20, Fig. S21, Fig. S10, and Fig. S11, Supporting Information) [49]. The salcolins A and B are diastereomers that contain threo vs. erythro configurations, respectively [50]. This was the first time that 4, the flavonolignans, salcolins A and B [49], and salcolin p-coumarate [51] were tentatively identified for T. pratense.

Fractions 12–45:

Compound 6 (Fig. 5) was enriched in fractions 12 to 18 (3.76 to 21.6%), while 9 was present in fractions 17 and 26 (1.33 to 23.6%) and 10 in fractions 18 to 29 (0.24–36.9%). A major component in fractions 33 to 41 was 14 (2.68 to 74.24%). Isoflavone 13 is considered one of the major ingredients in RCE. It was distributed into fractions 20 to 36 (1.58 to 89.4%). Compound 16 was also quantified as a major component in fractions 42 to 45 with highly enriched fractions in the range of 21.6 to 25.2%. A highly enriched compound 13 (fraction 27) is shown in Fig. 6. However, closer inspection of the ¹H NMR baseline analysis of this fraction led to the identification of 8 (0.68%), 10 (0.71%) and 11 (3.37%) as minor components (Fig. 7). LC-MS confirmed these findings by comparison of the retention times and fragmentation patterns of 8, 10 and 11 with those of authentic standards. Additionally, small amounts of methanol (0.14%), acetic acid (0.01%), stearic acid (0.56%; tentatively representing aliphatic “fatty” impurities), and 1,2-dichloroethane (0.04%) were identified (Fig. 7). Residual water was quantified in each sample via EC-qHNMR. For example, fraction 27 contained 5.06% of water and exhibited an absolute purity of 89.4% for the main compound, 13. All purity results are compiled in the Supporting Information, Table S1.

Fig. 6.

EC-qHNMR spectrum of fraction 27 (400 MHz) along with a table showing the percentage of 13 and the other metabolites present in the material.

Fig. 7.

The qHNMR spectrum (400 MHz) of fraction 27, along with a list of residual solvents and their content.

Fractions 46–83:

Isoflavone 17 was another major ingredient in RCE and was found in fractions 46 to 83 (28.6 to 94.8%). Fraction 66 contained 17 at apparently high purity (Fig. S26, Supporting Information). However, a small amount of 12 (0.66%), 15 (0.39%), and 16 (0.47%) could be observed by careful examination of the qHNMR spectrum near the baseline level (Fig. S27, Supporting Information). Some fractions contained relatively high levels of residual water according to EC qHNMR (Table S1, Supporting Information). Small signals corresponding to trace solvent were also quantified for this fraction and tentatively determined as 0.03% of methanol, 0.12% acetic acid, 0.04% octane, 0.06% of 1,2-dichloroethane, 0.02% of 1,4-dioxane, and 0.06% of ethylene glycol.

An even closer inspection on the baseline of the Fr 66 spectrum revealed numerous small signals below S/N 2:1 (Fig. S27, Supporting Information). Quantitation by qHNMR spectroscopy via mass ratio requires that the molecular mass of each component is known or can be reasonably estimated. LC-MS may provide the molecular masses of each detectable compound at a higher sensitivity than NMR spectroscopy. Matching LC-MS-MS data with NMR spectroscopy analysis not only provides the molecular masses of residual metabolites, but also aids in the structure elucidation of these compounds. The unknown compounds listed as “impurities - residual complexities” along with their molecular weights were enriched from the RCE during fractionation in order to quantitate them by qHNMR spectroscopy. This must be performed before combining fractions for the generation of the DESIGNER extracts. The reason why the quantification of minor compounds “impurities - residual complexities” are crucial in each fraction is that these minor compounds can potentially be responsible for the observed biological activity. The information about the percentages of known and unknown minor impurities present in the fraction helps build more rigorous relationships between compound mass percentage and biological activity (see Supporting Information).

Combined fractions 84–150:

The combination of highly lipophilic extrusion fractions yielded 19.0 mg of material, containing 1.62% of 13, 1.91% of 17 and 8.1% chlorophylls plus a complex mixture of fatty acids (Fig. S25, Supporting Information). This indicates that the CPC method separated the lipophilic portion of the RCE efficiently from the isoflavones, with very little carryover of the main target compounds.

Insoluble RCE material:

Insoluble plant material (9.94 mg) was also analyzed by EC qHNMR and indicated the presence of 0.99% of 5, 0.77% of 6, 3.22% of 8, 0.35% of 9, 1.67% of 11, 30.4% of 13, 1.53% of 14, 0.23% of 16, and 25.4% of 17 (Fig. S23, Supporting Information). 11 and 17 are the two isoflavonoids most abundant in the insoluble RCE material analyzed by NMR. The similarity of this quantitative profile with that of the original RCE indicates that the insoluble materials have about the same composition as the crude extracts, therefore, no specific exclusion of compounds resulted from the CCS-based processing.

Switch Fraction (SF):

The “switch fraction” is defined as waste generated while the fraction collector moves from tube-to-tube during the CPC separation. A 13.4 mg sample of “waste” effluent was analyzed and yielded 5.26% of 13 and 4.45% of 17 (Fig. S24, Supporting Information). The SF represents 5.9% of the total RCE injected into the CPC instrument. In many investigations this mass would be ignored because it is deemed insignificant, but for a rigorous metabolomic investigation it must be included. Additionally, Fig. S24, Supporting Information, contains a table with the percentages of minor components found in the SF.

CPC recovery analysis:

Crude extract solubility, loading capacity, and recovery are important considerations when preparing extracts for biological evaluation. In this work, after injection of 234.2 mg of RCE, the CPC fractions yielded 180.8 mg (77.2%; fractions 8–83) and 19.0 mg (8.11%; fractions 84–150) of material, 9.94 mg (4.24%) of insoluble matter, as well as the SF 13.4 mg (5.72%). Only 11.1 mg (4.75%) of injected RCE went unrecovered. This can be explained by a variety of factors: (a) experimental weighing variation associated with weighing on sample versus many samples; (b) loss of material during injection including residue left in the syringe; and (c) residual remaining on the column after extrusion and methanol/water after the experiment (Fig. 8). Importantly, this loss is non-specific for any analytes, in part practically unavoidable considering the relatively large amount of sample handled, and an intrinsic property of CPC design, involving small ducts and convoluted liquid flow paths (virtually non-existent in HSCCC instruments).

Fig. 8.

Quantitative composition of the RCE based on the CPC metabolomic profile. A: shows yield (mg) of isoflavonoids and related compounds calculated for 100 mg based on EC qHNMR [6] (blue color). B: mg of isoflavonoids and related compounds after CPC spread out RCE (blue color), insoluble RCE (red color), CPC Switch Fraction RCE (green color) and combination of fractions 84–150 (purple color). Yields in mg were calculated using EC qHNMR methods.

4. Conclusions

The present investigation demonstrates that the isoflavone metabolome of RCE may be fractionated efficiently and in high resolution by CPC. The SS selection process was streamlined with the use of GUESS, a TLC-based SS selection method. A single chromatographic run yielded all major isoflavone target compounds, with their purities falling in the range of 89.4 to 94.8% w/w. Using off-line CCS-qHNMR, together with LC-MS-MS analyses, revealed not only the true purities of the target compounds, but also allowed for the identification and quantitation of minor (residual) components. The unavoidable RC of any preparative (and analytical) separation method has an effect on the observed biological activities of the resulting fractions. For subsequent investigations, including the preparation of knockout extracts (KOEs) for biological tests as performed previously, should include a metabolomic profile [52] of each fraction using off-line CCS-qHNMR, in order to determine and quantify both enriched (“isolated”) components and the RC of the materials they are embedded into.

Twenty isoflavonoids with LOQs below 10:1 could be recognized but not elucidated in this investigation. However, using 100% qHNMR, they could still be quantified under the reasonable assumption of them being isomeric analogues of 17 and being part of the structurally closely related isoflavonoid metabolome of RCE. Again, qHNMR demonstrated again its utility for the quantitation of not fully determined structures [53]. It should be noted that future insights about the exact identity of these minor metabolites will permit their “retroactive” precise quantification based on the data presented here.

Traditionally, rather than analyzing every fraction separately, it is customary in NP chemistry laboratories to combine fractions that contain compounds with similar TLC R_f values. However, as the present study illustrates, this practice may require re-thinking. The advantage of analyzing each fraction individually is that the residual complexity can be quantitated, overlapped chromatographic peaks or bands be correlated with NMR and/or LC-MS signals, and the (partial) structures of the compounds present be determined. When applying qHNMR, the lower threshold of detection can be improved by increasing the number of scans (thereby increasing signal-to-noise ratio, in a square-root relationship). In fact, the full diversity of the RC of a given material may not be evident and/or appreciated until fractions or “pure isolates” are subjected to a rigorous qHNMR analysis.

The study chose external calibrants (ECs) in qHNMR for absolute quantitation because the samples produced were ultimately destined for biological testing. Without contamination from external sources, EC-qHNMR quantitates target compounds and determines their purity with accuracy. In addition, the use of two external calibrants afforded high accuracy and reproducibility. The suitability of EC in qHNMR applications has recently been confirmed and specified [54].

The first-time proof of the presence of tricins, the diastereomeric salcolins A and B, and salcolin p-coumarate in red clover demonstrates that the RC of this supposedly well-studied botanical is by far not fully understood. Future investigations might include the isolation and exploration of the potential biological activity of these and other compounds, e.g. with respect to their estrogenic activity [49]. Collectively, the study demonstrates that the bioactive isoflavones from RCE, and by analogy other bioactive compounds from other plants, are indivisibly connected with a certain degree of RC. Unless characterized in detail, biological studies will fall short of their ability to correlate both expected and unexpected outcomes with components that “live in the dark” or RC. The outcomes presented here show how RCE and preparations derived from RCE, including chemical knock-out materials, can be studied for RC and provides the reference data for an efficient evaluation.

Fig. 1.

Schematic representation of the 17 metabolites identified and quantitated in this study. A single chromatographic step produced highly enriched fractions of the 13, 14, 16, and 17.

Questionnaire.

1. Does your work deal with natural products? YES

2. Does your manuscript fit with the molecular trait of the Journal? YES

3. Are the results of your investigation original and not predictable based on the literature? YES

4. Has the organism investigated in your manuscript been properly identified? YES

5. If you report the study of the bioactivity of an extract, has this been phytochemically characterized, and have its active constituents identified and biologically assayed? YES [n/a]

6. Has a positive (or a negative) reference compound be used in the bioassays you report? YES [n/a]

7. If your manuscript present docking experiments, have they been validated by biological assays. YES [n/a]

8. Have isolation and bioactivity evaluation been described in full details? YES

9. Have you included original NMR spectra in the SI for all novel compounds, making the discussion of structure elucidation easier to follow for readers? YES

10. Have all structural formulas been prepared according to the Journal’s instructions (ChemDraw, ACS Style)? YES

11. Has configuration at all stereocenters been clearly indicated with the wedge/broken line notation? YES [n/a]