Integrated standardization concept for Angelica botanicals using quantitative NMR

Abstract

Despite numerous in vitro/vivo and phytochemical studies, the active constituents of Angelica sinensis (AS) have not been conclusively identified for the standardization to bioactive markers. Phytochemical analyses of AS extracts and fractions that demonstrate activity in a panel of in vitro bioassays, have repeatedly pointed to ligustilide as being (associated with) the active principle(s). Due to the chemical instability of ligustilide and related issues in GC/LC analyses, new methods capable of quantifying ligustilide in mixtures that do not rely on an identical reference standard are in high demand. This study demonstrates how NMR can satisfy the requirement for simultaneous, multi-target quantification and qualitative identification. First, the AS activity was concentrated into a single fraction by RP-solid-phase extraction, as confirmed by an (anti-)estrogenicity and cytotoxicity assay. Next, a quantitative ¹H NMR (qHNMR) method was established and validated using standard compounds and comparing processing methods. Subsequent 1D/2D NMR and qHNMR analysis led to the identification and quantification of ligustilide and other minor components in the active fraction, and to the development of quality criteria for authentic AS preparations. The absolute and relative quantities of ligustilide, six minor alkyl phthalides, and groups of phenylpropanoids, polyynes, and poly-unsaturated fatty acids were measured by a combination of qHNMR and 2D COSY. The qNMR approach enables multi-target quality control of the bioactive fraction, and enables the integrated biological and chemical standardization of AS botanicals. This methodology can potentially be transferred to other botanicals with active principles that act synergistically, or that contain closely related and/or constituents, which have not been conclusively identified as the active principles.

1. Introduction

Angelica sinensis (AS; Apiaceae; syn. Dang Gui) is among the most popular Traditional Chinese Medicines (TCMs). It is used as a general blood tonic and for gynecologic indications. Recent research has focused on the evaluation of cardiovascular, hepatoprotective, hematopoietic, antioxidant, antispasmodic, and immunomodulatory properties of aqueous ethanolic or hot water extracts of AS roots [1]. AS is also a popular constituent of Botanical Dietary Supplements (BDSs) in the U.S., and is used for health benefits in menopausal women [1]. It has been the subject of studies at the NIH/ODS Botanical Center for Women’s Health at the University of Illinois at Chicago [2–6].

The major chemical constituents of the plant thus far identified are alkyl phthalides, (i.e., monomeric and dimeric congeners of ligustilide), ferulic acid derivatives (phenylpropanoids), polyynes, and linoleic acid (a poly-unsaturated fatty acid [7]) [1,8–10]. In addition to the published studies conducted at UIC, numerous related publications have appeared which describe the in vitro and in vivo activities of AS and its extracts [11–15]. However, the active constituent(s) that might serve as chemical and biological standardization markers of the botanical have not been conclusively identified. Recent studies in our laboratory have been using a panel of in vitro bio-assays, which emphasize estrogenic or chemo-preventive effects. The methods and have employed different chromatographic separation techniques which have consistently found that the active fractions contained ligustilide(s) [5,6,16].

The standardization of BDSs is generally based on occurrence of specific marker compounds and uses gas- (GC) or liquid-chromatography (LC) hyphenated techniques which require authentic and pure reference materials to establish a standard curve for quantification. Because standardization of BDSs depends on the availability of identical (authentic), pure phytochemical reference materials, it is, in practice, often based on the more abundant components in the BDS rather than the biologically most relevant constituents. For the present study with AS, a standardization method was sought that could focus on the biologically relevant constituents and still be integrated into chemistry-oriented quality control (QC) measures for AS botanicals. As it becomes increasingly recognized that botanicals exert their observed health effects as a whole [17,18]}, new analytical protocols have to cover mixtures of chemical components rather than a single chemical entity. Therefore, a new method capable of identifying and quantifying multiple analytes in the bioactive fraction of AS preparations would be particularly beneficial and could be an important step towards multi-target botanical standardization (Fig. 2). In the case of AS, ligustilide (1) is regarded as a bioactive constituent, but because of its chemical instability [19–21], an analytical method was required that does not rely on an identical reference standard for the qualitative and quantitative determination of 1.

Fig. 2.

Comparison of the proposed qHNMR-based standardization concept vs. conventional standardization paradigms. The latter typically rely on identical and authentic reference materials, which in case of AS are difficult to both obtain and maintain due to instability. The qHNMR method avoids identical calibrants and enables quantification of multiple components contained in the bioactive SPE fraction of AS.

NMR has been employed extensively for qualitative structure elucidation of natural products, as well as for quantitative determinations of purified target analytes [22–25]. More importantly, because a strict proportionality exists between sample concentration and signal intensity (integrals) in ¹H NMR [26], reference standards for quantitative analysis need not be identical with the target analytes. Instead, compounds of high chemical stability and purity may be selected as standards for quantitative ¹H NMR (qHNMR) calibration. Additional advantages of NMR are the non-destructive character, the universal nature of detection, thus enabling the targeting of compounds that are not fully discerned, and the possibility of simultaneously quantifying multiple compounds in a complex mixture [27,28]. By emphasizing the latter two points, quantitative ¹H NMR (qHNMR) assays may be developed using either internal (IS [29]) or external standards (ES, [30,31]) for calibration, and can also be used to evaluate sample composition through the 100% qHNMR mass balance method [32]. Accordingly, qHNMR was selected as the analytical tool because it does not rely on an identical reference sample of 1 and allows for multi-target quantitative analysis of additional and not fully identified constituents present in the bioactive fraction. This approach simplifies analyses and reduces the workload involved to perform multiple quantifications [33,34].

Its universal character and relatively low selectivity may be viewed as a limitation of NMR detection, especially when working with low abundance analytes in complex matrices, such as plant extracts. Because solid phase extraction (SPE) is an established method for the enrichment of minor analytes prior to quantitative analysis [35–37], and in order to address the dynamic range challenge, the present study utilized an SPE pre-fractionation methodology to concentrate the bioactive analytes prior to qNMR analyses.

The two key aspects of the present study were: (a) the development of a method capable of quantifying 1 along with other minor constituents in the biologically active fraction of aqueous ethanolic botanical preparations of AS; and (b) the identification of suitable QC markers, which are meaningful for both chemical and biological standardization of AS botanicals. In order to focus on the bioactive portion of the AS metabolome and match the dynamic range of the bioactive principles with that of 1D qHNMR methodology, an SPE pre-fractionation method was developed that concentrated the active components into a single fraction. This was verified by bioassay testing of the SPE fractions. Next, a calibration curve was generated using dimethylsulfone (2) as a highly pure and stable reference standard, calibrating against the residual solvent signal in the deuterated NMR solvent, which was used as an internal reference. The qHNMR method was evaluated for accuracy and precision using the following additional reference materials: caffeine (3) as an established pharmacopoeial reference standard; butylidene phthalide (4) as a chemically stable representative of the alkyl phthalides as a compound class; and diethyl phthalate (5) as a chemically stable and highly pure alkyl phthalide congener, to determine the SPE recovery rate of the active principles.

2. Experimental

2.1 Instruments and Materials

One dimensional (1D) and homonuclear 2D ¹H NMR experiments (gCOSY) were acquired on a Bruker (Karlsruhe, Germany) Avance 600 NMR spectrometer equipped with a TXI cryogenic probe, with the sample temperature maintained at 25.0 °C (298.0 K). The GC-MS chromatograms were acquired using an Agilent 7890A GC System (7000A GC/MS Triple Quad, 7683B Series Injector). LC-MS analyses were carried out using Positive ion electrospray detection on a Thermo Scientific (Waltham, MA) LTQ linear ion trap mass spectrometer. Volumes were measured using a 200 μL Eppendorf pipette (calibrated against H₂O; accuracy/precision for 600 μL: 587 μL ± 8.09 (n=22, dichloromethane) and 588 μL ± 9.99 (n=8, DMSO). For the preparation of NMR and GC samples, a 1 mL gas-tight syringe (Pressure-Lok^®, VICI^® Precision Sampling, Baton Rouge, LA; accuracy/precision for 600 μL: 603.9 μL ± 1.31, n=10, DMSO), or Drummonds Pipetting Aids glass capillaries of different volumes (for volumes ≤ 100 μL; accuracy/precision for 100 μL: 96.5 μL ± 1.2 (RSD=1.2%), dichloromethane; 10 μL: 8.9 μL ± 0.59 (RSD = 5%), dichloromethane; 2 μL: 2 μL ±0.38, n=10, butylidene phthalide reference material, 4) were used. Weights were measured on a Mettler Toledo balance (www.mt.com, XP205, max 220 g, d = 0.01 mg, Excellence Plus series). Accuracy and precision for the volumetric measuring devices used for sample preparation were determined by sample weight using the solvent’s specific weight for volume conversion at the respective room temperature (range: 20–25 °C). SPE fractions were carefully dried in a Thermo Savant SC250Exp speed-vac with RVT4164 Refrigerated Vapor Trap (−120 °C) using standardized conditions (7 h, full rotation speed, vacuum level 3). Unless otherwise stated, all chemicals and solvents were purchased from Fisher Scientific (Hampton, NH, USA) or Sigma–Aldrich (St. Louis, MO, USA).

2.2 Plant Materials and Extractions

Samples of dried root material of Angelica sinensis (Oliv.) Diels (Apiaceae) were purchased from Chinatown, Chicago, IL (BC440: Nam Bac Hang; BC426: SunSun Tong; BC107 and BC165: reference materials [2]), and Starwest Botanicals (Lot#32642, BC574), Rancho Cordova, CA. The plant materials were identified through a series of comparative macroscopic, organoleptic, and TLC analyses against an authentic AS voucher sample (BC165) deposited at the UIC/NIH Center for Botanical Dietary Supplements Research, Chicago, IL. In addition, 20 aqueous ethanolic extracts (hydro-alcoholic percolations at 1:3 (m:v) in 70% alcohol) of AS, produced between 1998 and 2008 (see Tables 1–3, higher P indicate younger extracts), were kindly provided by Vitality Work (Albuquerque, NM). A commercial aqueous ethanolic solution of AS (Herb Pharm (HP), lot#215300-9M37) was included in the study. Extracts were prepared using ground roots of plant materials BC426, BC107, BC165, BC574 (maceration with 75% ethanol (v/v) 2 × 1:2.5–10 (m:v)), and BC440 (percolation with 75% ethanol (v/v) 1:5.5 (m:v)), respectively. All extracts were carefully dried in a speed-vac (except: BC440: rotary evaporator, 40 °C, at reduced pressure), weighed, and re-dissolved in 20 mL (except BC440: 1L) of 75% ethanol (v/v) for storage at 0 °C. Aliquots of all extracts (self-prepared and commercially obtained, 3 × 1 mL each) were dried in the speed-vac, weighed, and re-dissolved in 75% ethanol such that approximately 150–200 mg of each extract could be applied as 500–600 μL of solution onto 2 g RP-18 cartridges. Fractions were collected using a methanol-H₂O step gradient with 5, 60 and 100% methanol (v/v) (Fr1, Fr2, Fr3, respectively). Each of the 100% methanol fractions (Frs3) was subject to qHNMR analysis after drying in the speed-vac, re-dissolving in 800 μL of CDCl₃, and transferring 600 μL of the clear solution into 5 mm NMR tubes (Norell XR-55). The same fractions were monitored by GC-MS.

Table 1.

Concentrations of AS constituents determined by ¹H NMR. Content of 1 in Fr3 determined by the modified 100% method and as absolute value using the calibration curve, as well as the deviation between the two methods. Absolute value content [ppm (m/m)] of the plant metabolites 1, 4, 6, 7, and 8 are given as percent of the amount of AS extract applied to an SPE cartridge. The S/N ratio was ≥ 150 for all signals except for most signals of E-4 and 6 in P130, P241, P31, P976, BC165, and P766. (n.d. = not determined)

	100%Ma)	% (m/m)		ppm (m/m)
AS sample	1	1 in Fr3		1c) in AS	4d) in AS	6e) in AS	7f) in AS	8g) in AS	receiver
code	content	CCb)	% DEV	CCb)					gain
HP	10.3	16.6	0.4	2875	514	1566	127	1961	256
P590	12.0	25.0	0.5	3063	594	1469	202	1220	256
P1241	0.6	24.1	1.0	4814	431	1452	111	1862	256
P130	10.4	9.6	−0.1	3148	428	1886	82	1668	256
P1341	13.5	13.4	0.0	2726	216	1341	93	1747	256
P1518	0.9	8.0	0.9	1394	405	1303	114	2031	256
P1731	1.2	7.4	0.8	1641	544	1536	177	2222	256
P1841	12.0	19.5	0.4	3442	688	2381	234	2243	256
P2041	3.3	5.1	0.4	1040	434	1657	54	3850	256
P2370	13.2	18.7	0.3	4581	759	2018	127	3404	256
P264	10.9	10.9	0.0	4258	561	1744	171	2464	256
P2712	1.7	6.6	0.7	739	306	731	261	778	256
P280	11.0	10.5	0.0	3542	486	1329	185	2039	256
P31	4.1	4.9	0.2	903	365	1089	58	1745	256
P471	6.6	7.1	0.1	1825	603	1827	133	2039	256
P561	10.0	9.6	0.0	2771	445	1580	79	1707	256
P661	1.4	n.d.	n.d.	n.d.	n.d.	n.d.	n.d.	n.d.	256
P689	1.4	37.8	1.0	2282	494	1205	147	945	256
P765	5.8	6.1	0.0	1329	192	738	125	516	128
P976	13.8	12.0	−0.2	2111	229	1147	39	848	256
BC107	9.9	8.1	−0.2	3080	1327	2533	279	5267	128
BC165	2.5	2.3	−0.1	218	112	308	16	688	256
BC426	25.3	25.7	0.0	8971	361	2478	286	4231	128
BC440 (a)	24.2	20.4	−0.2	5910	279	1222	280	3054	256
P766	8.0	12.7	0.4	1709	334	892	68	1622	128
StarW	11.9	13.3	0.1	1090	568	838	293	464	256
BC440 (d)	18.4	19.8	0.1	n.d.	n.d.	n.d.	n.d.	n.d.	32
BC440 (c)	4.1	4.2	0.0	n.d.	n.d.	n.d.	n.d.	n.d.	32
BC440 (b)	17.0	13.6	−0.3	n.d.	n.d.	n.d.	n.d.	n.d.	32

^a)modified 100% method, content of 1

^b)absolute amount determined using calibration curve (CC)

^c)H-6 used for quantification

^d)average of Hs-4, 7, 8 (cis) and Hs-7, 8 (trans) used for quantification

^e)average of Hs-2, 1b, 8, 3 used for quantification; M_r used: 260 mg/mmol

^f)olefinic H used for quantification; M_r used: 180 mg/mmol

^g)olefinic Hs used for quantification (calc. as linoleic acid, 6Hs); M_r used: 280 mg/mmol

Table 3.

Concentrations of AS alkyl phthalides determined by 2D COSY/qHNMR. Content of six alkyl phthalides [ppm (m/m)] in the AS extract applied to the SPE separation. Values were calculated based on COSY cross peak ratios relative to Z-1. (n.d. = not determined)

	1H NMR	b)
AS sample	Z -1	E -1	Z -4	E -4	UA1	UA2	UA3
code	ppma)	ppm (m/m) in AS
HP	2875	122	80	34	30	105	32
P590	3063	80	64	25	33	82	28
P1241	4814	123	87	21	42	46	15
P130	3148	29	24	21	12	103	17
P1341	2726	71	59	44	21	76	25
P1518	1394	57	53	34	14	54	16
P1731	1641	70	65	40	14	71	23
P1841	3442	133	67	31	30	92	22
P2041	1040	70	33	25	9	79	29
P2370	4581	120	162	73	33	281	144
P264	4258	106	116	49	43	242	85
P2712	739	36	38	23	19	62	12
P280	3542	42	37	26	21	127	27
P31	903	54	17	9	3	154	29
P471	1825	74	95	40	16	67	18
P561	2771	67	66	23	21	87	28
P661	n.d.	n.d.	n.d.	n.d.	n.d.	n.d.	n.d.
P689	2282	83	32	19	27	115	41
P765	1329	62	72	46	16	66	19
P976	2111	n.d.	n.d.	n.d.	n.d.	n.d.	n.d.
BC107	3080	148	229	92	55	821	391
BC165	218	11	12	2	1	20	7
BC426	8971	513	82	77	57	243	94
BC440	5910	274	77	63	48	148	67
P766	1709	82	46	18	12	75	25
StarW	1090	12	25	4	12	17	8

^a)Z -1: absolute value determined using qHNMR and calibration curve in [ppm (m/m)] of AS extract applied to SPE

^b)content of E -1, Z -4, E -4, UA1, UA2, and UA3 determined as COSY cross peak ratio with Z -1; in [ppm (m/m)] of AS extract applied to SPE; M_r = 190 mg/mmol used for all calculations

2.3 Sample Preparation and Monitoring

Previous results on the chemical stability of 1 were taken into account for sample preparation [19], such as drying the samples under nitrogen or in the vacuum, fast sample re-dissolution because of decreased stability of 1 in the dry state, the choice of solvent for storage and analyses, storage in the dark and minimizing daylight exposure of samples as much as possible.

2.3.1 SPE Method

RP-18 cartridges (2 g, Bakerbond^®, Octadecyl (C₁₈), 40 μm prep LC Packing) were preconditioned by washing with 2 × 1 mL 5% methanol, 2 × 1 mL methanol, and 2 × 1 mL 5% methanol. An aliquot (approx.150–200 mg in 500–600μL) of each aqueous ethanolic extract (n=29) was applied to a dry cartridge. Elution was carried out sequentially with 2 × 1 mL 5% methanol (Fr1), 2 × 1 mL 60% methanol (Fr2), and 1 × 1 mL methanol (Fr3). Before reusing the cartridge, it was washed with up to 5 × 1 mL methanol alternating with equal amounts of 5% methanol before re-conditioning. Fractions were dried in a speed-vac at room temperature (during 6–9hrs), weighed, and Fr3s were immediately re-dissolved, each in 800 μL CDCl₃. NMR spectra where usually acquired on the same day, to prevent chemical changes due to the instability of 1. Samples were stored at −20 °C, and GC-MS chromatograms for analytical fraction monitoring where also run on the same or the following day.

2.3.2 Cell culture conditions

The Ishikawa cell line, an endometrial adenocarcinoma cell line, was kindly provided by Dr. R. B. Hochberg (Yale University, New Haven, CT) and was maintained in Dulbecco’s Modified Eagle medium (DMEM/F12) containing sodium pyruvate (1%), non-essential amino acids (NEAA, 1%), glutamax-1 (1%), insulin (0.05%), and heat-inactivated fetal bovine serum (FBS, 10%). A day prior to treating the cells, the medium was replaced with phenol red-free DMEM/F12 medium containing charcoal/dextran-stripped FBS and supplements.

2.3.3 Induction/inhibition of alkaline phosphatase (ALP) activity in an endometrial adenocarcinoma cell line (Ishikawa)

The assay was performed based on the method of Pisha et al. [38]. For estrogenic activity, Ishikawa cells were plated in estrogen-free media in 96-well plates (5 × 104 cells/well) and preincubated overnight. Test samples (20 µg/mL) and controls were added to the media and incubated at 37 °C for 4 days. For the determination of antiestrogenic activity, 1 nM estradiol was added to the media. After 4 days, cells were washed, Tris/Triton solution (pH 9.8) added, and cells were lyzed by an additional freeze-thaw cycle at -80 °C. Plates were allowed to thaw and to reach 37 °C. An aliquot of p-nitrophenyl phosphate in 0.1 M Tris buffer (pH 9.8) was added to each plate. Alkaline phosphatase enzyme activity was determined by monitoring the release of p-nitrophenolate at 405 nm every 15 s for 16 readings using a Power Wave 200 microplate scanning spectrophotometer (Bio-Tek Instruments, Winooski, VT).

For the determination of cytotoxicity, a corresponding toxicity plate was prepared for each induction test plate. Ishikawa cells (1,500 cells/well) were incubated in estrogen-free media overnight. Cells were incubated with test samples and controls for 4 days. A day 0 plate, which was harvested at the time when the cytotoxicity plate was treated, was prepared as a separate plate. The cells were fixed with 50 % trichloroacetic acid (TCA), incubated at 4 °C for 30 minutes, washed and stained with sulforhodamine B (SRB). Non-bound dye was washed off with 1% acetic acid and bound dye was solubilized with Tris buffer (pH 10). Absorbance was measured at 515 nm using a Power Wave 200 microplate scanning spectrophotometer (Bio-Tek Instruments, Winooski, VT).

For estrogenic determination, the percent induction as compared with the estradiol control was calculated using equation 1. For antiestrogenic determination, the percent inhibition of estrogen induced ALP induction was calculated using equation 2.

• 1(slopesample − slopeDMSO) /(slopeestrogen − slopeDMSO) × 100 = % estrogenic induction
• 2(1 − ((slopesample − slopecells*) /(slopeDMSO− slopecells))) × 100 = % antiestrogenic induction

• * Cells mean control cells, which are not treated with estrogen containing media or DMSO.

2.3.4 GC-MS Conditions

Samples for GC analysis where prepared in dichloromethane (20 μL of NMR sample were added to 600 μL dichloromethane; MS source: EI 70 eV; column: HP-5MS 5% phenyl methyl siloxane; run time: 29 min; heating profile: 70 °C, then +20 °C/min to 130 °C, hold for 1 min, then +8 °C/min to 250 °C, hold for 10 min; flow: 3 mL/min; splitless injection volume: 1 μL or 2 μL). Several days of storage did not impact the GC-MS chromatograms.

2.3.5 TLC Conditions

Analytical thin layer chromatography (TLC) was performed in parallel on the whole range of SPE-methanol gradient fractions at room temperature on pre-coated silica gel 60 F₂₅₄ glass plates (20 × 20 cm; Merck, Darmstadt, Germany; solvent system: toluene-ethylacetate-formic acid (90:10:1) and others previously used as part of plant species authentication in our laboratory [2]). Fractions were monitored under UV light at 254 and 365 nm, before and after spraying with anisaldehyde reagent (spray with 1% anisaldehyde in methanol, dry; spray with 10% H₂SO₄ in methanol, heat until full color development).

2.3.6 LC-MS Conditions

Fractions were analyzed by LC-MS using positive ion electrospray on a Waters SYNAPT^© hybrid quadrupole/time-of flight mass spectrometer. Separations were carried out on a Thermo Fisher Hypersil GOLD 2.0 × 150 mm C₁₈ column (5 μm particle size) using a gradient from 10–95%MeCN/0.1% formic acid over 35 min at a flow rate of 0.2ml/min. The column temperature was kept at 30°C using a thermostat. Mass spectra were acquired over the range of m/z 120–900 at a resolution of 10,000 FWHM. Product ion tandem mass spectra were recorder at collision energy of 25eV using argon as collision gas.

2.3.7. Statistics

Two-tailed T-test was performed using GraphPad Prism version 4.00 for Windows, GraphPad Software, San Diego California USA, www.graphpad.com. Results were considered significant when p < 0.05.

2.4 NMR spectrometry

Samples were dissolved in CDCl₃ (Aldrich, lot# 04005HH, 99.8% D) or DMSO-d₆ (Cambridge Isotopes, lot#8L-052, 99.9% D) and transferred to NMR tubes (Norell, 5 mm × 7″, XR-55, lot # D080107CX and D030508CSX). Chemical shifts were expressed relative to the residual solvent signal (δ_H = 7.270 ppm, CHCl₃ or δ_H = 2.500 ppm, DMSO-d₅) as internal standard. NMR data were processed with Bruker TopSpin^™ 3.0 and NUTS (version 20100424, Acorn NMR Inc.; CA), using the following processing parameters for all spectra: and Lorentzian-Gaussian line resolution enhancement (line broadening factor: −0.3; Gaussian factor: 0.05), zero-filled to 130 k, manual phasing, a polynomial baseline correction routine. The digital resolution was better than 0.092 Hz/pt for ¹H, and chemical shifts are reported to three decimal places on the ppm scale.

2.4.1 Modified 100% method

Following the previously described method [23] for the present study working with plant extracts and fractions, the purest signal of the target analyte (determined using integration of 1D slices of overlapped ¹H,¹H-COSY cross-peaks) was integrated, multiplied by the total number of protons of the analyte, and quantitatively related to the integral of the whole spectrum (0.5–10 ppm). The total integral did not include the residual solvent signal and discounted the analytes’ ¹³C satellites at 1.1% of their integral.

2.4.2 Quantitative NMR

All NMR spectra were acquired and processed using a qHNMR standard parameters set (Table S1). A qHNMR calibration curve was generated using dimethylsulfone (DMSO₂, 2, lot# 099K1615) as external standard (ES, at concentrations of 0.28 mM to 15.13 mM) and the residual CHCl₃ signal was used as internal standard (IS; Table S2). The spin-lattice relaxation times (T₁) were determined for CHCl₃ (7.974 s) and 2 (2.480 s). To verify the calibration curve, a mixed sample of 2 and caffeine (3, lot# 00102EJ) was used. Purity of 2 and 3 standards were determined using the 100% qHNMR method [32], and these standards were analyzed independently for their water content to confirm the certificates of analysis (2: 99.4% (CofA); 99.3% (qHNMR), <0.09% H₂O (elemental analysis, Karl-Fischer); 3: 99.5% (CofA); 98.7% (qHNMR), < 0.9% H₂O (elemental analysis, Karl-Fischer)). Both reference standards (2 and 3) were chosen because of their availability in high purity, their chemical stability, and non-hygroscopic nature [30,39]. The ID sub-routine of NUTS software was used to retain the integral regions of qHNMR spectra in the file tailer (TA command), which was applied to all data, thus keeping integral regions consistent for quantitative applications. An R (language and environment for statistical computing and graphics, http://www.r-project.org/) based program, rNMR [40], was used to define regions of interest (ROIs) in the ¹H spectra. This allowed semi-automatic determination of integrals for these ROIs for all samples. The ROI integral sums were exported from rNMR as absolute areas using “no internal standard” or “signal to noise” for normalization. Microsoft Excel was used for all further calculations. rNMR was used to generate the calibration curve, as well as for the integration of plant metabolite signals, and results were compared with results in NUTS. The software tool Fityk (http://fityk.nieto.pl) [41] was employed for the fitting of CHCl₃ residual solvent and ligustilide signals for absolute determination of ligustilide in comparison to the signal integrals determined in rNMR or NUTS (Fig. 8). The rNMR software was also used to define cross-peak areas in COSY spectra and to obtain the volume integrals of the defined ROIs, in the same way the ¹H spectra ROI values were exported.

Fig. 8.

Comparison of modified 100% and absolute value qHNMR methods. Percent content of 1 in Fr3 of AS extracts as determined by qHNMR using the modified 100% method (x-axis) vs. the ligustilide amount [mg] in Fr3 determined as absolute value using the calibration curve in % of the amount of Fr3 [mg]. The center trend-line represents 100% correlation between both methods, the two outer lines are the 80% confidence intervals which contain the quality samples (n=26 [AS extracts])

3. Results and Discussion

3.1 Obtaining the active fraction

In an earlier study, it was observed that ¹H NMR spectra of the aqueous ethanolic extract of AS in DMSO were dominated by broadened signals in the high field region of the spectrum. In addition the signal to noise ratio (S/N) of the olefinic signals of 1 was found at <20 (Fig. 3.A), which is insufficient for direct quantification [42]. A primary separation scheme using SPE on RP-18 cartridges was developed in order to obtain the chemically unstable 1 in one quick step of chromatography. Because 1 has also been associated with the bio-active principle(s) of AS botanical preparations, the SPE elution scheme was optimized by chemical means using TLC/LC–MS profiling (Fig. S1) as well as using bio-activity guidance (ALP enzyme induction/inhibition and cytotoxicity assay in an endometrial adenocarcinoma cell line). This resulted in an SPE fractionation scheme with only 3 fractions eluted with 5% (fraction 1, Fr1), 60% (fraction 2, Fr2) and 100% methanol (fraction 3, Fr3). Fr3 contained 1 as the main constituent as determined by qHNMR (Fig. 3.B) and also held all the observed bio-activity (Fig. 4). The average weight profiles of the SPE fractions were determined from 26 samples and found to be 67% ± 22 for Fr1, 5% ± 4 for Fr2, and 3% ± 1 for Fr3, when the initially applied crude aqueous ethanolic extract was set to 100% (Fig. 4).

Fig. 3.

A: ¹H NMR spectra (CDCl₃, 600 MHz) of the aqueous ethanolic AS extract and B: of the bioactive SPE fraction 3 (Fr3).

Fig. 4.

Chemical and biological properties of the AS SPE fractions. Chemistry: Weight profile of SPE fractions Fr1 (67 ± 22%), Fr2 (5 ± 4%), and Fr3 (3 ± 1%). The crude aqueous ethanolic extract was set to 100% (n = 26; total weight recovery was 74 ± 11%). Biology: Bioactivity as inhibition of estradiol (1 nM) induced Alkaline Phosphatase (ALP) activity in an endometrial adenocarcinoma cell line; Samples: solvent control (DMSO) (n = 9), AS extract (n = 3), Fr1 (n = 6), Fr2 (n = 6), Fr3 (n = 9) were tested at 20 μg/mL, 4-hydroxytamoxifen (4-OHTAM, 5 μM) was used as positive control. Statistical analysis using T-test comparing DMSO and Fr3 showed a highly significant difference (P < 0.0001). The extract and all fractions showed a cytotoxicity of less than 20% at the tested concentration.

The total recovery of the crude extracts was determined as 74% ± 11 by weight. As part of the validation of quantitative methods, a spike-recovery study is used to determine the fate of the target analyte during sample preparation. Because in this case the target analyte is a fraction rather than a single compound, and the major constituent 1 is chemically unstable, a follow up study is necessary to assess the loss of Fr3 as a whole during sample preparation.

3.2 Evaluation of the qHNMR methodology

The identification of QC markers for AS botanicals along with 1, representing the major constituent in the active Fr3 was one study goal. Another goal was to quantify 1 and the congeneric minor components by ¹H NMR. The targeted lower limit of quantification was a signal-to-noise (S/N) ratio of ≥ 150 (error values ≤ 1%) [42]. Since the current study was aimed at the quantification of NMR signals in a rather complex mixture, the region of integration was chosen as narrow as possible to fit the signal while excluding neighboring signals. This approach differs from a strategy proposed for the evaluation of pure reference materials, where spectra are much less crowded and the accuracy can be optimized by choosing an integral region that is 5 times the width-at-half-height of the quantification-signal [42]. Using the modified integration approach for qHNMR analysis of mixtures, the rNMR software was an excellent tool to ensure consistency among integral regions between spectra: (a) by defining regions of interest (ROIs) and (b) integration ranges which could be saved and applied to other 1D or 2D spectra at the same time or any time in the future. The ROIs were adjusted for all ¹H spectra including the calibration standards. A calibration curve was generated using 2 in CDCl₃ (Fig. 5). The residual solvent peak served as the internal standard to which all integral regions were normalized. In order to determine the comparability of the calibration curve of the current approach with other quantitative methods, the concentrations of the standards 2 and 3 (by weight) were correlated with their concentrations determined by the following qHNMR methods (Fig. 6.A): (a) using an internal standard (IS); (b) using an external standard (ES; TopSpin^™/Eretic2); (c) using the calibration curve (CC). Additionally, the quantitative results obtained from the integration routines using different processing software packages (NUTS, rNMR, and TopSpin^™) were compared. The qHNMR precision using an IS generally approached the 95% confidence interval when compared with the sample concentrations by weight, while the methods using an ES or a CC were found closer to a 90% confidence level (Fig. 6.A). Expressing the accuracy for each qHNMR method as the % deviation from the concentration by weight (=100%) resulted in 100.1 ± 5.7% for the IS method (precision 94.3%), 98.7 ± 5.8% for the ES method (precision 94.2%), and 103.8 ± 14.8% for relying on the CC method (precision 85.2%). Compared with previous qHNMR reports [31,42,43], the present study found lower achievable accuracy and precision. This was mainly attributed to the complex nature of the analyzed plant extracts and fractions, which required a reduction in the size of the integral regions (ROIs) below the aforementioned 5ω̄_1/2 threshold which has been proposed for the qHNMR of primary reference standards.

Fig. 5.

Calibration curve for the qHNMR analyses. The reference standard was dimethylsulfone (2), the concentrations were 0.28 mM, 1.48 mM, and 15.13 mM.

Fig. 6.

Comparison of different qHNMR methods used to determine the concentrations of reference standards 2 and 3 (A), as well as 1 in AS samples (B). A: Correlation of standard concentrations by weight [mg standard/600 μL] with their respective concentrations determined by qHNMR: processing software (NUTS, TopSpin^™, rNMR); quantitative methods: calibration curve (CC) with residual solvent signal as internal standard; 2 [1.48 mM] as external standard (ES) in the Eretic2 routine (TopSpin^™); 2 and 3 were used as mutual internal standards (IS). Dotted lines represent 95% and 90% confidence levels. B: Concentration of 1 (using H-6) determined in four AS Fr3 samples (a–d) using TopSpin^™ for processing and three quantification methods: rNMR for integration of H-6 and using the residual solvent signal as IS; Fityk to fit the signals of H-6 and using the residual solvent signal as IS; and 2 [1.48 mM] as external standard (ES) using the Eretic2 function in TopSpin^™.

3.3 Butylidene phthalide (4) as a reference compound for qHNMR

Compound 4 was present in Fr3 of AS preparations. Representing the aromatic derivative of 1, 4 is chemically much more stable and commercially available as a mixture of Z- and E-isomers. The purity of 4 was determined as 51% using the modified 100% method (Z-4: 45.7%, E-4: 5.3%, ratio=9:1). Each H-8 signal of the 4 isomers was used for quantification as they gave rise to the purest signals in the AS Fr3 samples. Using 2 as the IS, the content of 4 was determined as 53.8% (Z-4: 48.0%, E-4: 5.8%, ratio=8:1); CC gave a similar value of 53.0% (Z-4: 47.3%, E-4: 5.7%, ratio=8:1). These results confirmed that the modified 100% method is a valuable tool and produces accurate results for the purity of major constituents in a fraction or impure sample. The peak area integration in the GC-MS spectrum of Z/E-4 resulted in a ratio between Z-4 (R_t=13.8 min) and E-4 (R_t=14.5 min) of 10:1.

Because only low amounts of 4 are present in bioactive Fr3, this compound was used to explore the accuracy and precision of the qHNMR method when applied to mixture samples. The amount of 4 was determined in a sample of Fr3 (“matrix”), in a “matrix” sample spiked with 4, and in a sample with the same amount of 4 (Fig. 7). The CC method was used to determine absolute amounts of 4 in the 3 samples. The difference between its concentration in the spiked sample and the sample of pure 4 was expected to equal the amount of 4 found in the matrix sample. The results of this calculation are shown in Fig. 7, yielding 25.2 mM of 4 in the spiked sample, 22.9 mM in the standard sample, and 1.95 mM of 4 in the matrix sample of AS Fr3. The calculated sum of concentrations in the standard and matrix sample resulted in a slightly lower concentration for 4 (24.9 mM), corresponding to a 1.2% error.

Fig. 7.

Quantitative ¹H NMR spectra for the analysis of 4: (A) an AS Fr3 “matrix” sample, containing 23.95 mg/mL of Fr3 of BC426 in CDCl₃; (B) reference material of 4 (2 μL); (C) the “matrix” sample spiked with 2 μL of 4. Reference material 4 was a mixture of E-4 and Z-4. ¹H,¹H-COSY was used to select the purest signal of 4 for qHNMR based on minimum cross peak overlap. The signals of H-8 of both isomers E,Z-4 were used for quantification in all samples. The determined concentrations of 4 in the spiked sample vs. the genuine matrix sample plus the reference sample match within the error margin of the measurement.

3.4 Diethyl phthalate (5) as a reference compound for SPE recovery and qHNMR

In order to evaluate recovery of the bioactive fraction (Fr3) from the SPE cartridge, 5 was chosen as a reference because of its structural similarity to 1, its commercial availability in high purity (99.5%, Sigma-Aldrich; 97.5%, 100% method qHNMR (disregarding HDO) or 94.3% otherwise), and its chemical stability. The recovery from the SPE cartridge material was determined as 39 ± 6% (n=10) by weight and 24.5% (mean, n = 2) by qHNMR, both relative to the applied amount 5 (= 100% ± 3, n =10). Analysis of the ¹H NMR spectra of recovered 5 showed additional signals in the high field region of the spectrum which was indicative of co-eluted SPE cartridge material, and an increased HDO signal (Fig. S2). These observations resulted in a purity of 90.1% (excluding CHCl₃ and HDO) or 84.3% (excluding residual solvent only), which explained the higher recovery rate of 5 by weight. Reference material of 5 was subjected to the same SPE fractionation as the AS preparations in order to determine the amount of the non-recovered portion of bioactive Fr3 (Fig. 2). Further studies will be necessary to more accurately determine the amount of absorbed material.

3.5 Quantification of 1 in Fr3 of four AS samples

To quantify 1 in Fr3, four aliquots of the same AS extract were fractionated in parallel, and the Fr3s were analyzed by qHNMR. The purest signal of 1 (H-6) was chosen for quantification and determined using the integration of 1D gCOSY slices of overlapping cross-peaks (signal purity = 3.64% [44]). The bioactive Fr3 still represents a relatively complex mixture. To validate the qHNMR part of the method, the quantitative results obtained by different means of integration (NUTS, rNMR, and TopSpin^™ software) were compared with the results obtained by (a) line fitting of the H-6 signals using the Fityk software tool, (b) by CC calibration, and (c) using the ES method in TopSpin^™ (Eretic2) (Fig. 6.B). None of the software methods appeared to be of superior precision, including the line fitting approach which suffered from intrinsic integration inaccuracy due to the complexity of the H-6 signal: while appearing as a doublet of triplets, its higher order spin-effects will require quantum mechanics procedures for accurate integration of fitted lines (see [2] for perch analysis). However, the modified 100% method determination of 1 found in each of the Fr3 samples correlated well with the mean qHNMR results (Table 1).

3.6 Quantitative ¹H and COSY determination of 1 and minor constituents in Fr3

Based on these findings, rNMR was selected as tool for the quantification of 1 in each Fr3 fraction obtained from the 26 samples of AS. In addition Fr3 is proposed as a candidate for the integrated chemical and biological standardization of AS BDSs, and was used to identify new markers of botanical integrity.

Results obtained using the modified 100% qHNMR method were obtained and compared with absolute value (CC) results. The correlation of results for both methods is shown in Table 1 and Fig. 8. By examining the ¹H NMR spectra of samples that showed larger deviations between both methods, it became obvious that the observed deviations for samples with small signals for 1 resulted from elevated baselines due to the presence of degradation products and multiple overlapping (broad) resonances. This unavoidable effect as a result of the chemical instability of the bioactive Fr3 fraction explains the artificially elevated absolute values for the content for 1 when using a CC. The modified 100% method produced more accurate values, which was ascribed to the fact that integration of the whole spectrum took the elevated baselines into account. Examination of the ¹H NMR spectra of samples for which both results correlated well (within 80% confidence intervals = outer dotted trend lines, Fig. 8) confirmed that these spectra did not suffer from elevated baselines and the signals of 1 and other minor constituents could be identified more clearly (Fig. S3). In summary, the correlation plot in Fig. 8 could be a useful tool for a quick and overall quality assessment of AS preparations as samples of high quality showed good correlation between both methods (80% confidence level).

NMR data from previous studies in our laboratory [2] was used to identify minor constituents by comparison with 1D and 2D NMR (¹H,¹H-COSY, HMBC, and HSQC) spectra of Fr3. In addition to Z-1, the following minor compounds were chosen for quantitative evaluation: E-1, E/Z-4, phenylpropanoids (olefinic Hs, 6), polyynes (7), and poly-unsaturated fatty acids (8). While the H-6 signal of 1 was identified as purest signal, it is important to consider that with decreasing extract quality its purity is likely to decrease as a result of increasing peak overlap. To determine the amount of signal impurity, COSY cross-peaks that overlap with the H-6 integral region (5.99–6.04 ppm) were integrated and their amount calculated as percent of the cross-peak between H-6 and H-7 (= 100%, Table 2). This COSY-derived impurity value was used to correct all Z-1 concentrations in Tables 2 and 3.

Table 2.

Relative content of AS alkyl phthalides as determined by 2D COSY qHNMR. Concentrations for Z-1 were determined by qHNMR using the respective calibration curve. Z-1 concentrations were used to calculate the concentrations of the 6 additional alkyl phthalides in Fr3 that possess an olefinic H-8, and are based on COSY cross-peak volume ratios (n.d. = not determined).

	1H NMR	1H NMR		c)
AS sample	Z -1	impuritya)	Z -1b)	E -1	Z -4	E -4	UA1	UA2	UA3
code	μM	%	μM	μM
HP	2235	9	2030	86	56	24	21	74	23
P590	2461	17	2045	54	43	17	22	55	18
P1241	6207	8	5739	147	104	25	51	55	18
P130	2938	13	2552	24	19	17	10	84	14
P1341	4759	9	4309	112	93	69	33	120	39
P1518	1808	10	1624	67	62	40	17	63	19
P1731	1809	11	1616	69	64	39	14	70	23
P1841	3920	7	3639	140	71	32	31	97	23
P2041	946	14	816	55	26	20	7	62	23
P2370	5061	8	4675	122	165	74	34	287	147
P264	4540	11	4051	101	110	47	41	230	81
P2712	828	21	651	32	34	20	17	55	11
P280	4096	12	3618	43	37	26	21	129	28
P31	740	17	618	37	12	6	2	105	20
P471	1640	11	1457	59	76	32	13	53	14
P561	3107	10	2793	68	67	24	21	87	28
P661	1905	22	1492	42	14	8	19	57	13
P689	1998	21	1580	57	22	13	19	79	28
P765	1017	40	606	28	33	21	7	30	9
P976	3841	n.d.	n.d.	n.d.	n.d.	n.d.	n.d.	n.d.	n.d.
BC107	5269	19	4242	203	315	126	75	1131	538
BC165	294	24	223	11	13	2	1	20	7
BC426	20717	0	20717	1184	188	179	132	560	216
BC440	13982	8	12851	596	168	137	104	322	147
P766	2485	18	2025	97	55	22	14	89	30
StarW	2062	5	1964	22	44	8	21	30	14

^a)signal impurity of ¹H NMR integral, determined using the COSY cross peak overlap (impurities are expressed relative to the main cross peak)

^b)corrected Z -1 concentration (corrected by signal impurity of H-6 using COSY)

^c)concentrations of E -1, Z -4, E -4, UA1, UA2, and UA3 determined as COSY cross peak ratios to Z -1 (cross peak H-8/olefinic H-9)

Based on COSY spectra, seven cross-peaks between olefinic H-8 and H-9 resonances of alkyl phthalides were identified in the spectral windows of 2.10–2.70 ppm and 4.90–5.90 ppm. The rNMR tool was used to determine their cross-peak volumes in 26 AS samples (Table 2). Three of these alkyl phthalides could be identified as E-1 and E,Z-4, and three additional remain unidentified alkyl phthalides (UA 1–3). The COSY data was used for quantification because COSY cross-peaks were less overlapped than ¹H NMR signals. Since the molecular structures and associated J coupling patterns for the alkyl phthalides with olefinic H-8s are very similar, the volume integrals are largely proportional and can be used to calculate the ratios of such congeners present in each Fr3. Because the absolute concentration for Z-1 had already been confirmed by qHNMR already (Table 1), the concentrations for the other alkyl phthalides could be deduced from their cross-peak volume ratios with Z-1. A plot of the data (Fig. 10.A) shows that a Z-1 concentration of ≥ 85% is indicative of high product quality, and led to the identification of five samples of low quality (P2041, P2712, P31, BC107, and BC165). As a next step, the absolute amounts for the six alkyl phthalide compounds were calculated as weight percent of the AS preparations applied to the SPE cartridge. Plotting this data (Fig. 10.B) indicated that the total amounts of alkyl phthalides in AS preparations can also be used as a marker for their quality. AS preparations with ≤ 0.3% (w/w) of alkyl phthalides were of lower quality, thus allowing the grading of twelve preparations (P1518, P1731, P2041, P2712, P31, P471, P689, P765, P976, BC165, P766, StarW).

Fig. 10.

Concentration of AS alkyl phthalides as determined by 2D COSY. (A) Relative content of six alkyl phthalides. The values are based on ratios of ROIs of the relative cross-peak volumes between protons H-8 and H-9 of the alkyl phthalides and are calculated as COSY cross peak ratios relative to Z-1. (B) Absolute amounts of Z-1, E-1, Z-4, E-4, and the three unidentified congeners, UA 1–3, expressed as [% (m/m)] of the AS extract applied to the SPE separation. Values were calculated as % ratios relative to Z-1. An Mr of 190 mg/mmol was used for all compounds. Absolute amounts for Z-1 were determined by qHNMR.

The absolute value content of the other compound classes, e.g., phenylpropanoids, polyynes, and poly-unsaturated fatty acids, was determined by qHNMR using the signals of the protons shown in their respective molecular drawings, unless otherwise stated in Table 1. Fig. 10 shows the plot of the relative mM concentrations of the metabolites 1, 4, 6, 7, and 8 in the 26 samples of AS preparations (Table 1). Pursuing the group integration of these additional three compound classes using COSY data, the cross-peaks of the phenylpropanoids in the spectral region bounded by 7.58–7.74 and 6.26–6.40 ppm were integrated as representative olefinic Hs (Fig. S4.A). Similarly, the collective integration for a group of polyynes was achieved by integrating the COSY spectra in the spectral region 5.90–6.05 and 4.92–4.98 ppm, which contained the cross-peaks between the olefinic H-2 and H-3 methine protons with a geminal OH-group (Fig. S4.B). The quantity of poly-unsaturated fatty acids could be assessed by integrating the cross-peaks between their allylic and olefinic Hs (2.70–2.85 and 5.28–5.42 ppm, Fig. S4.C). The ¹H NMR signal pattern indicated that the samples contained a mixture of several 18:Δ fatty acids such as γ-linolenic acid (18:3, n-6) or linoleic acid (18:2, n-6). It is important to note that COSY cross-peak volumes depend on several variables including compound specific parameters (e.g., proton T₁, magnitude of J coupling between protons giving rise to the cross peak, conformational dynamics, etc.). Thus, COSY integrals cannot directly be used for absolute quantification of the compound groups 6, 7, and 8, but require calibration with identical standards of each compound for accurate quantification. However, for the purpose of establishing the botanical standardization methodology, in particular when developing convention methods, it is a viable approach to compare the COSY-based concentrations for each compound class across the 26 AS samples. For the three compound classes that were detected in fractions Fr3 in the present study, no obvious patterns were observed for the presence and distribution of non-alkyl phthalide constituents that would constitute obvious differential criteria for the integrity of AS preparations (Table 1, Fig. 9).

Fig. 9.

Relative content of AS constituents as determined by qHNMR. Absolute value content using the calibration curve of metabolites 1, 4, 6, 7, and 8. The S/N ratio was ≥ 150 for all signals except for most E-4 and the signals of 6 in P130, P241, P31, P976, BC165, and P766.

3.7 Discussion of the accuracy and precision of the SPE method

The current study demonstrates that one advantage of the SPE method was its capability of concentrating the active principle(s) into a chemically defined Fr3, representing a reduced window of the total AS phytometabolome. One remaining challenge for future studies is to determine the non-recoverable portion of the bioactive Fr3 (Fig. 2). The recovery of standard compound 5 was found to be only 39 ± 6%, but it remains unclear how well 5 could represent the “chemical behavior” of the constituents in Fr3. Ongoing work focuses on translating the SPE methodology into a loss-free liquid-liquid separation scheme to obtain the total bioactive fraction. This will enable measurement of the SPE-irrecoverable portion of Fr3 and may further improve reproducibility of quantifying multiple marker compounds in the active fraction.

3.8 Discussion of the accuracy and precision of the qHNMR method

The accuracy and precision of the qHNMR method (using rNMR, CC) is rated suitable for the purpose of multi-target standardization. This assessment is based on determination with the reference materials 2, 3, 4, and 5, comparison of quantitative results obtained with different NMR processing software packages, and correlations of results when using an IS, the CC, and the modified 100% method for calibration. The slightly reduced accuracy and precision observed in the present study, when compared with previous qHNMR reports on pure reference materials, was attributed to the influence of unavoidable peak overlap in complex botanical mixtures and the resulting selection of sub-optimal integration regions. The modified 100% method was validated by using a reference standard of 4. Interestingly, it was found to be a very valuable method when determining the purity of major signals in a complex mixture: due to its simplicity, it can be used with any ¹H NMR data set and without any need for special sample preparation or calibration.

3.9 Discussion of the identified quality markers of AS preparations

Since 1 was repeatedly found to be (associated with) the active principle(s), including in the present study, the beginning hypothesis was to use 1 as product quality marker. The concentrations of Z-4 and E-4 as aromatic derivatives could be regarded as markers of reduced quality, as they indicate exposure of the AS preparations to oxidation. This second hypothesis was supported by the finding that pure reference material of Z-4 and E-4 did not exhibit biological activity, i.e., ALP inhibition in an endometrial adenocarcinoma cell line [16]. The alkyl phthalides, together with 1 as the major compound, have been reported to be chemically unstable. This is due in part to the formation of thermally produced Diels-Alder dimers [45] and cyclobutane dimers presumably formed from light exposure [46] of 1. This known instability has led to the third hypothesis that monomeric alkyl phthalides as (part of) the bioactive principle are also prone to chemical degradation. Thus, understanding the role of the antioxidant phytoconstituents in AS preparations and their potential to stabilize alkyl phthalide monomers could be key to AS QC, and to establishing new standardization procedures for AS BDSs that target the active components. Considering their documented antioxidant properties [47,48], the presence of phenylpropanoids (6) was regarded as a marker of stability and, thus, good quality. Their concentration in the botanical preparation is expected to decrease over time, with exposure to oxidative processes. While the quantitative results from 1D qHNMR or COSY analysis did not reveal distinctive patterns to support this hypothesis, additional factors that could not be detected or recognized in the present panel of samples may play a role, such as factors related to the source and handling of the plant material [49] or the extraction process [50], which together may impact the presence of antioxidant phenylpropanoids.

Because of their highly unsaturated nature, the compound classes of polyynes (7) and poly-unsaturated fatty acids (8) were added to the third working hypothesis and also regarded as antioxidants. Hence, their olefinic protons were used for the 1D qHNMR and COSY analyses. Again, no conclusive connection between the age of the AS preparation and the concentrations of 7 or 8 could be made. Overall, the concentrations of the three antioxidant classes, 6, 7, and 8 did not seem to correlate with each other, i.e., their concentrations were not consistently elevated or reduced in any particular extract.

4. Conclusions

The goal of this study was to develop a new method and to identify new quality standards for AS BDSs while targeting the concentration of phytoconstituents in the bioactive fraction. This can be considered an important first step toward the combined biological and chemical standardization of AS preparations.

The chosen tandem approach of (a) sample clean-up (removal of non-bioactive materials) and (b) qNMR analysis of the active fraction provides a viable and attractive alternative to GC-/LC-based methods, for the following reasons: (a) no authentic, identical reference materials are required, in particular, there is no need to obtain primary reference standards of unstable 1; (b) the overall accuracy of the quantitative method increased as the overall quantification accuracy did not depend on the quality of the available authentic reference materials, but utilized a single, universal calibrant (e.g., DMSO₂); (c) the method targets an extended window of phytoconstituents belonging to different compound classes (Fig. 2) rather than a single compound. The latter point considers the frequent observation made during bio-activity guided fractionation of AS, which predominantly yields one active fraction but leads to the isolation of more than one active compound. By combining these three aspects, the tandem approach presented here enables a combined chemical and biological standardization protocol, and also makes standardization applicable to unidentified but active components, multiple active components, and synergistically acting compounds.

The following compounds and compound classes were present in the bioactive fraction Fr3 and were evaluated as quality markers for AS botanicals: Z/E-1, Z/E-4 and 3 unidentified alkyl phthalides, phenylpropanoids (6), polyynes (7), as well as poly-unsaturated fatty acids (8). Based on the findings described here, AS product integrity was related to the composition of the alkyl phthalide complex, as well as the absolute amount of alkyl phthalides in the AS preparations. The following quality markers were identified: AS preparations of good quality had absolute amounts of alkyl phthalides of ≥ 0.3% (w/w) (Fig. 10.B), and the relative amount of 1 in the alkyl phthalide complex was ≥ 85% (Fig. 10.A). In addition, the relative amounts of 1 compared to the total amount of all compound classes were ≥ 50% in quality AS extracts. The fourth quality marker relates to the presence/absence of baseline elevations in the qHNMR spectra, which were attributed to multiple overlap of broad signals arising from degradation products (impurities) (Fig. 7). The other three compound classes (phenylpropanoids, polyynes, unsaturated fatty acids) did not appear to be directly correlated with product quality, and the factors that influence their content in the preparations need to be further evaluated, likely using an extended panel of extracts with a larger variability of production.

The availability of a qHNMR batch processing method, a feature present in the rNMR software tool, is an important factor in the botanical QC protocol. Further comprehensive studies on the chemical interactions of the four compound classes 1 (alkyl phthalides), 6, 7, and 8 in the bioactive fraction of AS preparations are required to better understand the influence of their respective concentrations on product quality and stability [8]. Obtaining the bioactive fraction (Fr3) first and subsequently performing qHNMR analysis is proposed as a model for such studies. In addition to studying their chemical interactions, performing bioassays of each compound class in different assay systems and using different combinations of agents will provide indications of potential biological interactions between the bioactive phytoconstituents present in AS. Ongoing work in our laboratory is aimed at measuring the amount of the non-recovered portion of the bioactive fraction more accurately. The accuracy of the qHNMR method, when applied to complex natural samples, is also being studied by exploring different experimental conditions (e.g., acquisition/processing techniques) in order to reduce errors that would result from inevitable signal overlap.

Fig. 1.

Chemical structures of reference standards and marker constituents in the bioactive fraction, Fr3, of AS [7].

Appendix A. Supplementary data

Supplementary data to this article can be found online at doi:10.1016/j.fitote.2011.xx.xxx.