Identifying the compounds that can distinguish between Saposhnikovia root and its substitute, Peucedanum ledebourielloides root, using LC-HR/MS metabolomics

Article abstract of DOI:10.1007/s11418-020-01409-6

Abstract: Previously, we established a ¹H NMR metabolomics method using reversed-phase solid-phase extraction column (RP-SPEC), and succeeded in distinguishing wild from cultivated samples of Saposhnikoviae radix (SR), and between SR and its substitute, Peucedanum ledebourielloides root (PR). Herein, we performed LC-HR/MS metabolomics using fractions obtained via RP-SPEC to identify characteristic components of SR and PR. One and three characteristic components were respectively found for SR and PR; these components were isolated with their m/z values and retention times as a guide. The characteristic component of SR was identified as 4′-O-β-d-glucosyl-5-O-methylvisamminol (1), an indicator component used to identify SR in the Japanese Pharmacopoeia. In contrast, the characteristic components of PR were identified as xanthalin (2), 4′-O-β-d-apiosyl (1 → 6)-β-d-glucosyl-5-O-methylvisamminol (3), and 3′-O-β-d-apiosyl (1 → 6)-β-d-glucosylhamaudol (4) based on spectroscopic data such as 1D- and 2D-NMR, MS, and specific optical rotation. Among them, 4 is a novel compound. For the correlation between the NMR metabolomics results in the present and our previous report, only 1 and 2 were found to correlate with the chemical shifts, and the other compounds had no correlation. As the chemical shifts for compounds 1, 3, and 4 were similar to each other, especially for the aglycone moiety, they could not be distinguished because of the sensitivity and resolution of ¹H NMR. Accordingly, combining NMR and LC/MS metabolomics with their different advantages is considered useful for metabolomics of natural products. The series of methods used in our reports could aid in quality evaluations of natural products and surveying of marker components. Graphic abstract: [Figure not available: see fulltext.]

Full text of DOI:10.1007/s11418-020-01409-6

Journal of Natural Medicines
https://doi.org/10.1007/s11418-020-01409-6
ORIGINAL PAPER
Identifying the compounds that can distinguish
between Saposhnikovia root and its substitute, Peucedanum
ledebourielloides root, using LC‑HR/MS metabolomics
Taichi Yoshitomi1 · Daigo Wakana2 · Nahoko Uchiyama1 · Takashi Tsujimoto1 · Noriaki Kawano3 · Tsuguo Yokokura4 ·
Yutaka Yamamoto5 · Hiroyuki Fuchino3 · Takashi Hakamatsuka1 · Katsuko Komatsu6 · Nobuo Kawahara3 ·
Takuro Maruyama1
Received: 4 November 2019 / Accepted: 5 April 2020
© The Japanese Society of Pharmacognosy 2020
Abstract
1
Previously, we established a H NMR metabolomics method using reversed-phase solid-phase extraction column (RP-
SPEC), and succeeded in distinguishing wild from cultivated samples of Saposhnikoviae radix (SR), and between SR and
its substitute, Peucedanum ledebourielloides root (PR). Herein, we performed LC-HR/MS metabolomics using fractions
obtained via RP-SPEC to identify characteristic components of SR and PR. One and three characteristic components were
respectively found for SR and PR; these components were isolated with their m/z values and retention times as a guide. The
characteristic component of SR was identifed as 4′-O-β-d-glucosyl-5-O-methylvisamminol (1), an indicator component used
to identify SR in the Japanese Pharmacopoeia. In contrast, the characteristic components of PR were identifed as xanthalin
(2), 4′-O-β-d-apiosyl (1→6)-β-d-glucosyl-5-O-methylvisamminol (3), and 3′-O-β-d-apiosyl (1→6)-β-d-glucosylhamaudol
(4) based on spectroscopic data such as 1D- and 2D-NMR, MS, and specifc optical rotation. Among them, 4 is a novel
compound. For the correlation between the NMR metabolomics results in the present and our previous report, only 1 and
2 were found to correlate with the chemical shifts, and the other compounds had no correlation. As the chemical shifts for
compounds 1, 3, and 4 were similar to each other, especially for the aglycone moiety, they could not be distinguished because
of the sensitivity and resolution of 1H NMR. Accordingly, combining NMR and LC/MS metabolomics with their diferent
Electronic supplementary material The online version of this
article (https://doi.org/10.1007/s11418-020-01409-6) contains
supplementary material, which is available to authorized users.
* Takuro Maruyama
t-maruya@nihs.go.jp
1
Division of Pharmacognosy, Phytochemistry and Narcotics,
National Institute of Health Sciences, 3-25-26 Tonomachi,
Kawasaki-ku, Kawasaki, Kanagawa 210-9501, Japan
2
Faculty of Pharmaceutical Sciences, Hoshi University, 2-4-41
Ebara, Shinagawa-ku, Tokyo 142-8501, Japan
3
Research Center for medicinal Plant Resources, National
Institutes of Biomedical Innovation, Health and Nutrition,
1-2 Hachimandai, Tsukuba, Ibaraki 305-0843, Japan
4
Nippon Funmatsu Yakuhin Co., Ltd, 71-3 Izumicho, Yao-shi,
Osaka 581-0813, Japan
5
Tochimoto Tenkaido Co., Ltd, Oniya Kaibara-Cho, Tamba,
Hyogo 669-3315, Japan
6
Department of Medicinal Resources Institute of Natural
Medicine, University of Toyama, 2630 Sugitani,
Toyama 930-0194, Japan
Vol.:(0123456789)
1 3

Journal of Natural Medicines
advantages is considered useful for metabolomics of natural products. The series of methods used in our reports could aid
in quality evaluations of natural products and surveying of marker components.
Graphic abstract
1
Keywords H NMR-based metabolomics · LC-HR/MS metabolomics · Reversed-phase solid-phase extraction · Crude
drugs · Saposhnikoviae radix (SR) · Peucedanum ledebourielloides root (PR)
pharmacological activities such as anti-infammatory and
antioxidant actions have been reported [2, 3]. On the other
hand, SR is known to have a variety of components. In
fact, Kim et al. and Batsukh et al. reported the chemical
diversity in S. divaricata roots from Mongolia, China, and
South Korea [4, 5]. Other apiaceous plants that naturally
grow in China have been traditionally used as substitutes
for SR in local regions [6–9]. Some of the SR available
in Shaanxi and the surrounding regions were found to be
derived from Peucedanum ledebourielloides root (PR)
[9]. Owing to this background, constructing an objective
method to clearly distinguish between SR and its substi-
tutes is desirable.
Introduction
In the 17th edition of the Japanese Pharmacopoeia (JP;
The Ministry of Health, Labor and Welfare, Japan, 2016),
Saposhnikoviae radix (SR) is defned as a crude drug
derived from the roots and rhizomes of Saposhnikovia
divaricata Schischkin (Umbelliferae; Apiaceae). SR is a
crucial crude drug in oriental medicine and is utilized in
many Kampo medicines (traditional Japanese medicines)
owing to its perspiration and antipyretic actions [1]. Cou-
marins, chromones, essential oils, and polyacetylene
compounds have been reported as ingredients in apia-
ceous plants [2] and are contained in SR. Besides, their
1 3

Journal of Natural Medicines
In our previous paper, we applied 1H NMR metabolomics
using a reversed-phase solid-phase extraction column (RP-
SPEC) to SR and PR samples, and successfully discrimi-
nated between wild and cultivated samples of SR, and SR
and PR samples [10]. In the report, saccharides, whose main
content is sucrose, were identifed as the components that
aided in discriminating between wild and cultivated samples
through comparison to authentic compounds. In contrast,
the components contributing to the discrimination between
SR and PR were assumed to be the secondary metabolites
based on their chemical shifts in the s-plot; these, however,
could not be identifed. This is because components in apia-
ceous plants have high structural similarity, and obtaining
the standard was difcult. Recently, Batsukh et al. investi-
gated the chemical diversity of SR from diferent regions of
Mongolia by multivariate analysis of LC/MS data. In the
study, isolation of the characteristic components was needed
to achieve unambiguous identifcation [5]. Furthermore,
since the solvents used in metabolomics difer from those
used in structural analysis, a direct comparison is difcult
to perform. Thus, as we could not resolve these factors, even
when 1H NMR metabolomics using RP-SPEC was adopted,
their identifcation is still warranted.
from Kanto Chemicals. Methanol-d4, chloroform-d, and
DMSO-d6 were purchased from Taiyo Nippon Sanso. Ultra-
pure water was prepared by Integral-5 (Merck). 4′-O-β-d-
glucosyl-5-O-methylvisamminol was purchased from Wako
Pure Chemical, and 4′-O-β-d-apiosyl (1→6)-β-d-glucosyl-
5-O-methylvisamminol was purchased from Chem Scene.
Falcarindiol was purchased from Wuhan ChemFaces. d- and
l-Apiose were purchased from Carbosynth. l-Glucose was
purchased from Sigma-Aldrich. d-Glucose was purchased
from Kanto Chemical Co., Inc.
Source and documentation
SR and PR samples were provided by the National Institutes
of Biomedical Innovation, Health and Nutrition, Nippon
Funmatsu Yakuhin Co. Ltd, and Tochimoto Tenkaido Co.
Ltd. Further details are provided elsewhere [10].
Sample preparation for LC‑HR/MSS metabolomics
Solution C (fraction eluted with 100% methanol using solid-
phase extraction) of SR and PR prepared in our previous
study [10] was diluted fvefold with methanol for HPLC
and fltered with a 0.45-μm flter to prepare the analytical
samples (approximately 1 mg/ml; crude drug weight/sol-
vent volume). In addition, reference solution for the align-
ment of chromatograms was prepared by mixing all sample
solutions.
In the present study, we performed LC-HR/MS metabo-
lomics using RP-SPEC to obtain fractions to identify the
compounds contributing to the discrimination between SR
and PR. We succeeded in identifying 4 components includ-
ing a novel compound. We also discussed the features of
LC-HR/MS and 1H NMR metabolomics.
Liquid chromatography/high‑resolution mass
spectrometry (LC‑HR/MS) analysis
Materials and methods
All sample solutions were analyzed using Ultimate 3000
HPLC with an Orbitrap high-resolution mass spectrometer
and the Cortecs 1.6 μm, C18, 2.1×100 mm column main-
tained at 40 °C. Autosampler temperature was set to 4 °C
and injection volume was 1 μl. Mobile phase consisting
of 0.1% formic acid (solvent A) and acetonitrile contain-
ing 0.1% formic acid (solvent B) was run at a fow rate of
0.35 ml/min under the following gradient condition: 5%
B, 3.5 min; 5–70% B, 10 min; 70–100% B, 4.5 min; and
100% B, 3.5 min. Electrospray ionization mass spectrom-
etry (ESI–MS) parameters were: ion polarity, positive;
scan range, m/z 100 to 1000; capillary temperature, 400 °C;
sheath gas, 50 l/min; AUX gas, 25 l/min; capillary voltage,
15 V; and MS resolution, 60,000. All samples were analyzed
in triplicate.
General experimental procedures
Nuclear magnetic resonance (NMR) spectra were obtained
using ECZ-600 and ECA-800 NMR Spectrometers (JEOL).
Sep-Pak® plus C18 environmental cartridges (Waters) and
vacuum manifold (GL Sciences) were used to pretreat the
crude drug extracts. Automatic fash purifcation and recy-
cling preparative HPLC were performed using Isolera one
system (Biotage) equipped with a SNAP Ultra fash chro-
matography cartridge (25, 50, and 100 g), and the LaboAce
LC-5060 system (Japan Analytical Industry) equipped with
JAIGEL-ODS-AP-L (20 mm ID×500 mm; Japan Analytical
Industry), respectively. LC-HR/MS analysis was performed
using the Ultimate 3000 HPLC system (Thermo Fisher Sci-
entifc) linked to an Orbitrap XL mass spectrometer with
electron-transfer dissociation (Thermo Fisher Scientifc)
and equipped with Cortecs C18, 1.6 μm, 2.1 × 100 mm
column (Waters). Specifc optical rotation was measured
by DIP-370 (JASCO). Methanol, n-hexane, chloroform,
ethyl acetate, acetonitrile, and n-BuOH were purchased
Multivariate statistical analysis
Progenesis QI (Waters) was used for alignment and peak
extraction from the chromatogram. For peak extraction,
base peak mode (0.01%) was used. The obtained peaks were
1 3

Journal of Natural Medicines
converted into a data matrix by EZinfo statistical package
(Umetrics). PCA and PLS analyses were carried out using
SIMCA-14 software (Umetrics, Umeå, Sweden). Pareto scal-
ing was applied as a preprocessing method and the number
of latent variables was fxed as 3.
PR powder (131.5 g) was extracted three times with chlo-
roform/methanol=1/1 (600 ml) by sonication for 30 min.
All extracts were then combined and evaporated under
reduced pressure to yield the crude dry extract (12.8 g). Liq-
uid–liquid partitioning of the water suspension was carried
out with ethyl acetate, followed by n-BuOH. The n-BuOH
extract (2.63 g) was suspended in methanol (2.5 ml) and
applied to a SNAP samplet HP-SIL (10 g; Biotage). After
methanol removal under reduced pressure, the samplet was
applied to SNAP Ultra 50 g (Biotage). Chloroform (A) and
methanol (B) were used as the mobile phase at a fow rate
of 100 ml/min under the following gradient condition: 12%
B, 0–12 min; 12–15% B, 10 min; 20% B, 7 min. Fraction A
(25 min) for 3 and fraction B (7 min) for 4 were obtained.
Fraction A was again fractionated using SNAP Ultra 100 g
(Biotage), with mobile phase, chloroform (A) and metha-
nol (B), at 100 ml/min under the following gradient condi-
tion: 12% B, 0–10 min; 12–14% B, 17 min; 20% B, 5 min.
Compound 3 was isolated in the 20% B fraction (15.6 mg)
which was further fractionated using SNAP Ultra C18 30 g
(Biotage). Mobile phase consisting of water (A) and acetoni-
trile (B) was used at 25 ml/min under the following gradient
condition: 18% B, 0–24 min; 18–20% B, 15 min; 4 (12.4 mg)
was obtained at 36 min.
Isolation of compound 1
SR powder (112.2 g) was extracted three times with metha-
nol (500 ml) by sonication for 30 min. All extracts were
combined and then evaporated under reduced pressure to
yield the crude dry extract (12.8 g). Liquid–liquid parti-
tioning of the water suspension was carried out with ethyl
acetate, followed by n-BuOH. The n-BuOH extract (6.35 g)
was then suspended in chloroform (3.0 ml), mixed with
silica gel (8.0 g), added to a SNAP empty Samplet (10 g;
Biotage), and used in fash chromatography. Mobile phase
consisting of chloroform (A) and methanol (B) was used at
a fow rate of 100 ml/min with A/B=88/12 for 10 min. The
fraction containing 1 was obtained at 5 min (276.4 mg), and
was applied to recycling preparative HPLC with acetoni-
trile/water=80/20 at a fow rate of 6.5 ml/min to isolate 1
(90.0 mg).
4′-O-β-d-Glucosyl-5-O-methylvisamminol (1): white
amorphous powder; [ꢀ]26 + 53.4, c 1.30, in methanol
4′-O-β-d-Apiosyl (1 → 6)-β-d-glucosyl-5-O-
D
([ꢀ]25 + 88.2, c 1.66, in methanol [19]), HR-ESI–MS m/z
453D.1771 [M + H]+ (Calcd. C22H29O10, 453.1761). The
chemical shifts from 1H and 13C NMR were identical to the
published data [12].
methylvisamminol (3): white amorphous powder;
[ꢀ]24 − 34.9, c 1.01, in methanol ([ꢀ]28 − 4.17, c 0.96, in
D
meDthanol [7]), HR-ESI–MS m/z 585.2175 [M+H]+ (Calcd.
C27H37O14, 585.2183), 1H NMR (DMSO-d6, 600 MHz) H:
6.67 (1H, s, H-8), 5.96 (1H, s, H-3), 4.83 (1H, t, J=9.0 Hz,
H-2′), 4.80 (1H, d, J = 3.0 Hz, H-1′’’), 4.40 (1H, d,
J=7.2 Hz, H-1′’), 3.82 (1H, d, J=9.6 Hz, H-4′’’), 3.85 (3H,
s, 5-OCH3), 3.68 (1H, d, J=3.0 Hz, H-2′’’), 3.66 (1H, br d,
J=9.6 Hz, H-6′’), 3.56 (1H, d, J=9.6 Hz, H-4′’’), 3.38 (1H,
m, H-3′), 3.32 (2H, m, H-5′’’), 3.29 (1H, m, H-6′’), 3.24
(2H, m, H-3′, H-5′’), 3.14 (1H, t, J= 9.0 Hz, H-3′’), 2.97
(1H, t, J = 9.0 Hz, H-4′’), 2.87 (1H, t, J = 8.1 Hz, H-2′’),
2.26 (3H, s, 2-CH3), 1.26 (3H, s, 4′-gem CH3), 1.23 (3H, s,
4′-gem CH3); 13C NMR (DMSO-d6, 150 MHz) δC: 175.3
(C-4), 164.0 (C-7), 163.3 (C-2), 158.9 (C-8a), 155.0 (C-5),
117.1 (C-6), 110.9 (C-4a), 110.7 (C-3), 109.1 (C-1′’’), 97.1
(C-1′’), 93.1 (C-8), 89.9 (C-2′), 78.6 (C-3′’’), 76.8 (C-4′),
76.7 (C-3′’), 75.8 (C-2′’’), 75.0 (C-5′’), 73.3 (C-2′’), 73.1
(C-4′’’), 70.0 (C-4′’), 67.4 (C-6′’), 63.1 (C-5′’’), 60.1 (C-5-
OCH3), 27.0 (C-3′), 22.6 (C-4′-gem CH3), 21.7 (C-4′-gem
CH3), 19.1 (2-CH3).
Isolation of compound 2
PR powder (8.5 g) was extracted three times with ethyl
acetate (100 ml) by stirring for 60 min followed by evapora-
tion under reduced pressure to yield the crude dry extract
(1.27 g). The extract was then suspended in ethyl acetate
(1 ml) and applied to a SNAP samplet HP-SIL (3 g; Bio-
tage). After the removal of ethyl acetate under reduced
pressure, the samplet was applied to fash chromatography.
Mobile phase consisting of n-hexane and ethyl acetate mix-
ture (4/1) was used at 75 ml/min for 5 min. The fraction
collected during the frst 3 min (23.0 mg) was again sepa-
rated using ODS fash chromatography with water (A) and
acetonitrile (B) at 25 ml/min under the following gradient
condition: 40% B, 18 min; 40–60% B, 22 min; 60–75% B,
10 min. The peak at 36 min was collected and 2 was isolated
(5.7 mg).
3′-O-β-d-Apiosyl (1 → 6)-β-d-glucosylhamaudol (4):
white amorphous powder; HR-ESI–MS m/z 571.2018
[M+H]+ (Calcd. C26H35O14, 571.2027). 1H NMR (DMSO-
d6, 600 MHz), δH: 13.1 (1H, s, 5-OH), 6.36 (1H, s, H-8),
6.17 (1H, s, H-3), 4.92 (1H, d, J=3.0 Hz, H-1′’’), 4.34 (1H,
d, J=7.8 Hz, H-1′’), 3.91 (1H, t, J=5.4 Hz, H-3′), 3.85 (1H,
dd, J=10.8, 1.2 Hz, H-6′’), 3.83 (1H, d, J=9.6 Hz, H-4′’’),
Xanthalin (2): colorless amorphous powder;[ꢀ]24 −91.9,
D
c 0.73, in ethanol ([ꢀ]20 − 164.2, c 0.96, in ethanol [20]),
HR-ESI–MS m/z 427D.1757 [M + H]+ (Calcd. C24H27O7,
427.1757). The chemical shifts from 1H and 13C NMR were
identical to the published data [9].
Isolation of compounds 3 and 4.
1 3

Journal of Natural Medicines
3.75 (1H, dd, J=6.0, 2.4 Hz, H-2′’’), 3.60 (1H, d, J=9.0 Hz,
H-4′’’), 3.44 (1H, dd, J = 10.8, 7.2 Hz, H-6′’), 3.34 (2H,
m, H-5′’’), 3.32 (1H, m, H-5′’), 3.18 (1H, ddd, J=8.7, 4.2,
4.2 Hz, H-3′’), 3.02 (1H, ddd, J =7.2, 5.4, 4.8 Hz, H-4′’),
2.94 (1H, ddd, J = 8.4, 5.4, 5.4 Hz, H-2′’), 2.84 (1H, dd,
J=17.2, 5.4 Hz, H-4′), 2.64 (1H, dd, J=17.2, 5.4 Hz, H-4′),
2.35 (3H, s, 2-CH3), 1.33 (3H, s, 2′-gem CH3), 1.31 (3H, s,
2′-gem CH3); 13C NMR (DMSO-d6, 150 MHz) δC: 181.9
(C-4), 167.8 (C-2), 158.7 (C-5, C-8a), 155.4 (C-7), 109.2
(C-1′’’), 107.7 (C-3), 103.4 (C-4a, C-6), 100.9 (C-1′’), 94.3
(C-8), 78.8 (C-3′’’), 77.9 (C-2′), 76.8 (C-3′’), 75.8 (C-2′’’),
75.6 (C-5′’), 73.3 (C-4′’’), 73.2 (C-3′), 73.1 (C-2′’), 70.3
(C-4′’), 67.5 (C-6′’), 63.3 (C-5′’’), 25.2 (C-2′-gem CH3),
22.0 (C-2′-gem CH3), 21.7 (C-4′), 20.0 (2-CH3).
were also applied to the authentic d- and l-glucose and api-
ose. The retention times of the authentic standard deriva-
tives were 15.78 min for d-glucose, 14.36 min for l-glucose,
20.97 min for d-apiose and 12.30 min for l-apiose.
Results and discussion
Quantitative range and reproducibility in LC‑HR/MS
analysis
First, we investigated the dynamic range and reproducibil-
ity of data acquisition for multivariate analysis. ESI was
selected as the MS interface because of its wide ionization
range from low to high polar compounds. The gradient pro-
gram of 5% B to 100% B was also adopted as the mobile
phase to cover the compounds’ wide polarity range. Repre-
sentative total ion current (TIC) chromatograms of SR and
PR are shown in Fig. S1; the peak with the higher inten-
sity (7.38 min; m/z 453.1798 of the pseudo-molecular ion)
and that with the relatively lower intensity (8.13 min; m/z
291.1264 of the pseudo-molecular ion) were selected. Cali-
bration curves were then prepared for peak areas and sam-
ple concentrations (Fig. S2). Good linearity was obtained at
0.001–10 mg/ml for SR and PR (Fig. S2). Therefore, 1.0 mg/
ml was used as the sample concentration for multivariate
analysis to ensure quantitative linearity in the concentration
variation of the components from one-tenth to tenfold.
To verify the repeatability of the data, samples were
repeatedly analyzed and variations in retention time were
assessed (Fig. S2). A great change in retention time was not
observed and the high reproducibility was also confrmed.
In LC-HR/MS metabolomics, the data matrix consists of the
qualitative parameter (retention time, m/z) and quantitative
parameter (peak area). Data acquisition with the appropriate
dynamic range and reproducibility is thus important to per-
form the correct analysis. Based on the above investigations,
we deduced that the quality of the qualitative and quantita-
tive parameters was sufcient for multivariate analysis.
Hydrolysis of compound 4
We added 1,4-dioxane (1 ml) and 2 M HCl (1 ml) to 4
(19.2 mg), and refuxed the mixture for 3 h with a mantle
heater. The reaction mixture was then extracted with water/
ethyl acetate = 1/1 (2 ml), and the ethyl acetate layer was
used in recycling preparative HPLC with acetonitrile at a
fow rate of 6.25 mL/min to obtain compound 4a (1.8 mg).
Hamaudol (4a): white amorphous powder; [ꢀ]25 −39.8,
D
c 0.625, in ethanol ([ꢀ]24 – 32.0, c 0.13, in ethanol [8]),
D
1
ESI–MS m/z 277.1 [M + H]+. H NMR (chloroform-d,
600 MHz) δH: 13.0 (1H, s, 5-OH), 6.30 (1H, s, H-8), 5.97
(1H, s, H-3), 3.86 (1H, br-d, J=4.8 Hz, H-3′), 2.95 (1H, dd,
J=17.4, 4.8 Hz, H-4′), 2.73 (1H, dd, J=17.4, 5.4 Hz, H-4′),
2.31 (3H, s, 2-CH3), 1.37 (3H, s, 2′-gem CH3), 1.33 (3H, s,
2′-gem CH3); 13C NMR (chloroform-d, 150 MHz) (Table 1).
Determination of d, l‑sugar confguration
for compounds 1, 3 and 4
The confguration of sugar moiety was determined accord-
ing to the report of Tanaka et al.and Wang et al. [21, 22].
The compounds 1, 3 and 4 were hydrolyzed by heating with
2 M HCl at 75 °C for 6 h and neutralized with 3.5 M ammo-
nium hydroxide. After the evaporation of reaction mixture,
the residue was dissolved in pyridine (2 mL) and added to
l-cysteine methyl ester hydrochloride (15 mg), and heated
at 80 °C for 1 h. Then, phenyl isothiocyanate (30 μL) was
added to the reaction mixture, and was further heated at
80 °C for 1 h. The reaction mixture was directly analyzed
under the following conditions by LC-HR/MS: Acquity
UPLC BEH C18, 1.7 μm, 2.1×100 mm (Waters) was used
as the analytical column at 40 °C. Mobile phase consisting of
0.1% formic acid (solvent A) and acetonitrile/methanol/iso-
propanol (50:25:25) containing 0.05% formic acid (solvent
B) was run at a fow rate of 0.3 ml/min under the following
gradient condition: 14%–16.5% B (0–23 min) and then 100%
B (4 min hold). These derivatizing reactions and analyzes
Peak extraction using parameters of PLS‑DA
Subsequently, we determined the peak numbers in LC-HR/
MS metabolomics via two extraction methods: (1) automatic
extraction by software (normal mode and fewer mode; con-
ditions A and B, respectively), (2) base peak-based extrac-
tion where peaks with intensities greater than 0.1% and
0.01% by the base peak, were extracted (conditions C and
D). The respective peak numbers under conditions A to D
were 9421, 7903, 230, and 1387. To estimate the optimum
peak numbers, we determined the multiple correlation coef-
fcient (R2) and predictive ability parameter (Q2) (Table S1).
Values of R2 and Q2 were the same degree in any conditions,
1 3

Journal of Natural Medicines
Table 1 Comparison of the 13
NMR ( ppm in DMSO-d6 or
chloroform-d, 150 MHz) data
C
DMSO-d6
Chloroform-d
Position
Compound 4
sec-O-glucosyl- Hydrolyzate of
Hamaudol 2)
Visamminol 2)
hamaudol1
Compound 4 (4a)
for compounds 4 and 4a to the
literature values
1
2
3
4
4a
5
6
7
–
167.8
107.8
181.9
103.4
158.7
103.4
155.5
94.4
–
–
166.8
108.4
182.6
104.4
159.8
102.9
159.0
94.8
–
166.8
108.3
182.5
104.3
159.7
102.9
159.0
94.8
–
165.8
108.9
182.6
105.5
158.3
108.7
166.5
91.7
167.9
107.9
181.9
103.5
158.7
103.4
155.5
94.3
158.8
–
8
8a
1’
158.7
–
156.3
–
156.2
–
156.7
–
2’
3’
4’
77.9
73.2
21.7
25.1
22.1
20.0
Missing
72.7
21.5
25.3
21.7
20.0
78.6
68.8
25.4
24.9
22.1
20.6
78.5
68.7
25.4
25.0
22.1
20.6
88.9
20.4
71.9
25.9
26.8
23.9
Gem-CH3
Gem-CH3
2-CH3
Glucose
1′’
2′’
3′’
4′’
5′’
6′’
100.9
73.1
76.8
70.3
75.7
67.6
100.6
73.4
76.9
70.3
76.9
61.4
Apiose
1′’’
2′’’
3′’’
4′’’
109.2
75.8
78.8
73.2
63.3
5′’’
1Chem Pharm Bull, 2001, 49(2), 154–160
2Nat Pro Sci, 2017, 23(2), 97–102
and peak numbers greatly increased when conditions A and
B were selected. Excessive peak extraction could induce
the picking of the carry-over peaks derived from previous
analyses. Therefore, condition D with a moderate number of
peaks was selected as the peak extraction method.
was confirmed between the sum of all peak areas obtained
by LC/MS data and PC1 score (Fig. S3). Therefore, PC1
likely captured the total amount of the detected variables.
We then performed PLS-DA to search for components
that could help to distinguish SR from PR. The PLS-
DA score plot had a clearer separation of the SR and
PR samples (Fig. 1b). Furthermore, the s-plot indicated
the following four characteristic peaks: m/z 453.1750
(C22H28O10; 1) for SR, and m/z 427.1757 (C24H26O7; 2),
571.2071 (C26H34O14; 4), and 585.2175 (C27H36O14; 3) for
PR (Fig. 2). These components were at the edge of the ver-
tical and horizontal axes, and were explanatory variables
that strongly contributed to the objective variable.
PCA and PLS‑DA in LC‑HR/MS metabolomics
PCA was carried out using the resultant data matrix with
the above-mentioned conditions, and the entire data were
reviewed. SR (green circles) and PR (blue circles) sam-
ples were roughly separated on the PC1 vs PC2 plane
1
(Fig. 1a); such discrimination was similar to that of H
NMR metabolomics [10]. For PC1, a positive correlation
1 3

Journal of Natural Medicines
Fig. 1 Structures of the components used to distinguish between SR and PR in LC-HR/MS metabolomics
database search using the compositional formula obtained
Identifcation of the characteristic component in SR
and the keyword "Saposhnikovia". With the dramatic
advance made using mass spectrometers, the database
constructed using exact mass was remarkably developed,
As the characteristic compound 1 for SR had C22H28O10 as
the molecular weight due to its exact mass, we attempted a
1 3

Journal of Natural Medicines
(a)
(b)
Factor 2
PC2, 16.4%
SR
PR
SR
PR
Factor 1
PC1, 32.2%
Fig. 2 Score plots of PCA (a) and PLS-DA (b) for LC-HR/MS metabolomics. Green and blue circles indicate SR and PR samples, respectively
Fig. 3 HMBC correlation of compound 4 (only correlations for the anomeric position are described for the sugar parts)
allowing a more efcient search [11]. By searching sev-
eral databases, 1 was estimated as 4′-O-β-d-glucosyl-5-O-
methylvisamminol (Fig. 3), and when we isolated 1 from
the methanol extract of SR, this compound was confrmed
by NMR and MS [12] (Fig. S4a). In addition, we also deter-
mined the d, l-confguration of the sugar moiety. The sugar
generated from the hydrolyzation of 1 was converted to the
diastereomeric derivative, and its confguration was deter-
mined by comparing the retention times of the obtained
derivative in LC with those of the authentic d- and l-glucose
derivatives [21, 22]. As the results, the retention time of
sugar derivative from 1 was identical to the d-glucose. In
the 17th edition of JP, 1 is used as an indicator component in
the test to identify SR [13].
for 1 in the SR and PR samples, and verifed that mean con-
tent of 1 in SR samples was much higher than in PR samples
(p<0.005 by t test) (Fig. S4b). We think that the signifcant
diference of 1 content between SR and PR is the reason for
the picking of 1 as characteristic compound for SR.
Identifcation of characteristic components in PR
Three characteristic compounds of PR were found using
s-plot: m/z 427.1757 (C24H26O7; 2), 571.2071 (C26H34O14;
4), and 585.2175 (C27H36O14; 3). Thus, we performed a
search for the compositional formulae and used the keyword,
"Peucedanum". As a result, 2 was estimated to be a cou-
marin derivative, and was identifed as xanthalin by MS and
NMR (Fig. 3) [9]. Furthermore, we compared the retention
time of 2 and the corresponding peak from sample solutions,
Compound 1 was detected in both SR and PR from the
TICs (Fig. S4a). We calculated the average of all peak areas
1 3

Journal of Natural Medicines
1
and observed a good alignment (Fig. S5). In supplement II of
the 17th edition of JP, PR was defned as an impurity of SR
and a purity test using xanthaline as the marker compound
by TLC was designed [14].
In the H NMR analysis of 4, the characteristic signals
were: 1) aromatic or olefnic protons at δH 6.17 (1H, s) and
δH 6.36 (1H, s), 2) hydrogen-bonded phenolic hydroxyl
group at δH 13.1 (1H, s), 3) methyl group on the sp2 carbon
at δH 2.35 (3H, s), and 4) anomeric protons at δH 4.34 (1H, d,
J=7.8 Hz) and 4.92 (1H, d, J=3.0 Hz). The aglycon moiety
of 4 was considered to be hamaudol or visamminol based on
the literature values (Table 1) [8]. In addition, the structure
of the sugar moiety was similar to that of 3 based on the 1H
and 13C NMR spectra and thus, were presumed to also be
glucose and apiose. To confrm the structure of the aglycone
moiety, we hydrolyzed compound 4 and obtained aglycone,
4a. 13C NMR data of 4a were compared to the literature
values of hamaudol and visamminol [15]; 4a coincided with
hamaudol (Table 1). To determine the confguration of the
3′ position, we measured the specifc optical rotation of 4a.
This property showed a good agreement with the literature
value [8]. Based on the above data, 4a was identifed as
hamaudol. We also confrmed reasonable correlation in the
aglycone moiety for the heteronuclear multiple bond coher-
ence (HMBC) of 4 (Fig. 4). The correlation between the 3′
position of aglycone and anomeric position of glucose (1′’),
and the 6-position of glucose (6′’) and anomeric position of
apiose (1′’’) were also confrmed (Fig. 4). We determined
that each sugar has D-confguration in the same way of 1 and
For the identifcation of 3 and 4, these compounds were
isolated by normal and RP column chromatography. In the
1H NMR of 3, the following signals were observed: (1) aro-
matic or olefnic protons at δH 5.96 (1H, s) and 6.67 (1H,
s), (2) methoxy group at δH 3.85 (3H, s), and (3) methyl
group on the sp2 carbon at δH 2.26 (3H, s). These patterns
were similar to the visamminol skeleton of 1 which was the
characteristic component of SR. Additionally, two anomeric
protons were found at δH 4.40 (1H, d, J=7.2 Hz) and 4.80
(1H, d, J=3.0 Hz). In 2D-NMR and the coupling pattern,
glucose and apiose were identifed as the sugars. Hence, we
assumed that 3 was 4′-O-β-d-apiosyl (1→6)-β-d-glucosyl-
5-O-methylvisamminol. By comparing LC, MS pattern,
and NMR data for standard 3 to those of isolated 3, this
compound was identifed as 4′-O-β-d-apiosyl (1→6)-β-d-
glucosyl-5-O-methylvisamminol (Fig. 3). Furthermore, we
also confrmed that each sugar has D-confguration in the
same way as 1. Compound 3 has already been isolated from
Libanotis laticalycina (Umbelliferae/Apiaceae) and this
plant has been previously used in the Henan and Shanxi
provinces as a substitute plant for SR [7].
Fig. 4 S-plots of PLS-DA for
LC-HR/MS metabolomics. The
characteristic variables are high-
lighted in blue, SR and red, PR;
yellow, falcarindiol
m/z 571.2071 (4)
PR
m/z 585.2175 (3)
m/z 427.1757 (2)
Falcarindiol,
m/z 284.2017, 485.3417, 727.5091
m/z 453.1750 (1)
SR
1 3

Journal of Natural Medicines
3. Thus, 4 was identifed as the novel compound, 3′-O-β-d-
apiosyl (1→6)-β-d-glucosylhamaudol (Fig. 3).
17]. We analyzed the PR sample and standard falcarindiol
with APCI and confrmed that cluster ions did not occur
(Fig. S8). As APCI is suitable for analyzing hydrophobic
compounds, it would be very efective in molecules such
as falcarindiol [18]. Compared to NMR metabolomics, the
need for optimized measurement conditions is one of the dis-
advantages of LC/MS metabolomics. Thus, combining NMR
metabolomics which is a more comprehensive measurement
method and LC/MS metabolomics with its high sensitivity
and resolution, is considered to be a very useful approach
for quality evaluation of natural products containing various
components.
Correlation between NMR and LC‑HR/MS
metabolomics with RP‑SPEC
Finally, we reviewed the mass chromatograms of 2–4 from
the SR and PR samples. Although only trace amounts of
those compounds were detected in SR, the contents in PR
were much higher than in SR (Fig. S6). Thus, 2–4 were
indeed characteristic components of PR.
In the previous report [10], we were unable to adequately
identify the variables contributing to the discrimination
between SR and PR in 1H NMR metabolomics. Herein, we
investigated the correlation between the components identi-
fed by LC-HR/MS metabolomics and the chemical shift
obtained with 1H NMR metabolomics.
Conclusions
We conducted LC-HR/MS metabolomics to identify the
components used to distinguish between SR and PR which
In the s-plot for 1H NMR metabolomics, the signals, 1.34,
2.30, 4.58, 6.02, and 6.57 ppm were confrmed at the edge
on the SR side (blue circles, Fig. S7). These chemical shifts
corresponded to those of 1 which is the characteristic com-
ponent of SR in LC-HR/MS metabolomics (Table S2, Fig.
S7). Thus, the variables obtained by the s-plot in 1H NMR
metabolomics may have been derived from this compound.
For PR, the signals, 1.50–5.90 ppm, were confrmed at
the edge of the s-plot (Fig. S7); some chemical shifts (1.50,
1.74, 1.86, 5.58 ppm) corresponded to those of 2 (red cir-
cles, Fig. S7), but most of the other shifts were identifed as
falcarindiol analogs (falcarindiol; yellow circles, Fig. S7) as
described in our previous report [10]. Signals correspond-
ing to 3 and 4 were not confrmed. We considered that this
observation may be attributed to the structural similarity
among 1, 3, and 4. Most chemical shifts derived from these
compounds were similar among each other, particularly the
aglycone moiety which is indistinguishable due to the sen-
1
were found in the H NMR metabolomics carried out in
our previous study [10]. As a result, 1 was identifed as the
characteristic component of SR, and its data correlated with
the chemical shift obtained by 1H NMR metabolomics. In
contrast, 2–4 were identifed as characteristic components
of PR, and 4 was identifed as the novel compound, 3′-O-β-
d-apiosyl (1→6)-β-d-glucosylhamaudol. Compounds 1 and
2 were found to correlate with the chemical shifts observed
1
in the H NMR metabolomics, but a correlation was not
found with the other compounds. The latter result could
be attributed to the structural similarity of the components
contained in both SR and PR, and the generation of cluster
ions from falcarindiol in the ESI interface. It is important to
note that compounds with high structural similarities such as
coumarins and chromones are found in apiaceous plants. In
this study, we succeeded in capturing the subtle component
diference between SR and PR using a series of methods.
1
sitivity and resolution of H NMR. Therefore, 1, which is
present in a high amount in SR, could be considered as the
characteristic component of SR. Furthermore, 2, which has a
diferent skeleton, was recognized in 1H NMR metabolomics
as the characteristic component of PR, supporting the specu-
lation mentioned above.
Acknowledgements This study was supported by a research grant from
the Japan Agency for Medical Research and Development (the grant
number: 17ak0101074j2001, 15ak0101017h0003).
Compliance with ethical standards
The ions, m/z 284.2017, 485.3417, and 727.5091, in the
s-plot of LC-HR/MS metabolomics may belong to falcar-
indiol (yellow circles, Fig. 2). Furthermore, m/z 243.1737
was assumed to be an ion where protons were added after
the elimination of a water molecule [M-H2O + H]+; m/z
485.3417 and 727.5091 were expected to be the cluster
ions (e.g., [2(M-H2O) + H]+, [3(M-H2O) + H]+) [16, 17].
Falcarindiol may not have been positioned at the edge of
the s-plot for LC-HR/MS metabolomics because of the dis-
persed intensity of each ion by the generation of cluster ions.
Falcarindiol is generally analyzed by APCI as the interface
of MS, and cluster ions are not generated in this method [16,
Conflict of interest The authors declare that they have no conficts of
interest.
References
1. Tai J, Cheung S (2017) Anti-proliferative and antioxidant activi-
ties of Saposhnikovia divaricata. Oncol Rep 18:227–234
2. Liao H, Li Q, Liu R, Liu J, Bi K (2014) Fingerprint analysis and
multi-ingredient determination using a single reference standard
for Saposhnikoviae Radix. Anal Sci 30:1157–1163
3. Kong X, Liu C, Zhang C, Zhao J, Wang J, Wan H, Zhu H, Zhang
P, Chen W, Xiao Y, Lin N (2013) The suppressive efects of
1 3

Journal of Natural Medicines
Saposhnikovia divaricata (Fangfeng) chromone extract on rheu-
matoid arthritis via inhibition of nuclear factor-κB and mitogen
activated proteinkinases activation on collagen-induced arthritis
model. J Ethnopharmacol 148:842–850
4’-O--D-glucosyl-5-O-methylvisamminol from Radix saposh-
nikoviae by high speed countercurrent chromatography. J Liq
Chromatogr Rel Technol 29:751–759
13. The Ministry of Health, Labor and Welfare, Japan (2016) The
Japanese pharmacopoeia seventeenth edition
4. Kim HS, Choi G, Lee AY (2018) Ultra-performance convergence
chromatography method for the determination of four chromones
and quality control of Saposhnikovia divaricata (Turcz.) Schischk.
J Sep Sci 41:1682–1690
14. The Ministry of Health, Labor and Welfare in Japan (2019) Sup-
plement II to the Japanese Pharmacopoeia seventeenth edition.
The MHLW Notifcation No. 49
5. Batsukh Z, Toume K, Javzan B, Kazuma K, Cai S, Hayashi S,
Kawahara N, Maruyama T, Komatsu K (2020) Metabolomic
profling of Saposhnikoviae Radix from Mongolia by LC–IT–
TOF–MS/MS and multivariate statistical analysis. J Nat Med
74(1):170–188
15. Kwon Y, Kim H, Kim M, Chun W (2017) Acetylcholinest-
erase Inhibitors from Angelica polymorpha Stem. Nat Pro Sci
23(2):97–102
16. Kramer M, Mühleis A, Conrad J, Leitenberger M, Beifuss U,
Carle R, Kammerer DR (2011) Quantifcation of polyacetylenes in
apiaceous plants by high-performance liquid chromatography cou-
pled with diode array detection. Z Naturforsch C 66(7–8):319–327
17. El-Houri R, Kotowska D, Christensen KB, Bhattacharya S, Oks-
bjerg N, Wolber G, Kristiansen K, Christensen LP (2015) Polya-
cetylenes from carrots (Daucus carota) improve glucose uptake
in vitro in adipocytes and myotubes. Food Funct 6:2135–2144
18. Bruins AP (1994) Atmospheric-pressure-ionization mass spec-
trometry: II. Applications in pharmacy, biochemistry and general
chemistry. Trends Anal Chem 13(2):81–90
6. Wang J, Materia LZ (1989) Medica research of Saposhunikoviae
radix (provisional translation). Zhongguo Zhong Yao Za Zhi
14:579–581
7. Baba K, Qing X, Taniguchi M, Kozawa M, Fujita E (1991) Studies
on Chiniese Medicine Fang-Feng (III) Constituents of Shui-Fang-
Feng. Shoyakugaku Zasshi 45(2):167–173
8. Okuyama E, Hasegawa T, Matsushita T, Fujimoto H, Ishiba-
shi Yamazaki M (2001) Analgesic Components of Saposh-
nikovia Root (Saposhnikovia divaricata). Chem Pharm Bull
49(2):154–160
19. Sasaki H, Taguchi H, Endo T, Yoshioka I (1982) The Constituents
of Ledebouriella seseloides WOLFF. I. Structures of Three New
Chromones. Chem Pharm Bull 30(10):3555–3562
9. Maruyama T, Ezaki M, Shiba M, Yamaji H, Yoshitomi T, Kawano
N, Cheng SZX, Yokokura T, Yamamoto Y, Fuchino H, Sun H,
Komatsu K, Kawahara N (2018) Botanical origin and chemi-
cal constituents of commercial SR and its related crude drugs
available in Shaanxi and the surrounding regions. J Nat Med
72:267–273
20. Zheleva A, Soine TO, Bubeva-Ivanova L (1972) Natural Cou-
marins V: Isolation of Xanthalin and a New Pyranocoumarin,
Peuarenine, from Peucedanum arenarium. W K J Pharm Sci
61(10):1643–1644
10. Yoshitomi T, Wakana D, Uchiyama N, Tsujimoto T, Kawano N,
Yokokura T, Yamamoto Y, Fuchino H, Hakamatsuka T, Komatsu
21. Tanaka T, Nakashima T, Ueda T, Tomii K, Kouno I (2007) Facile
discrimination of aldose enantiomers by reversed-phase HPLC.
Chem Pharm Bull 55(6):899–901
1
K, Kawahara N, Maruyama T (2020) H NMR based metabo-
22. Wang YH, Avula B, Fu X, Wang M, Khan IA (2012) Simultane-
ous determination of the absolute confguration of twelve mono-
saccharide enantiomers from natural products in a single injection
by a UPLC-UV/MS method. Planta Med 78:834–837
lomic analysis coupled with reversed-phase solid-phase extraction
for sample preparation of Saposhnikovia roots and related crude
drugs. J Nat Med 74:65–75
11. Afendi F, Okada T, Yamazaki M, Hirai-Morita A, Nakamura Y,
Nakamura K, Ikeda S, Takahashi H, Altaf-Ul-Amin M, Darusman
L, Saito K, Kanaya S (2012) KNApSAcK family databases: inte-
grated metabolite-plant species databases for multifaceted plant
research. Plant Cell Physiol 53:e1–12
Publisher’s Note Springer Nature remains neutral with regard to
jurisdictional claims in published maps and institutional afliations.
12. Sun A, Feng L, Liu R (2006) Preparative isolation
and purification of prim-O-glucosyl-cinmifugin and
1 3