Determination and Identification of a Specific Marker Compound for Discriminating Shrub Chaste Tree Fruit from Agnus Castus Fruit Based on LC/MS Metabolic Analysis

Abstract

Shrub Chaste Tree Fruit (SCTF) is defined as the fruits of Vitex rotundifolia L. f. and V. trifolia L. and has been used as a component of some traditional Japanese medicines (Kampo formulations). Agnus Castus Fruit (ACF) is defined as the dried ripe fruits of V. agnus-castus L.; it is used in traditional European medicines, but is becoming popular in Japan as both an over-the-counter drug and as an ingredient in health foods for treating premenstrual syndrome (PMS). To ensure the efficacy and safety of both SCTF and ACF products, it is important to precisely authenticate their botanical origins and to clearly distinguish between SCTF and ACF. Therefore, we tried to identify SCTF-specific marker compounds based on LC/MS metabolic analysis. The multivariate analysis of LC/MS data from SCTF and ACF samples furnished candidate marker compounds of SCTF. An SCTF-specific marker was isolated from SCTF crude drugs and identified as 3-O-trans-feruloyl tormentic acid on the basis of spectroscopic data from NMR and MS. Since avoiding contamination from closely related species is a significant requirement for pharmaceuticals of natural origin, this information will be valuable for the quality control of both SCTF and ACF products from the viewpoint of regulatory science.

Dried fruits of some Vitex species (Verbenaceae) have been used as crude drugs for different purposes in East Asian and European traditional medicines. Shrub Chaste Tree Fruit (SCTF), called Mankeishi （蔓荊子） in Japanese, is defined as the fruits of V. rotundifolia L. f. and V. trifolia L. in the Japanese Standards for Non-Pharmacopoeial Crude Drugs 2015¹⁾ (Non-JPS 2015). Because of its analgesic, sedative, and anti-inflammatory effects, SCTF is a component of some traditional Japanese medicines (Kampo formulations), and ca. 0.39 tons of SCTF have been annually sold in the Japanese crude drug market.²⁾ On the other hand, the European Pharmacopoeia categorizes the dried ripe fruits of V. agnus-castus L. as Agnus Castus Fruit (ACF).³⁾ In Europe, ACF products are distributed as over-the-counter (OTC) drugs to treat gynecological disorders such as premenstrual syndrome (PMS), and in Japan one of them has recently become available as an OTC drug alongside health foods.^4–6)

SCTF and ACF products originating from Vitex fruits have been used for different purposes and used as traditional Japanese and European medications, respectively. To ensure the efficacy and safety of these products, the proper use of the correct original plant for each SCTF and ACF is essential. However, due to the similar macroscopic appearances of Vitex fruits, a high degree of expertise may be required to distinguish SCTF and ACF by their morphological characteristics; moreover, SCTF and ACF products are in powdered form, it is difficult to identify the original species.⁷⁾ Considering the international trade of Vitex fruits, this situation could lead to the misidentification of the botanical origins of these crude drugs, resulting in unexpected adverse reactions as well as decreased pharmaceutical efficacy. Therefore, an appropriate discrimination method other than morphological examination is necessary for crude drugs derived from Vitex fruits.

In this study, we analyzed the extracts of commercially available SCTF and ACF products by LC/MS and applied the data to multivariate analysis to find marker compounds with which to discriminate SCTF from ACF. We then isolated and identified an SCTF-specific triterpene derivative, which had never before been isolated from a Vitex plant.

Experimental

Materials

SCTF of non-JP crude drug standards grade and ACF health food products were purchased from their manufacturers in Japan and from online stores, respectively. ACF crude drugs of EP8.0 grade were purchased from pharmacies in Europe (Table 1). The original plant species of SCTF and ACF crude drugs were previously identified based on their DNA sequences.⁸⁾

Table 1. SCTF and ACF Samples Used in This Study

Sample ID	Category	Acquisition year	Form	Contents per capsule
SCTF-1	Crude drug	2008	Dried fruits
SCTF-2	Crude drug	2008	Dried fruits
SCTF-3	Crude drug	2012	Dried fruits
SCTF-4	Crude drug	2012	Dried fruits
ACF-1	European OTC drug	2008	Dried fruits
ACF-2	European OTC drug	2013	Dried fruits
ACF-3	Health food	2012	Capsule	Dry extracts 225 mg
ACF-4	Health food	2012	Capsule	Dry extracts 200 mg+powdered fruits 300 mg

General Procedures

For column chromatography, Silica gel N60 (Kanto Chemical, Tokyo, Japan), YMC GEL ODS-A S-150 (YMC, Kyoto, Japan), and MCI GEL CHP-20 (Mitsubishi Chemical, Tokyo, Japan) were used as chromatographic carriers. The preparative HPLC system was composed of a JASCO PU-2089 Plus, JASCO MD-2010 Plus, and JASCO CO-2060 Plus equipped with a COSMOSIL 5C₁₈-MS-II (20 mm i.d.×250 mm; Nacalai Tesque, Kyoto, Japan), COSMOSIL πNAP (10 mm i.d.×250 mm, Nacalai Tesque), or Shodex GF310HQ (4.6 mm i.d.×250 mm; Showa Denko, Kawasaki, Japan). LC/MS analysis for isolation study was performed on an LCMS-2010EV (Shimadzu, Kyoto, Japan) equipped with a COSMOSIL 5C₁₈-MS-II (2.0 mm i.d.×150 mm; Nacalai Tesque), and was eluted at a flow rate of 0.2 mL/min using a gradient ranging from 50 to 100% mobile phase B in a time span of 20 min (mobile phase A, 0.1% CH₃COOH; mobile phase B, CH₃CN). The injection volume was 10 µL. The eluent was scanned by an electrospray ionization (ESI⁻) interface in the negative mode. ¹H- and ¹³C-NMR spectra were recorded on an ECA-800 spectrometer (JEOL, Tokyo, Japan), and chemical shifts are expressed in δ (ppm) with tetramethylsilane (TMS).

LC/MS Analysis

Each powdered sample (200 mg) was extracted with CHCl₃ (10 mL) by sonication for 3 h at room temperature, and the mixture was filtered through filter paper No. 2 (Advantec, Ehime, Japan). An aliquot (1 mL) of the filtrate was evaporated on the centrifugal evaporator and redissolved in 1 mL MeOH. The resulting 20-fold-diluted solution was filtrated through a 0.45 µm Ultrafree-MC centrifugal filter unit (Millipore, Billerica, MA, U.S.A.) and an aliquot (5 µL) of each filtrate was injected onto ultra performance LC-photo diode array-quadrupole-time-of-flight (UPLC-PDA-QTOF)/MS for analysis. LC/MS analyses were performed on a Waters Acquity UPLC I-class FL coupled to a Waters Synapt G2-Si (Waters, Milford, MA, U.S.A.) equipped with an ESI source. An ACQUITY UPLC HSS C₁₈ column (2.1 mm i.d.×100 mm, 1.8 µm, Waters) was used for the chromatography at a flow rate of 0.4 mL/min and a column temperature of 40°C. The mobile phase was composed of A (water) and B (MeOH) with a gradient elution: 0–2 min, 50% B; 2–3 min, 50–80% B; 3–7 min, 80–98% B; 7–8 min, 98% B. The QTOF mass spectrometer was operated in ESI⁻ resolution mode (>20000 FWHM) with a capillary voltage of 0.5 kV, and the cone voltage was set to 20 V. The desolvation and cone gas flow were set to 800 and 50 L/h, respectively. The source temperature was set to 120°C and the desolvation temperature was set to 350°C. The full-scan mass spectra were collected in continuum mode from m/z 100 to 1200, in which the data acquisition rate was set to 0.2 s, with a 0.1 s interscan delay. All MS data were acquired using the reference lock mass of leucine–enkephalin ([M−H]⁻=554.2615) via a lock spray interface to ensure accuracy and reproducibility. An acquisition with alternating low energy (collision cell energy of 4 V) and elevated energy (collision cell energy ramped from 15 to 40 eV) in the MS^E approach was created to obtain the simultaneous acquisition of an exact mass precursor and fragment ion data. All the analyses and acquisitions were performed using MassLynx 4.1 software (Waters).

Multivariate Data Analysis

The LC/MS raw data were processed by MarkerLynx XS software (Waters) using the following parameters: initial retention time 1 min, final retention time 8 min, mass tolerance 0.02 Da, intensity threshold 1000 counts, mass window 0.02 Da, retention time window 0.1 min, noise elimination level 6. The resulting multivariate data matrix was normalized with Pareto scaling and analyzed by orthogonal partial least-squares discriminant analysis (OPLS-DA) using EZinfo software (Umetrics, Umea, Sweden). The preliminary molecular formula of marker candidates was calculated from the high resolution (HR) masses using carbon, hydrogen, nitrogen, and oxygen with mass tolerance <10 ppm and ring double bond (RDB) equivalent <50. The definitive elemental compositions were determined by isotope patterns and fragment ions in MS/MS spectra (Table 2).

Table 2. Candidate Marker Compounds for SCTF Obtained from the OPLS-DA Analysis

Peak ID	RT (min)	HR-mass (m/z)	Molecular formula	Tolerance (ppm)	Intensity trend*
					SCTF	ACF
5278	3.90	607.2149	C33H35O8	2.85	0.584	0.1470
2837	4.03	492.0464	C10H18NO19	−0.35	0.714	0.0857
3173	4.03	506.0280	C13H16NO20	1.98	0.815	0.0637
5816	4.59	633.3807	C39H53O7 (a)	2.12	4.290	0.0000
6436	4.59	663.3920	C40H55O8 (b)	2.85	0.582	0.0000
4420	5.66	565.1344	C29H25O12	0.35	1.420	0.5140
3680	5.87	316.9475	C2H5O18	0.46	4.860	0.1970
2928	5.89	494.9106	C5H5NO26	1.17	0.776	0.2130
6136	5.89	394.9007	—		0.940	0.0648
1326	6.17	721.1230	C31H29O20	−1.67	0.610	0.0732
2969	6.20	297.0446	C9H13O11 (c)	−0.64	2.200	0.0000
545	6.22	685.0326	C29H17O20 (d)	1.83	0.497	0.0000
5772	6.22	631.1136	C25H27O19 (e)	−0.50	0.494	0.0000
5228	6.33	367.1852	C15H29NO9	1.52	4.840	0.5120
3662	6.60	849.0623	C16H33O39	2.95	0.646	0.0302
2131	6.66	759.0525	C28H23O25	0.21	0.795	0.0280
2279	6.66	769.0828	C41H21O16 (f)	3.90	0.398	0.0000
942	6.69	705.1356	C24H33O24	−0.03	1.000	0.4430
4603	6.94	347.1838	C20H27O5	−1.50	0.366	0.0435
6472	6.94	403.2064	C30H27O	0.76	1.860	0.4320
5415	6.96	371.2187	C23H31O4	−2.99	1.820	0.6660
3796	7.40	535.5208	C39H67	−2.93	0.680	0.1610
4567	7.41	571.4940	C34H67O6	0.78	7.050	1.0200
5051	7.41	598.5143	C36H70O6	−2.39	0.637	0.1080
4752	7.42	581.5289	C40H69O2	−0.31	13.000	3.2500
6301	7.66	1036.1670	C39H42NO32	−1.44	0.621	0.2270
502	7.72	1093.2120	C43H49O33	−3.06	0.600	0.0326
4201	7.87	885.5516	C54H77O10 (g)	0.48	0.412	0.0000

(a)–(g): Most characteristic ions for SCTF. —: Molecular formula was not predicted. *Intensity trend shows the average of integrated intensities from SCTF and ACF samples.

Extraction and Isolation of Compound 2

The SCTF-2 (150 g) was extracted with CHCl₃ (7.5 L). The solvent was evaporated under reduced pressure to obtain CHCl₃ extract (27.28 g). This extract was fractionated by silica gel column chromatography (C. C.) with hexane–EtOAc (100 : 0→0 : 100, v/v) into 13 fractions. Three fractions (Frs. 8 to 10; total 1508.4 mg) containing the target compound were combined and subjected to octadecyl silica (ODS) C. C. with CH₃OH–H₂O (50 : 50→100 : 0) to yield 12 fractions. Fractionation of the eighth fraction (308.9 mg) by CHP-20 C. C. with CH₃OH–H₂O (50 : 50→100 : 0, v/v) gave 4 fractions. The repeated prep-HPLC of the third fraction (173.0 mg) using a COSMOSIL 5C₁₈-MS-II with CH₃OH–0.1% trifluoroacetic acid (TFA) (88 : 12, v/v), a COSMOSIL πNAP with CH₃OH–0.1% TFA (85 : 15, v/v), and Shodex GF310HQ with CH₃OH–H₂O (85 : 15, v/v) afforded 2 (1.1 mg).

3-O-trans-Feruloyl Tormentic Acid (2)

Amorphous colorless powder. ESI-MS (negative mode) m/z 663 [M−H]⁻. HR-ESI-MS (negative mode) m/z 663.3919 (Calcd for C₄₀H₅₅O₈, 663.3891). ¹H-NMR spectral data, see Table 3 and ¹H-NMR data including undescribed signals are consistent with literature’s data.^{9) 13}C-NMR spectral data, see Table 4.

Table 3. Comparison of ¹H-NMR Data between Compound 2 and Ref. in Methanol-d₄ (800 MHz, δ in ppm, J in Hz)

Position	2	Ref.a)
	δ	δ
2	3.84 (1H, ddd, J=4.0, 10.4, 11.2)	3.85 (1H, ddd, J=4.4, 9.9, 11.0)
3	4.64 (1H, d, J=10.4)	4.64 (1H, d, J=9.9)
12	5.29 (1H, t, J=3.0)	5.30 (1H, br s)
18	2.51 (1H, s)	2.51 (1H, br s)
23	0.9 (3H, s)	0.90 (3H, s)
24	0.95 (3H, s)	0.96 (3H, s)
25	1.05 (3H, s)	1.06 (3H, s)
26	0.82 (3H, s)	0.81 (3H, s)
27	1.36 (3H, s)	1.97b) (3H, s)
29	1.20 (3H, s)	1.20 (3H, s)
30	0.93 (3H, d, J=6.4)	0.93 (3H, d, J=6.6)
2′	6.43 (1H, d, J=16.0)	6.43 (1H, d, J=16)
3′	7.62 (1H, d, J=16.0)	7.63 (1H, d, J=16)
2″	7.20 (1H, d, J=1.6)	7.21 (1H, br s)
5″	6.81 (1H, d, J=8.0)	6.81 (1H, d, J=7.7)
6″	7.07 (1H, dd, J=1.6, 8.0)	7.08 (1H, br d, J=7.7)
OCH3	3.89 (3H, s)	3.90 (3H, s)

a) Zhao Q.-C., Cui C.-B., Cai B., Yao X.-S., Osada H., Molbank, M328 (2003). b) May be mistyped.

Table 4. ¹³C-NMR Data of Compound 2 in Methanol-d₄ (200 MHz, δ in ppm)

Position	2		Ref.a)	Position	2		Ref.a)
	δ	DEPT	δ		δ	DEPT	δ
1	48.2	CH2	48.55	21	27.5	CH2	27.30
2	67.8	CH	67.65	22	39.3	CH2	39.22
3	85.8	CH	85.64	23	29.4	CH3	29.25
4	40.8	C	40.65	24	18.5	CH3	18.32
5	56.6	CH	56.46	25	17.3c)	CH3	16.60
6	19.8	CH2	19.58	26	17.7c)	CH3	17.07
7	34.2	CH2	34.03	27	25.0	CH3	24.83
8	42.9	C	42.68	28	182.9	C	182.29
9	48.8	CH	8.71b)	29	27.2	CH3	27.05
10	39.4	C	39.00	30	16.8c)	CH3	17.50
11	24.9	CH2	24.75	1′	169.8	C	169.61
12	129.2	CH	129.14	2′	116.3	CH	116.16
13	140.5	C	140.16	3′	146.7	CH	146.50
14	41.3	C	41.14	1″	128.1	C	127.90
15	29.8	CH2	29.59	2″	111.8	CH	111.58
16	26.8	CH2	26.60	3″	149.6	C	149.40
17	49.3	C	48.72	4″	150.7	C	150.53
18	55.3	CH	55.09	5″	116.7	CH	116.49
19	73.8	C	73.57	6″	124.3	CH	124.04
20	43.3	CH	43.09	OCH3	56.7	CH3	56.51

a) Zhao Q.-C., Cui C.-B., Cai B., Yao X.-S., Osada H., Molbank, M328 (2003). b) May be mistyped. c) Assigned from HMBC correlations.

Results and Discussion

Exploring Specific Marker Compounds for SCTF

Metabolomics has been adopted for the classification of variables and the identification of marker compounds based on the characteristics of their metabolic profiles. Liquid chromatography coupled with tandem mass spectrometry (LC-MS/MS) is often applied to plant metabolomics, because it can detect broad metabolites with high sensitivity and provide useful information for identification. Chemometric analyses such as principal component analysis (PCA), partial least-squares discriminant analysis (PLS-DA), and orthogonal partial least-squares discriminant analysis (OPLS-DA) are frequently used as multivariate tools for discrimination analysis. Recently, phytochemical studies have been reported on the classification of plant samples and determining their marker compounds with differences involving geographical factors, species, and processing treatments.^10–14)

From the LC/MS data set of four SCTF and four ACF samples, a total 6508 peaks were picked up using MarkerLynx software and subjected to OPLS-DA to identify compounds that could be used to discriminate SCTF from ACF. The OPLS-DA model explained 95.3% of the total variance (R²X) with a 94.3% prediction goodness parameter (Q²). The resulting score plots showed obvious separation between the SCTF and ACF clusters by Factor 1 on the X-axis (Fig. 1A). The S-plots highlighted the most influential variables with exact mass retention time (EMRT) pairs that contribute to the segregation of SCTF and ACF samples, and 28 ions in the upper-right area were found as characteristic marker candidates by the arbitrary cutoff values of p>0.01 and p (corr)>0.8 (Figs. 1B, C). According to the preliminary molecular formula of each candidate and the ion intensity trends, the specific variables were restricted to 7 ions (a–g), the details of which we investigated (Table 2). The signal-to-noise ratio (S/N) of ion d was too small to obtain MS/MS information. The molecular formula of ions c and e were revised as C₁₂H₂₁OSi₃S and C₃₂H₂₄N₈O₂Br from the isotopic patterns, and their corresponding compounds did not represent natural products. The preliminary formulae of ions a, b, f, and g were confirmed by their MS/MS spectra, but no corresponding compounds of ions f and g were found in the natural products database. Ions a and b were finally predicted to correlate to compounds 1 (C₃₉H₅₄O₇) and 2 (C₄₀H₅₆O₈). Then, visual confirmation of these peaks on extracted mass chromatograms showed that ACF1 and ACF3 contained compound 1 (4.59 min, m/z=633.40, data not shown), whereas compound 2 is specific to SCTF. The usefulness of compound 2 as a specific marker for distinguishing ACF and SCTF was verified when additional products were analyzed on LC/MS by focusing compound 2 (Table S1, Fig. S1). Therefore, we targeted compound 2 to isolate and identify the chemical structure as the primary marker compound of SCTF.

Fig. 1. OPLS-DA/O2PLS Score Plot (A) and S-Plot (B, C) Based on LC/MS Data of SCTF and ACF Products

Black circles indicate ACF; open squares indicate SCTF. Figure 1C is an enlargement of the upper-right quadrant of the plot in Fig. 1B. EMRTs of ions a to g are shown in Table 2.

Identification of Marker Compound

To isolate the marker compound, SCTF was extracted with CHCl₃. The extract was fractionated using normal and reversed-phase column chromatography and prep-HPLC, monitoring the marker compound on LC/MS as a guide to obtain 2.

Compound 2 was isolated as an amorphous colorless powder, and its molecular formula was determined as C₄₀H₅₆O₈ by HR-ESI-MS, with 13 degree of unsaturation. The ¹H-NMR chemical shifts of 2 (Table 3) indicated the presence of a 1,2,4-trisubstituted aromatic ring at δ_H 6.81 (1H, d, J=8.0 Hz, H-5″), 7.07 (1H, dd, J=1.6, 8.0 Hz, H-6″), and 7.20 (1H, d, J=1.6 Hz, H-2″); a trans-olefin at δ_H 6.43 (1H, J=16.0 Hz, H-2′), 7.62 (1H, J=16.0 Hz, H-3′); a single olefin at δ_H 5.29 (1H, t, J=3.0 Hz, H-12); a methoxy at δ_H 3.89 (3H, s, OCH₃-3″); and seven methyls at δ_H 0.82 (3H, s, H-26), 0.90 (3H, s, H-23), 0.93 (3H, d, J=6.4 Hz, H-30), 0.95 (3H, s, H-24), 1.05 (3H, s, H-25), 1.20 (3H, s, H-29), 1.36 (3H, s, H-27).

Analysis of the ¹³C-NMR (Table 4) and distortionless enhancement by polarization transfer (DEPT) spectra revealed 8 methyls, one of which is methoxy; 8 methylenes; 12 methines; and 12 quaternary carbons, 2 of which are carbonyl carbons. All protonated carbons were assigned by heteronuclear multiple quantum coherence (HMQC).

In addition to the above-mentioned data, heteronuclear multiple bond connectivity (HMBC) analysis indicated that 2 was 3-O-trans-feruloyl tormentic acid. The comparison of physical and spectral data with corresponding values in the literature⁹⁾ supported this structure. However, the assignment of ¹³C-NMR data at C_25,26,30 in the literature was revised on the basis of HMBC correlations. Namely, HMBC analysis of 2 showed long-range correlations between H-25 and δ_C 39.4 (C-10), 48.2 (C-1), 48.8 (C-9), and 56.6 (C-5); H-26 and δ_C 34.2 (C-7), 41.3 (C-14), 42.9 (C-8), and C-9; H-30 and δ_C 27.5 (C-21), and 73.8 (C-19), respectively. Thus, we identified 2 as 3-O-trans-feruloyl tormentic acid. Compound 2 was isolated from a Vitex plant for the first time.

Fig. 2. Chemical Structure of Compound 2

Conclusion

In the present study, we analyzed the extracts of four SCTF crude drugs, two ACF crude drugs, and two ACF health foods by LC/MS. Subsequent multivariate analyses were performed to find marker compounds that distinguish between SCTF and ACF. An SCTF-specific marker was isolated using repeated column chromatography followed by prep-HPLC and was identified as 3-O-trans-feruloyl tormentic acid on the basis of spectroscopic data from NMR and MS. This compound was isolated from a Vitex plant for the first time. It will be valuable for establishing a method for the quality control of SCTF.

Acknowledgments

This work was supported by a Grant for Research on Health Sciences Focusing on Drug Innovation and by a Grant for Research on Development of New Drugs from the Japan Health Sciences Foundation.

Conflict of Interest

The authors declare no conflict of interest.

Supplementary Materials

The online version of this article contains supplementary materials.

References

• 1) “The Japanese Standards for Non-Pharmacopoeial Crude Drugs (Non-JP Crude Drug Standards),” Ministry of Health, Labour, and Welfare, Japan, 2015.
• 2) “The 3rd Report on Consumption of Raw Materials for Crude Drugs in 2011 and 2012 Fiscal Years,” Crude Drugs Committee of Japan Kampo Medicines Manufacturers Association, Tokyo, 2015, p. 8.
• 3) The European Directorate for the Quality of Medicines, “The European Pharmacopoeia,” 8th edition 8.0, Strasbourg, Council of Europe, 2014, pp. 1137–1138.
• 4) European Medicines Agency, “Community herbal monograph on Vitex agnus-castus L., fructus.”: ‹http://www.ema.europa.eu/docs/en_GB/document_library/Herbal_-_Community_herbal_monograph/2011/08/WC500110099.pdf›, cited Aug. 10, 2015.
• 5) Fukahori M., Kobayashi S., Naraki Y., Sasaki T., Oka H., Seki M., Masada-Atsumi S., Hakamatsuka T., Goda Y., Chem. Pharm. Bull., 62, 379–385 (2014).
• 6) Momoeda M., Sasaki H., Tagashira E., Ogishima M., Takano Y., Ochiai K., Adv. Ther., 31, 362–373 (2014).
• 7) Sakai E., Iida O., Kawahara N., Murata J., Sasaki T., Shoyakugaku Zasshi, 69, 48–52 (2015).
• 8) Oshima N., Masada S., Suzuki R., Yagi K., Matsufuji H., Suenaga E., Takahashi Y., Yahagi T., Watanabe M., Yahara S., Iida O., Kawahara N., Maruyama T., Goda Y., Hakamatsuka T., Planta Med., 82, 147–153 (2016).
• 9) Zhao Q.-C., Cui C.-B., Cai B., Yao X.-S., Osada H., Molbank, M328 (2003).
• 10) Putri S. P., Nakayama Y., Matsuda F., Uchikata T., Kobayashi S., Matsubara A., Fukusaki E., J. Biosci. Bioeng., 115, 579–589 (2013).
• 11) Gad H. A., El-Ahmady S. H., Abou-Shoer M. I., Al-Azizi M. M., Phytochem. Anal., 24, 1–24 (2013).
• 12) Zhu J., Fan X., Cheng Y., Agarwal R., Moore C. M., Chen S. T., Tong W., PLoS ONE, 9, e87462 (2014).
• 13) Duan L. X., Qi X., Chen M., Huang L. Q., “Plant Metabolomics,” Chap. 10, Springer, Netherlands, 2015, pp. 227–244.
• 14) Wakana D., Maruyama T., Zaima K., Takeda H., Sugimura K., Anjiki N., Iida O., Kawahara N., Goda Y., Hosoe T., Jpn. J. Food Chem. Safety, 21, 135–138 (2014).