Comparative Study of the Chemical Constituents and Bioactivities of the Extracts from Fruits, Leaves and Root Barks of Lycium barbarum

The fruits, leaves and root barks of L. barbarum plant are widely used as functional foods and as ingredients in traditional Chinese prescriptions and patent medicines. They are considered to have different pharmacological activities and health benefits because of their diverse constituents. Here, the chemical constituents of the extracts from fruits, leaves and root barks of L. barbarum were compared by ultra-high performance liquid chromatography coupled with high resolution mass spectrometry (UPLC-HR-MS). A total of 131 compounds were identified and seven of them were quantified. Among them, 98, 28 and 35 constituents were detected in fruits, leaves and root barks respectively. Dicaffeoylspermidine/spermine derivatives were the most detected compounds (74/131); among them, dicaffeoylspermine isomers and propionyl-dicaffeoylspermidine were found in root barks in very large amounts (e.g., kukoamine B = 10.90 mg/g dry powder); dicaffeoyl-spermidine isomers were detected in fruits/leaves in a high amount, and many of their glycosylated derivatives were mainly detected in fruits. In addition, six saponins from L. barbarum fruits were reported for the first time, and 5,6-dihydrosolasonine was reported for the first time in plants. The activity assays showed that the root bark extract possessed the strongest antioxidative activity and cytotoxicity, which was presumed due to the large amount of dicaffeoylspermine/spermidines in root barks. Fourteen potential bioactive components from fruits were identified by a target cell-based screening method. These results will help to understand the different biological activities of these three parts of L. barbarum plant and will benefit the discovery of new functional components.


Introduction
Lycium barbarum L. (L. barbarum), known as "goji" or Chinese wolfberry, belongs to the family of Solanaceae and is widely cultivated in China. Three parts of L. barbarum plant including fruits, leaves and root barks, have been used as functional foods and traditional Chinese medicinal herbs in China for centuries [1][2][3] and nowadays are being widely consumed all over the world. The fruits (goji berries, Chinese name: gouqizi) are reported to have multiple effects, such as anti-aging, neuroprotection, anti-fatigue, hypoglycemic, antiproliferative activity and cytoprotection, immunomodulation and antioxidant properties [3][4][5] and are being most widely used in foods and traditional medicines. The This result suggests that the new compound has a similar structure to solasonine except for two more hydrogens on the aglycone. Based on the fragmentation ions of both compounds and the This result suggests that the new compound has a similar structure to solasonine except for two more hydrogens on the aglycone. Based on the fragmentation ions of both compounds and the reported fragmentation pathway of solasonine [19], the new compound could be 5,6-dihydrosolasonine, which differs from solasonine by the C-C single bond at position 5,6 in the B-ring of the steroidal skeleton. The proposed aglycone of 5,6-dihydrosolasonine (named as soladulcidine) has been reported in other two glycoalkaloids (soladulcine A/B) isolated from Solanum dulcamara, which consist of chacotriose/lycotetraose and soladulcidine joined through a β-glycosidic bond [20]. The proposed fragmentation pathway for 5,6-dihydrosolasonine is shown in Figure 1. Furthermore, 5,6-dihydrosolasonine (20 mg, white powder, UV λ max : 225 nm) was isolated from dried fruit of L. barbarum (5 kg) and characterized by MS and 13 C-NMR. The NMR data (Table S1) showed that the peaks at 140.6 ppm and 121.4 ppm in the 13 C-NMR spectrum of solasonine disappeared in the 13 C-NMR spectrum of 5,6-dihydrosolasonine, while two new peaks appeared at 43.06 ppm and 28.77 ppm, which correspond to the change of the double bound to a C-C single bond at position 5,6 in the B-ring of the steroidal skeleton. In addition, soladulcidine (22R, 25R) has a stereoisomer, tomatidine (22S, 25S), the aglycone of α-tomatine found in the stems and leaves of tomato plants. The 13 C-NMR peak at 33.73 ppm (C23) and 45.87 ppm (C26) further confirmed that the aglycone of 5,6-dihydrosolasonine is soladulcidine [21].
Glycoalkaloids are nitrogen-containing steroidal glycosides, generally found in plants of the Solanaceae, such as tomato, potato, and aubergine [22]. Solasonine and solamargine are two major steroidal glycoalkaloids, which have been found in 200 Solanum species [23][24][25]. They are water soluble triglycosides with the same aglycone (solasodine) and different trioses (solatriose and chacotriose) [25]. Solasodine is one of the main aglycone of glycoalkaloids, and has been used as raw material for steroidal drugs. Although L. barbarum belongs to the Solanaceae family, solasonine has not been previously reported in L. barbarum. A ring E-opened dihydro-derivative of solasonine has been reported by Weissenberg et al. [26,27] which has the same molecular weight as 5,6-dihydrosolasonine, but should have different fragmentation ions from those shown in Figure 1. As far as we know, 5,6-dihydrosolasonine has not been previously identified in any plant. The isolated 5,6-dihydrosolasonine was used as a standard substance for the following quantitative study.

Multi-Component Analysis of Extracts by UPLC-HR-MS
Under the optimized UPLC-HR-MS experimental conditions, the accurate mass and composition for the precursor ions and product ions from the extracts of fruits, leaves and root barks were analyzed respectively using Xcalibur TM 3.0 (Thermo Fisher) software in both positive and negative ionization modes. Internal calibration by infusion of a calibrant achieved a typical mass accuracy within 10 ppm. The identification of the compounds in extracts was performed based on the retention time, high resolution MS/MS data, isotope abundance, fragment product ions, literature data, databases (Reaxys, PubMed, Mass Bank, Chemspider, etc.) and standard substances. The fragmentation patterns ( Figure 1 and Figures S2-S6) of seven standards were proposed based on their high resolution MS/MS spectra, which were further used to assist the identification of constituents in extracts. Finally, a total of 131 compounds were detected based on our analytical strategy ( Figure 2). Based on their chemical structures, the detected compounds were classified into six groups, including phenylpropanoids, dicaffeoylspermidine/spermine derivatives, phenolic amides, flavones, saponins and others ( Figure 2). The detailed information of the 131 compounds found in fruits, leaves and root barks of L. barbarum are presented in Table 1.  Figure S4).
Dicaffeoylspermidine and dicaffeoylspermine derivatives are conjugates of caffeoyl groups and spermidine or spermine via amide bonds, which mainly contain characteristic fragments at m/z 310/308, 293/291, 222/220 or 165/163. Compound B9 displaying a [M + H] + ion at m/z 531.3220 (C28H43O6N4, Cal. 531.3177, mass error 1.41 ppm) was confirmed to be the dicaffeoylspermine derivative kukoamine B by a standard ( Figure S5). The fragment ion at m/z 367.2736 was formed by neutral loss of one caffeoyl unit, and the fragment ion at m/z 165.0559 was formed by further neutral loss of the spermine unit. In the same manner, compounds B1, B5, B7 and B10 were identified to be isomers of kukoamine B.
Compounds    Figure S5). The fragment ion at m/z 367.2736 was formed by neutral loss of one caffeoyl unit, and the fragment ion at m/z 165.0559 was formed by further neutral loss of the spermine unit. In the same manner, compounds B1, B5, B7 and B10 were identified to be isomers of kukoamine B.
Compounds     ) and fragment ions at m/z 310, 220 and 163 were identified to be five positional isomers of diglycosyl-caffeoyl spermidine. Their common fragment ion at m/z 472 is formed by the cleavage of one glucosyl (162 Da, Figure S11). Zhou et al. reported 15 dicaffeoylspermidine derivatives in the fruit of L. barbarum [32]. Here, we further found more dicaffeoylspermidine and dicaffeoylspermine derivatives; for example, compounds B46, B50 and B54 found in fruit of L.  (Figure 1 and Figure S2). Compounds E2 and E4 were identified to be solasonine and 5,6-dihydrosolasonine by the standards and fragment ions as described above.  Figure S16). It is worth mentioning that four of the six identified saponins, E1 (gracillin), E2 (solasonine), E4 (5,6-dihydrosolasonine) and E6 (lycioside B) have the same sugar moiety, solatriose, but other steroidal glycoalkaloids from Solanum plants consist of the same aglycones and different sugar moieties, such as solamargine (chacotriose + solasodine) and soladulcine A/B (chacotriose/lycotetraose + soladulcidine), which were not detected in the three parts of L. barbarum, suggesting that solatriose might be a characteristic sugar unit in L. barbarum. Saponins are rarely reported in L. barbarum [1,47]. All the saponins (E1-E6) reported here were found in L. barbarum for the first time.
Eleven other compounds were also detected, but their diagnostic ions were not included in the characteristic ion database. Consequently, these compounds were not classified.
In summary, a total of 131 compounds including 13 phenylpropanoids, 74 dicaffeoylspermidine or dicaffeoylspermine derivatives, 21 phenolic amides, six flavonoids, six saponins and 11 others were detected. Among them, 98 compounds were found in fruits, 28 compounds were found in leaves and 35 compounds were found in root barks of L. barbarum, suggesting that the fruits extract contains many more components than root barks and leaves ones. As shown in the total ion chromatograms of UPLC-HR-MS (Figure 3), the distribution of these constituents in the extracts of fruits, root barks and leaves are obviously different. A lot of dicaffeoylspermidine or dicaffeoylspermine derivatives were detected in fruits, leaves and root barks. Among them, kukoamine A (B7) and kukoamine B (B9) (dicaffeoylspermine isomers, main peaks around 20 min in positive/negative modes) and propionyl-dicaffeoylspermidine (B69, main peak around 30 min in negative mode) were found in the extract of root barks in a very large amount, but not detected in leaves and fruits. Dicaffeoylspermidine isomers (B52, B63 and B70) were detected in a high amount in fruits and leaves, but were much lower in root barks. However, some glycosylated derivatives of kukoamine A/B isomers (such as B2, B3, B4, B6, B8, B66 and B72) and many glycosylated derivatives of dicaffeoylspermidine (such as B13-20, B24-42, B30, B38, B44, B46-51, B53-62, B64 and B67) were mainly found in fruits, suggesting that dicaffeoylspermidines and dicaffeoylspermines were glycosylated in the fruit of L. barbarum. The difference of the distribution of dicaffeoylspermine and dicaffeoylspermidine derivatives suggests low chemical similarity between root barks and fruits/leaves. Based on Table 1, phenylpropanoids were mainly found in the fruits and leaves, but not detected in the root barks. Phenolic amides were found in all the three parts of L. barbarum plants. Flavonoids were mainly found in the fruits and leaves in a high amount, e.g., rutin (D5). Saponins were only detected in the fruits of L. barbarum.

Molecules 2019, 24, x FOR PEER REVIEW 2 of 23
Eleven other compounds were also detected, but their diagnostic ions were not included in the characteristic ion database. Consequently, these compounds were not classified. In summary, a total of 131 compounds including 13 phenylpropanoids, 74 dicaffeoylspermidine or dicaffeoylspermine derivatives, 21 phenolic amides, six flavonoids, six saponins and 11 others were detected. Among them, 98 compounds were found in fruits, 28 compounds were found in leaves and 35 compounds were found in root barks of L. barbarum, suggesting that the fruits extract contains many more components than root barks and leaves ones. As shown in the total ion chromatograms  Table 1.

Quantitative Analysis of Seven Compounds in the Fruits, Leaves and Root Barks
The quantitative analysis of seven compounds in the extracts was performed by UPLC-Qtrap-MS with MRM mode owing to its high sensitivity, specificity, and selectivity in the quantitation of trace compounds in complex matrices [18]. The method validation data are listed in Table S2. The calibration curves of the seven standards showed good linearity (r from 0.989 to 0.998). The limit of detection (LOD) and the limit of quantification (LOQ) for each standard were 0.08-0.26 µg/mL and 0.02-0.07 µg/mL. Both intra-day (n = 3) and inter-day (n = 6) precision was evaluated and the RSDs of the seven standards were less than 4.57% and 3.86%. The recoveries obtained in this study were in the range of 92-112% (Table S3) with low RSDs (<7%) of all standards, demonstrating that the analytical method developed has high accuracy and good reproducibility.
The quantitative results were shown in Table 2. Kukoamine B was only detected in the root barks in a very high amount (10.9 mg/g dry powder). The amount of rutin (D5) in leaves (663.45 µg/g dry leaves) was much higher than that in fruits (93 µg/g dry fruits). However, it was not found in root barks, which was consistent with the results of the above qualitative assay (Figure 3). Chlorogenic acid was only observed in leaves in a high amount (1577 µg/g dry leaves). N-p-trans-coumaroyltyramine was found in both fruits and leaves in a little amount (<15 µg/g dry powder). Dihydrosolasonine was only found in fruits in an amount of 43 µg/g dry fruit. The major steroidal glycoalkaloid in Solanaceae, solasonine, was only found in fruits in a very low amount (2.16 µg/g dry fruits), which is much lower than that found in Solanum xanthocarpum (800 µg/g) [48].

Antioxidative Activity Assays
Since a lot of phenolic compounds (phenylpropanoids, dicaffeoylspermidine/dicaffeoyl-spermine derivatives, phenolic amides, flavonoids) were found in the extracts, their antioxidative activities were compared using 2,2-diphenyl-1-picrylhydrazyl (DPPH), 2,2'-azinobis-(3-ethyl-benzthiazoline-6-sulphonate) (ABTS) and ferric reducing ability of plasma (FRAP) assays [49,50]. The results demonstrated that all the extracts showed antioxidative activities (Table 3). Both DPPH and ABTS assays showed that root bark extract possessed the strongest free radical-scavenging capacity, leaf extract was the second, and fruit extract showed much lower free radical-scavenging capacity than either leaf or root bark extracts. The FRAP assay also showed that the fruit extract had the weakest reducing ability. The strong antioxidative activity of the root bark extract could be explained by the huge amount of kukoamine A/B and propionyl-dicaffeoylspermidine in the root barks [51]. The higher amount of chlorogenic acid and rutin in leaves than those in fruits might explain the higher antioxidative activity of the leaf extract. The overall much higher antioxidant ability of leave and root bark extracts than that of the fruit extract is different from our common understanding that the fruits have strong antioxidant ability.  4 11.2 ± 0.5 --Trolox 5 -6 ± 1 37.7 ± 0.3 1 DPPH (IC 50 ) represents the extract concentration scavenging 50% of DPPH radical; 2 ABTS (IC 50 ) represents the extract concentration scavenging 50% of ABTS radical; 3 FRAP (RC 50 ) represents the extract concentration providing 50% reduction of Fe 3+ to Fe 2+ ; 4,5 represent the positive control; Results are expressed as means ± SD, n = 3.

Protective Effects of Extracts on H 2 O 2 -Induced Oxidative Stress in Cells
H 2 O 2 triggers oxidative damage through an increase of intracellular ROS. Therefore, the effects of extracts on the production of ROS in H 2 O 2 -exposed L02 cells were measured by 2',7'-dichloro-dihydrofluorescein diacetate (DCFHDA) assay ( Figure 4). ROS in cells can oxidize DCFH (no fluorescence) to form high fluorescent DCF. The results revealed that all the three extracts caused a dose-dependent attenuation of the H 2 O 2 -induced ROS production in L02 cells. The leaf and root bark extracts showed significantly higher antioxidative activity than that of fruit extract (Figure 4a-c and Figure S17).  4 11.2 ± 0.5 --Trolox 5 -6 ± 1 37.7 ± 0.3 1 DPPH (IC50) represents the extract concentration scavenging 50% of DPPH radical; 2 ABTS (IC50) represents the extract concentration scavenging 50% of ABTS radical; 3 FRAP (RC50) represents the extract concentration providing 50% reduction of Fe 3+ to Fe 2+ ; 4,5 represent the positive control; Results are expressed as means ± SD, n = 3.

Protective Effects of Extracts on H2O2-Induced Oxidative Stress in Cells
H2O2 triggers oxidative damage through an increase of intracellular ROS. Therefore, the effects of extracts on the production of ROS in H2O2-exposed L02 cells were measured by 2',7'-dichlorodihydrofluorescein diacetate (DCFHDA) assay ( Figure 4). ROS in cells can oxidize DCFH (no fluorescence) to form high fluorescent DCF. The results revealed that all the three extracts caused a dose-dependent attenuation of the H2O2-induced ROS production in L02 cells. The leaf and root bark extracts showed significantly higher antioxidative activity than that of fruit extract (Figures 4 a-c and S17). The ROS in cells treated by 0.25 mg/mL fruit extract was only reduced to 86%, while the ROS in cells treated by 0.2 mg/mL leaf and root extracts were reduced to 44% and 48% respectively. The confocal imaging of cells (Figure 4d) also showed that the three extracts greatly reduced the fluorescence intensity in cells, and the fruit extract showed the weakest effect. This set of results was consistent with the results of the above antioxidative activity analysis in solutions. The ROS in cells treated by 0.25 mg/mL fruit extract was only reduced to 86%, while the ROS in cells treated by 0.2 mg/mL leaf and root extracts were reduced to 44% and 48% respectively. The confocal imaging of cells (Figure 4d) also showed that the three extracts greatly reduced the fluorescence intensity in cells, and the fruit extract showed the weakest effect. This set of results was consistent with the results of the above antioxidative activity analysis in solutions.

In Vitro Assays of Cytotoxicity
Glycoalkaloids are reported to have anticarcinogenic activity because of their cytotoxicity (IC 50 = 50 µg/mL) [52]. Thus, the cytotoxicities of the fruit, leaf and root bark extracts to L02 cells were measured by a CCK-8 assay. L02 cell line was derived from the human hepatic and have been widely used in the evaluation of basic cytotoxicity profiles of drug candidates. As shown in Figure 5, the fruit extract did not show cytotoxicity at 1 mg/mL but showed cytotoxicity at 2 mg/mL. The leaf extract showed weak cytotoxicity at 1 mg/mL. However, the root bark extract showed weak cytotoxicity at 0.5 mg/mL, and strong cytotoxicity at 2 mg/mL. The strongest cytotoxicity of the root bark extract may come from the very large amount of dicaffeoylspermidines (kukoamine A/B and propionyl-dicaffeoylspermidine). To prove this hypothesis, the cytotoxicity of kukoamine B was tested. The results showed that kukoamine B (Figure 5b) had strong toxicity at 0.2 mg/mL, which was equivalent to the level of kukoamine B in~6 mg/mL root bark extract. Considering the similar amount of kukoamine A and propionyl-dicaffeoylspermidine existed together with kukoamine B in root bark extract, the higher cytotoxicity of root bark extract could mainly due to the high content of dicaffeoylspermidine/spermine derivatives. Although glycoalkaloids were only detected in the fruit extract, the extremely low levels cannot cause cytotoxicity.

In vitro Assays of Cytotoxicity
Glycoalkaloids are reported to have anticarcinogenic activity because of their cytotoxicity (IC50 = 50 µg/mL) [52]. Thus, the cytotoxicities of the fruit, leaf and root bark extracts to L02 cells were measured by a CCK-8 assay. L02 cell line was derived from the human hepatic and have been widely used in the evaluation of basic cytotoxicity profiles of drug candidates. As shown in Figure 5, the fruit extract did not show cytotoxicity at 1 mg/mL but showed cytotoxicity at 2 mg/mL. The leaf extract showed weak cytotoxicity at 1 mg/mL. However, the root bark extract showed weak cytotoxicity at 0.5 mg/mL, and strong cytotoxicity at 2 mg/mL. The strongest cytotoxicity of the root bark extract may come from the very large amount of dicaffeoylspermidines (kukoamine A/B and propionyl-dicaffeoylspermidine). To prove this hypothesis, the cytotoxicity of kukoamine B was tested. The results showed that kukoamine B (Figure 5b) had strong toxicity at 0.2 mg/mL, which was equivalent to the level of kukoamine B in ~6 mg/mL root bark extract. Considering the similar amount of kukoamine A and propionyl-dicaffeoylspermidine existed together with kukoamine B in root bark extract, the higher cytotoxicity of root bark extract could mainly due to the high content of dicaffeoylspermidine/spermine derivatives. Although glycoalkaloids were only detected in the fruit extract, the extremely low levels cannot cause cytotoxicity. Figure 5. The cytotoxicities of different concentrations of extracts from fruits, leaves, root barks (a) and kukoamine B (b) towards L02 cells. Results were expressed as means ± S.D., n = 3. ** p < 0.01, *** p < 0.001, **** p < 0.0001.

Potential Active Compounds Assay in Cells
Since the fruit of L. barbarum is mostly used as a functional food and medicinal source, and contains much more components than leaves and root barks, the potential active compounds assay in the extracts will provide helpful information for understanding the action mechanism of fruits of L. barbarum. In general, to exert an effect, bioactive molecules should bind receptors or enzymes on cell membranes or enter into cells to interact with their molecular targets. Although a large number of compounds are presented in the extract of plant herbs, only a few of them can bind or enter into cells. Therefore, the cell-based screening has been applied for identification of potential bioactive components in plant herbs [53]. In order to identify the potential active compounds, a compound database of fruits of L. barbarum containing 91 chemicals was successfully established based on the quasimolecular ions in Q1 and the characteristic fragment ions in Q3 as ion pairs in MRM mode for the first time. The MRM ion pairs and corresponding declustering potential (DP) and collision energy (CE) of each constituent were optimized and presented in Table S4.
To screen bioactive molecules, L02 cells were incubated with the fruit extracts for 4 h and then extracted with methanol. The methanol extract of cells was analyzed by UPLC-Qtrap-MS in MRM mode under the same condition with the construction of database. The typical total ion chromatography (TIC) of the extracts of L02 cells treated and untreated with fruit extracts are shown in Figure 6. More than 14 compounds were detected with the retention time in the range of 27-36 Figure 5. The cytotoxicities of different concentrations of extracts from fruits, leaves, root barks (a) and kukoamine B (b) towards L02 cells. Results were expressed as means ± S.D., n = 3. ** p < 0.01, *** p < 0.001, **** p < 0.0001.

Potential Active Compounds Assay in Cells
Since the fruit of L. barbarum is mostly used as a functional food and medicinal source, and contains much more components than leaves and root barks, the potential active compounds assay in the extracts will provide helpful information for understanding the action mechanism of fruits of L. barbarum. In general, to exert an effect, bioactive molecules should bind receptors or enzymes on cell membranes or enter into cells to interact with their molecular targets. Although a large number of compounds are presented in the extract of plant herbs, only a few of them can bind or enter into cells. Therefore, the cell-based screening has been applied for identification of potential bioactive components in plant herbs [53]. In order to identify the potential active compounds, a compound database of fruits of L. barbarum containing 91 chemicals was successfully established based on the quasimolecular ions in Q1 and the characteristic fragment ions in Q3 as ion pairs in MRM mode for the first time. The MRM ion pairs and corresponding declustering potential (DP) and collision energy (CE) of each constituent were optimized and presented in Table S4.
To screen bioactive molecules, L02 cells were incubated with the fruit extracts for 4 h and then extracted with methanol. The methanol extract of cells was analyzed by UPLC-Qtrap-MS in MRM mode under the same condition with the construction of database. The typical total ion chromatography (TIC) of the extracts of L02 cells treated and untreated with fruit extracts are shown in Figure 6. More than 14 compounds were detected with the retention time in the range of 27-36 min, suggesting that the compounds entered cells have relatively high hydrophobicity. In total, at least eight dicaffeoylspermidine derivatives (e.g., B31 and B37) were detected in the cells. Two flavones (rutin (D5) and D6) were observed in L02 cells. Interestingly, three saponins (E2, E4 and E6) were detected in cells with higher relative content compared to that detected in the extract. Because E2 (solasonine) and E4 (5,6-dihydrosolasonine) were detected in the extract in a very low concentration, their observation in cells suggests that solasonine and 5,6-dihydrosolasonine were enriched in cells. These compounds detected in cells may have potential biological activities, which should be clarified in the future research.
Molecules 2019, 24, x FOR PEER REVIEW 6 of 23 min, suggesting that the compounds entered cells have relatively high hydrophobicity. In total, at least eight dicaffeoylspermidine derivatives (e.g., B31 and B37) were detected in the cells. Two flavones (rutin (D5) and D6) were observed in L02 cells. Interestingly, three saponins (E2, E4 and E6) were detected in cells with higher relative content compared to that detected in the extract. Because E2 (solasonine) and E4 (5,6-dihydrosolasonine) were detected in the extract in a very low concentration, their observation in cells suggests that solasonine and 5,6-dihydrosolasonine were enriched in cells. These compounds detected in cells may have potential biological activities, which should be clarified in the future research. Figure 6. TIC of the extract of L02 cells treated and untreated with fruit extract. All the compound numbers were same with those shown in Table 1 and Table S4.

Materials and Reagents
The dried fruits, leaves and root barks of L. barbarum were purchased from Ningxia Province. Rutin, kukoamine B, chlorogenic acid, scopolin, solasonine and N-p-trans-coumaroyltyramine were purchased from Baoji Herbest Bio-Tech Co., Ltd. (Baoji, China). 5,6-Dihydrosolasonine were isolated from the fruit of L. barbarum and identified by HR-ESI-MS and NMR techniques. Purities of all compounds were above 96% by HPLC analysis. HPLC grade methanol, acetonitrile, and MS grade formic acid were purchased from Thermo Fisher Scientific (Pittsburgh, PA, USA). Other chemicals and solvents were of analytical reagent grade.   Table 1 and Table S4.

Materials and Reagents
The dried fruits, leaves and root barks of L. barbarum were purchased from Ningxia Province. Rutin, kukoamine B, chlorogenic acid, scopolin, solasonine and N-p-trans-coumaroyltyramine were purchased from Baoji Herbest Bio-Tech Co., Ltd. (Baoji, China). 5,6-Dihydrosolasonine were isolated from the fruit of L. barbarum and identified by HR-ESI-MS and NMR techniques. Purities of all compounds were above 96% by HPLC analysis. HPLC grade methanol, acetonitrile, and MS grade formic acid were purchased from Thermo Fisher Scientific (Pittsburgh, PA, USA). Other chemicals and solvents were of analytical reagent grade.

Sample Preparation
For bioactivity analysis, the fruits, leaves and root barks (50 g of each) of L. barbarum were pulverized into powder and extracted thrice by ultrasound with ethanol/water (70:30, v:v) (500 mL, 1 h; 400 mL, 1 h; 400 mL, 1 h), respectively. After filtration and freeze-drying, the crude products were 20.5 g, 14.3 g, 14.5 g, respectively. The stock solutions of the extracts were prepared by dissolving the freeze-dried extracts in DMSO (200 mg/mL). For UPLC-MS analysis, the powder of fruits, leaves and root barks (100 mg of each) were extracted by ultrasound with 1 mL of ethanol/water (70:30, v:v) for 1 h. After centrifugation, the supernatants were applied for UPLC-MS analysis.

Isolation and Identification of 5,6-dihydrosolasonine from Fruit of L. barbarum
The dried fruit of L. barbarum (5 kg) was extracted twice by ultrasound with 25 L of 70% ethanol/water for 1 h each time. After filtration, the ethanol was removed under reduced pressure to yield a concentrated solution. The solution was passed through a macroporous resin column (AB-8) and successively eluted with 0, 20, 80% ethanol/water. Finally, the 80% ethanol/water fraction was concentrated under reduced pressure, then extracted with butanol for three times and concentrated under vacuum. After dissolving this fraction with methanol, the target compounds were purified by preparative high performance liquid chromatography. The HR-ESI-MS data were recorded on a Orbitrap mass spectrometer (Thermo Fisher Scientific, Bremen, Germany). NMR spectra were acquired with Bruker AV 600 spectrometers (Bruker BioSpin Group, Faellanden, Switzerland) using the solvent signals (in C 5 D 5 N) as internal standards.
For identification of the components in the extracts, both positive and negative full scan modes within the mass/charge (m/z) ratio range of 150-1500 at a resolution of 120,000 were used for acquisition of accurate molecular ions. The other parameters were as follows: spray voltage, + 3.0 kV in the positive mode and − 2.0 kV in the negative mode; sheath gas flow rate, 35 Arb; aux gas flow rate, 10 Arb; sweep gas flow rate, 2 Arb; ion transfer tube temperature, 325 • C, vaporizer temperature, 275 • C. The fragment ions in MS/MS data obtained by higher energy collision dissociation (HCD, collision energy: 35 eV) were further utilized for confirmation of the structures of components. In addition, standards were also used for assistance of component identification. Xcalibur TM 3.0 software was used for UPLC-HR-MS control and data handling.

Compound Database Construction by UPLC-Qtrap-MS
An AB Sciex Qtrap ® 4500 tandem MS (Foster City, CA, USA) equipped with an ESI source connected to the UPLC system (I-class Acquity UPLC, Waters, Framingham, MA, USA) was used to construct the compound database. Firstly, an instrument method in MRM (Q1 = Q3) information-dependent acquisition (IDA)-enhanced product ion (EPI) mode [18] for the analysis of fruit extract was established on the basis of the identified compounds from fruits by UPLC-HR-MS. Further, product ion scanning experiments were conducted and the DP and CE was optimized for each analyte to generate the most abundant product ions. The product ion spectra were further used to select the precursor-product ion pairs for the development of MRM assays. Finally, a compound database of fruit of L. barbarum was established based on the quasimolecular ion in Q1 and its characteristic fragment ions in Q3 as ion pair in MRM mode. This database was used for the next screening of active compounds.

Quantitative Analysis of Compounds in Extracts
UPLC-Qtrap-MS was used for quantitative analysis of seven compounds in the extracts of fruits, leaves and root barks by MRM mode. The liquid chromatographic conditions were the same as those of UPLC-Orbitrap-MS analysis. Method validation was carried out for seven standards in terms of linearity, sensitivity, intra/inter-day precisions and recovery [54]. The linearity was obtained by preparing a series of concentrations of standards solution with at least five appropriate concentrations in duplicate. The LOD and the LOQ for each analyte were acquired while the S/N was 3 and 10, respectively. The precision (inter and intra-day precision) was analyzed using the standard solutions with six replicates, and the RSD of the peak area for each standard was calculated. A spike recovery test was used to evaluate the accuracy of these methods. Three concentrations (high, middle and low) of mixed standard solutions were added to fruit extract respectively, then quantitative analysis was performed as described above. Each standard was tested at each concentration in triplicate. The spike recoveries were calculated using the following equation: Recovery% = [(measured amount-original amount)/amount added] × 100, RSD = SD/mean × 100 [55]. Additionally, quantification of seven compounds using UPLC-Qtrap-MS was also established and the MRM pairs, DP and CE were optimized based on the standards.

DPPH Radical Scavenging Activity
The effect of the extracts against 2,2-diphenyl-1-picrylhydrazyl (DPPH) radical was tested according to a previous report [56] with slight modifications. In brief, 100 µL of 0.2 mM DPPH radical solution in ethanol was added to 100 µL of extract solutions at different concentrations. After incubation for 30 min at room temperature in the dark, the absorbance was read at 517 nm. Ascorbic acid was used as a positive control and all measurements were done in triplicate. The extract concentration providing 50% inhibition (IC 50 ) was calculated by plotting inhibition percentages against the concentrations of extracts. The DPPH radical scavenging rate (S%) was calculated as follows: S% = [(A 0 −A 1 )/A 0 ] × 100 (A 1 and A 0 are the absorbance of DPPH radical solution after incubation with and without extracts, respectively).

ABTS Radical Scavenging Assay
ABTS •+ scavenging activity was measured according to the defined method with slight modifications [57]. In brief, the radical cations were prepared by reacting 7 mM aqueous ABTS with 2.45 mM potassium persulphate. The mixture was allowed to stand in the dark at room temperature for 16 h before use and the ABTS •+ solution was diluted with methanol to an absorbance of 0.700 ± 0.020 at 734 nm. Different concentrations of extracts in methanol (40 µL) were added to 160 µL of ABTS •+ solution and the absorbance was recorded after 4 min. The IC 50 and percentage inhibition of absorbance at 734 nm were calculated. All measurements were done in triplicate. Inhibition of ABTS •+ in percent, I (%) was calculated as follows: I (%) = (A b −A s /A b ) × 100, where A b was the absorbance of the control and A s was the absorbance of tested samples.

FRAP Assay
The principle of the FRAP assay is based on the reduction of ferric-tripyridyltriazine complex to its ferrous (colored form) in the presence of antioxidants. The FRAP assay was performed as described previously [58]. Briefly, the FRAP reagent contained 5 mL of 10 mM TPTZ (2,4,6-tripyridy-s-triazine) solution in 40 mM HCl and 5 mL of 20 mM FeCl 3 ·6H 2 O in 300 mM acetate buffer (pH = 3.6). The mixture was freshly prepared and warmed at 37 • C for 30 min. In parallel, a solution containing 5 µL of ultrapure water and 155 µL of FRAP solution was prepared as the negative control. Different concentrations of extracts in methanol (5 µL) were mixed with 155 µL of FRAP solution and kept for 30 min in the dark. The ferrous tripyridyltriazine complex (coloured product) was measured by reading the absorbance at 593 nm. The antioxidative capacity of test samples was given by the RC 50 value, the concentration (µg/mL) necessary for a 50% reduction of Fe 3+ . Trolox was used as the positive control with a concentration ranging from 0 to 150 µg/mL.

Reactive Oxygen Species Measurement in L02 Cell
In cells, reactive oxygen species (ROS) were determined using a fluorescent dye protocol [59]. Cells were treated with different concentrations of each extract for 1 h and then incubated with H 2 O 2 (100 µM) for 1 h. The DCF fluorescence intensity was detected on a SpectraMax M5 microplate reader (Molecular Devices, Sunnyvale, CA, USA) with excitation at 488 nm and emission at 535 nm. The confocal imaging was performed on an OLYMPUS FV3000-IX81 confocal microscope (Olympus Corporation, Tokyo, Japan). Confocal images were processed by Olympus FV10-ASW 4.2 viewer software (Olympus Corporation, Tokyo, Japan).

Cytotoxicity Assay
L02 cells were cultured in 10% FBS-supplemented DMEM and 1% gentamicin, and kept in a humidified atmosphere of 5% CO 2 at 37 • C. For cytotoxicity assay, L02 cells in logarithmic growth phase were plated in 96-well plates at a density of 5 × 10 3 cells per well in 100 µL of culture medium and were allowed to adhere for 24 h before treatment. Serial concentrations of each sample (fruit, leaf and root bark extracts and kukoamine B) were then added (100 µL per well). After treated for 24 h, 10 µL of CCK-8 solution was added to each well and incubated at 37 • C, 5% CO 2 for 1 h. The absorbance at 450 nm was measured using a SpectraMax M5 microplate reader (Molecular Devices).

Target Cell-Based Screening of Potential Active Compounds
The target cell-based screening of potential active compounds in the fruit of L. barbarum was performed as described in previous study with slight modifications [15]. Specifically, L02 cells in the logarithmic growth phase were seeded into cell culture dish at a density of 1.0 × 10 6 cells/mL, and were cultured in DMEM medium at 37 • C for 24 h. The culture medium was replaced by 3 mL of fruits extract of L. barbarum diluted in DMEM (free of serum) at a final concentration of 10 mg/mL, and incubated at 37 • C for 4 h. The incubation solution was discarded and the cells were washed five times with phosphate-buffered saline to remove free components. Finally, the cells were collected and extracted with 200 µL of methanol and centrifuged at 12,000 rpm for 10 min. The obtained supernatant was used for UPLC-Qtrap-MS analysis. The control sample without extract treatment was prepared by the same procedures as above.

Data Handling and Presentation/Statistical Analysis
Data for quantification were acquired from individual experiments repeated at least three times, and expressed as the means ± SD. Statistical significance was calculated by GraphPad Prism 6 software (GraphPad Software, Inc., San Diego, CA, USA) with unpaired two-tailed t-tests and accepted by p < 0.05 (*), p < 0.01 (**), p < 0.001 (***), p < 0.0001 (****). The IC 50 or RC 50 were calculated using the GraphPad Prism 6 software according to the inhibition rates or reduction rates (y) plotted against the sample concentrations (x).

Conclusions
In this study, a total of 131 compounds were identified in extracts (70% ethanol) from fruits, leaves and root barks of L. barbarum by UPLC-Orbitrap-MS and seven of them were quantified by UPLC-Qtrap-MS. The distribution of these compounds in the three parts of L. barbarum was significantly different. The fruit extract contained the most compounds. A very large amount of kukoamine A/B (dicaffeoylspermine isomers) and propionyl-dicaffeoylspermidine were found in the extracts of root barks, and a high amount of dicaffeoylspermidine isomers were detected in the fruits and leaves. Many glycosylated derivatives of dicaffeoylspermine/dicaffeoylspermidine were mainly detected in the fruits. The bioactivity assays showed that the fruit extracts had the lowest antioxidant activity and cytotoxicity. The root bark extracts showed the strongest antioxidant activity and cytotoxicity, which was caused by the large amount of dicaffeoylspermidine/spermine derivatives. Six saponins were found in L. barbarum plants for the first time, and they were only detected in the fruits; among them, 5,6-dihydrosolasonine, a new glycoside alkaloid (saponin) was isolated and characterized. In addition, 14 potential bioactive compounds were detected in L02 cells after treated with the extracts of fruits. All these results will provide important information for understanding the different biological activities of the three parts of L. barbarum plants and will be beneficial for drug discovery from L. barbarum plants.
Supplementary Materials: The following are available online, Figure S1: Total ion chromatograms (TICs) of 7 standard substances in the positive ion mode; Figure S2: Proposed fragmentation pathway for solasonine, based on HR-Orbitrap MS/MS spectra; Figure S3. The HR-Orbitrap MS/MS spectra and Proposed fragmentation pathway of chlorogenic acid; Figure S4. The HR-Orbitrap MS/MS spectra and proposed fragmentation pathway of scopolin; Figure S5. The HR-Orbitrap MS/MS spectrum and the proposed fragmentation pathway of kukoamine B; Figure S6. The HR-Orbitrap MS/MS spectrum and the proposed fragmentation pathway of N-p-trans-Coumaroyl tyramine; Figure S7. The HR-Orbitrap MS/MS spectrum and the proposed fragmentation pathway of rutin; Figure S8. HR-Orbitrap MS/MS spectrum of A1; Figure S9. HR-Orbitrap MS/MS spectrum of B52; Figure S10. HR-Orbitrap MS/MS spectra of B63; Figure S11. HR-Orbitrap MS/MS spectrum of B18; Figure S12. HR-Orbitrap MS/MS spectra of compounds B46, B50 and B54; Figure S13. HR-Orbitrap MS/MS spectra of B51; Figure S14. HR-Orbitrap MS/MS spectra of C15; Figure S15. HR-Orbitrap MS/MS spectrum of D6; Figure S16. HR-Orbitrap MS/MS spectrum of E3; Figure S17. Protection Effect of the extracts from fruits (0.25 mg/mL), leaves and root barks (0.2 mg/mL) on 100 µM H 2 O 2 -induced intracellular ROS production in L02 cells. Table S1.13C NMR spectral data for aglycone and glycoside moieties of solasonine (SS) and 5.6-dihydrosolasonine (2HSS) in pyridine (D5); Table S2. The regression equation, LOD, LOQ, intra-day and inter-day of the 7 standards using the optimized method for calibration; Table S3. The recoveries of 7 standards (n = 3); Table S4. MS/MS parameters for the construction of the database of the 91 compounds in fruit of L. barbarum using UPLC-Qtrap-MS in the positive ion mode.
Author Contributions: X.X., W.R., Z.Z. and D.S. designed the experiment; X.X., W.R. performed the experiments with the help of N.Z., T.B., X.L. and X.X., W.R., D.S. and Z.Z. analyzed the data; X.X., D.S., W.R., and Z.Z. wrote the paper; D.S. conceived and directed the overall project. All authors have given approval to the final version of the manuscript.

Conflicts of Interest:
The authors would like to declare no conflict of interest in the publication of this research.