Comparison of Fruits of Forsythia suspensa at Two Different Maturation Stages by NMR-Based Metabolomics

Forsythiae Fructus (FF), the dried fruit of Forsythia suspensa, has been widely used as a heat-clearing and detoxifying herbal medicine in China. Green FF (GF) and ripe FF (RF) are fruits of Forsythia suspensa at different maturity stages collected about a month apart. FF undergoes a complex series of physical and biochemical changes during fruit ripening. However, the clinical uses of GF and RF have not been distinguished to date. In order to comprehensively compare the chemical compositions of GF and RF, NMR-based metabolomics coupled with HPLC and UV spectrophotometry methods were adopted in this study. Furthermore, the in vitro antioxidant and antibacterial activities of 50% methanol extracts of GF and RF were also evaluated. A total of 27 metabolites were identified based on NMR data, and eight of them were found to be different between the GF and RF groups. The GF group contained higher levels of forsythoside A, forsythoside C, cornoside, rutin, phillyrin and gallic acid and lower levels of rengyol and β-glucose compared with the RF group. The antioxidant activity of GF was higher than that of RF, but no significant difference was observed between the antibacterial activities of GF and RF. Given our results showing their distinct chemical compositions, we propose that NMR-based metabolic profiling can be used to discriminate between GF and RF. Differences in the chemical and biological activities of GF and RF, as well as their clinical efficacies in traditional Chinese medicine should be systematically investigated in future studies.

NMR-based metabolomics was performed to compare the chemical differences between GF and RF. In vitro biological activities, including antioxidant and antibacterial effects, were also assayed. The results may facilitate rediscovery or determination of the differences between GF and RF, which will positively contribute to the protection of FF wild resources and appropriate clinical use of this herb.

NMR Metabolic Profiling
The samples were extracted by the CHCl3-MeOH-H2O system and then divided into chloroform and aqueous methanol fractions. Based on preliminary experimental results, more valuable chemical information was obtained from the 1 H-NMR spectra of aqueous methanol fractions than from those of chloroform fractions. In the 1 H-NMR spectra of chloroform fractions, the dominant signals originated from glycerol esters, as well as saturated and unsaturated fatty acids; multivariate data analysis showed no obvious difference between GF and RF (Figures S1 and S2). Thus, the 1 H-NMR spectra of aqueous methanol fractions were further studied. 1 H-NMR spectra signals were assigned based on comparisons with the chemical shift of authentic standards, metabolites in the Biological Magnetic Resonance Data Bank [17] and NMR data in the literature (Table 1). Table 1. Relative levels of metabolites detected by NMR in green Forsythiae Fructus (FF) (GF) and ripe FF (RF) (mean ± SD, n = 10). VIP, variable influence on projection.   Representative 1 H-NMR spectra of aqueous methanol fractions of GF and RF are shown in Figure 1, with metabolites indicated based on their chemical shifts and coupling constants. A total of 27 metabolites were identified (Table 1). These metabolites consisted of amino acids, including leucine, isoleucine, valine, threonine, alanine and phenylalanine; organic acids, including 3-hydroxybutyric acid, acetic acid, pyruvic acid, succinic acid, malic acid and formic acid; sugars, including sucrose, β-glucose, α-glucose, β-xylose and raffinose; and other compounds, such as choline, gallic acid, forsythoside A, forsythoside C, cornoside, rengyol, phillyrin, adoxosidic acid and rutin ( Figure S3). The characteristic signals of phenylethanoid glycosides were also prominent in the 1 H-NMR spectra. For example, one of the major compounds was identified as forsythoside A, with resonances at δ 6.   Table 1.
Based on the assignments of 1 H-NMR spectra, chemical classification of all samples was performed by multivariate data analysis, which aimed to highlight the differences between GF and RF. An unsupervised approach consisting of principal component analysis (PCA), a nonparametric method of classification, was used to reduce the dimensions of multivariate problems. PCA score plots showed that GF and RF were clearly clustered into two groups, with the first two components explaining 55.0% of total variance (Figure 2A). This finding indicated that GF and RF were significantly different in their metabolites. However, the two samples diverged from other RF samples possibly because of different harvest times. The partial least square discriminant analysis (PLS-DA) model was also validated by a permutation test with 200 permutations ( Figure 2B). R 2 = (0.0, 0.353) and Q 2 = (0.0, −0.111) indicated that the PLS-DA pattern combined with the loading plot could determine potential biomarkers. To further identify the significant metabolites contributing to distinction between GF and RF, orthogonal partial least square discriminate analysis (OPLS-DA) of these NMR data was further performed. The OPLS-DA score plots demonstrated a clear separation between the GF and RF groups (R 2 X = 0.863, R 2 Y = 0.995, Q 2 = 0.900) ( Figure 2C). The OPLS-DA model was validated by cross-validation analysis of variance (CV-ANOVA) with a p-value of 2.36 × 10 −5 . The results indicated that the S-plot ( Figure 2D) of the OPLS-DA pattern and the variable influence on projection (VIP) values were suitable for finding the metabolites responsible for separation [29]. A total of eight metabolites were identified as being significantly different between GF and RF and were signed in the S-plot of the OPLS-DA model ( Figure 2D) through independent-sample t-test combined with the VIP values (VIP > 1). When GF grew into RF, the levels of cornoside (p = 1.14 × 10 −9 ), forsythoside A (p = 4.16 × 10 −6 ), forsythoside C (p = 6.55 × 10 −6 ), rutin (p = 1.33 × 10 −6 ), phillyrin (p = 1.27 × 10 −4 ) and gallic acid (p = 5.56 × 10 −4 ) decreased significantly, whereas the levels of rengyol (p = 3.53 × 10 −8 ) and β-glucose (p = 1.36 × 10 −5 ) increased significantly.

Antioxidant Activity
The antioxidant activity as measured by the ability to scavenge DPPH and hydroxyl radical were compared between 50% methanol extracts of GF and RF. As summarized in Figure 3, GF extracts exhibited lower IC50 values for DPPH (0.063 ± 0.007 mg·mL −1 ) and hydroxyl radical levels (1.207 ± 0.246 mg·mL −1 ), which indicated that GF showed higher DPPH and hydroxyl radical scavenging activities. In addition, FF contains many phenolic compounds, many of which show high antioxidant properties in vitro. Total phenolic content of GF and RF extracts were measured by the Folin-Ciocalteu method (Supplementary Material). The correlation coefficients between antioxidant activity and total phenolic content of GF and RF extracts were calculated. Significant positive correlations were observed between the total phenolic content and 1/IC50 values for DPPH (r = 0.956, p < 0.05) and hydroxyl radical (r = 0.917, p < 0.05) scavenging capacity, which indicated the significant contribution of phenolic compounds to the antioxidant activities of GF and RF. Comparison of antioxidant capacities measured by DPPH and hydroxyl radical scavenging assay between GF and RF. Results are based on independent-sample t-test and presented as the mean ± SD (n = 10). *** The RF group compared with the GF group, p < 0.001.

Antibacterial Activity
The preliminary antibacterial activity of 50% methanol extracts from all samples was performed by the agar well diffusion method. Results showed that GF and RF were active against four tested bacterial strains. To confirm and quantitatively determine the antibacterial activity, the minimum inhibitory concentration (MIC) was determined for four bacterial strains. As shown in Table 2, the MICs of GF and RF against Staphylococcus aureus were more sensitive, with values of 3.592 and 3.903 mg·mL −1 , respectively. However, no obvious difference was observed between GF and RF in their MIC values against the four tested bacterial strains (p > 0.05).

Discussions
GF and RF, which can be collected in a month's interval in autumn, are fruits of F. suspense at different maturity stages. They are the official sources of FF. FF undergoes a complex series of physical and biochemical changes, including shape, color, pathogen and resistance, during the process of fruit ripening. Biochemical changes can certainly lead to variation in biological activities, but GF and RF are not distinguished in current clinical use. The quality of F. suspensa 70% ethanol extracts was evaluated by HPLC coupled with a photodiode array detector, and the results suggested that the contents of forsythoside A and phillyrin were all higher in GF than in RF [8]. The 12 major constituents of F. suspensa were also determined by HPLC, and the findings showed that forsythoside is the dominant compound in FF. The contents of bioactive compounds, such as forsythoside A, rutin and phillyrin, were also higher in GF than in RF [9]. Furthermore, the amino acid and trace element contents of GF and RF were compared; the amino acid content was higher in GF than in RF, whereas the trace element content of RF was greater than that of GF [30]. However, obtaining the metabolic profiles of chemical constituents with all of the above analytical techniques is arduous. Hence, to comprehensively compare the chemical compositions of GF and RF, 1 H-NMR-based metabolomics coupled with HPLC and UV spectrophotometry was adopted in the current study, and various metabolites were found. The antioxidant and antibacterial activities of 50% methanol extracts of GF and RF were also evaluated in vitro.

Metabolic Profiling of Fruits at Different Maturity Stages of F. suspensa (GF and RF)
Many studies have suggested that phenylethanoid glycosides, lignans and flavonoids are the main bioactive components detected by HPLC, which are responsible for the various biological activities of FF [1,2,7,8]. In the present study, metabolic profiling was performed based on HPLC at 235 nm ( Figure S4). A total of 26 characteristic peaks, including six characteristic peaks in the chromatograms of GF and RF, were obtained. PCA score plots showed that GF and RF samples could be clearly separated along PC1 (with PC1 and PC2 accounting for 62.6% of total variance; Figure S5A). The corresponding loading plot of PCA ( Figure S5B) showed that the contents of forsythoside A, forsythoside C, rutin and phillyrin were higher in GF than in RF, which was consistent with previous reports in the literature [8,9].
Second, NMR-based metabolomics coupled with multivariate data analysis was applied. Metabolomic profiling suggested that the main differences were observed in aqueous methanol fractions of GF and RF, but no obvious difference existed in chloroform fractions. The most relevant variable contributions to the separation of GF and RF are shown in Table 1 and Figure 2D. What is more, more chemical information, including secondary metabolites, such as cornoside and rengyol, and primary metabolites, such as amino acids, sugars and organic acids, were found by NMR compared with HPLC. Cornoside is also a characteristic ingredient of GF, whereas rengyol is a characteristic component of RF. Therefore, NMR-based metabolomics is better-suited for identifying GF or RF in TCM prescriptions.

In Vitro Antioxidant and Antibacterial Activities of Fruits in Different Maturation Stages of F. suspensa (GF and RF)
In this study, antioxidant and antibacterial activity assays were used to evaluate the biological activities of FF in vitro. Phenolic compounds, such as forsythoside A, forsythoside C, phillyrin and rutin, which exhibit free radical scavenging activities, were mostly found in GF. When GF had grown to RF, the contents of forsythoside A, forsythoside C, phillyrin and rutin decreased, which corresponded to a reduction in total phenolic content and antioxidant activity. Forsythoside A, forsythoside C, rutin, phillyrin and gallic acid were all found to be present at significantly different levels in metabolites of GF compared with those of RF. However, comparison of their relative contents showed that forsythoside A content was higher than those of other phenolic compounds in GF, which was consistent with previous reports [8,9]. Thus, the higher antioxidant activity of GF compared to that of RF may be attributed to the greater contents of forsythoside A in GF than in RF. The high DPPH and hydroxyl radical scavenging activities of GF suggest that antioxidant drugs can be sourced from GF.
FF also exhibits antibacterial activities in phenylethanoid glycosides, specifically forsythoside A, which is mainly responsible for the antibacterial activities of GF [31]. However, in the current study, no significant difference was observed in the antibacterial activities of GF and RF, although GF was found to have a higher level of forsythoside A than RF. Thus, other RF compounds responsible for its antibacterial activity should be further investigated. In addition, neither GF nor RF can affect fungi. To determine the differences in biological activities between GF and RF more effectively, some in vivo pharmacological models that conform to FF against the main TCM clinical symptoms should be used in future research.

Putative Biosynthetic Pathways of the Major Secondary Metabolites in F. suspensa
The metabolic profiles of FF performed in this study using NMR-based metabolomics coupled with HPLC showed that phenylethanoid glycosides, lignans, cyclohexylethanol derivatives and flavonoids are the major secondary metabolites in FF. These findings are consistent with those of previous studies on the chemical composition of FF. The biosynthesis of these above kinds of compounds begins with the generation of phenylalanine and tyrosine precursors via the shikimic acid pathway (Figure 4) [32,33]. The biosynthesis pathways of lignans and flavonoids have been systematically studied, and the enzymes involved, such as pretycinnamate-4-hydroxylase (C4H), 4-coumarate: coenzyme A ligase (4CL), coumarate 3-hydroxylase (C3H) and caffeic acid O-methyltransferase, have been confirmed [32,34,35]. However, the biosynthetic pathways of phenylethanoid glycosides and cyclohexylethanol derivatives have not been studied thoroughly, and the enzymatic steps remain unclear. The caffeoyl moiety of phenylethanoid glycosides might be synthesized from phenylalanine via a cinnamic acid pathway. The hydrotyrosol moiety of phenylethanoid glycosides might be derived from tyrosine, either through tyramine or dopamine, based on the biosynthesis of phenylethanoid glycosides in other plants, such as verbascoside and echinacoside [36,37]. Hu et al. reported that tyrosine is a better precursor than phenylalanine for phenylethanoid glycoside accumulation and that the rate-limiting enzymes may be in the tyrosine branches [38]. The biosynthetic pathways of cyclohexylethanol derivatives may start from tyrosol and lead via salidroside to cornoside before further transforming to rengyol [25]. Endo et al. [26,39,40] have elucidated the biogenetic transformations from salidroside via cornoside to rengyol and the related natural cyclohexylethanoids by oxidation and successive reduction. However, all of the above findings are speculative, and real biosynthetic details remain unknown.
Our results showed that GF contained high levels of secondary metabolites, such as forsythoside A, cornoside, rutin and phillyrin, whereas RF contained a high level of rengyol (Table 1). This suggests that the enzymes catalyzing conversion of cornoside to rengyol may be upregulated during ripening. Phenylethanoid glycosides, such as forsythoside A and forsythoside C, may also be significantly decomposed by the enzyme and degraded finally into rengyol. Moreover, the contents of phillyrin and rutin, which have been shown to mediate various biological activities of FF [7], were decreased from GF to RF stages. Hence, cornoside may be accumulated in the early mature stage of the fruit (GF), whereas cyclohexylethanol derivatives, such as rengyol, may be accumulated in the completely mature stage of the fruit (RF). RF is also commonly used in Japan, where FF is referred to as 'rengyo' and RF contains higher levels of rengyol than GF [33,39]. Thus, some contradictions remain in the exact clinical usage of FF, and further studies are needed for clarification on this issue.

Plant Materials
Up to 20 commercial FF samples (Table 3) consisting of 10 GF and 10 RF were collected from 10 different counties located in Southeast Shanxi Province of China, which are considered the main producing areas of FF. The samples were authenticated by Prof. Xue-Mei Qin of Shanxi University. All voucher specimens were deposited in the herbarium in the Institute of Molecular Science of Shanxi University, China. All samples were ground to a fine powder and sieved with a bolt (20 meshes). The powder was then kept in an airtight container at −80 °C until use.

1 H-NMR Data Acquisition and Analysis
Two different extraction procedures were used for NMR analysis [21]. All powder samples (200 mg each) were transferred into 10-mL glass centrifuge tubes. About 6 mL of CHCl3-MeOH-H2O (2:1:1, v/v/v) mixture was added to the tube, followed by vortexing for 1 min and ultrasonication for 30 min.
The material was then centrifuged at 3500 rpm for 25 min. Chloroform and aqueous methanol fractions were transferred separately into a 25-mL round-bottomed flask and dried using a rotary vacuum evaporator. Chloroform fractions were dissolved in 800 µL CDCl3, and aqueous methanol fractions were dissolved in 800 µL mixture (1:1) of CD3OD and KH2PO4 buffer in D2O (adjusted to pH 6.0 by 1 N NaOD) containing 0.05% TSP. The samples were then centrifuged for 10 min at 13,000 rpm, and supernatants (600 μL) were transferred into 5-mm tubes for NMR analysis. 1 H-NMR was recorded at 25 °C on a Bruker 600MHz AVANCE III NMR spectrometer (600.13 M proton frequencies). CD3OD and CDCl3 were used for internal lock purposes. Each 1 H-NMR spectrum consisted of 64 scans, which required 5 min of acquisition time with the following parameters: 0.18 Hz/point, pulse width = 30° (12.7 μs) and relaxation delay = 5.0 s. A presaturation sequence was used to suppress the residual H2O signal with low-power selective irradiation at H2O frequency during the recycle delay. Free induction decays (FIDs) were Fourier transformed with Luria-Bertani (LB) = 0.3. A mixture of CHCl3-MeOH-H2O (2:1:1) was selected as the extraction system for FF. The resulting spectra were manually phased, baseline-corrected and calibrated to TSP at 0.00 ppm for aqueous methanol fractions and TMS at 0.00 ppm for chloroform fractions.
The 1 H-NMR spectra were processed using MestReNova software (Version 8.0.1, Mestrelab Research, Santiago de Compostela, Spain). For aqueous methanol fractions, spectral intensities were scaled to TSP and reduced to integrated regions of equal width (0.04 ppm) corresponding to the region of δ 0.50-10.02. The regions of δ 4.70-5.02 and δ 3.30-3.38 were excluded from the analysis because of residual signals of D2O and CD3OD, respectively. For the chloroform fractions, spectral intensities were scaled to TMS and reduced to integrated regions at 0.04 ppm corresponding to the region of δ 10.02-0.50. The region between δ 7.22 and δ 7.30 was removed from the analysis, because of the residual signal of CHCl3. The 1 H-NMR data obtained after mean-centering and Pareto scaling were imported into SIMCA-P 13.0 (Umetrics, Umeå, Sweden), which was then used for multivariate statistical analyses, including PCA, PLS-DA and OPLS-DA. The corresponding variables with variable importance on projection (VIP) values were calculated in the OPLS-DA model. The OPLS-DA, S-plot and VIP value were used for selecting the metabolites responsible for sample differentiation. Metabolites were chosen as differentiating metabolites when their VIP values were larger than 1.0 [41,42]. Statistical analysis was also performed using an independent-sample t-test (SPSS 16.0, SPSS Inc., Chicago, IL, USA). In comparisons of differences between two groups, p < 0.05 was considered statistically significant.

Antioxidant Activity
All powders (0.10 g each) were accurately weighed and mixed with 10 mL of 50% methanol in centrifuge tubes and ultrasonically extracted (KQ-250DB ultrasonic bath, Kunshan, Jiangsu, China) for 30 min. The resultant mixture was adjusted to the original weight and centrifuged at 3000 rpm for 10 min. The collected supernatant was used as a stock solution to determine antioxidant activity.

DPPH Radical Scavenging Assay
The antioxidant activity of the above extract was determined by the DPPH radical scavenging assay as described before with slight modifications [43,44]. The stock solution (10 mg·mL −1 ) was diluted with methanol to final concentrations ranging from 0.010 mg·mL −1 to 0.50 mg·mL −1 . Methanol was used as a negative control, while ascorbic acid served as a positive control. Half inhibitory concentration (IC50) values were also calculated.

Antibacterial Activity
All powders (2.0 g each) were accurately weighed and extracted with 50% methanol (1:10 w/v) in a 50-mL round-bottom flask by refluxing for 1 h. The extract solvents were evaporated under reduced pressure using a rotary vacuum evaporator at 50 °C. Residues were dissolved in sterile water at 200 mg·mL −1 of dried herbs, then sterilized before use.
The minimum inhibitory concentration (MIC) was determined by the microdilution bioassay in 96-well microplates [44,46,47] with some modifications. Briefly, two-fold serial dilutions of each extract were added to LB broth to yield volumes of 100 μL, with final concentrations ranging from 0.20 mg·mL −1 to 100 mg·mL −1 , and about 100 μL of bacterial culture (approximately 10 5 CFU/mL) were added into the microplate wells. The microplates were covered and incubated overnight at 37 °C. About 40 μL of 0.2 mg·mL −1 p-iodonitrotetrazolium chloride were added to the microplate wells and incubated at 37 °C for 30 min to induce bacterial growth. The bacterial suspension of the microplate wells where bacterial growth occurred changed to red. Streptomycin (100 mg·mL −1 ) was used as a positive control for each bacterium, while solvent and bacterium-free wells served as negative controls. MIC was calculated as the lowest concentration that completely inhibited bacterial growth. Samples were tested against two Gram-negative bacteria (Escherichia coli and Pseudomonas aeruginosa) and two Gram-positive bacteria (Staphylococcus aureus and Bacillus subtilis), all of which were obtained from Jun Xu of the College of Life Science, Shanxi University, China. All experiments were performed in triplicate.

Data Analysis
All data are presented as means (mean ± SD) of three measurements. The IC50 values were calculated using Origin 8.0 (Microcal Software, Inc., Northampton, MA, USA) by plotting percentage inhibition against concentration. The correlation values were examined using Pearson correlation. All statistical analysis of the data was performed by the independent-sample t-test, and Pearson's correlation coefficient (r) was calculated using SPSS 16.0 (SPSS Inc., Chicago, IL, USA). Differences at p < 0.05 were considered significant.

Conclusions
In this study, we first conducted aqueous methanol metabolic profiling using NMR for the analysis of F. suspensa fruits. In vitro biological activities, such as antioxidant and antibacterial effects, were then measured. Total phenolic content in GF and RF were also determined. A total of 27 metabolites were identified from our NMR data, and eight of them were different between the GF and RF groups. The GF group contained a higher level of forsythoside A, forsythoside C, cornoside, rutin, phillyrin and gallic acid, as well as a lower level of rengyol and β-glucose compared to the RF group. In addition, antioxidant activity was much higher in GF, and no significant difference was found between the antibacterial activities of GF and RF.
Despite our knowledge of chemical differences between GF and RF, no distinctions in their clinical efficacy have been established to date. In order to guarantee the efficacy and safety of FF in clinical applications, a systematic investigation of differences in the chemical and biological activities of GF and RF is warranted. Furthermore, since FF is used commonly in TCM, further study on the difference in effectiveness of GF and RF in TCM prescriptions is also necessary.