An Integrated Approach to Characterize Intestinal Metabolites of Four Phenylethanoid Glycosides and Intestinal Microbe-Mediated Antioxidant Activity Evaluation In Vitro Using UHPLC-Q-Exactive High-Resolution Mass Spectrometry and a 1,1-Diphenyl-2-picrylhydrazyl-Based Assay

Intestinal bacteria have a significant role in metabolism and the pharmacologic actions of traditional Chinese medicine active ingredients. Phenylethanoid glycosides (PhGs), as typical phenolic natural products, possess wide bioactivities, but low oral bioavailability. The aim of this work was to elucidate the metabolic mechanism underlying PhGs in the intestinal tract and screen for more active metabolites. In this study, a rapid and reliable method using an effective post-acquisition approach based on advanced ultra-high-performance liquid chromatography (UHPLC) coupled with hybrid Quadrupole-Orbitrap high resolution mass spectrometry (Q-Exactive-HRMS) provided full MS and HCD MS2 data. Thermo Scientific™ Compound Discoverer™ software with a Fragment Ion Search (FISh) function in one single workflow was developed to investigate the intestinal microbial metabolism of four typical PhGs. Furthermore, antioxidant activity evaluation of PhGs and their related metabolites was simultaneously carried out in combination with a 1,1-diphenyl-2-picrylhydrazyl (DPPH) assay to understand how intestinal microbiota transformations modulate biological activity and explore structure–activity relationships (SARs). As a result, 26 metabolites of poliumoside, 42 metabolites of echinacoside, 42 metabolites of tubuloside, and 46 metabolites of 2′-acetylacteoside were identified. Degradation, reduction, hydroxylation, acetylation, hydration, methylation, and sulfate conjugation were the major metabolic pathways of PhGs. Furthermore, the degraded metabolites with better bioavailability had potent antioxidant activity that could be attributed to the phenolic hydroxyl groups. These findings may enhance our understanding of the metabolism, pharmacologic actions, and real active forms of PhGs.

Intestinal bacteria have a significant role in metabolism and the pharmacologic actions of traditional Chinese medicine active ingredients. Phenylethanoid glycosides (PhGs), as typical phenolic natural products, possess wide bioactivities, but low oral bioavailability. The aim of this work was to elucidate the metabolic mechanism underlying PhGs in the intestinal tract and screen for more active metabolites. In this study, a rapid and reliable method using an effective post-acquisition approach based on advanced ultra-highperformance liquid chromatography (UHPLC) coupled with hybrid Quadrupole-Orbitrap high resolution mass spectrometry (Q-Exactive-HRMS) provided full MS and HCD MS 2 data. Thermo Scientific™ Compound Discoverer™ software with a Fragment Ion Search (FISh) function in one single workflow was developed to investigate the intestinal microbial metabolism of four typical PhGs. Furthermore, antioxidant activity evaluation of PhGs and their related metabolites was simultaneously carried out in combination with a 1,1-diphenyl-2-picrylhydrazyl (DPPH) assay to understand how intestinal microbiota transformations modulate biological activity and explore structure-activity relationships (SARs). As a result, 26 metabolites of poliumoside, 42 metabolites of echinacoside, 42 metabolites of tubuloside, and 46 metabolites of 2′-acetylacteoside were identified. Degradation, reduction, hydroxylation, acetylation, hydration, methylation, and sulfate conjugation were the major metabolic pathways of PhGs. Furthermore, the degraded

INTRODUCTION
Recently, more and more attention has focused on the nutraceutical industry and preventive medications in the quest for natural antioxidants from plant materials. Polyphenolic compounds are well known to possess various pharmacologic effects, especially antioxidant activity (Mikulic-Petkovsek et al., 2016;Zhao et al., 2017;Kschonsek et al., 2018). Phenylethanoid glycosides (PhGs) are a class of polyphenolic glycoside compounds that are widely distributed in many plants. To date, various PhGs have been isolated and identified (Song et al., 2016a;Song et al., 2016b;Zou et al., 2017) In recent decades, pharmacologic studies have shown that the bioactivities of PhGs are diverse, including neuroprotective, cardioactive, hepatocyte protective, antioxidative, anti-inflammation, and immunomodulatory effects (Wang et al., 2012;Li et al., 2016c;Gu et al., 2016;Xue and Yang, 2016;Fu et al., 2018). These outstanding PhG activities in diverse diseases have proven importance in medicinal chemistry research.
Most herbal medicines are administered orally, and the components are inevitably metabolized before absorption from the gastrointestinal tract; however, poor oral absorption of PhGs was observed in a Caco-2 cell monolayer model, suggesting poor intestinal permeability (Gao et al., 2015;Zhou et al., 2018). A low blood drug concentration and relatively rapid metabolism of PhGs were observed after dosing in previous pharmacokinetic studies (Qi et al., 2013;Deng et al., 2014;Li et al., 2015a;Cui et al., 2016a;Li et al., 2016a;Su et al., 2016;Cui et al., 2017;Feng et al., 2018;Qian et al., 2018); however, these existing results are inadequate to fully understand the metabolic process and the mechanism underlying the pharmacologic activities. An in vitro digestion model provides a useful platform for fast and reproducible assessment of herbal medicine metabolism (Payne et al., 2012;Kang et al., 2013;Xu et al., 2017;Cui et al., 2018;Zhang et al., 2018;Feng et al., 2019). Several in vitro studies involving the metabolism of PhGs (echinacoside, acteoside, isoacteoside, and 2′-acetylacteoside) showed that PhGs are easily transformed into degradation products, such as caffeic acid (CA) and hydroxytyrosol (HT), by human or rat intestinal bacteria or intestinal enzymes (Li et al., 2015b;Cui et al., 2016b;Li et al., 2016b;Li et al., 2017). These degradation products are involved in further metabolism, which is reasonable for the low oral bioavailability of PhGs. These results also suggested that PhGs may serve as a prodrug and is transformed by intestinal flora leading to the occurrence of new metabolites with increased activity. Consequently, more attention should be given to these metabolites and the biological properties to help explain the health effects of PhGs that are not easily absorbed through the gut barrier.
In the current study, we systematically characterized the metabolic process of four typical PhGs (echinacoside, poliumoside, 2′-acetylacteoside, and tubuloside A) using a UHPLC-HRMS Orbitrap™ instrument with high specificity, sensitivity, and accuracy. A full scan with data-dependent MS 2 (ddMS 2 ) acquisition is enabled in one run, providing complementary information for the structural elucidation of metabolites. Compound Discoverer software is the only smallmolecule analysis solution able to make full use of the rich highresolution accurate-mass (HRAM) data produced by Orbitrap mass spectrometers. The application of LC-HRAM with postacquisition data processing using new Compound Discoverer software was verified in accurate and batch metabolite identification. Transformation sites and pathways of the four components were also proposed. Although we found that these PhGs are extensively metabolized by intestinal bacteria, little is known about the role of intestinal bacteria in bioactivity modification of PhG types. Accordingly, these PhGs and related metabolites should be simultaneously taken into account when assessing bioactivities. Previous studies have shown that PhGs as powerful natural antioxidants relevant to the phenolic group and some of the intestinal microbial metabolites still bear a phenolic group and have been suggested to possess strong antioxidant activities (Dueñas et al., 2015;Jiang et al., 2016;Ydjedd et al., 2017;Wang et al., 2017a). In this regard, the antioxidant activities were compared with the 1,1-diphenyl-2-picrylhydrazyl (DPPH)based assay in vitro for the first time. Some selected metabolites have been used to carry out structure-activity relationship (SAR) studies to better understand which structural features are essential for biological activity. Based on the SARs, this is also the first report to elucidate these intestinal microbe-mediated biotransformation modulations on antioxidant activities in vitro to date. Hence, our results may provide helpful information to better understand in vivo metabolism, mechanism of action of PhGs, and drug discovery in clinical development research. Co., Ltd. (Chengdu, China). The structures are shown in Figure 1 and the purity of each reference standard was determined to be >98% by normalization of the peak areas detected by HPLC-DAD-TOF/MS and the reports of 13 C-NMR analysis. DPPH, L-ascorbic acid, and trolox were purchased from Sigma-Aldrich (St. Louis, MO, USA). The solvents, acetonitrile and methanol, were of HPLC grade and obtained from Merck (Darmstadt, Germany). Formic acid (MS grade) was purchased from Fisher Scientific (Madrid, Spain). Deionized water (18 MΩ) was prepared by distilled water through a Milli-Q system (Millipore, Bedford, MA, USA). All of the other chemicals and reagents were of analytical grade. AnaeroPack rectangular jars were purchased from Mitsubishi Gas Chemical Company, Inc. (Tokyo, Japan). General anaerobic medium (GAM) was purchased from Shanghai Kayon Biological Technology Co., Ltd. (Shanghai, China). Corning (Costar) Inc. (Tewksbury, MA, USA) was the vendor for 96-well plates.
Ten male Sprague-Dawley rats (7-8 weeks old and weighing 220 ± 20 g) were obtained from Weitonglihua Company (Beijing, China) and housed with free access to standard food and water. The animals were kept under controlled conditions (temperature, 22-24°C; relative humidity, 50 ± 10%) with a 12-h light/12-h dark cycle and acclimated to the housing environment for 1 week prior to the experiment with free access to food and distilled water. All of the experiments were performed according to the National Institutes of Health Guidelines for Animal Research and approved by the Ethics Committee of the Institute of Medicinal Plant Development, CAMS & PUMC.

Incubation of Four PhGs With Rat Intestinal Bacteria
GAM content (g/L) was as follows: 15.0 g of peptone; 10.0 g of tryptone; 3.0 g of soya peptone; 5.0 g of yeast extract; 13.5 g of digestibility serum powder; 1.2 g of beef extract; 3.0 g of glucose; 2.5 g of KH 2 PO 4 ; 3.0 g of NaCl; 0.3 g of soluble starch; 0.3 g of L-cysteine hydrochloride; 0.15 g of sodium thioglycolate; and 6.0 g of anaerobic broth powder. Thirty grams of GAM was dissolved in 1,000 mL of distilled water, filtered while hot, treated with an anti-bacterial process with high pressure (0.15 MPa) and temperature (121°C) for 20 min, and cooled to 45°C, and 1 mg of vitamin K1 and 6 mg of hematin chloride were dissolved in the GAM broth solution.
Fresh fecal samples were obtained from 10 male Sprague-Dawley rats. The fresh rat feces were mixed immediately and homogenized with 4 volumes of aseptic physiologic saline using a vortex mixer. The suspension was centrifuged at 2,000 rpm for 10 min, and the suspension was inoculated into 9 volumes of GAM broth, which was then incubated at 37°C in an anaerobic pack for 24 h. The resulting mixture of bacteria was centrifuged at 4000 rpm for 10 min, and the residue was suspended in aseptic physiologic saline to be used as the intestinal bacterial mixture.
One milliliter of GAM broth containing 0.1 mL of intestinal bacterial mixture and the 4 PhGs (1 mM) were separately transferred to 2-mL Eppendorf tubes and incubated at 37°C for 24 h in anaerobic cultivation bags sealed with anaerobic airbags. Incubation of intestinal bacteria in medium, but lacking the PhG solutions, was used to monitor metabolites arising from basal metabolism (control groups).

Sample Preparation
The cultured mixture was removed and extracted with watersaturated n-butanol (1:1, v/v) after incubation for 24 h. The mixture was centrifuged at 4,800 rpm for 20 min and then the supernatant was concentrated under a stream of nitrogen at room temperature. The residue was dissolved in 0.5 mL of methanol and centrifuged at 12,000 rpm for 10 min and then the supernatant was analyzed by UHPLC-Q-Exactive Orbitrap HRMS.
Tandem mass spectrometry was performed with a Q Exactive Orbitrap MS (Thermo Fisher, Waltham, MA, USA) using a heated electrospray ionization source for the ionization of the target compounds in the negative ion (NI) mode. The operating parameters were as follows: spray voltage, 2.00 kV; sheath gas pressure, 30 psi; auxiliary gas pressure, 10 arb; capillary temperature, 320°C; auxiliary gas heater temperature, 350°C; scan modes, full MS (resolution 70,000) and ddMS 2 [resolution 17,500, with stepped collision energy (20, 40, and 60 eV)]; and scan range, m/z 80-1200. All data were acquired using the Xcalibur 3.1 software (Thermo Scientific).

DPPH Radical Scavenging Assay
The antioxidant activity of PhG and some metabolite-related analogs was assessed using a feasible and rapid DPPH radical scavenging assay in vitro. Each sample stock solution was diluted to final concentrations of 0.05-2,000 μg/mL in MeOH. The DPPH radical standard solution (1 × 10 −4 mol/L) was prepared fresh by immediately dissolving an accurately weighed DPPH sample in ethanol while protected from light on the day of each test. One milliliter of each sample solution at various concentrations was mixed with 1 mL of 1 × 10 −4 mol/L DPPH solution (in ethanol). All samples were shaken and allowed to stand in the dark at room temperature for 30 min. The reduction in DPPH-free radicals was measured by reading the absorbance at 517 nm against a blank on a microplate reader (Tecan, Switzerland). One milliliter of ethanol plus 1 mL of sample solution was used as a blank, while 1 mL of DPPH solution plus 1 mL of MeOH was used as a negative control. Trolox and L-ascorbic acid were used as positive controls. The radical scavenging activity (% inhibition) of the tested samples was expressed as the DPPH scavenging percentage and calculated using the following formula: % inhibition = [(A control − A sample)/A control] × 100. The antioxidant activity was calculated by plotting the percent inhibition against the sample concentration and represented as the sample concentration required to scavenge 50% of the DPPH radical (IC 50 ). All tests were carried out in triplicate, and the IC 50 values were reported as the mean ± SD.

Data Processing
The raw MS data files of the blank matrix and the control and sample groups were imported into Compound Discoverer™ software (v.2.1; Thermo Scientific, Fremont, CA, USA) to identify metabolites of the four PhGs. Compound Discoverer TM software can quickly find and identify metabolites with background removal from the blank matrix. A list of potential metabolites was generated depending on the HRAM measurement, as follows: each within ±5 ppm of mass error; retention time tolerance of ±0.1 min; an ion ratio tolerance within ±30%; and fine isotopic pattern matching >90% of the precursor and the characteristic product ions. Furthermore, structural elucidations and transformations were suggested for each chromatographic peak by the Fragment Ion Search (FISh) function in Compound Discoverer TM software. The FISh coverage score was calculated and fragments are auto-annotated with structure, molecular weight, and elemental composition on MS/MS spectra. The expected compounds table was filtered by background and maximum area ≥ 10 5 or FISh coverage score ≥ 50.

Fragmentation Studies
As reported, the common chemical structure of PhGs is characterized by a phenethyl alcohol (C6-C2) moiety, such as HT or p-tyrosol, attached to a β-glucopyranose/β-allopyranose via a glycosidic bond, side-chain aromatic acids (e.g., CA or coumaric acid), and sugar groups, including rhamnose (Rha) and glucose (Glu) Han et al., 2012;Zheng et al., 2014). In this work, eight reference standards were used for the fragmentation patterns study using UHPLC-Q-Exactive Orbitrap HRMS, which is helpful in further metabolite characterization. The detailed fragmentation patterns of echinacoside, acteoside, isoacteoside, 2′-acetylacteoside, and tubuloside B have been reported in our previous study by UPLC-ESI-QTOF/MS n (Wang et al., 2017b). In the NI mode of Q-Exactive Orbitrap HRMS, the fragmentation patterns of other parent compounds, such as poliumoside and tubuloside A, were first proposed in our study. By way of comparison with echinacoside, poliumoside is also a trisaccharide glycoside; however, there is a Rha, not a Glu unit, at the C-6′ position of the central Glc unit. Therefore, the other fragmentation patterns of poliumoside were similar to those of echinacoside, apart from the different neutral loss of the Rha or Glu moiety. Tubuloside A is also a trisaccharide glycoside. The structure of tubuloside A has an additional Glc group conjugated at the C-6′ position of the central Glc unit, and when compared with 2′-acetylacteoside, the fragmentation of tubuloside A can also be predicted. The proposed fragment ion structures and stepwise elucidations on the fragmentation patterns are illustrated in Figure 2. In the present study, the common fragmentation patterns (Figure 3) with the MS n data in the NI mode were systematically summarized as follows: 1) the major and typical neutral losses corresponding to the loss of Ac (

Metabolite Identification of Four PhGs by Rat Intestinal Bacteria
In this study, UHPLC-Q-Exactive Orbitrap HRMS with Compound Discoverer software were used to investigate the   Table S1.
Metabolites were classified into the following three categories: 1) degradation products of parent compounds, 2) phase I and II metabolites of parent compounds, and 3) phase I and II metabolites of degradation products. The metabolite identification strategy and results are elucidated in detail, as follows. were identified as poliumoside, echinacoside, tubuloside A, and 2′-acetylacteoside, respectively, by comparing the retention times and mass spectra with those of authentic reference standards ( Figure 2). The fragment ion of the parent compound resulting from in-source by high energy collision dissociation (HCD) has exactly the same structure of one of the main degradation products because the glycosidic or ester bonds of the compound that can be hydrolyzed easily by intestinal bacteria in the alimentary tract are also the more likely to be broken during in-source fragmentation ( Figure  3). Therefore, the degradation pathways, including de-rhamnose, deglycosylation, de-caffeoyl, de-HT, and deacetylation, are always similar to the mass fragmentation patterns. Moreover, the identification process of these compounds is simplified with automatic batch FISh annotations via direct fragment match.

Degradation Products of Parent Compounds
Accordingly, we found that degradation products were formed by a combination of stepwise cleavage of Rha (146 Da), Glu (162Da), CA (162 Da), HT (136 Da), or Ac (42 Da) moieties via the "fragmentation-degradation" comparisons with the parent compound. Moreover, the detected diagnostic product ions originated from the following are also the fragment characteristic of the core chemical groups within the parent molecule: a) HT; b) CA; and c) Glu or Rha moieties. As discussed, all degradation products have been found to characterize the quasi-molecular ions with the same m/z and calculated elemental compositions to that of the parent compound fragments. The results are illustrated in detail as follows.
M1  ions associated with CA at m/z 179.03, 161.02, and 135.04, suggesting that M1-21 was the de-Rha and de-HT product of M1. M1-2 [m/z 341.08850 (C 15 H 17 O 9 ), eluted at 7.97 min] was 146 Da (C 6 H 10 O 4 ) less than M1-21, and the diagnostic fragmentation from the CA moiety was found at m/z 179.03, 161.02, and 135.04, suggesting that M1-2 was the de-HT and de-2Rha product of M1. M1-3 eluted at 9.94 min and was characterized with the quasimolecular ion at m/z 607.22443 (C 26 H 39 O 16 ), which is really close to the de-CA fragment ion (m/z 607.22437) of M1 and is proposed to be the de-CA product of M1. The major fragment ions of M1-3 (m/z of 461.17 and 315.11) were also the same as from M1. The diagnostic fragmentation from the HT moiety was found at m/z 153.06 and 123.04. Based on this rule, M1-22 [m/z 461.16632 (C 20 H 29 O 12 ), eluted at 8.07 min] was identified as the de-CA and de-Rha product of M1.
For M2, there is a Glu, not a Rha moiety, substituted at the C-6′ position by comparison with M1. Therefore, M2-15 or isomers [m/z 623.20 (C 29 H 35 O 15 ), eluted at 23.14 and 26.08 min, respectively] were assigned as C-6′ de-Glu products of M2 and also confirmed as acteoside and isoacteoside, respectively.  42 Da over CA and the presence of a fragment ion at m/z 177.06 (C 10 H 9 O 3 ) formed by a CO 2 loss. Similarly, the acetylated product of 3,4-dihydroxybenzenepropionic acid also showed an added mass of 42 Da over 3,4-dihydroxybenzenepropionic acid at m/z 223.06 (C 11 H 11 O 5 ) and the successive CO 2 loss ion at m/z 179.07 (C 10 H 11 O 3 ). In the same way, the sulfated product of 3,4-dihydroxybenzenepropionic acid [m/z 261.01 (C 9 H 9 O 7 S), 9.3 min] showed the added mass of 80 Da over 3,4-dihydroxybenzenepropionic acid and the presence of characteristic fragment ions at m/z 181.05 (C 9 H 9 O 4 ) and 137.06 (C 8 H 9 O 2 ) formed by losing a SO 3 and CO 2 , respectively.

Phase I and II Metabolites
In this study, we proposed a strategy for rapidly characterizing the phase I and II metabolites of both parent compounds and their degradation products (see the section Degradation Products of Parent Compounds). The direct phase I and II metabolites of these compounds were relatively easy to characterize. First, with the benefit of HRAM data acquired by Orbitrap mass, these metabolites were compared by the mass and elemental composition differences of the precursors. Subsequently, the reasonable formula change could be determined based on the knowledge of transformations. Finally, the fragmentation pattern comparisons helped to confirm the characterizations and the distinctive fragmentation ions contributed to determining metabolites or isomers.
Moreover, most metabolites can go through degradation and further phase I and/or phase II metabolism. Such metabolites are always produced from multiple metabolic reactions from the parent compound. The proposed multiple crossover "fragmentation-degradation" comparisons between parent compound and their phase I and II metabolites help to rapidly characterize such metabolites (García-Reyes et al., 2007;Wang et al., 2009;López et al., 2016).
The identification process of these metabolites is also automatically achieved by the FISh function in Compound Discoverer™ software. To better characterize the metabolites, those metabolites with the same metabolic pathway which had similar fragmentation patterns and diagnostic ions could be discussed together. Furthermore, some structural characterization of the metabolites can be achieved and described in detail as follows.       M1-10 [m/z 801.28125 (C 36 H 49 O 20 ), eluted at 24.24 min] was 14 Da (+CH 2 ) higher than M1-9 in molecular weight, indicating that M1-10 was the methylated product of M1-9. In the MS/MS spectrum of M1-10, the fragment ion at m/z 769.26 (C 35

Sulfated Metabolites (+ SO 3 , +80 Da)
M1-11a (eluted at 22.77 min), M1-11b (eluted at 12.27 min), and M1-11c (eluted at 23.53 min) exhibited the same molecular formula of C 35 H 45 O 22 S (m/z 849.21) and were 80 Da (+SO 3 ) higher than M1. M1-11a, M1-11b, and M1-11c were determined to be the sulfated products of M1. In the MS/ MS spectra, M1-11a, M1-11b, and M1-11c showed different fragment ions, indicating that M1-11a, M1-11b, and M1-11c were sulfated at different sites of the structure. In the MS/MS spectra of M1-11c, the fragment ion at m/z 769.26 was formed by a loss of SO 3 from the quasi-molecular ion. Additionally, other fragment ions from M1-11c were found (m/z of 607. 22, 461.17, 179.03, and 161.02), which were consistent with the characteristic fragment ions of M1, indicating that the sulfation site was at the C-2′ position. In the MS/MS spectra of M1-11a, the fragment ion at m/z 215.00 (C 8 H 7 O 5 S) was presumed to be a CO 2 loss from the sulfated CA [m/z 258.99 (C 9 H 7 O 7 S)] as evidenced by the 80-Da (+SO 3 ) difference from CA [m/z 179.03 (C 9 H 7 O 4 )] and the presence of characteristic fragment ions of CA (m/z 179.03, 161.02, and 135.04) formed by a loss of SO 3 , indicating that the sulfation site of M1-11a was at one of the hydroxyls of the CA moiety. In the MS/MS spectra of M1-11b, the fragment ion at m/z 687.18 (C 26 H 39 O 19 S) was a CA moiety loss from the quasi-molecular ion, which suggested that the sulfate site was not at the CA moiety. In addition, the fragment ion at m/z 215.00 (C 8 H 7 O 5 S) was presumed to be a H 2 O loss from sulfated HT [m/z 233.01 (C 8 H 9 O 6 S)], as evidenced by the 80-Da (+SO 3 ) difference from HT [m/z 153.06 (C 8 H 9 O 3 )], and confirming that the sulfation site of M1-11b was at one of the hydroxyls of the HT moiety. Similarly, M4-24a (22.44 min), M4-24b (27.76 min), and M4-24c (16.69 min), with a similar quasi-molecular ion at m/z 745.16 (C 31 H 37 O 19 S), were deduced as the sulfated products of M4, and the sulfation sites of M4-24b, 24a, and 24c were at the C-2′ position, one of the hydroxyls of the CA moiety, and one of the hydroxyls of the HT moiety, respectively.

Acetylated Metabolites (+ C 2 H 2 O, +42 Da)
M1-13a (eluted at 30.41 min), M1-13b (eluted at 24.73 min), and M1-13c (eluted at 29.52 min) exhibited the same molecular formula of C 37 H 47 O 20 (m/z 811.26) and was 42 Da (+C 2 H 2 O) higher than M1. M1-13a, M1-13b, and M1-13c were determined to be the acetylated products of M1. In the MS/MS spectra, M1-13a, M1-13b, and M1-13c showed different fragment ions, indicating that M1-13a, M1-13b, and M1-13c were acetylated at different sites of the structure. In the MS/MS spectra of M1-13a, the fragment ion at m/z 769.26 was formed by a loss of C 2 H 2 O from the quasi-molecular ion. Additionally, other fragment ions from M1-13a were demonstrated (m/z of 607.22, 179.03, 161.02, and 153.06), which were consistent with the characteristic fragment ions of M1, thus indicating that the acetylation site was at the C-2′ position. In the MS/MS spectra of M1-13c, the fragment ions at m/z 203.03 (C 11 H 7 O 4 ) and 177.06 (C 10 H 9 O 3 ) were presumed to be due to a loss of H 2 O and CO 2 from the acetylated CA (C 11 H 9 O 5 ) and other characteristic fragment ions of CA (m/z 179.03, 161.02, and 135.04) formed by a loss of C 2 H 2 O, indicating that the acetylation site of M1-13c was at one of the hydroxyls of the CA moiety. In the MS/MS spectra of M1-13b, the presence of fragment ion at m/z 195.07 (C 10 H 9 O 4 ) and other characteristic fragment ions of CA (m/z 179.03, 161.02, and 135.04) also suggested that the acetylation site of M1-13b was at one of the hydroxyls of the HT moiety. Similarly, M2-27 or isomers [m/z 827.26 (C 37 H 47 O 21 -16 [m/z 545.19 (C 24 H 33 O 14 ), eluted at 15.41 min] were also deduced as the acetylated C-2′ position product corresponding to the detected degradation products of M2 and M4, respectively. M2-28 [m/z 845.27271 (C 37 H 49 O 22 ), eluted at 17.92 min] was 18 Da (+H 2 O) higher than M2-27 or isomers, indicating that M2-28 was the hydrated product of M2-27 or isomers. In the MS/MS spectrum of M2-28, the fragment ion at m/z 827.26 was formed by a loss of H 2 O from the quasi-molecular ion. Additionally, the MS/MS spectrum showed a series of fragment ions at m/z 665. 23, 623.22, 477.16, 315.11, 179.03, and 153.06, which were the same as the characteristic fragment ions of M2-27 or isomers. Thus, M2-28 could be deduced as the hydrated CA product of M2-27 M2-29a and M2-29b [m/z 907.22 (C 37 H 47 O 24 S), eluted at 24.41 and 20.84 min, respectively] were 80 Da (+SO 3 ) higher than M2-27 or isomers, indicating that M2-29a and M2-29b were the sulfated products of M2-27 or isomers. In the MS/MS spectra of M2-29a, the fragment ion at m/z 215.00 (C 8 H 7 O 5 S) presumed to be a CO 2 loss from the sulfated CA [m/z 258.99 (C 9 H 7 O 7 S)], indicating that M2-29a was one of the sulfated products of M2-27 and the sulfation site was at the CA moiety. In the MS/MS spectra of M2-29b, the fragment ion at m/z 215.00 (C 8 H 7 O 5 S) was presumed to be a H 2 O loss from sulfated HT [m/z 233.01 (C 8 H 9 O 6 S)] and confirmed that the sulfation site was at the HT moiety. M2-30 [m/z 843.25507 (C 37 H 47 O 22

Intestinal Microbial Metabolism Characterization of Four PhGs
Based on the above-identified metabolites, it was shown that PhGs go through extensive metabolism by intestinal microbiota in vitro, leading to low oral bioavailability. The main proposed metabolite pathways are shown in Figure 6, indicating that degradation was the main pathway, along with reduction, acetylation, sulfate conjugation, hydroxylation, methylation, dihydroxylation, and hydration. Particularly, the degradation pathways occurring in glycosidic or ester bonds could be simplified by reference on the mass fragmentation patterns.
Employing the effective data processing approach described above, the direct match and transformation-shifted fragments were rapidly identified and automatically annotated using the predicted fragmentation patterns and spectral annotation functions, thus facilitating elucidations on structures and localizations on metabolite sites. In the MS/MS spectra, our results suggested that the α′, β′-double bond and the phenolic hydroxyl group of the CA or HT moiety are the main metabolic sites, and the fragmentation patterns also have some regularities after being metabolized in the transformed moieties. Apart from the neutral losses and series of diagnostic ions corresponding to transformed CA or HT, other identical fragment ions were formed by the same fragmentation patterns as the parent compounds (see the section Metabolite Identification of four PhGs by Rat Intestinal Bacteria). A detailed summary of the information pertaining to metabolism is in Table  S2. Using these fragmentation features, we rapidly and confidently identified more PhG-related metabolites or isomers. This study extended the current understanding of PhG metabolisms.

Antioxidant Effect Evaluation Using the DPPH Assay in vitro and Structure-Antioxidant Activity Relationships of Typical PhGs and Related Metabolites
It is worth noting that PhGs have been shown to possess effective antioxidant activity. The degradation products of PhGs, especially CA, HT, and the derivatives, are also known potent antioxidants with many beneficial effects on human health (Selma et al., 2009;Peyrol et al., 2017;Alson et al., 2018;Shahidi and Yeo, 2018;Wu et al., 2018). The intestinal microbial metabolism characterizations provide valuable information to understand the efficacy and mechanism of PhG action and trace the potential therapeutic effects of metabolites. To better understand the mechanism of PhG action in vitro, SAR studies are warranted to identify the structural features essential for the biological activities and screen for more active components.
In the present study, antioxidant activity evaluations on a) PhGs, b) CA-related metabolites, and c) HT-related metabolites were evaluated using the DPPH assay (Table 1), and some SARs were obtained (Figure 7). The DPPH radical scavenging activities of the tested compounds were expressed as an IC 50 value, which is the effective compound concentration that resulted in 50% of scavenged radicals. Lower IC 50 values indicate stronger antioxidant activities.
Based on the structures and IC 50 values of these compounds, we theorize the following: 1) With respect to HT or CA and the related further metabolites, degradation products, such as HT, CA, and the reduction product of CA {3,4-dihydroxyphenylpropionic acid [dihydrocaffeic acid (DCA)]}, show the strongest radical scavenging activity (1.312 ± 0.06 to 1.601 ± 0.08 µg/mL) compared to trolox (VE) (2.638 ± 0.1 µg/mL), indicating that reduction on the α′, β′-double bond has little effect on the activity. Methylation at one of the hydroxyls resulted in an approximate 2.5-fold loss of activity among ferulic acid and isoferulic acid. In addition, because of dehydroxylation at one of the hydroxyls, the radical scavenging activity decreased significantly among p-tyrosol, m-hydroxycinnamic acid, p-hydroxycinnamic acid, 3-hydroxyphenylpropionic acid, and 3-(4-hydroxyphenyl) propionic acid (IC 50 > 100 µg/mL). Cinnamic acid and hydrocinnamic acid are inactive (IC 50 > 1,000 µg/mL), indicating that the ortho-dihydroxyphenyl structure in the HT and CA group is critical for the effects of DPPH-scavenging and antioxidants. 2) HT or CA glycosylated to PhGs reduce activity. Regarding PhGs, trisaccharide glycoside showed weaker activity than disaccharide glycoside, indicating that the sugar groups result in lower radicalscavenging activity due to steric hindrance, but the type of substituted sugar group has little effect on the activity. In addition, the radical scavenging activity significantly increased and the hydroxyl groups on the C-2′ of the β-glucopyranose were acetylated and the CA moiety linked to C-4′ of the central Glu was stronger than the moieties linked to C-6′, referring to 2′-acetylacteoside with the lowest IC 50 (1.155 ± 0.05 µg/mL), followed by tubuloside B, acteoside, isoacteoside, tubuloside A, poliumoside, and echinacoside (2.469 ± 0.1 to 4.799 ± 0.2 μg/mL), and close to the positive control [trolox (VE); 2.638 ± 0.1 µg/mL], while osmanthuside B (11.103 ± 0.6 µg/mL) and saildroside (> 100 µg/mL) had only one hydroxyl group in the CA or HT group, exhibiting weaker activity compared to the tested PhGs and the other positive control [L-ascorbic acid (VC); 5.292 ± 0.5 µg/mL]. The SAR study may provide new insight into the development of lead compounds with enhanced pharmacologic activity; however, further studies are warranted to elucidate antioxidant activity in vivo, as well as the mechanism of action related to pharmacologic activity.

CONCLUSIONS
In summary, systematic research on intestinal microbemediated metabolism and pharmacologic activities of four typical bioactive PhGs has been reported for the first time. The metabolite profiles using a rat intestinal bacteria incubation system, including a total of 26, 42, 42, and 36 metabolites, were characterized using UHPLC-Q-Exactive Orbitrap HRMS with automated data analysis software (Compound Discoverer). The accurate masses of precursor and fragment ions, automatic structural annotation, and transformation localization provided in this process showed more confidence in the metabolic characterization.
More metabolites were first identified in the current study, revealing that PhGs go through extensive metabolism, mainly involving deglycosylation, reduction, hydroxylation, hydration, methylation acetylation, and sulfate conjugation. The main metabolic sites were the α′, β′-double bond and phenolic hydroxyl group of the CA or HT moiety, and characteristic fragmentation patterns of PhGs and their metabolites were summarized in detail. Moreover, antioxidant activity evaluations focusing on both parent compounds and the metabolites were compared with the rapid DPPH-based assay in vitro and explored in SAR studies for the first time. Some metabolites still had potent antioxidant activity (superior DPPH radical scavenging capacities than VE and VC), which provides a scientific basis on elucidation of effective forms of PhGs with poor bioavailability by oral administration. In this review, we showed evidence of interactions with gut microbiota that PhG conversion into active and bioavailable metabolites, which help understand the dietary health effects of PhGs. Furthermore, SAR studies suggested that ortho-dihydroxyphenyl, C-2′-acetyl, and steric hindrance have an influence on the activity; thus, active metabolites formed by hydroxylation, deglycosylation, and C-2′ acetylation. Above all, it seemed that intestinal microbe-mediated metabolism plays an important role in regulating not only the pharmacokinetic but also the pharmacologic effects of PhGs. Therefore, the proposed integrative strategy was powerful for further exploration of the role of intestinal bacteria in the metabolism and mechanism of action underlying other natural products.

DATA AVAILABILITY
The raw data supporting the conclusions of this manuscript will be made available by the authors, without undue reservation, to any qualified researcher.

AUTHOR CONTRIBUTIONS
YS and XW participated in research design. XW, XC, XL, MS, RX, and JC were responsible for performing the experiments. XW, YD, and YS performed data analysis. XW and YS contributed to