LC-MS Profile, Gastrointestinal and Gut Microbiota Stability and Antioxidant Activity of Rhodiola rosea Herb Metabolites: A Comparative Study with Subterranean Organs

Golden root (Rhodiola rosea L., Crassulaceae) is a famous medical plant with a one-sided history of scientific interest in the roots and rhizomes as sources of bioactive compounds, unlike the herb, which has not been studied extensively. To address this deficiency, we used high-performance liquid chromatography with diode array and electrospray triple quadrupole mass detection for comparative qualitative and quantitative analysis of the metabolic profiles of Rhodiola rosea organs before and after gastrointestinal digestion in simulated conditions together with various biochemical assays to determine antioxidant properties of the extracts and selected compounds. R. rosea organs showed 146 compounds, including galloyl O-glucosides, catechins, procyanidins, simple phenolics, phenethyl alcohol derivatives, (hydroxy)cinnamates, hydroxynitrile glucosides, monoterpene O-glucosides, and flavonol O-glycosides, most of them for the first time in the species. The organ-specific distribution of compounds found for catechins, procyanidins, and cinnamyl alcohols and glucosides was typical for underground organs and flavonoids and galloylated glucoses concentrated in the herb. Extracts from rhizomes, leaves and flowers showed high phenolic content and were effective scavengers of free radicals (2,2-diphenyl-1-picrylhydrazyl (DPPH•), 2,2′-azino-bis(3-ethylbenzothiazoline-6-sulfonic acid) (ABTS•+), O2•−, •OH) and protected β-carotene in a bleaching assay. Digestion in the gastric and intestine phase influenced the composition of R. rosea extracts negatively, affecting the content of catechins, procyanidins, and galloyl glucoses, and therefore, the antioxidativity level. After gut microbiota treatment, the antioxidant capacity of rhizome extract was lower than leaves and flowers due to the aglycone composition found in the colonic phase of digestion. Our study demonstrated that the herb of R. rosea is a rich source of metabolites with high antioxidant properties and could be a valuable plant for new bioactive products.

This paper aims to estimate the chemical and biomedical prospects of R. rosea herb as a possible future remedy. Thus, we realized the first comparative analysis of the metabolic profiles of R. rosea organs using the high-performance liquid chromatography with diode array and electrospray triple  [17,22]. Equipment used for UV-Vis spectrophotometry was SF-2000 UV-Vis-spectrophotometer (OKB Specter, St. Petersburg, Russia).

Total Extract Preparation from R. rosea Organs
For the preparation of the total extract of R. rosea organs, the dry and powdered sample of the organ (100 g) was extracted twice with stirring in a glass flask (2 L) with 70% methanol (1 L) using an ultrasonic bath (80 min, 50 • C, ultrasound power 100 W, frequency 35 kHz). The extracts were passed through a cellulose filter, concentrated under reduced pressure until dryness, and stored at 4 • C before using for the chemical analysis of biological activity study. The yields of total extracts of R. rosea were 32.5 g (roots), 22.5 g (rhizome), 20.0 g (leaves), 25.0 g (flowers), 9.0 g (stems).

High-Performance Liquid Chromatography with Diode Array Detection and Electrospray Ionization Triple Quadrupole Mass Spectrometric Detection (HPLC-DAD-ESI-QQQ-MS)
Reversed-phase high-performance liquid chromatography with diode array detection and electrospray ionization triple quadrupole mass spectrometric detection (HPLC-DAD-ESI-QQQ-MS) was used for phenolic profiling. Experiments were performed on an LCMS 8050 liquid chromatograph coupled with diode-array-detector and triple-quadrupole electrospray ionization detector (Shimadzu, Columbia, MD, USA) coupled with GLC Mastro C18 column (150 × 2.1 mm, Ø 3 µm; Shimadzu, Kyoto, Japan) at the column temperature 35 • C. Eluent A was 0.5% formic acid in water and eluent B was 0.5% formic acid in acetonitrile. The injection volume was 1 µL, and elution flow was 100 µL/min. B. The DAD acquisitions were performed in the range of 200-600 nm. MS detection was performed in negative and positive ESI mode using the parameters as follows: temperature levels of ESI interface, desolvation line, and heat block were 300 • C, 250 • C, and 400 • C, respectively. The flow levels of nebulizing gas (N2), heating gas (air) and collision-induced dissociation gas (Ar) were 3 L/min, 10 L/min and 0.3 mL/min, respectively. The MS spectra were recorded in negative (-3--5 kV source voltage) and positive mode (+3-+4 kV source voltage) by scanning in the range of m/z 100-1900 at the collision energy of 5-40 eV. The system was operated under LabSolutions workstation software with the internal LC-MS library. The identification of compounds was done by the analysis of their retention time, ultraviolet, and mass-spectrometric data comparing the same parameters with the reference samples and/or literature data.

Flavonol O-Glycosides Hydrolysis and HPLC-DAD-ESI-QQQ-MS Analysis
Powdered total extract of R. rosea organs (100 mg) was mixed with acetone (20 mL) and hydrochloric acid (20%; 7 mL), and boiled under a reflux condenser for 20 min. The mixture was filtered through absorbent cotton into the separating funnel, mixed with water (20 mL) and extracted with diethyl ether (5 × 20 mL). The ether layer was dried by filtration over anhydrous sodium sulfate and concentrated in vaccuo until dryness. The dry residue was dissolved in methanol (5 mL), filtered through a

HPLC-MS Quantification
Quantification of compounds 1-146 was realized using HPLC-MS data (MS peak area) in conditions described in Section 2.5. To prepare the stock solutions of reference compounds, 44 standards (Tables S1 and S2) were accurately weighed (10 mg) and individually dissolved in methanol in a volumetric flask (10 mL). The external standard calibration curve was generated using six data points, 100, 50, 25, 10, 5, 1 µg/mL. The calibration curves were created by plotting the MS peak area vs. the concentration levels and the validation criteria (correlation coefficients, r 2 ; standard deviation, S YX ; limits of detection, LOD; limits of quantification, LOQ; linear ranges) was calculated using the previous recommendations [29] ( Table S2). All the analyses were carried out in triplicate and the data were expressed as mean value ± standard deviation (SD). For the preparation of sample solution, an accurately weighted powdered plant of R. rosea organs (100 mg) or total extract of R. rosea organs (40 mg) were placed in an Eppendorf tube, 2 mL of 70% methanol was added. Then the sample was extracted twice in an ultrasonic bath for 30 min at 30 • C and centrifuged (3000× g, 15 min). Combined supernatants were transferred to the volumetric flask (5 mL) and the final volume was reduced to 5 mL. The resultant extract was filtered through a 0.22 µm PTFE syringe filter before injection into the HPLC system for analysis. Caffeine (final concentration 500 µg/mL in acetonitrile) and benzoic acid (final concentration 250 µg/mL in acetonitrile) were used as the internal standards for analysis of Sep-Pak C 18 eluates as well as picein (final concentration 250 µg/mL in methanol) for polyamide eluates I, scopoletin-7-O-neohesperidoside (final concentration 250 µg/mL in 40% methanol) and isorhamnetin (final concentration 125 µg/mL in methanol) for polyamide eluates II, and 4-O-caffeoylquinic acid (final concentration 200 µg/mL in methanol) for polyamide eluates III.

Simulated Gastrointestinal Digestion and Gut Microbiota Incubation
The assays previously described used to simulate gastrointestinal digestion [32] and gut microbiota incubation [33]. For the simulation, the samples of R. rosea dry extracts (rhizomes, leaves, flowers extracts; 500 mg) were incubated with simulated gastric fluid [32] (25 mL, pH 2.0) in a shaking water bath (37 • C, 167 rpm, 60 min) followed by neutralization (1 M NaOH) up to pH 7.0 (gastric phase) and HPLC-DAD analysis (Section 2.9, conditions HPLC-DAD-1; samples were filtered through 0.22 µm syringe filters before injection into the HPLC system). To prepare simulated gastric fluid, aliquots of 61.0 mL NaCl (200.0 g/L), 11.7 mL NaH 2 PO 4 (88.8 g/L), 35.8 mL KCl (89.6 g/L), 70.0 mL CaCl 2 ·2H 2 O (22.2 g/L), 39.0 mL NH 4 Cl (30.6 g/L), and 32.5 mL HCl (37%) were mixed in a volumetric flask and the total volume was adjusted to 250 mL by distilled water. Then the solution was supplemented by HCl up to pH 2.0 (solution I). The simulated gastric fluid was prepared before use by mixing pepsin (400 mg; Sigma-Aldrich; 3200-4500 units/mg protein) with the 25 mL of solution I (stored at 4 • C).
The gastric fluid treated sample was transferred to the dialysis bag and mixed with bile solution (1 mL) and simulated intestinal fluid (4 mL) [32] and continuously stirred (4 h) in a clear simulated intestinal fluid without of pancreatin addiction (1000 mL, pH 7.0, 37 • C) (intestinal phase). Non-dilalyzed retentate after 4 h incubation used for HPLC-DAD analysis after filtration (Section 2. The retentate after intestinal phase subjected to lyophilic drying and the dry residue dissolved in 5 mL of distilled water and neutralized if necessary. The resultant solution (5 mL) was mixed with fecal slurries (1 g in 5 mL of brain heart infusion) donated by healthy volunteers as described previously [33] and pure brain heart infusion (15 mL). The samples were incubated under anaerobic conditions (37 • C, 48 h) in BD GasPak TM EZ anaerobe container system sachets (New Jersey, NJ, USA) then centrifuged (6000× g, 20 min), mixed with acetonitrile (1:1), passed through 0.22 µm syringe filters and analyzed by HPLC-DAD assay (Section 2.9, conditions HPLC-DAD-2). The 13 reference standards for HPLC-MS quantification (Section 2.7) were used for HPLC-DAD quantifying after calibration curve generating (Tables S1 and S3). All the analyses were carried out in triplicate and the data were expressed as mean value ± standard deviation (SD). The samples of total extracts of R. rosea organs treated with gastric and intestinal fluids were chromatographed without any pretreatment after filtering through a 0.22 µm PTFE syringe.

Incubation of R. rosea Extracts with Digestive Enzymes Mixture
The sample of R. rosea extract from rhizome or flowers (500 mg) was dissolved in water (25 mL), mixed with pepsin (400 mg) and pancreatin (40 mg) and incubated in a shaking water bath (37 • C, 167 rpm, 5 h). An aliquot (1 mL) of the incubated mixture was vigorously shaken with 2 mL of acetonitrile in the Eppendorf tube than filtered through a 0.22 µm PTFE syringe filter and analyzed using HPLC-DAD assay (conditions 1; Section 2.10).

Trolox-Equivalent Content in Simulated Gastric, Intestinal and Gut Microbiota Media
The Trolox-equivalent content in digestive media and gut microbiota medium was found using bromine radical scavenging assay based on the coulometric titration method with electrogenerated bromine radicals [31]. The measurements were carried out using Expert-006 potentiostat (Econics Expert Ltd., Moscow, Russia) with four-electrode two-compartment electrochemical cell. A bare platinum foil with 1 cm 2 surface area was used as the working electrode, and a platinum wire separated from the anodic compartment with a semipermeable diaphragm-as the auxiliary electrode. The time of titration was used for the total antioxidant capacity calculation that was expressed in units of the quantity of electricity (Coulombs (C)) spent for titration of the full probe of digestive media. The reference compound Trolox solutions (500, 250, 100, 50, 10 µg/mL in methanol) was titrated coulometrically, and a calibration curve was plotted in coordinates "concentration (µg/mL)-the quantity of electricity (C)". Finally, the value of Trolox-equivalent content was calculated as mg Trolox equivalents per probe. Values are expressed as mean obtained from five independent experiments. The gastric, intestinal, and gut microbiota fluids (media) showed zero or traces Trolox-equivalent content.

Statistical Analysis
Statistical analyses were performed using a one-way analysis of variance (ANOVA), and the significance of the mean difference was determined by Duncan's multiple range test. Differences at p < 0.05 were considered statistically significant. The results are presented as mean values ± SD (standard deviations) of the three-five replicates.

Phenolic Composition and Antioxidant Activity of Rhodiola rosea Extracts
To better understand if there was any prospect for studies and practical application of R. rosea herb compared to traditional rhodiola subterranean organs, we must undertake a reassessment of our knowledge of the chemistry and bioactivity of R. rosea underground roots and rhizomes and aerial parts (leaves, flowers and stems). In this study, we will focus mostly on the phenolic compounds in R. rosea due to the greater scientific weight and a better understanding of their mode of action. Seven groups of phenolics were chosen and quantitatively analysed ( Table 3) based on known Rhodiola plants phenolome data [2]. Non-trace levels were found for flavonoids, catechins, procyanidins, phenylpropanoids and gallotannins; ellagitannins and anthocyanins were treated as non-essential. The total phenolic content was high in all R. rosea organs, especially in rhizomes (140.60 mg/g), leaves (122.98 mg/g), and flowers (112.79 mg/g) followed by roots (44.87 mg/g) and stems (21.84 mg/g). Flavonoids were at the highest level in the aerial parts, such as flowers (46.36 mg/g) and leaves (16.71 mg/g), and lower in roots and rhizomes (0.75-1.89 mg/g). The present data did not confirm the earliest records of low flavonoid content (<2 mg/g in the dry plant) in samples of R. rosea herb, as opposed the high amount of flavonoids in roots and rhizomes (12-19 mg/g in the dry plant) [34,35]. In the absence of systematic data about biochemical features of the Siberian populations of R. rosea, it is possible to talk about the regional variation of flavonoid content resulting in conflicting data, particularly as that was previously described for the European populations of R. rosea [36,37].
Catechins were the most significant group of phenolics in underground organs, with 10.84 and 61.30 mg/g in roots and rhizomes, respectively. Biogenetically close to catechins, procyanidins showed high contents in rhizomes (31.37 mg/g) and roots (9.21 mg/g). The concentration of catechins and procyanidins in aerial organs was not more than 2 mg/g. There is no previous comparative data of catechins and procyanidins distribution in R. rosea plant except the information that catechin content in the root may vary from 4.6 mg/g in Polish samples [38] to 20 mg/g in samples of Indian origin [39].
Phenylpropanoids of R. rosea that are mostly derivatives of cinnamyl alcohol (rosavins) and salidroside [2] were highest in rhizomes (46.45 mg/g) and roots (21.89 mg/g) and lowest in leaves (8.27 mg/g), stems (5.11 mg/g), and flowers (1.14 mg/g). In early research, rhizomes of R. rosea traditionally enriched by phenylpropanoids rosavins at 2.5-3.5 times more than roots [36] and aerial parts showed no detectable amounts of salidroside and rosavins [37]. The variation of phenylpropanoid content in roots may reach from nil in Russian and Chinese samples to 86 mg/g in Norwegian plants [40].
Gallotannins, quantified in R. rosea for the first time, were the basic phenolic group in leaves (97.53 mg/g), flowers (63.11 mg/g) and stems (12.35 mg/g). The roots and rhizomes contents of gallotannins were low (0.73-1.04 mg/g).
Already at this stage of our study, it seems clear that underground organs tend to concentrate catechins, procyanidins, and phenylpropanoids while aerial organs accumulated flavonoids and gallotannins. These differences in organ chemistry would affect the bioactivity potentials of R. rosea extracts.
As referred to in many papers, the antioxidant potential of R. rosea roots and/or rhizome extracts and remedies are already well known [2,3]. We, therefore, compared the bioactivities of five rhodiola extracts derived from roots, rhizomes, flowers, leaves and stems. The antioxidant properties of mentioned extracts were studied in five assays; the scavenging capacity against 2,2-diphenyl-1picrylhydrazyl radical (DPPH • ), 2,2 -azino-bis(3-ethylbenzothiazoline-6-sulfonic acid) cation radical (ABTS •+ ), superoxide radical (O 2 •− ), hydroxyl radical ( • OH) and carotene bleaching assay and expressed activity were found in all cases (Table 4). Table 4. Antioxidant activity of R. rosea extracts. The reference compound trolox showed close or less power of activity. The most active extracts were from R. rosea rhizomes, leaves and flowers followed by the stem and root extracts. This trend is not surprising at all, the high content of known strong antioxidants, such as catechins, procyanidins, gallotannins, and flavonoids, in R. rosea extracts should have led to the same results, but in fairness, the high antioxidant potency of R. rosea herb is shown for the first time. However, the lack of precise data about metabolites of R. rosea herb, as well as the quantitative content and comparative assessment of metabolic profiles of underground organs and aerial parts of the plant makes it hard to bring to a close the issue of R. rosea antioxidants.

Procyanidins
The common satellite compounds to catechins are their oligomers, procyanidins, and they were found in roots, rhizomes, flowers, and stems of R. rosea; only leaves did not contain procyanidins. The mass spectrometric data were typical for n-flavan-3-ols [47] and suggested that the fourteen procyanidins found were oligomers of epigallocatechin and/or epigallocatechin gallate including: Previous data for the procyanidin profile of R. rosea rhizomes harvested in Norway gave close results [59], but roots, flowers and stems of R. rosea were studied for the first time.

Phenethyl Alcohol Derivatives
Tyrosol (p-hydroxyphenethyl alcohol) (50) and its O-glucoside salidroside (48) are the most frequently reported components of rhodiola plants. Tyrosol, first prepared in 1911 [65] and later found in R. rosea roots [48], is still the subject of many studies from different directions [66]. In our research, both compounds 48 and 50 were identified not only in roots and rhizomes but also in the aerial part, a previously unknown source. Compound 47, described as tyrosol O-hexoside-O-pentoside (m/z 431 [M − H] − →299, 137), was most likely an O-arabinosyl analogue of salidroside that was not yet characterized, but its O-methyl ester mongrhoside was found in R. rosea [67]. An unknown tyrosol O-desoxyhexoside (49) and three acylated derivatives of salidroside with one (65, 66) and two (67) acetyl groups were the components in the aerial part.

(Hydroxy)Cinnamates
Seventeen (hydroxyl)cinnamates of various structures were found in the whole R. rosea plant. Cinnamyl alcohol (83) and its typical for R. rosea O-glucosides, rosarin (70), rosavin (71) and rosin (73) were successfully identified using reference standards only in roots and rhizomes. Glucoside 72 was isomeric to 70 and 71 (m/z 427 [M − H] − ) and its most probable structure was cinnamyl alcohol O-(6 -O-xylosyl)-glucoside, the known R. rosea glucoside [52]. Two triglucosides of cinnamyl alcohol 68 and 69 from the underground part have close to 70-72 mass spectral behaviours but were 132 a.m.u. larger, pointing to additional pentosyl residue in their structures. The O-pentosyl derivatives of 70-72 are still unknown. Triandrin (sachaliside I) or p-hydroxycinnamyl alcohol O-glucoside (55) was described early as a component of roots [48] but its presence in the aerial part was shown for the first time. The O-pentoside of p-hydroxycinnamyl alcohol O-glucoside 54 in roots and rhizomes and two monoacetates 75 and 76 from the aerial part of R. rosea have no close structures in known plant metabolites. The known R. rosea hydroxycinnamates p-methoxycinnamyl alcohol (85) and its O-glucoside vimalin (82) [52], together with two p-methoxycinnamyl alcohol O-hexoside-O-pentosides 80 and 81, were discovered in roots and rhizomes. Two isomeric glucosides of p-methoxycinnamyl alcohol with (6"-O-arabinopyranosyl)-glucoside and (6"-O-arabinofuranosyl)-glucoside fragments were characterized in R. rosea previously [2,52]. Two acids, cinnamic acid (84) and 3-O-feruloylquinic acid (128), were first found in R. rosea organs. Hence, although the chemodiversity of R. rosea (hydroxyl)cinnamates is well studied, it is still possible to find new compounds in both roots and the herb.

Hydroxynitrile Glucosides
Three hydroxynitrile glucosides 57-59 were the components of R. rosea organs, including the known rhodiocyanoside A (59) [59] identified using the reference standard. The isomeric compound 58 with close chromatographic mobility was tentatively determined as rhodiocyanoside D (59), also found in R. rosea roots [68]. Both compounds were detected primarily in R. rosea herb. The roots/rhizome component 57 gave a deprotonated ion with m/z 390, demonstrating the presence of extra pentosyl units in 57 or 59 skeletons. The closest known analogue is rhodiocyanoside F from R. sacra roots [69].

Flavonol O-Glycosides
To separate the diversity of R. rosea flavonoids, we successfully used the polyamide solid-phase extraction technique to isolate two fractions of flavonol O-glycosides as neutral (with no uronide-and/or acyl-substituents) and acidic (uronide-and/or acyl-containing compounds). Finally, sixty flavonol O-glycosides (86-146) were found both in neutral (29 compounds; 86-114) and acidic (31 compounds; 115-146) flavonoid fractions. The HPLC-MS profiles also allowed concluding that the neutral flavonol O-glycosides were the trace components of root/rhizome and leaf extracts and dominated in the flowers extract opposite the acidic flavonol O-glycosides, which were the major flavonoids of root/rhizome and leaf extracts and minor for the flowers extract. It is also important to note that the previous data on R. rosea flavonoids refers to only the neutral flavonol O-glycosides [2,[13][14][15][53][54][55][56]70] (Table S7), so this is the first report describing the acidic flavonol O-glycosides as components of R. rosea.
The preliminary analysis of hydrolysis products of flavonoid O-glycosides indicated the presence of five flavonol aglycones in detectable levels ( Figure S5). Four compounds were identified using the reference standards as kaempferol, quercetin, herbacetin, and gossypetin and were the usual aglycones of R. rosea [13][14][15][53][54][55][56]. An additional minor flavonol with lower HPLC mobility than gossypetin was found in hydrolysates of leaf and flower extracts. The UV pattern of the minor flavonol aglycone was typical for 8-hydroxylated flavonols [13,15,57] and the compound also gave a deprotonated ion with m/z 333, showing its similarity to gossypetin with an extra-hydroxyl group, such as a hibiscetin (5 -hydroxy-gossypetin). Hibiscetin derivatives were not detected in Rhodiola genus previously but the systematically close genus Sedum contains hibiscetin in free [71] and glucosylated forms [72,73].
The features of UV spectra or spectral pattern of flavonoids were used to facilitate the identification and determine the aglycone substitution type (Tables S5 and S6). This was particularly true in the case of herbacetin and gossypetin di-, tri-, and tetra-glucosides that have many more numbers of substitutions than the kaempferol or quercetin derivatives.
Kaempferol O-glucosides.  Compounds 103, 111, and 114 were detected previously in R. rosea roots [2,54,56] and 101 was found in herb [15]. The unknown kaempferol glucosides were described Finally, sixty compounds were found in R. rosea, including 10 known flavonoids. Fifty compounds were described for R. rosea for the first time, of which about thirty members may be the new compounds that need additional study.

Comparative Chemodiversity of Various Organs of R. rosea
One hundred and forty-six compounds found in R. rosea demonstrated various distributions in organs. The total numbers of components detected were 69 in stems, 87 in leaves, 90 in roots and rhizomes, and 100 in flowers of R. rosea. The maximal numbers of known discovered compounds were from nil in stems to 25 in rhizomes; therefore, the amounts of newly found components were 65 in rhizomes, 69 in roots and stems, 84 in leaves, and 95 in flowers. This was the first time that HPLC-DAD-tQ-ESI-MS profiling of R. rosea metabolites had taken place. In this regard, it is also relevant to bear in mind a large number of compounds with potentially new structures were identified in present work. Is it possible to talk about the organ-specific distribution of compounds in R. rosea organs? It is possible in part because some groups of compounds were found only in separate plant parts. Good examples are the poly-galloyl glucoses (tetra-to octa-substituted) found only in the aerial part of R. rosea in contrast to cinnamyl alcohol O-glucosides and catechins, which were the typical markers of underground organs. In other cases, the organ-specificity of plant metabolites was not evident.
The application of quantitative analysis of chromatographic data demonstrated more clear deviations in the chemical profiles of R. rosea organs in which about 90 compounds were quantified ( Table 6, Table S8).
In the analysis process, among the metabolites in underground organs, the catechins and cinnamyl alcohols glucosides were the basic compounds of roots (10.58 and 19.89 mg/g) and rhizomes (57.40 and 55.76 mg/g) in contrast to aerial organs with zero to the low content of the mentioned groups. Marked levels of procyanidins (10.58-57.40 mg/g) and p-hydroxyphenethyl alcohol glucosides (1.97-20.12 mg/g) are also worthy of note for the underground organs. Typically, the levels of all compounds were higher in rhizomes, except flavonoids and galloylated glucoses with relatively more content in roots. The galloylated glucoses were found as phenolics with the highest content in R. rosea leaves (100.63 mg/g) and flowers (68.19 mg/g) and in both organs, poly-galloyl glucoses (penta, hexa, hepta) were predominant (12.21.02-37.79 mg/g). Flavonol glucosides were highest in flowers (55.55 mg/g) followed by leaves (26.85 mg/g) and stems (6.30 mg/g). The distribution of neutral and acidic types of flavonol glucosides in R. rosea herb was different with the domination of neutral compounds in flowers (42.88 mg/g vs. 12.67 mg/g) and acidic in leaves (26.85 mg/g vs. traces) and stems (4.56 mg/g vs. 1.74 mg/g). The basic flavonol derivatives were herbacetin glucosides (5.51-33.74 mg/g) and a smaller content was found for gossypetin glucosides (0.72-20.60 mg/g). Hydroquinone glucosides found in all plants showed maximal concentrations in flowers (24.04 mg/g) and leaves (16.62 mg/g) and lowest in the underground organs (4.64-8.87 mg/g). Hydroxynitrile glucosides and rosiridol glucosides were found in all organs as low to medium amount compounds. R. rosea underground (roots/rhizomes) and aboveground parts (herb) chemically are two different plant sources accumulating various groups of compounds. The diversity of metabolites in both R. rosea parts requires a wide spectrum of bioactivities and this has worked in the case of antioxidant activity of all R. rosea organs. However, due to the lack of scientific information about digestive and gut microbiota transformation of the majority of compounds found in rhodiolas, we studied the influence of the digestive processes on the chemical composition of R. rosea extracts that allow us to find possible bioavailable compounds with potential high bioactivity.

Stability of Metabolites and Antioxidant Activity of Rhodiola rosea Extracts in Simulated Gastrointestinal Digestion Model
Digestion is a complex physical and chemical process can result in changes in the metabolic profiles of plant remedies. The key impacts are linked to the enzyme-dependent depolymerization of oligo-and polymers, destruction of acid-or base-instable compounds, and formation of protein complexes with various molecules. Ultimately, these benefits would lead to drastically changed concentrations of the selected components and changes in the bioactivity power. The R. rosea extracts were no exceptions and demonstrated variation in the content of some compounds after simulated gastrointestinal digestion (Tables S9-S12). Changes were made mainly to the content of procyanidins and catechins; rhizome extract lost cc. 95% and 97% of these two groups, respectively (Table 7).
A similar phenomenon was also found in leaf and flower extracts, which showed a 90.20-94.17% decrease of galloyl glucoses content. The changes observed for the hydroquinone glucoside arbutin and p-hydroxyphenethyl alcohol glucoside salidroside was less pronounced and cinnamyl alcohol glucosides and flavonol glucosides were highly stable. Consequently, digestion affected mostly the tannin-like compounds, such as procyanidins, catechins and galloyl glucoses, that are relatively resistant in digestive pH variation but can form complexes with protein molecules, such as enzymes [78]. Chromatographic experiments of model mixtures of R. rosea extracts with enzymes confirmed the enzyme-caused "de-tannisation" of plant matrices, demonstrating the direct impact of digestive fluids on the selected R. rosea compounds ( Figure 2). A similar phenomenon was also found in leaf and flower extracts, which showed a 90.20-94.17% decrease of galloyl glucoses content. The changes observed for the hydroquinone glucoside arbutin and p-hydroxyphenethyl alcohol glucoside salidroside was less pronounced and cinnamyl alcohol glucosides and flavonol glucosides were highly stable.
Consequently, digestion affected mostly the tannin-like compounds, such as procyanidins, catechins and galloyl glucoses, that are relatively resistant in digestive pH variation but can form complexes with protein molecules, such as enzymes [78]. Chromatographic experiments of model mixtures of R. rosea extracts with enzymes confirmed the enzyme-caused "de-tannisation" of plant matrices, demonstrating the direct impact of digestive fluids on the selected R. rosea compounds ( Figure 2). In the face of a strong decline of phenolics in rhodiola extracts after gastrointestinal reactions, it would be logical to expect a reduction of bioactivity power. This emerges from a study on the antioxidant capacity of the extract after the gastric and intestinal stage of digestion, demonstrating the gradual decrease in trolox-equivalent content in reaction mixtures (Figure 3a). The rhizome extract showed antioxidant capacity level cc. 38% less after gastric phase than in the treated sample and the final greatest reduction at 80%. A close tendency was found for the leaf extract, with 53 and 74% activity decreases. The highest residual activity, at cc. 46% of the original level, was from the flower extract, indicating it had the best antioxidant potency. Obvious reasons for a dramatic fall of rhizome extract activity include the low residual content of antioxidants and the low activity of the residual compounds after two phases of gastrointestinal digestion. Most catechins and procyanidins, the basic radical scavengers in R. rosea rhizomes, were protein-bonded; in contrast, the hydroquinone glucosides (arbutin), p-hydroxyphenethyl alcohol glucosides (salidroside) and cinnamyl alcohol glucosides (rosavins) showed good concentrations but low activity (Figure 3b). In the case of flower extract, the residual level of the galloyl glucoses, as the strongest antioxidants of R. rosea herb, was extra low; however, there was the high content of herbacetin and gossypetin glucosides (rhodiolgidin, rhodionidin) that possessed good antioxidant activity (Figure 3b). The results thus obtained argued that the dynamic character of the mechanism of antioxidant protection of the R. rosea extracts changes in time and depending on the digestion phase. In the face of a strong decline of phenolics in rhodiola extracts after gastrointestinal reactions, it would be logical to expect a reduction of bioactivity power. This emerges from a study on the antioxidant capacity of the extract after the gastric and intestinal stage of digestion, demonstrating the gradual decrease in trolox-equivalent content in reaction mixtures (Figure 3a). The rhizome extract showed antioxidant capacity level cc. 38% less after gastric phase than in the treated sample and the final greatest reduction at 80%. A close tendency was found for the leaf extract, with 53 and 74% activity decreases. The highest residual activity, at cc. 46% of the original level, was from the flower extract, indicating it had the best antioxidant potency. Obvious reasons for a dramatic fall of rhizome extract activity include the low residual content of antioxidants and the low activity of the residual compounds after two phases of gastrointestinal digestion. Most catechins and procyanidins, the basic radical scavengers in R. rosea rhizomes, were protein-bonded; in contrast, the hydroquinone glucosides (arbutin), p-hydroxyphenethyl alcohol glucosides (salidroside) and cinnamyl alcohol glucosides (rosavins) showed good concentrations but low activity (Figure 3b). In the case of flower extract, the residual level of the galloyl glucoses, as the strongest antioxidants of R. rosea herb, was extra low; however, there was the high content of herbacetin and gossypetin glucosides (rhodiolgidin, rhodionidin) that possessed good antioxidant activity (Figure 3b). The results thus obtained argued that the dynamic character of the mechanism of antioxidant protection of the R. rosea extracts changes in time and depending on the digestion phase.  Groups: 1-initial group; 2-group after gastric phase; 3-group after intestinal phase. (b) Radical scavenging activity of selected compounds (as IC 50 , µM). 1-1,6-di-O-galloyl glucose; 2-1,2,3,4,6penta-O-galloyl glucose; 3-procyanidin B1 (as an example of procyanidin dimer); 4-procyanidin C1 (as an example of procyanidin trimer); 5-epicatechin; 6-epigallocatechin gallate; 7-gallocatechin gallate; 8-arbutin; 9-salidroside; 10-rosarin; 11-rosavin; 12-rosin; 13-rhodiolgidin; 14-rhodionidin; 15-herbacetin-3-O-glucoside-8-O-glucuronide; 16-trolox.

Gut Microbiota Metabolites of R. rosea Extracts and Their Antioxidant Activity
The colonic phase of digestion, as a final step of gastrointestinal transformation, involves gut microbiota, leading to the formation of metabolites absent in the original metabolome. The extracts of R. rosea were not exceptions and 24 h incubation of intestinal phase samples with gut bacteria from healthy volunteers induced strong changes in HPLC patterns. In all extracts studied, the simplification of metabolic profiles or aglyconation was observed, notably cinnamyl alcohol, tyrosol, and hydroquinone were detected in gut-treated rhizome extract but leaf and flower extracts gave herbacetin, gossypetin, tyrosol, and hydroquinone ( Figure 4). gossypetin were detected in leaf and flower extracts for the same reason. The aglyconation of glucosides during fermentation with human faecal flora was previously postulated as a primary mechanism of flavonoid, salidroside, and other compound utilization [79][80][81].
Distinct changes in the chemical composition of R. rosea extracts have had an impact on the antioxidant activity, as demonstrated in trolox-equivalents content variation in gut microbiota medium. The R. rosea rhizome extract lost a great amount of its antioxidant potential, showing only 20% of its intestine phase level, which accounted for 4% of the original value. The most obvious reason for the disastrous decline in the bioactivity of rhizome extract is a low index of antioxidant protection for the cinnamyl alcohol (IC50 > 200 μM) and tyrosol (IC50 > 200 μM). A radically different situation was found for leaf and flower extracts, which already have good antioxidant potential parameters that rise after incubation with the gut microbiome. The difference values of activity between intestinal and colonic phase levels were 40% (leaf extract) and 45% (flower extract), in favour of the latter. The gut-induced deglucosylation of R. rosea flavonol glucosides resulted in the removal of substituents from positions C-3, C-7, and C-8 and the formation of the more active aglycones herbacetin (IC50 = 28.63 ± 0.56 μM) and gossypetin (IC50 = 18.45 ± 0.35 μM) with free hydroxy-function. Thus, the colonic phase of digestion can influence the bioactivity of R. rosea extracts in different ways, depending on the original chemical profiles.
The tendency of plant phenolics to reduce content after incubation with simulated gastrointestinal fluids was demonstrated for gallotannins, catechins, procyanidins, and flavonol glycosides in several works. In early study, the members of gallotannins (pentagalloyl glucose) and catechins (epigallocatechin gallate) showed pH instability, self-association, and complex formation with digestive enzymes resulted to the lowering effect and poor recovery after digestion end [82]. Mono-, di-, and trigalloyl glucoses found in Terminalia and Emblica gave 3-16% degradation after the intestine phase of digestion [32]. The same instability was showed for the tetra-and pentagalloyl These results were attributable to intensive hydrolytic reactions in the gut microbiome resulting in the formation of aglycones from glycosidic compounds. The major compound in rhizome extract, cinnamyl alcohol, was a breakdown product of rosarin, rosavin, and rosin, as well as tyrosol and hydroquinone, which originated from salidroside and arbutin, respectively. Herbacetin and gossypetin were detected in leaf and flower extracts for the same reason. The aglyconation of glucosides during fermentation with human faecal flora was previously postulated as a primary mechanism of flavonoid, salidroside, and other compound utilization [79][80][81].
Distinct changes in the chemical composition of R. rosea extracts have had an impact on the antioxidant activity, as demonstrated in trolox-equivalents content variation in gut microbiota medium. The R. rosea rhizome extract lost a great amount of its antioxidant potential, showing only 20% of its intestine phase level, which accounted for 4% of the original value.
The most obvious reason for the disastrous decline in the bioactivity of rhizome extract is a low index of antioxidant protection for the cinnamyl alcohol (IC 50 > 200 µM) and tyrosol (IC 50 > 200 µM). A radically different situation was found for leaf and flower extracts, which already have good antioxidant potential parameters that rise after incubation with the gut microbiome. The difference values of activity between intestinal and colonic phase levels were 40% (leaf extract) and 45% (flower extract), in favour of the latter. The gut-induced deglucosylation of R. rosea flavonol glucosides resulted in the removal of substituents from positions C-3, C-7, and C-8 and the formation of the more active aglycones herbacetin (IC 50 = 28.63 ± 0.56 µM) and gossypetin (IC 50 = 18.45 ± 0.35 µM) with free hydroxy-function. Thus, the colonic phase of digestion can influence the bioactivity of R. rosea extracts in different ways, depending on the original chemical profiles.
The tendency of plant phenolics to reduce content after incubation with simulated gastrointestinal fluids was demonstrated for gallotannins, catechins, procyanidins, and flavonol glycosides in several works. In early study, the members of gallotannins (pentagalloyl glucose) and catechins (epigallocatechin gallate) showed pH instability, self-association, and complex formation with digestive enzymes resulted to the lowering effect and poor recovery after digestion end [82]. Mono-, di-, and trigalloyl glucoses found in Terminalia and Emblica gave 3-16% degradation after the intestine phase of digestion [32]. The same instability was showed for the tetra-and pentagalloyl glucoses of Myrciaria trunciflora during in vitro gastrointestinal digestion [83]. Some catechins (epigallocatechin gallate, epigallocatechin, epicatechin gallate) significantly degraded in in vitro digestion fluids and able to form homo-and heterocatechin dimers due to autoxidation [84]. Non-galloylated catechins as catechin and epicatechin were comparatively stable. Wine procyanidins react with salivary proteins that influenced negatively on the concentration of phenolics [85], and, also, the procyanidins may significantly degrade after pancreatic digestion to give low-molecular-weight compounds [86]. In the vast majority of cases, all parameters of antioxidant protection decline rapidly after the intestinal phase of simulation since a reduction of bioavailable antioxidants content in digestive fluids. Unlike previous compounds, the flavonol glycosides of green tea [87] or Ginkgo biloba [88] demonstrated good stability up to the colonic phase of digestion and can keep the good level of antioxidativity. The final stage of digestion realized in colonic conditions leading to the destruction of most gastrointestinally stable compounds. The flavonoid glycosides are not exceptional and usually gave corresponding aglycones as in the case of cosmosiin [89], baicalin [90], rutin, hesperidin, and naringin [79]. Considering the fact that flavonoid aglycones are the better antioxidants than the glycosides [91], the antioxidant potential of colonic fluid is expected to increase. Rhodiola rosea extracts demonstrating the close mechanism of chemical transformation and antioxidant activity variation during gastrointestinal and colonic digestion. This highlights the general path of digestive utilization of plant phenolic mixtures.

Conclusions
In this paper, the whole Rhodiola rosea plant was shown as a natural accumulator of diverse metabolites of phenolic and non-phenolic nature, not only in roots or rhizomes, but also in leaves, flowers and stems with an unexpectedly high content of galloyl glucoses, flavonol glycosides, and other compounds in the herbal parts of the plant. These rich phenolomes of roots/rhizomes and herb made it possible to express values of antioxidant activity and identify the aerial organs of R. rosea as a new source of antioxidants. In an attempt to understand the evolution of R. rosea metabolites and bioactivity during the digestion process, in simulated conditions, we found the decreasing tannin content in rhizome, leaves, and flower extracts caused by digestive enzyme precipitation and consequently, the decrease of antioxidant capacity. The further gut microbiota assisted in transforming of R. rosea glucosides generated aglycones that eventually lowered the antioxidant capacity of rhizome extract but increased the bioactivity of leaf and flower extracts. Thus, the chemical features of R. rosea extracts have various impacts on the bioactivity due to the different methods of digestive utilization. If asked about preferences between R. rosea roots or herb as good antioxidants, it should be noted that the herb extracts keep antioxidant properties longer during digestion, but additional data is needed to close the issue. This study demonstrated the good biomedical prospects of R. rosea herb as a future plant remedy or a source of new functional products.