The Polyphenolic Profile and Antioxidant Activity of Five Vegetal Extracts with Hepatoprotective Potential

Oxidative stress is among the major triggers for many important human functional disorders, which often lead to various metabolic or tissue diseases. The aim of the study is to obtain five standardized vegetal extracts (Cynarae extractum—CE, Rosmarini extractum—RE, Taraxaci extractum—TE, Cichorii extractum—CHE, and Agrimoniae extractum—AE) that contain active principles with an essential role in protecting liver cells against free radicals and quantify their antioxidant actions. The compounds of therapeutic interest from the analyzed extracts were identified and quantified using the UHPLC–HRMS/MS technique. Thus, the resulting identified compounds were 28 compounds in CE, 48 compounds in RE, 39 compounds in TE, 43 compounds in CHE, and 31 compounds in AE. These compounds belong to the class of flavonoids, isoflavones, phenolic acids and dicarboxylic acids, depsides, diterpenes, triterpenes, sesquiterpenes, proanthocyanidins, or coumarin derivatives. From the major polyphenolic compounds quantified in all the extracts analyzed by UHPLC–HRMS/MS, considerable amounts have been found for chlorogenic acid (619.8 µg/g extract for TE–2032.4 µg/g extract for AE), rutoside (105.1 µg/g extract for RE–1724.7 µg/g extract for AE), kaempferol (243 µg/g extract for CHE–2028.4 µg/g extract for CE), and for naringenin (383 µg/g extract for CHE–1375.8 µg/g extract for AE). The quantitative chemical analysis showed the highest content of total phenolic acids for AE (24.1528 ± 1.1936 g chlorogenic acid/100 g dry extract), the highest concentration of flavones for RE (6.0847 ± 0.3025 g rutoside/100 g dry extract), and the richest extract in total polyphenols with 31.7017 ± 1.2211 g tannic acid equivalent/100 g dry extract for AE. Several methods (DPPH, ABTS, and FRAP) have been used to determine the in vitro total antioxidant activity of the extracts to evaluate their free radical scavenging ability, influenced by the identified compounds. As a result, the correlation between the content of the polyphenolic compounds and the antioxidant effect of the extracts has been demonstrated. Statistically significant differences were found when comparing the antiradical capacity within the study groups. Although all the analyzed extracts showed good IC50 values, which may explain their antihepatotoxic effects, the highest antioxidant activity was obtained for Agrimoniae extractum (IC50ABTS = 0.0147 mg/mL) and the lowest antioxidant activity was obtained for Cynarae extractum (IC50ABTS = 0.1588 mg/mL). Furthermore, the hepatoprotective potential was evaluated in silico by predicting the interactions between the determined phytochemicals and key molecular targets relevant to liver disease pathophysiology. Finally, the evaluation of the pharmacognostic and phytochemical properties of the studied extracts validates their use as adjuvants in phytotherapy, as they reduce oxidative stress and toxin accumulation and thus exert a hepatoprotective effect at the cellular level.


Preparation of Vegetal Extracts
The obtained extracts were in the form of dry powders, with a uniform dispersion by homogenization, and they sum up the organoleptic particularities (color, smell, and taste) specific to the plant source from which they were obtained. The extraction yield varied depending on the vegetable raw material used and the concentration of alcohol used for the extraction. Thus, the yield was 12.54% for CE (Cynarae extractum), 12.84% for RE (Rosmarini extractum), 24.31% for TE (Taraxaci extractum), 12.51% for CHE (Cichorii extractum), and 13.23% for AE (Agrimoniae extractum).

Quantitative and Qualitative Chemical Analysis
The results obtained from the quantitative chemical determinations are presented in Table 1. Total phenolic acid content (TPA), total flavonoid content (TF), total phenolic content (TP). Results were expressed as Mean ± SD (n = 5).
The quantitative chemical analysis showed that the extracts have a variable amount of secondary metabolites. AE is the richest in TPA, total phenolic acid content (expressed in chlorogenic acid), with a concentration of 24.1528 ± 1.1936 g phenolic acids/100 g dry extract, while CE has a concentration of only 1.7389 ± 0.0904 g chlorogenic acid/100 g dry extract. It was also observed that RE, TE, and CHE have appreciable amounts in these derivatives. Regarding the total flavonoid content, RE has the highest concentration of 6.0847 ± 0.3025 g flavones/100 g dry extract expressed in rutoside equivalents and TE has the lowest concentration of only 1.9019 ± 0.1080 g rutoside/100 g dry extract. The total phenolic content varies between 5.7627 ± 0.6946 g tannic acid equivalent for CE and  31.7017 ± 1.2211 g tannic acid equivalent for AE, which is the richest in total polyphenols, with a strong antioxidant role (Table 1).

UHPLC-HRMS/MS Analysis
The spectrum of the identified and quantified compounds is quantitatively dependent on the type of the extract used, the identified compounds being in the class of flavonoids, phenolcarboxylic acids, and sesquiterpene derivatives. The analysis showed that 28 compounds were identified in CE ( Table 2, Figures 1   flavonoids, phenolcarboxylic acids, and sesquiterpene derivatives. The analysis showed that 28 compounds were identified in CE (  Figure S1), 48 compounds in RE (   According to our results, rutoside was identified in all extracts except AE; apigenin and kaempferol in all types of extracts; chlorogenic acid and caffeic acid in all extracts except CHE (where only chlorogenic acid was identified with one of its isomers); vitexin was identified in CE, TE, CHE, and AE; cynarine, scolimoside, and cynaropicrin were identified in CE; cichorin in CE, RE, and CHE; cynaroside and cynarotrioside in TE, CHE, and AE; catechin and epicatechin in CHE and AE; azelaic acid in all extracts; biochanin A in RE, CHE, and AE; genistin and daidzin in CHE; genistin in TE and AE; and condensed procyanidins (proanthocyanidins) in TE, CHE, and AE.      The dendrogram of the rosemary extract's chemical composition (Supplementary Materials Figure S5) reveals the disposition of the identified compounds in five clusters. The first cluster contains a group of 12 compounds of the flavonoid class (the heteroside forms); the second one contains a group of 29 compounds represented by flavonoid aglycones and glycosides, phenolic acids, and derivatives with an isoflavan nucleus; the third one contains a group of 5 compounds; and the fourth and fifth ones contain 1 compound each.      According to our results, rutoside was identified in all extracts except AE; apigenin and kaempferol in all types of extracts; chlorogenic acid and caffeic acid in all extracts except CHE (where only chlorogenic acid was identified with one of its isomers); vitexin was identified in CE, TE, CHE, and AE; cynarine, scolimoside, and cynaropicrin were identified in CE; cichorin in CE, RE, and CHE; cynaroside and cynarotrioside in TE, CHE, and AE; catechin and epicatechin in CHE and AE; azelaic acid in all extracts; biochanin A in RE, CHE, and AE; genistin and daidzin in CHE; genistin in TE and AE; and condensed procyanidins (proanthocyanidins) in TE, CHE, and AE.
The artichoke extract composition's dendrogram ( Figure 2) shows the disposition of the identified compounds in four clusters. The first cluster contains a group of 9 compounds of the flavonoid class, the second one contains a group of 14 compounds represented by flavonoid aglycones (in this group there are 2 diterpene derivatives-carnosol and rosmanol, and 1 sesquiterpene compound-cynaropicrin), the third one contains a group of phenolic compounds such as chlorogenic acid and 2 coumarin derivatives, and the last one contains 2 phenolic acids-caffeic and azelaic acid.
From a quantitative point of view, 13 compounds were quantified in CE, 17 in RE, 18 in TE, 19 in CHE, and 19 in AE (Table 7).   The dendrogram of the dandelion extract's chemical composition (Supplementary Materials Figure S8) shows the disposition of the identified compounds in five clusters. The first cluster contains a group of 15 compounds from the flavonoid class, the second one contains 16 compounds represented by flavonoid aglycones or glycosides and phenolic acids, the third one contains a group of 3 compounds, and the fourth and fifth ones contain compounds from the category of phenolic acids.
The agrimony extract dendrogram (Supplementary Materials Figure S14) shows the disposition of the identified compounds in five clusters, the most representative being the third one, in which 16 compounds of the polyphenol class are grouped.
From a quantitative point of view, 13 compounds were quantified in CE, 17 in RE, 18 in TE, 19 in CHE, and 19 in AE (Table 7). A horizontal analysis of the results shows that AE is rich in catechin (11,854.8 µg/g extract), while appreciable amounts (but half the concentration for AE) were also quantified in CHE (5885.7 µg/g extract) and TE (5883.4 µg/g extract). Additionally, we observed that caffeic acid was not quantifiable in CHE; genistin, an isoflavone, was quantified in appreciable amounts in AE (5514.9 µg/g extract); chlorogenic acid was quantified in all the types of extract, with concentrations ranging from 619.8 µg/g TE extract to 2032.4 µg/g extract in AE; the hyperoside was quantified in an appreciable amount in AE (15,431.3 µg/g extract), but was not quantified in CE and RE; apigenin, rutoside, ellagic acid, and naringenin were quantified in different amounts in all the types of extract; and quercetol was predominant in AE (4958.05 µg/g extract).

Antioxidant Activity
The antioxidant effects induced by the tested extracts were directly correlated with the concentration of secondary metabolites (Table 1). Comparing the obtained results, it was found that the most intense antioxidant actions were induced by AE (lowest IC50 value by all three methods, compared to the other extracts), which is justified by the higher content of polyphenols in this type of extract (Table 8). It can also be noted that, of all the extracts analyzed, the IC50/EC50 values obtained for AE are much closer to the antioxidant values of the standard used (ascorbic acid), which accentuates the superior antioxidant action of AE compared to the other samples. All the analyzed extracts contain significant amounts of total polyphenols, with high concentrations for CHE, RE, and AE, and moderate concentrations for CE and TE. In addition, high values of phenolcarboxylic acids were identified in their composition: high concentrations for RE and AE and moderate concentrations for the other extracts (Table 8). Therefore, it is particularly important to evaluate the correlation between the total polyphenol content and the antioxidant effect of the extracts and the correlation between TPA and antioxidant action.
Given that the antioxidant effect was determined by three methods (DPPH, ABTS, and FRAP), it was analyzed using One-way ANOVA if there were statistically significant differences between the results of the antioxidant action obtained by the three methods (Supplementary Materials Table S1). It was observed that there are no statistically significant differences at the 95% significance threshold between the three methods applied to evaluate the antioxidant effect (p = 0.1170, p > 0.05- Table 9). The data in Supplementary Materials Table S2 show the correlation between the methodologies used in this study (DPPH, ABTS, and FRAP) for the evaluation of the antioxidant action in the analyzed extracts. The Pearson coefficient (r) calculated for each set of data pairs obtained by the different methods shows values above 0.900, which presents a very strong correlation between the experimentally applied methodologies (ABTS vs. DPPH: r = 0.995, p = 0.000; ABTS vs. FRAP: r = 0.964, p = 0.008; DPPH vs. FRAP: r = 0.982, p = 0.003).
The boxplot diagram showing the distribution and spread of IC50 values between different methods is presented in Supplementary Materials Figure S15.
Besides revealing a very well-correlated antioxidant activity of the extracts, the Pearson coefficient (r > 0.900) as well as the coefficients of determination (R 2 > 0.900) show that the results are not significantly influenced by methodological errors or other interferences related to the principles used in the determinations (Table 10). Subsequently, the statistically significant differences within the same method were analyzed by dispersion analysis. For the DPPH method, the results obtained by statistical validation using One-Way ANOVA showed statistically significant differences (p < 0.05; Table 11) between the free radical reduction effects of the plant extracts. After comparing the averages between groups for the DPPH method with the post hoc tests, statistically significant differences were observed, expressed by the p value as follows: Cynara vs. Rosmarinus  Table S3). The boxplot diagram showing the distribution and spread of the DPPH data set for every vegetal extract group is presented in Supplementary Materials Figure S16.
For the ABTS method, a test was conducted to determine whether the difference between the groups occurs randomly or is statistically significant, and it was observed that there are statistically significant differences (p < 0.05; Table 12) when comparing the antioxidant capacities of the studied plant extracts (One-Way ANOVA). Thus, after performing the Games-Howell post hoc test for unequal variances, it was found that there was a statistically significant difference between: Cynara vs. Agrimonia   Table S4 from the Supplementary Materials details the p values for each comparison between the analyzed groups, as well as the degree of significance.
The boxplot diagram showing the distribution and spread of the ABTS data set for every vegetal extract group is presented in Supplementary Materials Figure S17.
For the FRAP method, the Tukey HSD post hoc test (for equal variances) showed that there are statistically significant differences between the analyzed extracts (p < 0.05; Table 13), in terms of obtained optical density.  The values of the coefficients of determination (R 2 ) are very high (91.3936%, 86.4900%, 95.8441%), which shows the association between the analyzed data and emphasizes the strong link between the polyphenol content of the extracts and the anti-radical effect exerted by them in the human body (Table 14). Thus, at least 86% of the variation in the antioxidant effect of the studied extracts is explained by the quantified polyphenol concentration in the plant extracts, demonstrating a direct influence on the chelating capacity of free radicals.  The Pearson correlation analysis also indicates that there is a statistically strong and inverse correlation between the concentration of phenolcarboxylic acids and the antioxidant effect of the plant extracts; as the IC50 value is lower, the TPA concentration is higher, and the extract is a better antioxidant (negative Pearson coefficient).
These results explain the special antiradical action of the studied plant extracts, as well as their annihilation power against reactive oxygen species at the tissue and cellular levels (TPA vs. DPPH: r = −0.973, p = 0.005, TPA vs. ABTS: r = −0.980, p = 0.003, TPA vs. FRAP: r = −0.965, p = 0.008). The results of the Pearson correlation analysis are presented in Supplementary Materials Table S7. The determination coefficient R 2 is mathematically related to the Pearson correlation coefficient (r) and shows the magnitude of the association between the antioxidant action of the extracts evaluated by the three methodologies and the total phenolic acid concentration for each plant extract (Table 15).

Molecular Docking
A total of 23 identified polyphenolic compounds were subjected to molecular docking simulations to predict their binding affinities and molecular interactions with potential targets involved in hepatoprotection. The docking protocol was successfully validated by redocking the positive controls into the active sites, with the redocked ligands showing only slight variations in pose conformations. The CYP2E1 inhibitor showed a binding energy of -7.776 kcal/mol and 0.1704 Å RMSD after superposition on the initial conformation. The TNF-α inhibitor exhibited a docking score of -8.867 kcal/mol and 0.1325 Å RMSD, while redocking the GPx4 allosteric activator yielded a binding energy of -6.978 kcal/mol and 0.2691 Å RMSD (Supplementary Materials Figure S19).
Docking results for the screened polyphenols are shown in Table 16. Out of the 23 docked compounds, 2 ligands did not fit into the active site of CYP2E1, and the binding affinity could not be determined due to their large molecular volume (naringin and rutin). The binding energies after docking on CYP2E1 ranged from -8.942 kcal/mol to -0.884 kcal/mol, with a mean value of -7.128 ± 1.855 kcal/mol. Pinocembrin showed the highest binding affinity, while hyperoside exhibited the lowest affinity for CYP2E1. The highest ligand efficiency was observed for cinnamic acid (0.7417, -8.159 kcal/mol). Ten polyphenolic compounds showed higher affinities than the positive control (pinocembrin, chrysin, apigenin, formononetin, epicatechin, naringenin, catechin, cinnamic acid, p-coumaric acid, and abscisic acid). The binding energies for TNF-α ranged from -8.998 to -5.673 kcal/mol, with a mean value of -7.382 ± 0.998 kcal/mol. The lowest binding energy was observed for rutin, and the highest energy for syringic acid. Moreover, cinnamic acid showed the highest ligand efficacy for TNF-α as well (0.5436, -5.980 kcal/mol). Only one screened compound exhibited higher binding affinities than the positive control (rutin), while three polyphenols showed slightly lower but comparable affinities (genistin, hyperoside, and naringin).
The molecular docking experiment targeted at the GPx4 allosteric binding site yielded binding energies ranging between -7.023 and -4.637 kcal/mol (-6.198 ± 0.789 kcal/mol). Naringin showed the highest affinity for the GPx4 allosteric site, while syringic acid showed the lowest affinity. Only two ligands had binding energies lower than the positive control (naringin and genistin), while six other compounds had similar predicted activity values (apigenin, rutin, chlorogenic acid, hyperoside, epicatechin, and kaempferol).
The predicted molecular interactions are further discussed for the top scoring phytochemicals on each target. We chose to comment on the interactions between CYP2E1 and cinnamic acid since this particular compound showed a strikingly high ligand efficiency. Cinnamic acid is involved in hydrogen bonding with Asn206 through its carboxylic moiety. The protein-ligand complex is further stabilized by a carbon-hydrogen bond with Val239, pi-pi stacked interactions with Phe298, and van der Waals interactions with nine other residues within the active site ( Figure 8A,B).
Rutin showed the highest predicted binding affinity for the TNF-α binding site. Rutin acted as a hydrogen bond donor for four residues (Ser60, Gln61, Tyr119) through several hydroxyl groups and formed a carbon-hydrogen bond with Leu120 ( Figure 8C,D). Moreover, nonpolar interactions such as pi-alkyl (Tyr59) and van der Waals interactions were also responsible for binding to the active site. Apigenin had the third highest binding affinity for GPx4 among the screened phytochemicals, also showing a good ligand efficiency value. The binding potential of apigenin onto the GPx4 allosteric binding site is supported by a hydrogen bond with Met102, 2 pi-anion interactions with Asp21 and Asp23, 2 pi-alkyl interactions with Val27 and Lyes90, and 11 van der Waals interactions ( Figure 8E,F).
Moreover, nonpolar interactions such as pi-alkyl (Tyr59) and van der Waals inte were also responsible for binding to the active site. Apigenin had the third highest affinity for GPx4 among the screened phytochemicals, also showing a good efficiency value. The binding potential of apigenin onto the GPx4 allosteric bindi supported by a hydrogen bond with Met102, 2 pi-anion interactions with As Asp23, 2 pi-alkyl interactions with Val27 and Lyes90, and 11 van der Waals inte ( Figure 8E,F).

Discussion
In the present paper, five types of plant extracts were analyzed: Cynarae extractum, Rosmarini extractum, Taraxaci extractum, Cichorii extractum, and Agrimoniae extractum. The qualitative chemical profile (UHPLC-HRMS/MS method) and quantitative analysis (spectrophotometric and UHPLC-HRMS/MS methods) were established. The antioxidant activity was determined in the acellular system (DPPH, ABTS, and FRAP), and docking studies were performed to predict the potential hepatoprotective action induced by the extracts. The chemical profile is dependent on the type of plant raw material from which the extracts were obtained, so the most ennobled extract in polyphenolic compounds was Agrimoniae extractum.
Qualitative and quantitative chemical analyses were used for the integrated characterization and comparative analysis of the polyphenolic profile and in vitro antioxidant action.
The polyphenol content in plant extracts is directly correlated with their antioxidant action. The large variety of chemical compounds identified in the study by the UHPLC-HRMS/MS technique emphasizes the need to evaluate the antioxidant action of plant extracts using at least three different methods of determination for the comparative analysis to be accurate.
Several phenolic compounds with high antioxidant activity were identified in the vegetal extracts using the UHPLC-HRMS/MS method. Most of them, mainly phenolic acids such as chlororgenic acid, p-coumaric acid, and ferulic acid, were identified in all the analyzed extracts.
A Hierarchical Cluster Analysis (HCA) was performed as a statistical technique that identifies groups of samples that behave similarly or show similar characteristics and thus quantifies the structural characteristics of the samples or variables. The procedure of hierarchical clustering involves the construction of a hierarchy of treelike structures. As it uses a hierarchical configuration-a tree called a dendrogram-to structure the data, cluster analysis is a quantitative form of classification in which classes are subclassified into groups [48,49].
We implemented the hierarchical clustering approach to identify, in each extract, the groups of identified compounds in relation to the response variables: exact mass, adduct ion (m/z)/monitored negative ion, and retention times.
The cluster analysis performed on the five plant extracts showed a heterogeneous distribution in a variable number of clusters for each type of extract; the results obtained were in direct correlation with the compounds identified by the UHPLC-HRMS/MS method. Thus, in CE, the 28 identified compounds were distributed into four groups; in RE, TE, and AE, although a variable number of compounds were identified, they were distributed into five groups each; and in CHE, where 43 compounds were identified, they were distributed into six clusters. The distributed compounds belong to the class of flavonoids, isoflavones, depsides, diterpenes, triterpenes, sesquiterpenes, phenolic acids, proanthocyanidins, and coumarin derivatives.
Furthermore, the comparative analysis of the antioxidant capacity of some vegetal extracts was assessed. Different antiradical activities were obtained for each extract due to the different types and content of their polyphenolic compounds and other phytochemical constituents. According to the results, the order of IC50 values in this assay was AE < RE < CHE < TE < CE. Thus, considering that a lower IC50 value signifies a higher total antioxidant activity, agrimony extract was the most potent antioxidant of all, followed by rosemary, chicory, dandelion, and artichoke.
The statistical analysis performed on the data sets, but also the statistically significant p values (p < 0.05) for the antioxidant results obtained by dispersion analysis within the same method of determination, support the obvious differences between the antiradical effects of the studied extracts.
Moreover, after calculating the Pearson coefficient (denoted r) and the coefficient of determination (denoted R 2 ), a very strong correlation can be demonstrated not only between different antioxidant methodologies but also between the active principles and total antioxidant activity of the extracts. The obtained results for the phenolcarboxylic acids are very close to the ideal scientific value of 100% (R 2 = 94.6729%, R 2 = 96.0400%, R 2 = 93.1225%). This confirms the existence of a relevant association between the data, and emphasizes, together with the Pearson correlation analysis, the interdependence of the data sets and the conclusion of this study. Although the two coefficients are calculated differently, they describe the relationship between data sets and measure the magnitude of the relationship between the analyzed data. In all determinations, the Pearson correlation coefficient values are greater than 0.900, which reveals the highest correlation between the antioxidant activity of the extracts determined using three different methodologies and analyzed by pairs of data. As the coefficient of determination expressed as a percentage can have a major impact and can also be more informative (intuitively) than other statistical measures, it is more robust for poor data set matches and is therefore more accurate.
The antioxidant activity of the analyzed extracts, associated with a high polyphenolic content, could explain their hepatoprotective potential. Some important elements link hepatotoxicity and oxidative stress, particularly the production of reactive oxygen species and other reactive species, which can damage hepatocyte integrity and functionality and contribute to the pathogenesis of many liver diseases.
The antioxidant activity and hepatoprotective potential for the studied herbal products and extracts were confirmed in other studies as well. In a clinical study reported in the literature, the benefits of artichoke extract supplementation were shown in patients with metabolic syndrome, where a significant decrease in LDL cholesterol was observed, based on the antioxidant effects induced by polyphenolic compounds [50]. In addition, an important decrease in malondialdehyde (MDA) levels in the liver was observed in preclinical studies [51].
According to the studies of Lombardo et al. (2010) [52], and Nouraei et al. (2018) [53], the concentration of polyphenols in Cynara scolymus is influenced by the harvest time, species genotype, soil, and climatic conditions [12]. According to the studies reported by Allahdadi M., and Farzaneh P. in 2018, the use of soil fertilizers leads to a decrease in the content of TPA, TF, and TP in artichokes [54]. We assume that the low concentration of total polyphenols in the extract obtained in our study is due to an inadequate collection of the plant product. These statements are based on the results obtained during the evaluation of the plant's raw materials.
The compounds quantified in CE plant extracts fit within the profile of polyphenols cited in the literature [55,56]. The identification of pinocembrin in large quantities (43.1 µg/g) in RE is consistent with the literature data, as this compound was also identified in rosemary honey [57,58]. The chemical profile of the compounds identified in TE is also consistent with studies reported in the literature [59,60].
According to Gordon's studies [61], the polyphenolic derivatives of rosemary extract act as primary antioxidants, annihilating lipid and hydroxyl radicals. Fang et al. [62] consider that they may act as chelating agents, especially on metal ions such as Fe 2+ , thereby reducing the generation of free radicals.
Carnosic acid and carnosol act as potent peroxyl radical scavengers since they are strong inhibitors of lipid peroxidation in liposomal and microsomal systems. Thus, carnosic acid specifically scavenges H 2 O 2 and can also act as a substrate for peroxidase. Additionally, rosmanol has been reported to induce an antioxidant effect four times higher than various synthetic compounds used in tests as a reference for its antioxidative effectiveness (BHTbutylhydroxytoluene, BHA-butylhydroxyanisole) [22].
Ivanov [63] shows that the presence of sinapic and chicoric acid, correlated with the high number of polyphenols, justifies the good antioxidant effect of the dandelion extract. Atef et al. [64] highlighted the role that an aqueous extract obtained from chicory played in increasing the level of reduced glutathione (GSH) in damaged liver tissue, following CCl 4 poisoning in experimental animal models. According to the studies of Correia et al. [65], Ivanova et al. [40], and Santos et al. [66], Agrimonia extracts have a significant capacity to capture reactive oxygen species (ROS), an effect which is correlated with an increased polyphenol content.
A previous study [67] regarding the hepatoprotective and radical scavenging activity of rosemary on alcoholic liver disease reported the remarkable capacity of Rosmarinus officinalis to diminish the level of serum hepatic enzymes such as ALT (alanine aminotransferase) and ACP (acid phosphatase). The hepatoprotective effect in the rat model of alcoholic liver disease is due to its mechanism of action and the antioxidant effect, which leads to the reduction of liver tissue damage.
According to other scientific studies, dandelion showed a significant hepatoprotective action in animal studies, which was validated by the reduction of all the determined liver markers. These findings support its effective antioxidant activity in toxin-induced oxidative stress or damage [68].
In addition, the administration of chicory extract in experimental rats demonstrated hepatoprotective and antihepatotoxic effects against dexamethasone-induced alterations. These effects are due to the content of polyphenols and other antioxidants that can reduce the activity of liver enzymes [69].
Furthermore, the superior antioxidant activity of agrimony over artichoke and other vegetal extracts is stated in Kuczmannová et al.'s work [70]. This article suggests a relationship between the hepatoprotective effects and the antioxidant properties induced by polyphenols, which are the dominant compounds in the extract.
Free radical scavenging activity is by no means the sole molecular mechanism responsible for the hepatoprotective effects of some phytochemicals. The hepatoprotection exerted by plant extracts can be achieved by the interaction of phytoconstituents with biological targets that either promote or alleviate toxicity at the hepatocellular level. For instance, inhibition of CYP2E1 activity impairs the conversion of xenobiotics and polyunsaturated fatty acids to toxic metabolites that can promote liver damage [45]. Inflammation and oxidative stress can be alleviated by down-regulating the expression and activity levels of cytokines (TNF-α, interleukins) or by up-regulating key enzymes with antioxidant and anti-inflammatory roles, such as GPx4 [46,47]. Moreover, gomisin A, a bioactive compound found in Schisandra chinensis, was reported to ameliorate carbon tetrachlorideinduced hepatotoxicity in rats by inhibiting the activation of nuclear factor kappa B (NF-κB), down-regulating inducible nitric oxide synthase (iNOS) expression, and reducing fibrogenesis [71].
The hepatoprotective potential of the studied herbal extracts can be supported not only by the aforementioned findings but also by our molecular docking results. Some of the detected polyphenols exhibited satisfying predicted binding affinities and ligand efficiencies for the selected biological targets (CYP2E1, TNF-α, and GPx4). Therefore, the phytochemical constituents have the potential to exert hepatoprotective effects by either inhibiting cytochrome P450 2E1 isoform and tumor necrosis factor alpha or activating glutathione peroxidase 4 through an allosteric mechanism. The docking results of the screened phytochemicals can also be correlated with previously published studies. For instance, chrysin, catechin, and apigenin have been shown to inhibit CYP2E1 activity and improve redox balance, while cinnamic acid and syringic acid have been reported to ameliorate hepatotoxicity in laboratory animal studies [72,73]. Other studies revealed that catechin, hyperoside, rutin, genistein, and naringin suppress TNF-α mediated inflammatory responses in various in vitro and in vivo experiments [74][75][76][77][78]. Furthermore, another study highlighted that apigenin activates glutathione peroxidase 4 and alleviates ferroptosis and oxidative stress, which is in concordance with our molecular docking results [79]. Furthermore, it has been shown that naringenin alleviates myocardial ischemia/reperfusion injury by regulating the Nrf2/System xc-/GPx4 axis [80], while rutin induces glutathione peroxidase activity in cadmium-induced oxidative stress [81].
Following the in vitro determination of the antioxidant effect of the studied extracts, the extracts' in vivo action against oxidative stress at the cellular and tissue level can be quantified in future research. by Chromeleon 7.2 Software (Thermo Fisher Scientific, Waltham, MA and Dionex Softron GMbH Part of Thermo Fisher Scientific, Germany).

MS parameters
A HESI (Heated ElectroSpray Ionization) ion source was used for the ionization in the negative mode. The ion source parameters were optimized as follows: the nitrogen as sheath and the auxiliary gas flow rate were set to 8 and 6 units, respectively. The source heater temperature was set to 300 • C, the capillary temperature was set to 300 • C, the auxiliary gas heater temperature was set to 300 • C, the electrospray voltage was set to 2800 V, and the S-lens RF level was set to 50.
The full-scan HRMS analysis of the compounds was performed using a Q-Exactive Mass Spectrometer. Full-scan data in negative mode were acquired at a resolving power of 70,000 FWHM at m/z 200. A scan range of m/z 100-1000 Da was chosen; the Automatic Gain Control (AGC) was set at 3 × 10 6 and the injection time was set to 200 ms. Scan-rate was set at 2 scan/s. External calibration was performed by the calibration solution in positive and negative mode.
For structural information, a vDIA (variable Data Independent Acquisition) approach was selected for untargeted HRMS/MS analysis. A total of six scan events were combined: one full scan event with the mentioned parameters and five MS-MS events. In the MS2 scan events, the precursor ion ranges from m/z 95-205, 195-305, 295-405, 395-505, and 500-10,005 were consecutively selected, fragmented in the HCD cell, and measured in five separate Orbitrap scans at a resolving power of 35,000 FWHM. The fragmentation events were performed at 30, 60, and 80 NCE (Normalized Collision Energy). The C-trap parameters for all scan events were as follows: the Automatic Gain Control (AGC) target was 1 × 10 6 and the injection time was 100 ms.
Data were processed by the Quan/Qual Browser Xcalibur 2.3 (Thermo Fisher Scientific, Germany). The mass tolerance window was set to 5 ppm.
Validation parameters of the LC-HRMS analytical method are presented in Supplementary Materials Table S8.
In MS-MS analysis, detection of at least two fragment ions was performed by comparing them to the reference standards. For those compounds without available references, the structures were presumed based on high-accuracy analysis of deprotonated precursors and fragment ions of specific components. The chemical elemental composition for each target peak was assigned within a mass error of 2 ppm using the chemical Chemspider database  Table S9).
The optimization of the UHPLC and MS conditions and the validation of the quantitative method are presented in the Supplementary Materials.

Determination of Antioxidant Activity (AA) 4.2.1. DPPH Free Radical Scavenging Activity
The free radical scavenging activity of the plant extracts was measured by using the diphenyl-picrylhydrazyl (DPPH) assay, according to Çelik S.E., where the antioxidant activity is influenced by both the characteristics of the substrate and the polarity of the solvent [90]. Equal amounts of 0.25 g dry extracts were dissolved in 25 mL of 50% ethanol for CE, RE, AE, and CHE, and 20% ethanol for TE, depending on the solvent used for the formulation of each dry extract. Volumes of 10 µL, 20 µL, 30 µL, 40 µL, 50 µL, 60 µL, 70 µL, 80 µL, 90 µL, and 100 µL of each obtained solution were poured into 10 mL volumetric flasks and adjusted to 10 mL by adding the same solvent as above. An amount of 0.5 mL of each dilute solution was mixed with 3 mL of 0.1 mM DPPH radical solution (Sigma-Aldrich, Germany) [91]. The solutions were held in the dark for 30 min, and the absorbance was then measured at 515 nm using a spectrophotometer (Jasco, Japan). Ascorbic acid (Sigma-Aldrich, Germany) was used as a reference for the calibration curve in the range of concentration between 2-22 µg/mL (Supplementary Materials Figures S23 and S24).
The percentage of inhibition of DPPH • was calculated using the formula below: ref. [92], where: A (blank) = blank absorbance of 0.1 mM DPPH solution in the absence of extracts (1.00 ± 0.10); A (sample) = sample absorbance of the DPPH solution in the presence of extracts after 30 min.
Based on the established values, inhibition curves (%) were constructed depending on the concentration (mg/mL). Using the linear equations, the IC50 values (mg/mL) were determined for each extract (for the value y = 50).

ABTS Method of Total Antioxidant Capacity Assessment
The ABTS assay is considered one of the most sensitive techniques to measure the antioxidant activity of both hydrophilic and lipophilic compounds due to the response of antioxidants involving faster reaction kinetics in a pH-independent manner [93,94].
The ABTS radical cation (ABTS •+ ) was obtained by reacting the ABTS (Sigma-Aldrich, Germany) 7.4 mM solution with 2.6 mM potassium persulfate (K 2 S 2 O 8 -Sigma-Aldrich, Germany) and keeping the mixture in the dark at room temperature for 16 h before use [95].
Equal amounts of 0.25 g dry extracts were dissolved in 25 mL of 50% ethanol for CE, RE, AE, and CHE, or 20% ethanol for TE, depending on the solvent used for the formulation of each dry extract. Volumes of 10 µL, 20 µL, 30 µL, 40 µL, 50 µL, 60 µL, 70 µL, 80 µL, 90 µL, and 100 µL of each obtained solution were poured into volumetric flasks and adjusted to 10 mL by adding the same solvent as above. An amount of 0.5 mL of each dilute solution was mixed with 3 mL of ABTS •+ solution diluted with ethanol (Sigma-Aldrich, Germany). The solutions were stirred and held in the dark for 6 min [96]. The absorbance was then measured at 734 nm, relative to absolute ethanol, using a spectrophotometer (Jasco, Tokyo, Japan).
The percentage of inhibition of ABTS •+ was calculated using the formula below: A (t = 0 min) = absorbance of the blank sample (ABTS •+ sol in the absence of tested compounds: 0.70 ± 0.02); A (t = 6 min) = absorbance of the vegetal extract (ABTS •+ sol in the presence of tested compounds). IC50 value, the concentration of sample required to scavenge 50% of the ABTS •+ free radical, was calculated from the plotted graph of radical scavenging activity against the concentration of extracts (IC-inhibition). The lower the IC50 value for an extract, the stronger the antioxidant activity.

Antioxidant Activity Using FRAP Assay (Ferric Reducing Antioxidant Power Assay)
The ferric reducing power of plant extracts was determined using a modified FRAP assay [94]. The assay measures the antioxidant potential through the reduction of ferric iron (Fe 3+ ) to ferrous iron (Fe 2+ ) by antioxidants present in the samples. Following the reduction of ferric iron (Fe 3+ ) to ferrous iron (Fe 2+ ), a blue color develops.
Equal amounts of 0.25 g dry extracts were dissolved in 25 mL of 50% ethanol for CE, RE, AE, and CHE, or 20% ethanol for TE, depending on the solvent used for the formulation of each dry extract. Volumes of 50 µL, 60 µL, 70 µL, 80 µL, 90 µL, 100 µL, 200 µL, 300 µL, 400 µL, and 500 µL of each obtained solution were poured into volumetric flasks and adjusted to 10 mL by adding the same solvent as above. An amount of 2.5 mL of each dilute solution was mixed with phosphate buffer pH 6.6 (Sigma-Aldrich, Germany) and 2.5 mL of K 3 (FeCN) 6 1% (Sigma-Aldrich, Germany) before being heated to 50 • C for 20 min. Then, 2.5 mL of trichloroacetic acid (Sigma-Aldrich, Germany) was added to each sample. To 2.5 mL of each of the resulting solutions, 2.5 mL of distilled water and 0.5 mL of 0.1% FeCl 3 (Sigma-Aldrich, Germany) were added, then left to stand for 10 min. The change in the absorbance at 700 nm was measured relative to a blank sample obtained by mixing 5 mL of distilled water with 0.5 mL of FeCl 3 0.1%.
The antioxidant capacity was determined using the EC50 value (mg/mL) as the concentration of the solutions at which the absorbance has a value of 0.5 (at half the antioxidant effect; EC-effective).
Due to the variability of plant properties and the nonuniformity of pharmacognostic profiles of vegetal extracts, different extract volumes were tested to reach the absorbance value of 0.5 (the more the experimental values obtained are around the point to be determined (EC50 for y = 0.5), the more accurate its approximation is). The optimized values have been set as above to perform a relevant comparative analysis within the same method and between different methods of assessing the antioxidant capacity.
The SMILES codes of the selected phytochemicals for docking were retrieved from Pub-Chem database. The ligand preparation protocol consisted of the generation of corresponding 3D structures with DataWarrior 5.2.1 [104], energy minimization using MMFF94s+ force-field, and protonation at pH 7.4. Phytochemicals and positive controls were docked using AutoDock Vina v1.1.2 [105] algorithm within YASARA. The search space (25 × 25 × 25 Å) was centered around the co-crystallized ligands within the binding sites and 12 docking runs were performed for each ligand.
Docking results were retrieved as the binding energy or docking score (∆G, kcal/mol) and ligand efficiency (LE, ∆G\no. of heavy atoms) of the best binding pose for each compound. The conformations of the predicted protein-ligand complexes and molecular interactions were analyzed using BIOVIA Discovery Studio Visualizer (BIOVIA, Discovery Studio Visualizer, Version 17.2.0, Dassault Systèmes, 2016, San Diego, CA, USA).

Statistical Analysis
All statistical analysis was performed using IBM SPSS Statistics software version 28.0 (IBM Corporation, Armonk, NY, USA). For each set of experimental data, the essential conditions for the application of statistical tests were evaluated, such as the normality of data and the homogeneity of variances. The normal distribution of the data was assessed by the Shapiro-Wilk test and by histograms. To detect significant differences between the data groups, One-Way ANOVA (Single-factor ANOVA) and Post Hoc Tests were applied for mean comparison: Tukey HSD for equal variances, or Games-Howell for unequal variances, depending on data distribution, sampling dispersion, and the number of studied groups. The Levene's test was used to verify the homogeneity of variances for the experimental data sets. The Welch ANOVA analysis was run as a robust test when the condition of the homogeneity of variances was violated. When certain experimental data did not follow a Gaussian distribution, they were transformed (by logarithm in base 10), so that they became normally distributed and could be subjected to statistical tests. The correlation between certain groups of analyzed experimental data was also established by calculating the Pearson correlation coefficient. Interpretations were made after the mandatory application criteria were met (continuity of variables, independence of measurements, normality and linearity of data, and absence of outliers). In all cases, the significance level was set at 0.05. When p < 0.05, the obtained results were considered statistically significant.
Clustering analysis was also performed. The purpose of a clustering analysis is to construct groups or classes with two properties; within a group, observations behave similarly, and there must be dissimilarities between the observations of two distinct groups [106]. We implemented the hierarchical clustering approach to identify in each extract the homogeneous groups of compounds in relation to the response variables of exact mass, adduct ion (m/z)/monitored negative ion, and retention time. Since the variables have different units of measurement, we first performed a scaling and then calculated the distance based on the "average" method. The optimal number of groups was identified by the "silhouette" method.

Conclusions
In this study, the polyphenolic profile and in vitro antioxidant activity of Cynarae extractum, Rosmarini extractum, Taraxaci extractum, Cichorii extractum, and Agrimoniae extractum were characterized and compared. According to our results, all the vegetal extracts exert antioxidant properties due to their different content of phytochemicals, with agrimony extract being the most potent antioxidant of all. Nevertheless, due to their synergistic effect, all the studied extracts can be combined into herbal formulas for future applications in the field of phytomedicine to prevent hepatic disorders and protect human liver cells against oxidative damage and inflammation.
Future research will focus on determining the cytotoxicity of the studied vegetal extracts and performing preclinical tests on experimental animals to highlight the in vivo effects. Moreover, the impact on key biochemical parameters will be quantified by applying histological tests to liver tissue.