Natural and Synthetic Xanthone Derivatives Counteract Oxidative Stress via Nrf2 Modulation in Inflamed Human Macrophages

Natural products have attracted attention due to their safety and potential effectiveness as anti-inflammatory drugs. Particularly, xanthones owning a unique 9H-xanthen-9-one scaffold, are endowed with a large diversity of medical applications, including antioxidant and anti-inflammatory activities, because their core accommodates a vast variety of substituents at different positions. Among others, α- and γ-mangostin are the major known xanthones purified from Garcinia mangostana with demonstrated anti-inflammatory and antioxidant effects by in vitro and in vivo modulation of the Nrf2 (nuclear factor erythroid-derived 2-like 2) pathway. However, the main mechanism of action of xanthones and their derivatives is still only partially disclosed, and further investigations are needed to improve their potential clinical outcomes. In this light, a library of xanthone derivatives was synthesized and biologically evaluated in vitro on human macrophages under pro-inflammatory conditions. Furthermore, structure–activity relationship (SAR) studies were performed by means of matched molecular pairs (MMPs). The data obtained revealed that the most promising compounds in terms of biocompatibility and counteraction of cytotoxicity are the ones that enhance the Nrf2 translocation, confirming a tight relationship between the xanthone scaffold and the Nrf2 activation as a sign of intracellular cell response towards oxidative stress and inflammation.


Introduction
Over the last decades, natural compounds have attracted strong interest, not only for their wide range of pharmacological applications but also for the possibility of chemical modification in order to improve their pharmacodynamics or pharmacokinetics. Among these molecules, the heat-stable xanthone scaffold provides advantageous therapeutic effects in different pathological conditions leading researchers to better investigate their potential [1,2]. The core nucleus is represented by the xanthene-9-one, which is characterized by different substitution patterns to elicit a wide range of biological activities, including antioxidant, anti-inflammatory, antimicrobial, and anti-cytotoxic responses. The antioxidant and anti-inflammatory effects have recently emerged as the most relevant, mainly for the treatment of skin inflammatory diseases [3]. However, the main mechanism of action of xanthones and their derivatives is still only partially disclosed. It has been reported that the biological effects of the xanthone core nucleus might be related to the modulation of various pro-inflammatory and anti-inflammatory cytokines as a sign of the recruitment of immune cells. In parallel, nuclear factor kappa-light-chain-enhancer of activated B cells (NF-κB), To investigate whether the polyphenolic and aminated synthesized xanthone derivatives have biological activity in an inflamed environment, the compounds were tested on LPS-stimulated macrophages. In parallel, cells were pre-incubated and exposed to BSO to trigger oxidative stress and stimulate a sustained Nrf2 activation, as reported in previous studies [32] and shown in Scheme 2. When the LPS and the BSO were added, macrophage metabolic activity dramatically increased (160% of metabolically active cells compared to untreated cells, as a sign of activation. For comparison, macrophages were also exposed to N-acetylcysteine (NAC) (5 and 10 mM,), a widely known antioxidant capable of decreasing Nrf2 activation in macrophages under oxidative stress conditions [32] and αand γ-mangostins. Since mangostins, as natural xanthones, represent the ideal controls for the newly synthesized xanthone derivatives, only data after mangostin exposure are shown.

Scheme 1.
Methodologies used for the synthesis of xanthones 2-11, 13-37, 39, and 40. Adapted from [15]. Compounds 2, 6 (Grover, Shah, and Shah), and 31 (Eaton's reagent) were synthesized via (a), while compounds 3, 4, 7, 11, 16, 17, 24, 26, and 28 were obtained via (b). The compound 13 was obtained through a dehydrative process (d) from commercial 2,2′,4,4′-tetrahydroxybenzophenone. The diaryl process led to the synthesis of compounds 9 and 14 via (e) and the compounds 5, 8, 10, and 15 via (f), respectively. Further structural modifications were performed on compounds 13, 26, and 28 in order to extend the structural diversity of this library of compounds as previously reported by other groups [23,28,31]. 1,7-Dihydroxyxanthone (12) was previously isolated from Cratoxylum maingayi (Guttiferae) [15]. To investigate whether the polyphenolic and aminated synthesized xanthone derivatives have biological activity in an inflamed environment, the compounds were tested on LPS-stimulated macrophages. In parallel, cells were pre-incubated and exposed to BSO to trigger oxidative stress and stimulate a sustained Nrf2 activation, as reported in previous studies [32] and shown in Scheme 2. When the LPS and the BSO were added, macrophage metabolic activity dramatically increased (160% of metabolically active cells To investigate whether the polyphenolic and aminated synthesized xanthone derivatives have biological activity in an inflamed environment, the compounds were tested on LPS-stimulated macrophages. In parallel, cells were pre-incubated and exposed to BSO to trigger oxidative stress and stimulate a sustained Nrf2 activation, as reported in previous studies [32]  To investigate whether the polyphenolic and aminated synthesized xanthone derivatives have biological activity in an inflamed environment, the compounds were tested on LPS-stimulated macrophages. In parallel, cells were pre-incubated and exposed to BSO to trigger oxidative stress and stimulate a sustained Nrf2 activation, as reported in previous studies [32]  To investigate whether the polyphenolic and aminated synthesized xanthone derivatives have biological activity in an inflamed environment, the compounds were tested on LPS-stimulated macrophages. In parallel, cells were pre-incubated and exposed to BSO to trigger oxidative stress and stimulate a sustained Nrf2 activation, as reported in previous studies [32] and shown in Scheme 2. When the LPS and the BSO were added, macrophage metabolic activity dramatically increased (160% of metabolically active cells To investigate whether the polyphenolic and aminated synthesized xanthone derivatives have biological activity in an inflamed environment, the compounds were tested on LPS-stimulated macrophages. In parallel, cells were pre-incubated and exposed to BSO to trigger oxidative stress and stimulate a sustained Nrf2 activation, as reported in previous studies [32] and shown in Scheme 2. When the LPS and the BSO were added, macrophage metabolic activity dramatically increased (160% of metabolically active cells To investigate whether the polyphenolic and aminated synthesized xanthone derivatives have biological activity in an inflamed environment, the compounds were tested on LPS-stimulated macrophages. In parallel, cells were pre-incubated and exposed to BSO to trigger oxidative stress and stimulate a sustained Nrf2 activation, as reported in previous studies [32] and shown in Scheme 2. When the LPS and the BSO were added, macrophage metabolic activity dramatically increased (160% of metabolically active cells To investigate whether the polyphenolic and aminated synthesized xanthone derivatives have biological activity in an inflamed environment, the compounds were tested on LPS-stimulated macrophages. In parallel, cells were pre-incubated and exposed to BSO to trigger oxidative stress and stimulate a sustained Nrf2 activation, as reported in previous studies [32] and shown in Scheme 2. When the LPS and the BSO were added, macrophage metabolic activity dramatically increased (160% of metabolically active cells When α-mangostin was added at 24 h, the cell metabolic activity significantly increased both at 1 and 10 µM with respect to the sample in the presence of LPS and BSO, which was chosen for 100% of the experimental procedure. This trend maintained after 48 h and was even greater in the presence of γ-mangostin (Tables 2 and 3). To verify if the increase in metabolic activity and, thus, the macrophage activation led to cytotoxicity or not, the LDH assay was performed on the supernatant from cells exposed to mangostins after 24 h ( Figure 1). As expected, the cells dramatically released a pool of LDH when stimulated with LPS and BSO (25.1-fold of the pure UC). It should be noted that cells only pre-incubated with BSO (UC) also abundantly released LDH (29.1-fold). When αand γ-mangostin were added to the cultures, the release of LDH was significantly decreased (6.6-and 4.8-fold, respectively). It is plausible to assume that the metabolic activation and the decreased cytotoxicity in the presence of mangostins allowed the authors to speculate that the presence of these natural xanthones could even amplify the Nrf2 activation and translocation to counteract the oxidative stress occurrence and, thus, the macrophage activation. In order to demonstrate this speculation, a Western blot analysis was performed ( Figure 2). The immunoblotting clearly revealed that Nrf2 translocates into the nuclei in the presence of mangostins under inflammatory conditions, mainly in the presence of the γ derivative. Despite the significant macrophage activation in terms of metabolic activity, it is plausible to assume that mangostins, amplifying the Nrf2 signaling, ameliorate the cell response towards LPS-and BSO-induced oxidative stress and thus counteract cytotoxicity occurrence in these cells. In this light, after having screened all the newly synthesized xanthones derivatives in the same experimental conditions with the cell metabolic activity test (Tables 2 and 3), the derivatives resembling the cell metabolic activity behavior in the presence of mangostins were chosen for further analyses (namely compounds 5, 6, 10, 16, 17, and 27). Int Compounds 2, 6 (Grover, Shah, and Shah), and 31 (Eaton's reagent) were synthesized via (a), while compounds 3, 4, 7, 11, 16, 17, 24, 26, and 28 were obtained via (b). The compound 13 was obtained through a dehydrative process (d) from commercial 2,2′,4,4′-tetrahydroxybenzophenone. The diaryl process led to the synthesis of compounds 9 and 14 via (e) and the compounds 5, 8, 10, and 15 via (f), respectively. Further structural modifications were performed on compounds 13, 26, and 28 in order to extend the structural diversity of this library of compounds as previously reported by other groups [23,28,31]. 1,7-Dihydroxyxanthone (12) was previously isolated from Cratoxylum maingayi (Guttiferae) [15]. compared to untreated cells, as a sign of activation. For comparison, macrophages were also exposed to N-acetylcysteine (NAC) (5 and 10 mM,), a widely known antioxidant capable of decreasing Nrf2 activation in macrophages under oxidative stress conditions [32] and α-and γ-mangostins. Since mangostins, as natural xanthones, represent the ideal controls for the newly synthesized xanthone derivatives, only data after mangostin exposure are shown.

Scheme 2.
In vitro experimental model used for the screening of xanthone derivatives 1-40 and controls (mangostins). Undifferentiated human monocytes (CRL-9855™) were stimulated with 100 ng/mL of PMA (phorbol-12-myristate-13-acetate) for 48 h to obtain differentiated macrophage-like cells. Next, macrophages were pre-incubated with 50 µ g/mL of BSO (L-buthionine sulfoximine) for 18 h. After that, cells were stimulated with 0.5 µ g/mL of LPS (Lipopolysaccharide) and exposed in parallel to BSO or xanthone derivatives (test compounds) or mangostins for up to 48 h.
When α-mangostin was added at 24 h, the cell metabolic activity significantly increased both at 1 and 10 µ M with respect to the sample in the presence of LPS and BSO, which was chosen for 100% of the experimental procedure. This trend maintained after 48 h and was even greater in the presence of γ-mangostin (Tables 2 and 3). To verify if the increase in metabolic activity and, thus, the macrophage activation led to cytotoxicity or not, the LDH assay was performed on the supernatant from cells exposed to mangostins after 24 h ( Figure 1). As expected, the cells dramatically released a pool of LDH when stimulated with LPS and BSO (25.1-fold of the pure UC). It should be noted that cells only pre-incubated with BSO (UC) also abundantly released LDH (29.1-fold). When α-and γ-mangostin were added to the cultures, the release of LDH was significantly decreased (6.6-and 4.8-fold, respectively). It is plausible to assume that the metabolic activation and the decreased cytotoxicity in the presence of mangostins allowed the authors to speculate that the presence of these natural xanthones could even amplify the Nrf2 activation and translocation to counteract the oxidative stress occurrence and, thus, the macrophage ac-Scheme 2. In vitro experimental model used for the screening of xanthone derivatives 1-40 and controls (mangostins). Undifferentiated human monocytes (CRL-9855™) were stimulated with 100 ng/mL of PMA (phorbol-12-myristate-13-acetate) for 48 h to obtain differentiated macrophagelike cells. Next, macrophages were pre-incubated with 50 µg/mL of BSO (L-buthionine sulfoximine) for 18 h. After that, cells were stimulated with 0.5 µg/mL of LPS (Lipopolysaccharide) and exposed in parallel to BSO or xanthone derivatives (test compounds) or mangostins for up to 48 h.
The selected xanthone derivatives underwent LDH release measurements to verify cytotoxicity occurrence. The assay revealed that 17 and 27 were the best compounds capable of decreasing the LDH release from inflamed macrophages (8.6-and 6.8-fold, respectively), and these data are comparable with that obtained in the presence of mangostins (Figure 1). In parallel, Western blot analysis on cell cytosols and nuclei was performed, revealing that 27 was the best compound in activating the Nrf2 translocation, followed by 16 and 10 ( Figure 2).
Due to a large number of derivatives and biological data, proper structure-activity relationship (SAR) studies must be extrapolated. Different SAR methods, which relate chemical structure to molecular properties, are frequently used to determine which derivatives should be synthesized based on the information that has been achieved in the optimization procedures. In recent years, matched molecular pairs (MMPs) have become widely used in medicinal chemistry, such as in SAR studies, activity profile analysis, and lead optimization [33]. In fact, a substituent group replacement or a modification in the structure scaffold can cause changes in the physico-chemical properties and activity profile of a molecule. MMPs are defined as pairs of compounds that only differ by a chemical change at a single site. The MMPs transformation can be used to explore changes in a molecule's dataset activities [34].  When studying SARs, the use of MMPs is useful in identifying structural features that affect activity/potency. The MMP protocol used identified crucial substitutions by identifying all possible MMPs and grouping them by common transformations. The frequency of the transformations indicates the probability that the structural change will give rise to a property variation, and outliers can reveal a means to circumvent the general tendency [35]. µM + LPS 0.5 µg/mL + compounds 10 µM. a = p < 0.05, b = p < 0.005, and c = p < 0.0005 between compounds and LPS; d = p < 0.05, e = p < 0.005, and f = p < 0.0005 between compounds and α-mangostin at the same concentration; g = p < 0.05, h = p < 0.005, and i = p < 0.0005 between compounds and γ-mangostin at the same concentration; m = p < 0.05, n = p < 0.005, and r = p < 0.0005 between the same compound at 10 µM and 1 µM.
A large library of natural and synthetic xanthone derivatives was studied by MMPs to identify structural trends that could influence activity. Fifty-eight MMPs were found spanning 13 structural substitutions (Figure 3). For each MMP, the activity change was classified as neutral, increased, or decreased. The activity effects are tabulated and sorted to show which MMPs have the largest effects in each category. Six types of substitution were responsible for an increase in activity [C(Br)Br >> H, C(Br)Br >> C(=O)OC, C=O >> C(=O)OC, C=O >> C(Br)Br, C(Br)Br >> C, and C=O > C]; one substitution caused an increase in activity in 66% of the cases, and had no effect in 33% of the cases (C=O >> H); four substitutions caused an increase, a decrease, or had no effect on activity (H >> C with BSO 50 µ M + LPS 0.5 µ g/mL + compounds 10 µ M. a = p < 0.05, b = p < 0.005, and c = p < 0.0005 between compounds and LPS; d = p < 0.05, e = p < 0.005, and f = p < 0.0005 between compounds and α-mangostin at the same concentration; g = p < 0.05, h = p < 0.005, and i = p < 0.0005 between compounds and γ-mangostin at the same concentration; m = p < 0.05, n = p < 0.005, and r = p < 0.0005 between the same compound at 10 µ M and 1 µ M. The selected xanthone derivatives underwent LDH release measurements to verify cytotoxicity occurrence. The assay revealed that 17 and 27 were the best compounds capable of decreasing the LDH release from inflamed macrophages (8.6-and 6.8-fold, respectively), and these data are comparable with that obtained in the presence of mangostins ( Figure 1). In parallel, Western blot analysis on cell cytosols and nuclei was performed, revealing that 27 was the best compound in activating the Nrf2 translocation, followed by 16 and 10 ( Figure 2). . LPS = cells stimulated with 0.5 µ g/mL LPS alone. The pure UC (cells exposed to complete RPMI for the entire experimental procedure) was set as 1 (not shown). c = p < 0.0001 between samples and UC; d = p < 0.01 and f = p < 0.0001 between samples and LPS. Representative images from experiments were obtained by optical microscopy (phase-contrast, magnification 100×). . LPS = cells stimulated with 0.5 µg/mL LPS alone. The pure UC (cells exposed to complete RPMI for the entire experimental procedure) was set as 1 (not shown). c = p < 0.0001 between samples and UC; d = p < 0.01 and f = p < 0.0001 between samples and LPS. Representative images from experiments were obtained by optical microscopy (phase-contrast, magnification 100×). Some conclusions could be reached from this MMP analysis, namely the groups that are more favorable for activity. Concerning the molecules studied by MMP analysis, methyl groups always increased activity, and methyl formate groups were typically important for activity (or at least did not affect activity). Conversely, formaldehyde groups were always prejudicial for activity. Other groups had variable behavior according to the group they were substituting. Moreover, -OH groups at positions R2, R3, R6, and/or R8; -OCH 3 groups at positions R1; prenyl groups at R3, R4, and/or R8; N,N,N ,N -tetramethylpropane-1,3-diamine groups at R1; piperidine and piperazine groups at R3 and/or R7; and -Cl groups at R8 presented the highest activities (Figure 4).
QSAR studies have been used for several years to point out properties of small molecules that are relevant for activity and to forecast the activity of new compounds [36]. Therefore, a QSAR model was built to highlight the descriptors that are relevant to the activity of the tested xanthone derivatives. Furthermore, a QSAR model allows efforts to be directed toward the synthesis of compounds that are more likely to have the desired activity [36]. In this work, a 2D-QSAR model was elaborated using the Comprehensive Descriptors for Structural and Statistical Analysis (CODESSA 2.7.2) software package, which calculates approximately 500 descriptors. The heuristic method performs a preselection of descriptors by eliminating descriptors that are not available for each structure, have a small variation in magnitude, are correlated pairwise, and have no statistical significance. The heuristic method is a convenient method for searching for the best set of descriptors, without restrictions on the dataset size [37]. Due to a large number of derivatives and biological data, proper structure-activity relationship (SAR) studies must be extrapolated. Different SAR methods, which relate chemical structure to molecular properties, are frequently used to determine which derivatives should be synthesized based on the information that has been achieved in the optimization procedures. In recent years, matched molecular pairs (MMPs) have become widely used in medicinal chemistry, such as in SAR studies, activity profile analysis, and lead optimization [33]. In fact, a substituent group replacement or a modification in the structure scaffold can cause changes in the physico-chemical properties and activity profile of a molecule. MMPs are defined as pairs of compounds that only differ by a chemical change at a single site. The MMPs transformation can be used to explore changes in a molecule's dataset activities [34].
When studying SARs, the use of MMPs is useful in identifying structural features that affect activity/potency. The MMP protocol used identified crucial substitutions by identifying all possible MMPs and grouping them by common transformations. The frequency of the transformations indicates the probability that the structural change will give rise to a property variation, and outliers can reveal a means to circumvent the general tendency [35].
A large library of natural and synthetic xanthone derivatives was studied by MMPs to identify structural trends that could influence activity. Fifty-eight MMPs were found spanning 13 structural substitutions (Figure 3). For each MMP, the activity change was classified as neutral, increased, or decreased. The activity effects are tabulated and sorted The correlation coefficient (R 2 ), standard error (S), and Fisher value (F) measures were used to evaluate the validity of the regression equation [38]. As the rules of QSAR establish that there must be one descriptor for every five molecules used to build the model, seven descriptors were used to build the QSAR equation. The multilinear regression analysis using the heuristic method for 36 compounds in the seven-descriptor model is shown in Figure 5. The compounds are uniformly distributed around the regression line, which indicates that the obtained model has a satisfactory predictive ability.
The best QSAR equation has an R 2 of 0.7346, a Fisher value of 11.07, and an S of 11.96, which reveals that the proposed model has statistical validity [39]. R 2 is greater than 0.7, which proves the high strength of the relationship between the QSAR model and the dependent variable [40]; it represents close to 70% of the total variance in the dependent variable (activity) shown by the test compounds. The QSAR model is significant at the 95% level, as shown by the Fisher's F-test value (F = 11.96), which is larger than the tabulated value (2.36), which is required for a statistically significant model [40]. The standard deviation S displays a small value (s = 11.96), showing that the model is significant and has small variation around the regression line [41]. The reliability of the resulting QSAR model was analyzed using two different types of validation criteria: external validation by using a test set and internal validation by leave-one-out (LOO) cross-validation [42]. The model was able to predict the activity of an external test set with an average difference of 9.00 from the experimental value [43]. Moreover, the cross-validated R 2 (Q 2 = 0.5824) from the LOO internal validation process is greater than 0.5 and smaller than the overall R 2 , as anticipated, and the difference between R 2 and Q 2 is lower than 0.3, which indicates that the model does not suffer from overfitting [44].
four substitutions caused an increase, a decrease, or had no effect on activity (H >> C, H >> O, O >> OC, OC >> H); and finally, two substitutions did not cause a relevant difference in activity [Cl >> H and C >> C(=O)OC] (Figure 3). Some conclusions could be reached from this MMP analysis, namely the groups that are more favorable for activity. Concerning the molecules studied by MMP analysis, methyl groups always increased activity, and methyl formate groups were typically important for activity (or at least did not affect activity). Conversely, formaldehyde groups were always prejudicial for activity. Other groups had variable behavior according to the group they were substituting. Moreover, -OH groups at positions R2, R3, R6, and/or R8; -OCH3 groups at positions R1; prenyl groups at R3, R4, and/or R8; N,N,N',N'-tetramethylpropane-1,3-diamine groups at R1; piperidine and piperazine groups at R3 and/or R7; and -Cl groups at R8 presented the highest activities (Figure 4). Each bar indicates a MMP transformation, and the color coding classifies the activity change: increases activity (green), decreases activity (red), and has neutral effect on activity (yellow). The labels on the left axis indicate the type and number of transformations. Additionally, a table on the right side contains the SMILES transformation and a sketch of the associated reaction. In order to abbreviate the large structures of the tested molecules, * represents the rest of the molecule. By interpreting the molecular descriptors in the regression model, it is possible to have some insight into structural characteristics that are likely to be responsible for the antioxidant activity of the studied compounds. Seven variables were found to have a significant influence on the potency of the compounds ( Figure 5).
Three of the descriptors-moment of inertia A, topographic electronic index, and minimum partial charge for an O atom-have a positive regression coefficient of 1613.2, 45.652, and 2158.9, which means that an increase in these descriptors will lead to an increase in the antioxidant activity of the xanthonic derivatives. On the other hand, minimum partial charge for a H atom, HACA-1, WNSA-1 weighted PNSA, and RNCS relative negative charged SA descriptors have negative regression coefficients of −1106.4, −2.3678, −0.2266, and −1.1013, which reveals that an increase in the values of these descriptors will lead to a decrease in the activity of the molecules.
The moment of Inertia A (I A ) is obtained from the 3D coordinates of the atoms in the given molecule and is defined as the product of the mass times the distance from the axis squared, I A = ∑ i m i r 2 ix , where m i is the atomic mass, and r ix denotes the distance of the i-th atomic nucleus from the main rotational axes, x. I A characterizes the mass distribution in the molecule. The high positive correlation coefficient of the moment of inertia A and the highest t-value (t-values define the statistical significance of a descriptor) suggest that the orientation behavior in relation to the size of the whole molecule is very important for activity. QSAR studies have been used for several years to point out properties of small molecules that are relevant for activity and to forecast the activity of new compounds [36]. Therefore, a QSAR model was built to highlight the descriptors that are relevant to the activity of the tested xanthone derivatives. Furthermore, a QSAR model allows efforts to be directed toward the synthesis of compounds that are more likely to have the desired activity [36]. In this work, a 2D-QSAR model was elaborated using the Comprehensive Descriptors for Structural and Statistical Analysis (CODESSA 2.7.2) software package, which calculates approximately 500 descriptors. The heuristic method performs a pre-selection of descriptors by eliminating descriptors that are not available for each structure, have a small variation in magnitude, are correlated pairwise, and have no statistical significance. The heuristic method is a convenient method for searching for the best set of descriptors, without restrictions on the dataset size [37].
The correlation coefficient (R 2 ), standard error (S), and Fisher value (F) measures were used to evaluate the validity of the regression equation [38]. As the rules of QSAR establish that there must be one descriptor for every five molecules used to build the model, seven descriptors were used to build the QSAR equation. The multilinear regression analysis using the heuristic method for 36 compounds in the seven-descriptor model is shown in Figure 5. The compounds are uniformly distributed around the regression line, which indicates that the obtained model has a satisfactory predictive ability. The best QSAR equation has an R 2 of 0.7346, a Fisher value of 11.07, and an S of 11.96, which reveals that the proposed model has statistical validity [39]. R 2 is greater than 0.7, which proves the high strength of the relationship between the QSAR model and the dependent variable [40]; it represents close to 70% of the total variance in the dependent variable (activity) shown by the test compounds. The QSAR model is signifi- Figure 5. QSAR model obtained with the heuristic method for 36 molecules with the CODESSA software (R 2 = 0.7346, F = 11.07, S = 11.96). X, ∆X, and t-test are the regression coefficients of the linear model, standard errors of the regression coefficient, and the student's t-test significance coefficient of the determination, respectively. The next implicated descriptor is the topographic electronic index, T E , which connects submolecular polarity parameters with molecular topography expressed by interatomic distances [45]. This index is calculated as a sum of ratios q i − q j /r 2 i,j over all the pairs of atoms, both connected and disconnected (q i and q j are the corresponding partial charges on atoms i and j in the pair, whereas r i,j is the interatomic distance) [46].
Minimal partial charge for a H atom and minimal partial charge for an O atom are electrostatic descriptors related to charge distribution. Electrostatic descriptors consider the electrostatic structure of the molecules characterized by the partial charge distribution or the electronegativity of the atoms. The partial charges in the molecule can be calculated using the method proposed by Zefirov [47], which takes molecular electronegativity as a geometric mean of atomic electronegativity. HACA1 (hydrogen bonding acceptor ability of the molecule) is a charged partial surface area (CPSA) descriptor that is determined by the equation ∑ A S A A ∈ X H−acceptor where S A stands for the solvent-accessible surface area of H-bonding acceptor atoms, selected by threshold charge.
WNSA-1 weighted PNSA (partial negative surface area) is a quantum-chemical descriptor that characterizes molecules by molecular shape and electron distribution and is defined as PNSA1×TMSA

1000
, where PNSA1 is the partial negatively charged molecular surface area, and the TMSA is the total molecular surface area. This descriptor is defined based on the total molecular surface area and charge distribution in the molecule, thus indicating the influence of charge distribution on antioxidant activity [48].
RNCS relative negative charged surface area (SAMNEG * RNCG) is an electrostatic descriptor that depends on the distribution of the charges on the molecule. The relative negative charge of the molecule and its surface area can thus influence activity.
The molecular descriptors used in the QSAR model demonstrate that the mechanism underlying the antioxidant activity of xanthones is related to the mass distribution, the polarity of the atoms in relation to their interatomic distances, the electronegativity of hydrogens and oxygens, the hydrogen bonding acceptor ability of the molecule, the molecular shape, and the electron and charge distribution of the molecule. The inspection of the molecular descriptors can result in a better comprehension of the relationship between the structure and activity of xanthones. The QSAR model developed in the present paper may be useful for increasing the awareness of the mechanisms modulating Nrf2 activity.