Flavonoids as tyrosinase inhibitors in in silico and in vitro models: basic framework of SAR using a statistical modelling approach

Abstract Flavonoids are widely distributed in plants and constitute the most common polyphenolic phytoconstituents in the human diet. In this study, the in vitro inhibitory activity of 44 different flavonoids (1–44) against mushroom tyrosinase was studied, and an in silico study and type of inhibition for the most active compounds were evaluated too. Tyrosinase inhibitors block melanogenesis and take part in melanin production or distribution leading to pigmentation diseases. The in vitro study showed that quercetin was a competitive inhibitor (IC50=44.38 ± 0.13 µM) and achieved higher antityrosinase activity than the control inhibitor kojic acid. The in silico results highlight the importance of the flavonoid core with a hydroxyl at C7 as a strong contributor of interference with tyrosinase activity. According to the developed statistical model, the activity of molecules depends on hydroxylation at C3 and methylation at C8, C7, and C3 in the benzo-γ-pyrane ring of the flavonoids.


Introduction
Flavonoids are secondary plant metabolites that can be chemically divided into groups based on their substitutions. Flavonoid moieties can be modified by glycosylation, hydrogenation, hydroxylation, acetylation and methylation, as well as by malonylation and sulfatation. The chemical and biological potentials of flavonoids and their derivatives are connected with the position of diverse substitutions on the molecule and the saturation of double bonds in the structure 1 . Thus, structure-activity relationship (SAR) studies can help predict the biological activities of compounds with related structures and help avoid off-target outcomes due to their toxicity and side effects. The substantial contribution of a treatment is based on selecting the right medicine for a given target. Available theoretical tools (e.g., molecular docking studies) can help predict potential inhibitory activity against enzymes, including tyrosinase, and virtual screening can be used to select compounds that target tyrosinase and to determine expected targets for well-known and newly discovered flavonoids 2,3 . Tyrosinase is common in mammals, fungi, bacteria, and plants, and it plays a critical role in melanin biosynthesis. Tyrosinase consists of two identical H subunits as a catalytic component and two identical L subunits. The H subunit includes four helices that contain the catalytic binuclear copper-binding site. Each Cu 2þ cofactor forms coordination bonds with three histidine residues (His61, His85, His94, and His259, His263, His296, respectively) located at the centre of two antiparallel a helix pairs 4,5 . His85 is covalently bound to Cys83 through a thioether bond, and these two residues are connected to each other via a threonine residue that forms a triad motif that is conserved among tyrosinases and is considered essential for catalytic activity 6,7 . The histidine ligands of the copper ions are stabilised via interactions with nearby residues such as Phe90 and Phe292 for catalytic activity; thus, interactions with the copper ions as well as their ligands and the nearby residues are required for effective inhibition 8,9 . In humans, abnormal melanin production or distribution leads to pigmentation diseases, such as overtanning, freckles, age spots, and melasma. Disorderly melanin production plays a key role in melanotic melanoma, and inhibiting tyrosinase activity may reduce melanin content and be a useful process in skin-whitening compounds 10,11 .

Tyrosinase inhibition assay
The inhibitory effect on mushroom tyrosinase (TYase) was evaluated using a method reported in the literature with some modifications 21 . This assay was performed in PBS (100 mM, pH ¼ 6.8 in 25 C). Test samples 1-44 (80 mL) were preincubated with TYase solution (40 mL; 250 U/mL) at 25 C for 10 min. Then, L-DOPA (80 mL; 0.19 mg/mL) was added, and after an additional 10 min at 25 C, the absorbance was measured at 492 nm. A blank analysis was performed using PBS instead of a sample, and the positive control was conducted with kojic acid. The inhibitory effect was calculated as follows: where B and C are the absorbances of the blank and the compounds, respectively. The compound concentration that inhibits 50% of tyrosinase activity (IC 50 ) was calculated. All tests were performed in triplicate.

Tyrosinase inhibition kinetic analysis
Based on IC 50 values, the seven most active compounds (4,17,26,32,35,41) were selected for kinetic analysis. The enzyme reaction kinetics of these compounds were measured by constructing Lineweaver-Burk plots of inverse velocities (1/V), contrary to the inverse of substrate concentration (1/S) 22,23 . Preincubations and measurement times were performed using the same protocol as described above. The enzyme concentration (250 U/mL) was kept constant in the presence of substrate solutions (L-DOPA) between 0.25 and 2 mM in all kinetic studies. The inhibitor concentrations for all test compounds were 0, 25, and 50 mM. The inhibitory types and inhibitory constant (K i ) values were described by 1/V versus inhibitor concentration plots (Dixon plot).

Molecular docking
The selected flavonoids were sketched in ACD/ChemSketch to build and generate their mol topology format.  26 . In this process, redundant molecules were removed, H atoms were added, partial charges were assigned, and ionisation, tautomeric states, and H bonds were set. The active site grids were generated using the Receptor Grid Generator panel of Maestro by setting the central coordinates as À0.62, 26.99, and À43.78, and the volume as 27.000 Å 3 . The ligand was docked to the active site using Glide (2021-2, Schr€ odinger LLC, New York, NY) in additional precision (XP) mode with 50 runs per ligand, and the results were visually evaluated 27 . Prior to docking of the ligands, the co-crystallized ligand in the PDB structure, tropolone, was removed and redocked to the active site, and the obtained binding mode was compared with the co-crystallized conformer by calculating the root-mean-square deviation (RMSD) to confirm the precision of the method. The predicted binding mode for tropolone was similar to that of the original conformer (RMSD: 1.33 Å).

Statistical analysis
Statistical analysis for IC 50 calculations was performed using nonlinear regression using GraphPad Prism 9 (Trail, GraphPad Software, La Jolla, CA, USA). Data were collected as mean ± SD (n ¼ 3), and the significance of differences was analysed using one-way analysis of variance (ANOVA). Statistical analysis of SARs was performed using Stata/IC 13.1 (StataCorp LP, TX 77845, USA).

Inhibitory effects of tested compounds on tyrosinase
The effects of different concentrations of selected flavonoids ( Figure 1) and kojic acid, a well-known inhibitor of L-DOPA oxidation by tyrosinase, were studied. The effects, expressed as IC 50 values, are shown in Table 1.
Although the rendered activity is accumulative of an entire molecule, a structure-activity relationship was recognised by examining the effect of different substitutions: the presence or absence of hydroxyl groups or their methylation/acetylation/glycosylation in all carbons of the B-ring, as well as in carbons 3, 5, 6, 7, and 8 of the benzo-c-pyrane ring, on the potential inhibitory activity.
Some flavonoids (4, 17, 35) exhibited IC 50 values that were comparable to the positive control, kojic acid (IC 50 ¼49.48 ± 0.23 mM). Compound 4 (IC 50 ¼58.43 ± 0.38 mM), tested in this study for the first time, was one of the most effective inhibitors of tyrosinase, along with compound 17 (IC 50 ¼58.88 ± 0.78 mM) and compound 35 (IC 50 ¼44.38 ± 0.13 mM). The distinction in position C3 in 35 and 36 (IC 50 ¼220.10 ± 1.14 mM) or 37 (IC 50 ¼141.67 ± 1.30 mM) is only a difference in the glycosylation of a hydroxyl group. However, this small difference in their structure causes immense differences in their inhibitory potentials. It is presumed that substitutions at C3 led to steric hindrances that prevent molecules from binding to enzymes. However, the most typical locations for C-glycosyl radicals are the C6 and C8 positions at the A-ring. It seems that aglycone 4 (IC 50 ¼58.43 ± 0.38 mM) is a more effective tyrosinase inhibitor than it glycoside: 20 (IC 50 ¼132.55 ± 2.32 mM). The exact location of the glycosidic residue (carbon 6 or 8) does not affect activity. Also, disaccharides linked with the flavonoid core in C8 (22,23,24,25) exhibit weak inhibitory potential with IC 50 values that are higher than those of the maximum tested concentration. When the antityrosinase activity of methylated 26, 27, and 28 molecules with IC 50 values 65.11 ± 1.09 mM, 290.46 ± 1.19 mM, and >500 mM, respectively, was compared, the presence of methyl groups likely prevents appropriate interactions with the enzyme active site 28 . The same SAR pattern was observed with compounds 4 (without methylation), 1 (one methyl group at C7), and 2 (two methyl groups at C7, C4'), with IC 50 values of 58.43 ± 0.38 mM, 441.92 ± 1.91 mM, and >500 mM, respectively. The links of the B-ring in the C-ring allow flavone and isoflavone activity to be compared. Isoflavones (3-B-ring) with IC 50 values >500 mM (43 and 44) exhibit no inhibitory potential. Additionally, the confrontation of the activity of 35 (IC 50 ¼44.38 ± 0.13 mM) with 42 (IC 50 ¼100.33 ± 1.86 mM) leads to the conclusion that the additional hydroxyl group in C5' reduces phenol inhibitory potency. It was not possible to determine the IC 50 values of flavonoids 2, 21,22,23,25,28,40,43 and 44 up to 500 mM, which was the highest tested concentration. Thus, based on the in vitro study, molecules containing more unsubstituted hydroxyl groups in C3, C5, and C7, as well as in C3' and C4', generally achieve highly potent activity against tyrosinase, this finding is consistent with previous   11 . Similar conclusions of SAR were presented using inhibitors of xanthine oxidase and elastase and scavengers of superoxide radicals 29,30 .

Kinetic analysis of tyrosinase inhibition
Kinetic analysis of compound-induced inhibition was performed to determine the type of inhibition of the most active constituents (i.e., IC 50 values below 100 mM). Lineweaver-Burk plots (Figure 2) in double-reciprocal form and Dixon plots ( Figure 3) were used to determine the inhibition types and evaluate the dissociation constants for the inhibitors (K i ) ( Table 2). The inhibitory type of isookanin (4) and robinetin (41) was tested for the first time in this study.
As shown in Figure 2(A,B,F), the straight lines intersected at the same point on the x-axis, implying that the K m value remained constant, while the maximum reaction rate (V max ) decreased with increasing concentrations of tested inhibitors (0, 25, and 50 mM).
Thus, isookanin (4), luteolin (17) and robinetin (41) likely caused non-competitive tyrosinase inhibition. Compounds 4, 17 and 41 produce allosteric regulation, which is a specific type of enzyme inhibition that is characterised by an inhibitor binding to an allosteric site, resulting in decreased enzyme efficacy. The luteolin type of inhibition was consistent with previous studies 31 . The inhibition constant K i was obtained from Dixon plots ( Figure  3(A,B,F)) as 18.64 ± 0.53 mM, 11.32 ± 0.77 mM and 27.42 ± 0.62 mM for compounds 4, 17, and 41, respectively.
In contrast, kaempferol (26), tiliroside (32) and quercetin (35) exhibit tyrosinase inhibition that are similar 11,32,33 . As shown in Figure 2(C,D,E), all straight lines crossed at the same point on the y-axis, which indicated that V max remained constant. As the x-axis intercepted increased with increasing inhibitor concentrations, K m increased because a higher concentration of substrate is required to overcome the inhibitory effects of a competitor. Thus, compounds 26, 32, 35 can bind to the active site and prevent binding to the real substrate. Based on Dixon plots (Figure 3(C-E)), K i for

Molecular docking studies
Active compounds were generally predicted to show high affinity to the active site, as represented by their docking scores (Table 3). Compounds 26 and 41 were particularly noteworthy based on their docking scores, which were based on a set of profitable interactions with the enzyme and its cofactors.
Isookanin's (4) binding mode lacked interactions with Cu 2þ and its ligands (Figure 4(A)), which resulted in a moderate docking score. This result was most likely due to the lack of aromaticity (i.e., a double bond between C2 and C3) of its flavonoid core, which, in the case of other compounds, was able to reach deeper into the copper zone and make p-stacks with the key histidine residues. The lack of aromaticity in isookanin's flavonoid core also cancelled out the compound's planarity and led to bending, which may have made it sterically difficult to approach the copper ions.
Conversely, compounds 17, 26, 32, 35, and 41 were found to be suitable chelators of Cu 2þ due to the common flavonoid core and the ionisable OH substituent at the 7th position ( Figure  4(B-F)). The ionised hydroxyl was effective even for double Cu 2þ engagement in the case of 26, 32 and 41, which were the best scoring compounds (see Table 3). Aromaticity of the flavonoid core enabled p stacking with the histidine residues serving as    These results highlight the importance of the flavonoid core with a free hydroxyl at C-7 as a key component for copper chelation and for interactions with the key residues of the tyrosinase active site. The exceptional binding mode of 4 in the catalytic site with respect to its activity can be explained by more effective binding to a possible allosteric site.

Statistical modelling approach
To create a statistical model, 26 attributes ( Table 4) that were consistent with all tested flavonoid compounds and their variables (IC 50 values) were identified. Table 5 shows the multivariate linear regression model for the ln IC 50 . Statistical analysis explained more than 50% of the variability of the IC 50 variable (adjusted R 2 ¼50.88%) with p¼ 0.0002. All 11 independent variables in this approach yield p< 0.2, while five of them (A4, A10, A18, A21 and A22) have a statistically significant influence on the dependent variable (IC 50 ).
Based on these results, the presence of methoxy substitution at C3'(A4) increases the mean IC 50 value by approximately 4.15 times, and increasing this value (by approximately 2.47 times) relates to methoxy substitution at C7 (A18). Methylation at C8 (A22) caused an increase in the IC 50 value, as shown by lnIC 50 (1.452888), which indicates that, in this case, the mean IC 50 increases by approximately 4.28 times. Conversely, the presence of hydroxy substitution at C3 (A10) and hydroxy substitution at C8 (A21) reduced the mean IC 50 by approximately 2.83 and 5.11 times, respectively. Overall, statistical modelling suggests that flavonoids with higher antityrosinase activity possess an OH group at C3 and at C8, while methylation of the hydroxyl group at carbon 3 and at carbon 8 notably reduces the inhibitory activity of the molecules.

Conclusions
Results from the literature that describe the effect of flavonoids on mushroom tyrosinase activity are dispersed and sometimes contradictory. This dissidence may relate to various experimental conditions (e.g., concentration of enzyme or substrate L-DOPA, wavelength), which led to different ranges of reported inhibitory activity. Furthermore, generally few flavonoids were included in such investigations, which yields restricted structure-activity relationships information. The statistical model developed in this study provides accurate SAR data in a comprehensive format. Many flavonoids were tested under the same conditions to obtain values that are as close to those in the real world as possible. In summary, a comparison of IC 50 values, kinetic reactions, molecular docking scores and statistical data allowed us to identify the characteristics of flavonoid structures that facilitate tyrosinase inhibition: the presence of hydroxyls at C3 and C7, Oand Cglycosylation, methylation and acetylation of OH groups. As in many other cases reported in the literature 34 , the natural products represent a gold mine for interesting biological activities which can be translated to potential biomedical applications.

Acknowledgement
The authors express thanks to J.W. Strawa and A.M. Juszczak from the Medical University of Białystok for isolation of compounds to in vitro studies.