Electronic-Structure Determinants of Flavonoid Inhibition at the Molybdenum Cofactor of Xanthine Oxidase: A QM/MM Study Authors: Mei-Ling Li, Cheng-Hong Hsieh, Shang-Ming Huang, Kuo-Chiang Hsu Published: 2026 ABSTRACT Flavonoid-mediated inhibition of xanthine oxidase (XO) is well established at the empirical level; however, while molecular docking, MD, and MM-PBSA/MM-GBSA methods have been widely applied, the quantum-mechanical origin of the structure–activity relationship at the molybdenum cofactor (MoCo) remains unresolved. Hybrid QM/MM calculations (B3LYP/def2-SVP) on five dietary flavonoids bound to the MoCo active site of bovine XO (50 snapshots total) revealed that all five exhibit net charge redistribution at the MoCo interface, with the magnitude varying nearly 9-fold. The charge-redistribution descriptor |CTlig| correlated with Ki (r = – 0.922, p = 0.026; Spearman ρ = – 1.000), and Hirshfeld charge analysis independently confirmed the Mulliken-derived ranking (r = 0.942, p = 0.017), while representative def2-TZVP calculations preserved the principal compound hierarchy (Spearman ρ = 0.900), supporting robustness to both charge partitioning and basis-set effects. Frontier orbital analysis links the observed structure–activity relationship to the saturated C2–C3 bond of naringenin, which interrupts π-conjugation and limits electronic communication toward the MoCo environment. The electrostatic interaction energy further correlated with Ki (r = – 0.890, p = 0.043), whereas total interfragment contact density showed no correlation, indicating that within this five-compound series, inhibitory potency was more closely associated with charge-redistribution magnitude and electrostatic interaction than with the interfragment contact-density descriptor. Introduction Xanthine oxidase (XO; EC 1.17.3.2) is a molybdenum-containing flavoprotein that catalyzes the final two steps of purine catabolism, namely the oxidation of hypoxanthine to xanthine and subsequently to uric acid. The enzyme is a homodimer of approximately 290 kDa, with each monomer containing three distinct redox domains: a molybdopterin cofactor (MoCo) center where substrate oxidation occurs, two iron–sulfur ([2Fe–2S]) clusters that mediate intramolecular electron transfer, and a flavin adenine dinucleotide (FAD) domain where molecular oxygen is ultimately reduced. , Inhibition of XO activity is a well-established therapeutic strategy for managing hyperuricemia and gout. While the clinical XO inhibitor allopurinol targets the MoCo active site as a competitive inhibitor, its use is associated with adverse effects including skin rashes and hypersensitivity reactions, motivating the search for natural alternatives. Flavonoids are a class of polyphenolic compounds widely distributed in fruits, vegetables, and grains that have demonstrated significant XO inhibitory activity in the low micromolar range. , In a companion study, we characterized the XO inhibitory activity of five flavonoids (kaempferol, quercetin, apigenin, morin, and naringenin) identified from corn byproduct extracts and demonstrated that all five exhibit mixed-type inhibition consistent with dual-site binding, with kaempferol showing the strongest MoCo-site preference (Ki = 2.43 μM). Multiscale molecular dynamics (MD) simulations and MM-GBSA calculations further revealed that kaempferol preferentially binds the MoCo site (ΔΔG = +3.01 kcal/mol), quercetin and naringenin favor the FAD domain (ΔΔG = – 6.02 and – 7.78 kcal/mol, respectively), and apigenin shows essentially no site preference (ΔΔG = +0.55 kcal/mol). The present study focuses exclusively on MoCo-bound flavonoid complexes, thereby isolating the electronic contribution of the MoCo binding mode from the overall inhibition mechanism, which may additionally involve FAD-site interactions. Notably, morin exhibited mixed-type inhibition despite a strong MoCo preference (ΔΔG = +5.46 kcal/mol), suggesting that absolute FAD binding competence rather than site-preference direction alone determines the inhibition mode. However, classical force-field-based approaches, including molecular docking, MD simulations, and MM-GBSA binding free energy calculations, treat atomic charges as fixed parameters that do not respond to changes in the electronic environment upon binding. As a consequence, these methods cannot capture charge redistribution, electron donation, or polarization effects that may play critical roles in enzyme inhibition mechanisms, particularly at metalloenzyme active sites where the redox state of the metal center directly determines catalytic competence. For XO specifically, the Mo­(VI) center at MoCo must accept electrons during the catalytic cycle; any perturbation to its electronic environment, such as charge redistribution induced by a bound inhibitor, could directly impair its oxidative capacity. Hybrid quantum mechanics/molecular mechanics (QM/MM) methods offer a rigorous framework for investigating these electronic effects by treating the chemically active region (the QM zone) with density functional theory (DFT) while embedding it within a molecular mechanical (MM) description of the protein environment. , QM/MM calculations have been successfully applied to elucidate reaction mechanisms and electronic properties of various metalloenzymes, but to the best of our knowledge, QM/MM methods have not been widely applied to investigate the charge-redistribution mechanism underlying flavonoid-mediated XO inhibition. The structure–activity relationship (SAR) of flavonoids toward XO is well established at the empirical level: the C2C3 double bond, the 4-oxo group, and hydroxyl groups at positions C-5 and C-7 are recognized as essential structural features for XO inhibitory activity. Flavonols (kaempferol, quercetin, morin) and flavones (apigenin) possess planar structures due to C2C3 conjugation, whereas flavanones (naringenin) adopt nonplanar configurations owing to the saturated C2–C3 bond. While these structural features have been correlated with inhibitory potency, the quantum-mechanical origin of this structure–activity relationship and how it enables or disrupts electronic communication with MoCo remains unresolved. The present study addresses this gap by performing exploratory QM/MM calculations on five structurally diverse flavonoids bound to the MoCo active site of bovine XO. The specific objectives were to (1) quantify ligand-derived charge redistribution in MoCo-bound QM/MM complexes using fragment-based Mulliken population analysis across multiple MD snapshots, (2) assess the association between charge-redistribution magnitude and experimental Ki as the primary mechanistic metric, while evaluating IC50 as a complementary measure of overall inhibitory potency, (3) compare Mo electronic environments in the gas phase versus protein-bound states, (4) analyze frontier molecular orbitals to visualize ligand-centered orbital character and the electronic communication pathway, and (5) provide a mechanistic rationale for why the C2C3 double bond is a key structural feature for XO inhibitory activity. Importantly, because our companion study demonstrated that not all five flavonoids bind preferentially at MoCo (quercetin and naringenin favor the FAD domain, while apigenin and morin also exhibited mixed-type inhibition despite MoCo-preferring or neutral site preferences), the present analysis should be understood as characterizing the electronic contribution to inhibition specifically for MoCo-bound configurations, not as a universal model for the overall IC50, which reflects the aggregate of MoCo-site, FAD-site, and dual-site inhibition mechanisms. Computational Methods System Preparation and MD Trajectories The starting structures were taken from previously reported 100 ns molecular dynamics simulations of five flavonoid–XO complexes at the MoCo active site of bovine XO (PDB: 3NRZ). Briefly, the MD simulations were performed using GROMACS 2025.2 with the Amber99sb-ildn force field for the protein and the General Amber Force Field 2 (GAFF2) for the ligands, parametrized via ACPYPE. The systems were solvated with TIP3P water in a cubic box with 0.15 M NaCl and equilibrated through energy minimization, NVT (100 ps, 300 K), and NPT (100 ps, 1 bar) stages before 100 ns production runs. Full simulation details are provided in our previous work. For QM/MM calculations, 10 snapshots were extracted at 10 ns intervals (t = 10, 20, 30, ..., 100 ns) from each of the five MoCo-bound trajectories (3NRZ conformation), yielding 50 QM/MM calculations in total. Each snapshot was extracted from the equilibrated portion of the trajectory, as confirmed by backbone RMSD convergence. QM/MM Model Construction The QM region was defined to include the flavonoid ligand and the MoCo center, encompassing the molybdenum atom (Mo), the coordinating sulfur atoms, oxo groups (=O), and phosphorus with its associated oxygen atoms, totaling 210–266 atoms depending on the flavonoid and snapshot (mean: 233 atoms). The boundary between the QM and MM regions was placed at covalent bonds connecting the QM side-chain residues to the protein backbone, treated using hydrogen link atoms. The remainder of the protein, including all water molecules and ions within 12 Å of the QM region, was treated at the MM level using the Amber99sb-ildn force field. The QM region thus included the flavonoid, the MoCo center, and all amino acid side chains with any heavy atom within 4 Å of MoCo, yielding a variable-size QM region that adapted to the local protein conformation at each snapshot. Prior to the QM/MM calculation, each snapshot underwent a short geometry prerelaxation using the GFN2-xTB semiempirical method with MoCo heavy atoms frozen, resolving any steric clashes introduced by the MoCo grafting procedure. QM/MM Single-Point Energy Calculations QM/MM single-point energy calculations were performed using ORCA 6.1.0. The QM region was treated at the B3LYP/def2-SVP level of density functional theory. , The D3BJ dispersion correction was applied to account for van der Waals interactions within the QM region. The MM region exerted an electrostatic embedding on the QM wave function through point charges derived from the Amber99sb-ildn force field. The resolution of identity with chain-of-spheres exchange (RIJCOSX) approximation and the def2/J auxiliary basis set were employed to accelerate the calculations. The def2-ECP effective core potential was applied to Mo. SCF convergence was set to TightSCF criteria. The calculations were performed in the unrestricted Kohn–Sham (UKS) formalism with a net charge of – 3. Spin multiplicity was assigned according to the total electron count of each snapshot-specific QM region: even-electron models were initialized with multiplicity 1, whereas odd-electron models were initialized with multiplicity 2. The net charge was computed dynamically by summing the formal charges of each QM fragment: MoCo (−2, reflecting Mo­(VI) with dithiolene, oxo, sulfido, and phosphate ligands), the flavonoid ligand (0, neutral), and the amino acid side chains at their standard pH 7 protonation states (GLU802 – 1, ARG912 + 1, GLU1261 – 1; all others 0), yielding a total of – 3. The lowest-spin state, singlet for even-electron QM regions and doublet for odd-electron QM regions, was used as the nominal reference state. Across the 50 production calculations, the UKS solutions frequently exhibited substantial open-shell character and spin contamination, with ⟨S2⟩ values of 3.72 ± 1.12 (range: 0.00–5.66). This behavior is consistent with the known redox noninnocence of the molybdopterin dithiolene ligands, which can support ligand-centered radical character and antiferromagnetically coupled spin density at the Mo–S interface. The ⟨S2⟩ distributions were comparable across the five flavonoids (mean values for each compound: 3.48–3.92), suggesting that the observed open-shell character primarily arises from the MoCo–side-chain subsystem rather than from the flavonoid ligand. To verify that CTlig is robust to spin-state assignment, a spin-state sensitivity test was performed on representative snapshots of kaempferol and naringenin (Table S3a). Four spin multiplicities (mult = 1, 3, 5, 7) were tested on the same QM/MM geometry. For both compounds, the quintet (mult = 5, S = 2) was the lowest-energy state, lying only 1.0–1.2 kcal/mol below the nominal singlet, consistent with near-degenerate spin states and possible multireference character. Critically, CTlig was insensitive to spin-state assignment across all physically accessible states: kaempferol yielded CTlig = – 0.663 to – 0.690 e (range: 0.027 e), while naringenin yielded CTlig = – 0.051 to – 0.055 e (range: 0.004 e). The ∼ 13-fold difference between the two flavonoids was maintained at every spin multiplicity tested (Table S3a). The kaempferol triplet state (mult = 3) was excluded from this comparative assessment because it lies +32.56 kcal/mol above the singlet reference and is negligibly populated at 300 K (Boltzmann factor <10–23). CTlig is therefore used throughout this study as a comparative descriptor that is robust to the underlying spin-state complexity of the MoCo QM region. Each of the 50 QM/MM calculations was performed using RIJCOSX acceleration, with individual calculations requiring approximately 2–8 h depending on QM region size. Gas-Phase Reference Calculations To isolate the effect of ligand binding on the electronic properties of MoCo, gas-phase single-point calculations were performed on the MoCo fragment alone (extracted from the same snapshots but without the flavonoid ligand or protein environment) at the identical B3LYP/def2-SVP level of theory. The gas-phase Mo Mulliken charge served as the unperturbed reference (+0.347) against which the bound-state charges were compared. Fragment-Based Charge-Redistribution Quantification Mulliken population analysis was performed on the converged QM/MM wave function for each snapshot. To distinguish the ligand-specific electronic contribution from the composite effects of protein side chains and electrostatic embedding on the Mo center, a fragment-based charge decomposition was employed. The QM atoms were partitioned into three fragments: (i) the flavonoid ligand, (ii) the MoCo center, and (iii) the protein side-chain residues included in the QM region. The net charge of the ligand fragment in the bound QM/MM complex, used here as a charge-redistribution descriptor, was defined as the sum of Mulliken atomic charges over all ligand atoms CTlig=Σqi(ligand⁡atoms⁡in⁡bound⁡QM/MM⁡complex) where qi represents the Mulliken atomic charge of ligand atom i. Because the isolated flavonoid is formally neutral, a nonzero ligand-fragment charge reflects charge redistribution between the ligand and the remainder of the QM subsystem. The absolute value, |CTlig|, was used as a comparative measure of redistribution magnitude without assigning a specific electron-transfer direction. This fragment-based definition avoids conflating ligand-derived charge redistribution with effects mediated by protein side chains or induced by electrostatic embedding, as discussed in Section . All charge-redistribution values are reported as mean ± SD across the 10 snapshots for each flavonoid. Frontier Molecular Orbital Analysis Gas-phase geometry optimizations and frontier molecular orbital calculations were performed for kaempferol, quercetin, and naringenin at the B3LYP/def2-SVP level. HOMO isosurface plots were generated using UCSF ChimeraX at an isovalue of 0.02 au The spatial distribution of the HOMO across the flavonoid rings was qualitatively assessed to compare the intrinsic π-conjugation patterns of the three compounds. Statistical Analysis Pearson and Spearman correlation analyses were performed among the compound-level charge-redistribution descriptors, experimental IC50 and Ki values, and Hirshfeld electrostatic interaction energies. Coefficients of determination were calculated for linear regressions. Leave-one-compound-out sensitivity analyses were conducted to assess whether the observed associations were driven by any single compound. All statistical analyses were performed using GraphPad Prism 11. Hirshfeld Charge-Redistribution Validation To verify that the CTlig ranking is independent of the charge-partitioning scheme, Hirshfeld charge analysis was performed independently on all 50 QM/MM snapshots using Multiwfn 2026.3.18. , The Hirshfeld |CTlig| was computed analogously to the Mulliken-based metric, as the absolute value of the sum of Hirshfeld atomic charges on the ligand fragment atoms. Additionally, the Hirshfeld electrostatic interaction energy (ΔECoul) between the ligand and receptor QM fragments was computed as the pairwise Coulombic sum using Hirshfeld atomic charges and coordinates from each snapshot (Table S4). IGMH Interaction Density Analysis To distinguish between total interfragment contact density and ligand-associated charge-redistribution effects, the independent gradient model based on Hirshfeld partition (IGMH) analysis was performed on five representative snapshots (one per compound, selected at the median |CTlig| value) using Multiwfn 2026.3.18. Two fragments were defined: fragment 1 (ligand atoms) and fragment 2 (receptor QM atoms). A medium-quality grid (∼512,000 points) was used for the δginter calculation, followed by atomic pair contribution analysis at high-quality radial (40) and angular (170) integration grids. The integrated δginter over the entire grid served as a quantitative measure of total interfragment electron density deformation, encompassing hydrogen bonds, van der Waals contacts, and steric repulsion. Isosurface visualization (isovalue = 0.003 au) was rendered using VMD 2.0.0, colored by sign­(λ2)­ρ (Table S5; Figure S2). Basis-Set Sensitivity Analysis To assess whether the principal conclusions depend on the basis-set size, single-point QM/MM calculations were repeated at the B3LYP-D3BJ/def2-TZVP level for one representative snapshot per compound (selected at the median |CTlig| value). All other computational parameters, including the QM region definition, point-charge embedding, and molecular geometry, were held identical to the def2-SVP calculations. The def2-TZVP Mulliken |CTlig| values were compared with the corresponding def2-SVP values from the identical representative geometries, and the compound-level rankings were additionally compared with the SVP ensemble-derived hierarchy (Table S3b). Additionally, leave-one-compound-out sensitivity analysis was performed on the |CTlig|−Ki and ΔECoul–Ki correlations to evaluate whether the observed associations were driven by any single compound (Table S6). Safety Statement No unexpected hazards or risks were encountered in this computational study. All calculations were performed in silico without the use of hazardous materials. Results QM/MM Model and Charge-Redistribution Overview Figure illustrates the QM/MM partitioning scheme employed in this study. The QM region (highlighted in color) encompasses the flavonoid ligand, the MoCo center (Mo atom and its coordinating oxo, sulfido, and phosphate ligands), and all amino acid side chains with any heavy atom within 4 Å of MoCo (∼17 residues, 210–266 atoms total depending on the snapshot). The surrounding protein backbone and solvent are treated at the MM level (gray cartoon) via electrostatic embedding. This partitioning ensures that the direct electronic interactions between the flavonoid, the catalytically essential Mo center, and the immediate protein environment are described at the DFT level. QM/MM model of the kaempferol–XO complex at the MoCo active site (PDB: 3NRZ). The QM region (colored sticks) encompasses kaempferol, the MoCo center (Mo, S, O, P atoms), and amino acid side chains within 4 Å of MoCo (∼17 residues, 210–266 atoms total), treated at the B3LYP/def2-SVP level of theory. The protein backbone and solvent (gray cartoon) are described by Amber99sb-ildn point charges via electrostatic embedding. Charge Redistribution in MoCo-Bound Complexes The magnitude of ligand-derived charge redistribution in each MoCo-bound complex, quantified by fragment-based Mulliken population analysis across 10 MD snapshots, is presented in Figure and Table . All five flavonoids exhibited nonzero ligand-fragment charges in the bound QM/MM complexes, reflecting charge redistribution between the ligand and the remainder of the QM subsystem; however, the magnitude varied substantially across compounds (Table S1). Apigenin showed the largest mean |CTlig| (0.468 ± 0.211 e), followed by kaempferol (0.353 ± 0.221 e), morin (0.218 ± 0.015 e), quercetin (0.207 ± 0.096 e), and naringenin (0.051 ± 0.020 e). Absolute ligand-derived charge redistribution (|CTlig|) for each flavonoid in MoCo-bound QM/MM complexes, calculated from fragment-based Mulliken population analysis of 10 MD snapshots (10–100 ns) per compound. Bars represent mean values; error bars represent SD. Individual data points are shown as open circles. Apigenin exhibited the largest mean |CTlig|, while naringenin showed the smallest charge-redistribution magnitude, consistent with its disrupted π-conjugation. 1Summary of QM/MM Charge-Redistribution and Mo Electronic Properties for Five Flavonoids Bound to the XO MoCo Active Site FlavonoidSubclassIC50 (μM) Ki (μM) Inhibition mode Site preference |CTlig| (e) qMo (e) ΔqMo (e) KaempferolFlavonol4.212.43Mixed-typeMoCo0.353 ± 0.221+0.257 ± 0.060–0.090ApigeninFlavone8.141.86Mixed-typeNo preference0.468 ± 0.211+0.288 ± 0.176–0.059QuercetinFlavonol10.484.76Mixed-typeFAD0.207 ± 0.096+0.288 ± 0.204–0.059MorinFlavonol12.187.77Mixed-typeMoCo0.218 ± 0.015+0.306 ± 0.062–0.041NaringeninFlavanone27.889.65Mixed-typeFAD0.051 ± 0.020+0.177 ± 0.061–0.170 aQM/MM calculations performed at B3LYP/def2-SVP (def2-ECP for Mo), UKS broken-symmetry singlet (charge = – 3, mult = 1), with electrostatic embedding (Amber99sb-ildn). QM region includes flavonoid + MoCo + side chains within 4 Å (210–266 atoms).bExperimental values from Li et al., 2026. cValues are reported as mean ± SD across 10 QM/MM snapshots sampled at 10 ns intervals from the 100 ns MD trajectory for each flavonoid.dΔMo charge = Mo charge­(bound) – Mo charge­(gas-phase reference, +0.347). A negative value indicates reduced Mo positive charge upon complex formation.eThe ΔMo charge for naringenin is the largest despite its smallest |CTlig| this reflects effects mediated by protein side chains and electrostatic embedding on Mo rather than ligand-derived charge redistribution (see Section ). Summary of QM/MM Charge-Redistribution and Mo Electronic Properties for Five Flavonoids Bound to the XO MoCo Active Site QM/MM calculations performed at B3LYP/def2-SVP (def2-ECP for Mo), UKS broken-symmetry singlet (charge = – 3, mult = 1), with electrostatic embedding (Amber99sb-ildn). QM region includes flavonoid + MoCo + side chains within 4 Å (210–266 atoms). Experimental values from Li et al., 2026. Values are reported as mean ± SD across 10 QM/MM snapshots sampled at 10 ns intervals from the 100 ns MD trajectory for each flavonoid. ΔMo charge = Mo charge­(bound) – Mo charge­(gas-phase reference, +0.347). A negative value indicates reduced Mo positive charge upon complex formation. The ΔMo charge for naringenin is the largest despite its smallest |CTlig| this reflects effects mediated by protein side chains and electrostatic embedding on Mo rather than ligand-derived charge redistribution (see Section ). The charge-redistribution values exhibited considerable variability among snapshots (Figure ), reflecting the conformational dynamics of the flavonoid–MoCo interaction sampled across the 100 ns MD trajectory. Despite this variability, a clear hierarchy was maintained: the four planar flavonoids (kaempferol, quercetin, apigenin, and morin) consistently showed |CTlig| values above 0.1 e, while naringenin remained below 0.1 e in 9 out of 10 snapshots (gray shaded region in Figure ). The consistently minimal charge transfer of naringenin is attributable to its saturated C2–C3 bond, which disrupts the π-conjugation that extends across the A–C–B ring system in flavonols and flavones. This electronic disconnection confines the HOMO to the A and C rings (Figure C), preventing effective electronic communication with the MoCo center and resulting in the attenuated charge-redistribution magnitude observed across all snapshots. Time-resolved |CTlig| from each flavonoid across the 100 ns MD trajectory (snapshots at 10 ns intervals). The gray shaded region indicates the range of naringenin |CTlig| values. The four planar flavonoids (kaempferol, quercetin, apigenin, morin) consistently show higher and more variable |CTlig| values compared to the nonplanar naringenin. Correlation between Charge Redistribution and Inhibitory Potency Figure B presents the correlation between mean |CTlig| and experimental Ki values determined in our previous work. A strong negative correlation was observed (r = – 0.922, R2 = 0.851, p = 0.026, n = 5), indicating that greater ligand-derived charge redistribution in the modeled MoCo-bound state was associated with a lower inhibition constant. Apigenin showed the highest mean |CTlig| (0.468 e) and the lowest Ki (1.86 μM), whereas naringenin showed the smallest |CTlig| (0.051 e) and the highest Ki (9.65 μM). Correlations between mean |CTlig| and experimental inhibition parameters. (A) The correlation with IC50 showed a negative trend (r = – 0.833, R2 = 0.693, p = 0.08, n = 5), suggesting that greater ligand-associated charge redistribution was associated with stronger XO inhibition. (B) The correlation with Ki was stronger and statistically significant (r = – 0.922, R2 = 0.851, p = 0.026). Dashed lines represent linear regression fits. Experimental IC50 and Ki values were obtained from Li et al. (2026). This figure was generated by the authors using original statistical analyses of the experimental data reported in ref . No previously published graphical material was reproduced or adapted. IC50 was evaluated as a complementary measure of overall inhibitory potency (Figure A). The |CTlig|−IC50 relationship showed a negative but nonsignificant trend (r = – 0.833, R2 = 0.693, p = 0.08, n = 5). Kaempferol exhibited the lowest IC50 (4.21 μM), whereas naringenin showed the highest IC50 (27.88 μM). Apigenin appeared as a partial outlier, combining the highest mean |CTlig| with an intermediate IC50 (8.14 μM). The weaker association with IC50 is consistent with the mixed-type inhibition of all five flavonoids, because IC50 reflects the combined contributions of MoCo-site and FAD-site interactions. Thus, charge redistribution at MoCo should be regarded as one component of a multifactorial inhibition mechanism. Molybdenum Electronic Environment: Gas Phase versus Bound State To assess whether flavonoid binding measurably perturbs the electronic environment of the catalytic Mo center, the Mo Mulliken charge was compared between gas-phase MoCo (no ligand, no protein) and the flavonoid-bound QM/MM complexes (Figure ; Table ; Table S2). In the gas phase, the Mo atom carried a Mulliken charge of +0.347 (red dashed line in Figure ), representing the unperturbed electronic state. The isolated MoCo cofactor in its resting oxidized state consists of a molybdenum center coordinated by a pyranopterin dithiolene ligand, an oxo group (O), a hydroxo group (−OH), and a sulfido group (S), adopting a square-pyramidal geometry. Mo Mulliken charge in the gas phase (gray bars; MoCo fragment alone) versus the flavonoid-bound QM/MM complex (blue bars; mean ± SD of 10 snapshots). The red dashed line indicates the gas-phase reference (+0.347). All bound complexes show reduced Mo charge; however, this decrease reflects composite contributions from the ligand, QM side chains, and MM embedding (see Section ). Upon flavonoid binding in the protein environment, the Mo charge decreased in all cases, consistent with electron density being redistributed within the QM subsystem. Kaempferol reduced the Mo charge to +0.257 ± 0.060, representing a decrease of 0.090 charge units. Quercetin and apigenin showed intermediate reductions (to +0.288 ± 0.204 and +0.288 ± 0.176, respectively), while morin showed the smallest decrease (to +0.306 ± 0.062). Notably, naringenin exhibited the lowest bound-state Mo charge (+0.177 ± 0.061, ΔMo = – 0.170) despite its minimal ligand-derived charge redistribution (|CTlig| = 0.051 e). This apparent discrepancy is explained by the composite nature of the Mo Mulliken charge in the QM/MM framework: it is influenced not only by the bound ligand but also by the ∼ 17 protein side-chain residues included in the QM region and the electrostatic embedding from the MM environment. The large Mo charge reduction for naringenin thus reflects effects mediated by protein side chains and electrostatic embedding rather than ligand-derived charge redistribution, underscoring the importance of using the fragment-based CTlig, which specifically quantifies charge redistribution at the ligand fragment, as the primary metric for assessing ligand electronic effects. The ΔMo charge should therefore be interpreted as a composite indicator of the MoCo electronic environment, not as a direct measure of ligand donation. Frontier Molecular Orbital Analysis To illustrate the intrinsic electronic communication pathway within the flavonoid scaffold, HOMO isosurfaces of gas-phase optimized kaempferol, quercetin, and naringenin were analyzed (Figure ). These compounds were selected to compare two conjugated flavonols with the flavanone naringenin, in which the C2–C3 bond is saturated. Because these calculations were performed on isolated molecules, the HOMO distributions reflect the intrinsic orbital topology of each flavonoid, allowing direct comparison among the three structures. HOMO isosurfaces of the gas-phase optimized structures of (A) kaempferol (flavonol), (B) quercetin (flavonol), and (C) naringenin (flavanone), calculated at the B3LYP/def2-SVP level. C2 and C3 are labeled in all panels for direct comparison with Figure . In kaempferol and quercetin, the orbital density is delocalized across all three rings (A, C, and B), consistent with continuous π-conjugation through the C2C3 double bond. In naringenin, the orbital density is confined to the A and C rings, with negligible contribution from the B ring because the saturated C2–C3 bond disrupts π-conjugation. For kaempferol (Figure A), the HOMO exhibits orbital density delocalized across all three rings of the flavonoid scaffold (A, C, and B rings). This extensive π-conjugation, enabled by the C2C3 double bond in the C ring, creates a continuous electron density distribution from the A ring through the C ring to the B ring. This delocalized orbital topology provides an intrinsic electronic pathway across the flavonoid scaffold and is consistent with the greater charge-redistribution magnitude observed for kaempferol in the separate QM/MM calculations. For quercetin (Figure B), the HOMO likewise exhibits delocalization across all three rings, closely resembling the kaempferol orbital topology. The presence of the additional 3′-hydroxyl group on the B ring does not disrupt the conjugation pathway maintained by the C2C3 double bond. This observation confirms that the HOMO delocalization pattern is a general feature of flavonols possessing the C2C3 bond and is not unique to kaempferol, consistent with the comparable charge-redistribution magnitudes of the two compounds (|CTlig| = 0.353 and 0.207 e for kaempferol and quercetin, respectively). In stark contrast, the HOMO of naringenin (Figure C) is predominantly localized on the A and C rings, with negligible orbital density on the B ring. The saturated C2–C3 bond in naringenin’s C ring interrupts π-conjugation, effectively disconnecting the B ring from the conjugated system. As a consequence, electronic communication across the flavonoid scaffold is disrupted, consistent with the smallest charge-redistribution magnitude (|CTlig| = 0.051 e) observed in the separate QM/MM calculations. Structural Basis: The Role of the C2C3 Bond The structural basis for the differential charge redistribution between flavonols/flavones and flavanones is summarized schematically in Figure , comparing kaempferol, quercetin, and naringenin to demonstrate that the observed pattern is a general feature of flavonoids possessing the C2C3 double bond. In kaempferol (Figure A) and quercetin (Figure B), the C2C3 double bond maintains full π-conjugation across the tricyclic scaffold, enabling π-delocalization (blue arrow) across the flavonoid scaffold. This conjugation pathway is the structural prerequisite for effective charge redistribution. The arrows in Figure represent the extent of electronic communication inferred from the HOMO comparison between kaempferol, quercetin, and naringenin in Figure , together with the continuous conjugated scaffold shared by kaempferol and quercetin. They do not represent directional electron flow. Schematic comparison of electronic communication in (A) kaempferol (flavonol, IC50 = 4.21 μM; |CTlig| = 0.353 e), (B) quercetin (flavonol, IC50 = 10.48 μM; |CTlig| = 0.207 e), and (C) naringenin (flavanone, IC50 = 27.88 μM; |CTlig| = 0.051 e). In kaempferol and quercetin, the C2C3 double bond maintains π-conjugation across the A–C–B ring system, enabling delocalized electronic communication across the flavonoid scaffold. In naringenin, the C2–C3 single bond disrupts π-conjugation, confines electron density largely to the A–C ring system, and attenuates charge redistribution. The arrows indicate the extent of electronic communication inferred from the HOMO topology and do not represent directional electron flow. Experimental IC50 and mean |CTlig| values are shown for comparison. In naringenin (Figure C), the C2–C3 single bond breaks this conjugation. The B ring, although structurally present, is electronically disconnected from the A–C ring system. The dashed gray arrow with the red “×” symbol indicates that the electronic communication pathway is disrupted. This structural difference provides a clear electronic-level explanation for the well-established empirical SAR observation that the C2C3 double bond is essential for flavonoid XO inhibitory activity. Detailed Charge Distribution at the Flavonoid–MoCo Interface Figure presents the Mulliken charge distribution at the kaempferol–MoCo interface for a representative snapshot (t = 10 ns, CTlig = – 0.690 e). The kaempferol hydroxyl groups carry substantial negative charges (−0.178 to – 0.376), functioning as electron-rich regions that facilitate hydrogen bonding and electrostatic interactions with the MoCo center. The Mo atom retains a net positive charge (+0.236), consistent with its role as an electrophilic center. The sulfur atoms of MoCo carry near-neutral charges (−0.020 and – 0.064, yellow circles), while the phosphorus atom is strongly positive (+0.943, orange), reflecting its coordination with three electronegative oxygen atoms (−0.443 to – 0.486). Mulliken charge distribution at the kaempferol–MoCo interface from a representative QM/MM snapshot (B3LYP/def2-SVP; CTlig = – 0.690 e). Red circles: negatively charged atoms (electron-rich); yellow circles: near-neutral atoms (S); blue circle: Mo (electron-poor, + 0.236); orange circle: P (+0.943). The blue arrow indicates the interface of net charge redistribution between kaempferol and MoCo. Hirshfeld Validation and Electrostatic Interaction Analysis To assess whether the Mulliken-derived |CTlig| ranking is an artifact of the Mulliken partition scheme, Hirshfeld charge analysis was performed independently on all 50 QM/MM snapshots (Table S4). The Hirshfeld |CTlig| values reproduced the Mulliken compound hierarchy: apigenin (0.521 ± 0.224 e) > kaempferol (0.362 ± 0.220 e) > quercetin (0.353 ± 0.096 e) > morin (0.262 ± 0.015 e) > naringenin (0.056 ± 0.023 e), with a strong intermethod correlation (Pearson r = 0.942, p = 0.017; Spearman ρ = 0.900, p = 0.037; Figure B). Hirshfeld absolute values were systematically higher than Mulliken values (mean ratio 1.21), consistent with the known differences between Hilbert-space and real-space partitioning schemes. Critically, the standard deviations across snapshot standard deviations were nearly identical between methods (e.g., morin: 0.015 e by both Hirshfeld and Mulliken), confirming that the observed variability reflects genuine conformational dynamics rather than partition-scheme artifacts. Correlation analysis of charge-redistribution and interaction metrics with inhibitory potency. (A) Hirshfeld |CTlig| versus Ki (r = – 0.925, p = 0.024; Spearman ρ = – 1.000), showing complete rank-order agreement within this five-compound series. (B) Cross-validation of Hirshfeld and Mulliken |CTlig| values (r = 0.943, p = 0.016), confirming partition-scheme independence of the charge-redistribution ranking. (C) Hirshfeld electrostatic interaction energy (ΔECoul) versus Ki (r = – 0.890, p = 0.043), supporting an electrostatic energetic component of the charge redistribution–potency association. (D) IGMH δginter versus Ki (r = – 0.062, p = 0.922, ns), showing that the representative-snapshot IGMH descriptor did not reproduce the Ki ranking. Solid lines represent linear regression fits with 95% confidence bands (gray shading); the dashed line in panel D indicates a nonsignificant relationship. In panels A–C, each data point represents the compound-level mean of 10 QM/MM snapshots (10–100 ns). In panel D, each point represents one median-|CTlig| representative snapshot per compound. Hirshfeld |CTlig| correlated significantly with Ki (r = – 0.925, p = 0.024; Spearman ρ = – 1.000; Figure A), showing complete rank-order agreement within this five-compound series. To explore whether the charge-redistribution descriptor contains an electrostatic energetic component, the Hirshfeld electrostatic interaction energy (ΔECoul) was computed from the pairwise Coulombic interaction between ligand and receptor fragments using Hirshfeld charges. ΔECoul also correlated significantly with Ki (r = – 0.890, p = 0.043; Figure C). In addition, |CTlig| was strongly associated with ΔECoul (r = +0.973, p = 0.005), and both descriptors were associated with Ki, suggesting that the charge-redistribution descriptor has an electrostatic energetic component. In contrast, the IGMH interfragment interaction density (δginter), which integrates all interfragment electron density deformation including hydrogen bonds, van der Waals contacts, and steric effects, showed no correlation with Ki (r = – 0.062, p = 0.922; Figure D). Morin exhibited the largest δginter (0.726 au) despite only moderate Ki (7.77 μM), while apigenin showed a smaller contact surface (δginter = 0.335 au) with superior potency (Ki = 1.86 μM). This dissociation indicates that, within this compound series, inhibitory potency is more closely associated with charge-redistribution magnitude than with total interfragment contact area (Table S5; Figure S2). To evaluate basis-set sensitivity, single-point QM/MM calculations were repeated at the def2-TZVP level for one representative snapshot per compound. The def2-TZVP calculations preserved the overall SVP-derived hierarchy, with only a swap between apigenin and kaempferol at the top two positions (Spearman ρ = 0.900, p = 0.037; Table S3b). This interchange is consistent with the broad and overlapping SVP ensemble distributions of both compounds (apigenin: 0.468 ± 0.211 e; kaempferol: 0.353 ± 0.221 e), whereas the ranking of morin, quercetin, and naringenin was unchanged. Notably, naringenin remained the lowest-ranked compound at both basis-set levels, preserving the separation between conjugated planar flavonoids and the nonplanar flavanone. Leave-one-compound-out sensitivity analysis preserved the negative direction of both the |CTlig|−Ki and ΔECoul–Ki associations in all five subsets, indicating that the observed trends were not driven by any single compound (Table S6). The color-coded scheme (red: electron-rich/negative; yellow: near-neutral; blue: electron-poor/positive) provides an intuitive visualization of the charge distribution landscape at the catalytic interface, illustrating the electrostatic complementarity between electron-rich regions of the flavonoid and the positively charged Mo-centered environment. Discussion Charge Redistribution as a Potential Electronic Contributor to Inhibition in MoCo-Bound Configurations The present QM/MM study provides an exploratory QM/MM perspective on how ligand-derived charge redistribution in MoCo-bound complexes relates to XO inhibitory potency. The negative correlation between |CTlig| and IC50 (r = – 0.833, R2 = 0.693; Figure ) suggests that flavonoids capable of exhibiting larger ligand-fragment charge-redistribution magnitudes tended to be more potent XO inhibitors, though this observation is based on a limited compound set (n = 5) and the correlation does not reach conventional statistical significance (p = 0.08). This finding extends the mechanistic understanding beyond what is achievable with classical force-field-based methods, which treat atomic charges as fixed parameters and therefore cannot capture the electronic reorganization that occurs upon ligand binding. It is important to note that the IC50 measured in the companion study reflects overall enzyme inhibition, which is a composite outcome of binding at MoCo, FAD, or both sites, and is not a direct readout of MoCo-specific electronic perturbation. The present CTlig analysis specifically characterizes the electronic interaction in MoCo-bound configurations. For kaempferol, which exhibits the strongest MoCo-site preference among the five flavonoids (Ki = 2.43 μM, ΔΔG = +3.01 kcal/mol favoring MoCo), the correspondence between large |CTlig| and low IC50 is the most mechanistically interpretable: kaempferol not only binds stably at MoCo but also exhibits substantial charge redistribution within the MoCo-bound QM subsystem (|CTlig| = 0.353 e), potentially perturbing the electronic environment of the catalytic Mo center and impairing its oxidative capacity. For the remaining flavonoids (apigenin, quercetin, morin, and naringenin), which show weaker MoCo-site preference, the CTlig at MoCo is less directly relevant to their overall inhibitory potency, because their inhibition involves both the MoCo and FAD domains. Taken together, the companion MD/MM-GBSA study and the present QM/MM analysis support a hierarchical model in which overall XO inhibition depends first on site preference and binding stability, and second, for compounds that productively occupy the MoCo active site, on the extent of ligand-derived electronic perturbation of the catalytic metal center. A related question is whether a minimum threshold of charge redistribution is required for effective inhibition. Based on our data from five flavonoids, all compounds exhibit measurable charge redistribution upon binding to MoCo, including naringenin (|CTlig| = 0.051 e), which is the weakest inhibitor in the series. The data therefore do not support a binary threshold model; rather, they suggest a continuous, monotonic relationship in which greater charge redistribution correlates with stronger inhibition. However, with only five compounds spanning limited structural diversity, the possibility of a minimum threshold for biologically relevant inhibition cannot be excluded. Establishing such a threshold would require expanding the compound library to include noninhibitory flavonoids or structurally related compounds that bind without measurable charge transfer. Electronic Basis of the C2C3 Structure–Activity Relationship The importance of the C2C3 double bond for flavonoid XO inhibitory activity has been recognized empirically for decades. Cos et al. classified this structural feature among the essential pharmacophoric elements, and the potency ranking of flavonoid subclasses (flavonols/flavones ≫ flavanones) has been repeatedly confirmed. However, the quantum-mechanical origin of this structure–activity relationship has remained unresolved. Our HOMO analysis (Figure ) provides a mechanistic rationale for this long-standing SAR: the C2C3 double bond maintains π-conjugation across the entire flavonoid scaffold, enabling the ligand-centered HOMO to delocalize across all three rings. This delocalization creates a continuous electronic communication pathway between the flavonoid and MoCo. When this conjugation is disrupted, as in naringenin where the C2–C3 single bond confines the HOMO to the A and C rings, the electronic communication pathway is severed, and charge-redistribution magnitude drops by approximately ∼ 7-fold (from 0.353 to 0.051 e). This insight has implications for rational design of flavonoid-based XO inhibitors: structural modifications that extend π-conjugation (e.g., additional unsaturation or aromatic substituents on the B ring) may enhance charge redistribution and thereby improve inhibitory potency, while modifications that disrupt conjugation should be avoided. The Apigenin Anomaly: Charge Redistribution Alone Does Not Predict IC50 Apigenin presented an instructive anomaly in the CTlig–IC50 correlation. Despite exhibiting the highest mean |CTlig| among the five flavonoids, apigenin showed only moderate inhibitory potency (IC50 = 8.14 μM), ranking behind kaempferol (4.21 μM). This observation underscores that charge redistribution at MoCo is one contributing factor rather than the sole determinant of overall XO inhibition. In our previous study, apigenin was confirmed to exhibit mixed-type inhibition (Ki = 1.86 μM) with essentially no site preference (mean ΔΔG = +0.55 kcal/mol) and showed relatively high ligand RMSD values at both binding sites, suggesting lower binding stability compared to kaempferol. The overall IC50 likely reflects a combination of factors: binding affinity and residence time at each site, charge-redistribution magnitude when bound at MoCo, and the relative population of MoCo-bound versus FAD-bound states under assay conditions. The R2 value of 0.693 is consistent with this multifactorial picture, and the limited sample size (n = 5) precludes definitive mechanistic partitioning. This discrepancy suggests that productive occupation of the MoCo site, rather than merely the magnitude of charge redistribution, is a prerequisite for CTlig to translate into observable inhibitory potency. Comparison with Previous Computational Studies Previous computational studies of flavonoid–XO interactions have predominantly employed molecular docking, classical MD simulations, and MM-GBSA or MM-PBSA binding free energy calculations. , These approaches have provided valuable insights into binding poses, hydrogen bond networks, conformational stability, and relative binding affinities. However, they fundamentally cannot address the electronic basis of inhibition because the fixed-charge approximation inherent in classical force fields precludes charge redistribution upon binding. Several DFT studies have examined the electronic properties of isolated flavonoids, − including HOMO–LUMO gaps and electrostatic potential surfaces, but these gas-phase calculations do not account for the protein environment, which significantly modulates the electronic structure through electrostatic embedding. The present study bridges this gap by embedding the QM calculation within the realistic protein environment via the QM/MM approach, ensuring that the calculated charge-redistribution values reflect the actual electronic perturbation occurring at the enzyme active site. An important methodological consideration is whether omitting the MM component would affect the current results. The MM environment provides electrostatic embedding that accounts for the protein electric field at the active site, which can influence the charge distribution at the flavonoid–MoCo interface. Omitting the MM component (i.e., performing gas-phase QM-only calculations) would remove the polarizing effect of surrounding protein residues (e.g., Glu802, Arg912, Glu1261) and solvent, potentially altering the absolute magnitude of the charge-redistribution values. Whether the compound ranking would be preserved cannot be established from the present calculations and would require a systematic paired comparison of QM-only and QM/MM calculations. Such an analysis is beyond the scope of the present study but represents a worthwhile future investigation. To contextualize our findings, we note that ligand-to-metal charge transfer is a recognized feature of molybdopterin enzyme chemistry. Matz et al. demonstrated that noninnocent dithiolene ligands in oxomolybdenum complexes can modulate metal-center redox properties through charge redistribution, a phenomenon directly relevant to the MoCo electronic environment examined here. Similarly, Leopoldini et al. showed through DFT calculations that polyphenol–metal interactions involve significant electron density redistribution at transition-metal centers, supporting the plausibility of the flavonoid-induced charge redistribution observed in our QM/MM calculations. This study extends prior computational investigations of flavonoid–XO interactions by introducing a QM/MM-based analysis of ligand-derived charge redistribution at the metalloenzyme active site and correlating the magnitude of this redistribution with experimental inhibitory activity. The exploratory nature of the present work, based on single-point calculations on 10 snapshots per compound, should be noted; it provides a snapshot-based electronic portrait rather than a fully converged free-energy-level description. Nonetheless, the demonstrated robustness of CTlig to spin-state assignment (Table S3a) lends confidence to the comparative trends reported here. Limitations Several limitations of the present study should be acknowledged. First, Mulliken population analysis is basis-set-dependent and may overestimate or underestimate absolute charge values; however, the relative trends across compounds calculated with the same basis set (def2-SVP) are internally consistent and suitable for comparative analysis. However, the Hirshfeld charge analysis performed in this study confirmed that the Mulliken-derived rankings are robust to the charge-partitioning scheme, and def2-TZVP calculations on representative snapshots preserved the principal compound hierarchy (Spearman ρ = 0.900, p = 0.037; Table S3b) (r = 0.942, p = 0.017; Table S4), with Hirshfeld |CTlig| showing complete rank-order agreement with Ki (Spearman ρ = – 1.000), addressing this limitation directly. Second, the QM/MM calculations were performed as single-point energy evaluations on fixed MD snapshots rather than as Born–Oppenheimer QM/MM molecular dynamics. While the use of 10 snapshots per flavonoid captures some of the conformational variability, the sampling is sparse. More densely sampled snapshots or dynamic QM/MM MD simulations would provide a more statistically robust description of the electronic properties. Third, although the QM region included ∼ 17 side-chain residues within 4 Å of MoCo, the variable composition of this region across snapshots (210–266 atoms) introduces a potential source of variability in the calculated charges. Future work could employ a fixed QM region across all snapshots to ensure strict comparability, at the cost of some configurations having suboptimal QM/MM boundaries. Fourth, the correlation analysis (n = 5, p = 0.08) is based on a limited number of compounds and does not reach conventional statistical significance. Expansion to additional flavonoids and nonflavonoid XO inhibitors would be necessary to confirm the generality of the CTlig–IC50 relationship. Fifth, the present study focused exclusively on MoCo-bound complexes. Because the companion study demonstrated that all five flavonoids exhibit mixed-type inhibition involving both MoCo and FAD domains, the CTlig values reported here characterize only the MoCo-bound electronic state, which may not be the predominant binding mode for these compounds under physiological conditions. QM/MM analysis of FAD-bound complexes would be a valuable complement. Finally, UKS converged to broken-symmetry solutions with significant open-shell character (⟨S2⟩ ≈ 3.72), and spin-state sensitivity testing revealed near-degenerate singlet–quintet energy gaps (∼1 kcal/mol), indicating multireference character. While CTlig was demonstrated to be robust across spin states (Table S3a), the absolute electronic structure description should be interpreted cautiously. Future studies employing multireference methods (e.g., CASSCF/NEVPT2) would provide a more rigorous treatment of the MoCo electronic structure. Conclusions This study provides exploratory quantum mechanical evidence that ligand-derived charge redistribution in MoCo-bound complexes is a plausible electronic contributor to flavonoid-mediated XO inhibition. The main findings are(1)All five dietary flavonoids examined exhibit net charge redistribution between the ligand fragment and the MoCo + protein subsystem upon binding, but the magnitude of charge redistribution varies approximately 9-fold, from 0.051 e (naringenin) to 0.468 e (apigenin).(2)The magnitude of charge redistribution correlates negatively with experimental IC50 values (r = – 0.833, R2 = 0.693, n = 5), suggesting that greater charge redistribution is associated with stronger inhibition, though the limited sample size warrants caution in generalizing this relationship.(3)Frontier molecular orbital analysis reveals that the C2C3 double bond is a key structural feature enabling charge redistribution: it maintains π-conjugation across the flavonoid scaffold, allowing the ligand-centered HOMO to delocalize to the B ring and thereby facilitating electronic communication with MoCo. Saturation of this bond (as in naringenin) confines the orbital density to the A–C ring system and markedly attenuates the calculated charge-redistribution magnitude.(4)The Mo Mulliken charge decreases from the gas-phase reference (+0.347) upon flavonoid binding; however, this decrease reflects composite contributions from the ligand, protein side chains, and electrostatic embedding, and should not be interpreted solely as ligand-derived donation.(5)Hirshfeld charge analysis independently confirmed the Mulliken-derived CTlig ranking (r = 0.942, p = 0.017), with complete rank-order agreement between Hirshfeld |CTlig| and Ki (Spearman ρ = – 1.000). The electrostatic interaction energy derived from Hirshfeld charges also correlated with Ki (r = – 0.890, p = 0.043), supporting an electrostatic energetic component of the observed association between charge redistribution and inhibitory potency. In contrast, total interfragment interaction density (IGMH δginter) showed no correlation with Ki (r = – 0.062, p = 0.922), suggesting that, within this limited compound series, Ki was more closely associated with charge redistribution and electrostatic interaction than with the representative-snapshot IGMH contact-density descriptor. These observations, combined with the binding affinity and site-preference data from our companion MD/MM-GBSA study, are consistent with a model in which inhibition by kaempferol at the MoCo site involves both restriction of the substrate access tunnel (bottleneck radius narrowing demonstrated by CAVER 3.0 analysis) and electronic perturbation of the Mo catalytic center through charge redistribution. For compounds that engage both binding sites, the MoCo-specific charge redistribution characterized here represents only one component of their overall inhibitory mechanism. Future studies incorporating QM/MM analysis of FAD-bound complexes and expanded compound sets will be needed to further delineate the relative contributions of MoCo-site electronic effects versus FAD-site binding to overall XO inhibition. (1)All five dietary flavonoids examined exhibit net charge redistribution between the ligand fragment and the MoCo + protein subsystem upon binding, but the magnitude of charge redistribution varies approximately 9-fold, from 0.051 e (naringenin) to 0.468 e (apigenin). All five dietary flavonoids examined exhibit net charge redistribution between the ligand fragment and the MoCo + protein subsystem upon binding, but the magnitude of charge redistribution varies approximately 9-fold, from 0.051 e (naringenin) to 0.468 e (apigenin). (2)The magnitude of charge redistribution correlates negatively with experimental IC50 values (r = – 0.833, R2 = 0.693, n = 5), suggesting that greater charge redistribution is associated with stronger inhibition, though the limited sample size warrants caution in generalizing this relationship. The magnitude of charge redistribution correlates negatively with experimental IC50 values (r = – 0.833, R2 = 0.693, n = 5), suggesting that greater charge redistribution is associated with stronger inhibition, though the limited sample size warrants caution in generalizing this relationship. (3)Frontier molecular orbital analysis reveals that the C2C3 double bond is a key structural feature enabling charge redistribution: it maintains π-conjugation across the flavonoid scaffold, allowing the ligand-centered HOMO to delocalize to the B ring and thereby facilitating electronic communication with MoCo. Saturation of this bond (as in naringenin) confines the orbital density to the A–C ring system and markedly attenuates the calculated charge-redistribution magnitude. Frontier molecular orbital analysis reveals that the C2C3 double bond is a key structural feature enabling charge redistribution: it maintains π-conjugation across the flavonoid scaffold, allowing the ligand-centered HOMO to delocalize to the B ring and thereby facilitating electronic communication with MoCo. Saturation of this bond (as in naringenin) confines the orbital density to the A–C ring system and markedly attenuates the calculated charge-redistribution magnitude. (4)The Mo Mulliken charge decreases from the gas-phase reference (+0.347) upon flavonoid binding; however, this decrease reflects composite contributions from the ligand, protein side chains, and electrostatic embedding, and should not be interpreted solely as ligand-derived donation. The Mo Mulliken charge decreases from the gas-phase reference (+0.347) upon flavonoid binding; however, this decrease reflects composite contributions from the ligand, protein side chains, and electrostatic embedding, and should not be interpreted solely as ligand-derived donation. (5)Hirshfeld charge analysis independently confirmed the Mulliken-derived CTlig ranking (r = 0.942, p = 0.017), with complete rank-order agreement between Hirshfeld |CTlig| and Ki (Spearman ρ = – 1.000). The electrostatic interaction energy derived from Hirshfeld charges also correlated with Ki (r = – 0.890, p = 0.043), supporting an electrostatic energetic component of the observed association between charge redistribution and inhibitory potency. In contrast, total interfragment interaction density (IGMH δginter) showed no correlation with Ki (r = – 0.062, p = 0.922), suggesting that, within this limited compound series, Ki was more closely associated with charge redistribution and electrostatic interaction than with the representative-snapshot IGMH contact-density descriptor. These observations, combined with the binding affinity and site-preference data from our companion MD/MM-GBSA study, are consistent with a model in which inhibition by kaempferol at the MoCo site involves both restriction of the substrate access tunnel (bottleneck radius narrowing demonstrated by CAVER 3.0 analysis) and electronic perturbation of the Mo catalytic center through charge redistribution. For compounds that engage both binding sites, the MoCo-specific charge redistribution characterized here represents only one component of their overall inhibitory mechanism. Future studies incorporating QM/MM analysis of FAD-bound complexes and expanded compound sets will be needed to further delineate the relative contributions of MoCo-site electronic effects versus FAD-site binding to overall XO inhibition. Hirshfeld charge analysis independently confirmed the Mulliken-derived CTlig ranking (r = 0.942, p = 0.017), with complete rank-order agreement between Hirshfeld |CTlig| and Ki (Spearman ρ = – 1.000). The electrostatic interaction energy derived from Hirshfeld charges also correlated with Ki (r = – 0.890, p = 0.043), supporting an electrostatic energetic component of the observed association between charge redistribution and inhibitory potency. In contrast, total interfragment interaction density (IGMH δginter) showed no correlation with Ki (r = – 0.062, p = 0.922), suggesting that, within this limited compound series, Ki was more closely associated with charge redistribution and electrostatic interaction than with the representative-snapshot IGMH contact-density descriptor. These observations, combined with the binding affinity and site-preference data from our companion MD/MM-GBSA study, are consistent with a model in which inhibition by kaempferol at the MoCo site involves both restriction of the substrate access tunnel (bottleneck radius narrowing demonstrated by CAVER 3.0 analysis) and electronic perturbation of the Mo catalytic center through charge redistribution. For compounds that engage both binding sites, the MoCo-specific charge redistribution characterized here represents only one component of their overall inhibitory mechanism. Future studies incorporating QM/MM analysis of FAD-bound complexes and expanded compound sets will be needed to further delineate the relative contributions of MoCo-site electronic effects versus FAD-site binding to overall XO inhibition. Supplementary Material