All of the reagents are analytical grade or chromatographically pure, and were purchased from commercial sources. All solutions were prepared using double-distilled water. Dihydromyricetin (DHM) was extracted by our laboratory from tender stems and leaves of Ampelsis grossedentata with more than 95% purity. Standard DHM was purchased from Aladdin Reagent, Shanghai (China). Metal ion reagents (copper, zinc, iron, nickel, and cobalt acetates) were also bought from Aladdin Reagent, Shanghai (China). Organic solvents were obtained from Kemiou reagent, Tianjin (China).

According to the literature (Tu et al., 2016; Maitera et al., 2018; Thangavel et al., 2018; Khater et al., 2019; Gençkal et al., 2020), the general synthetic procedure for DHM–M(II) complexes is as follows: solid DHM (0.5820 g, 1.82 mmol) was dissolved completely in ethanol (5 mL, 95%, W/W). Then, the mixture was transferred to 50 mL of an alkaline solution (sodium acetate dissolved in a water–ethanol solution (50%, v/v) at pH = 8.2) under stirring for 5 min. Afterward, metal ion reagents (2 mmol) (copper, zinc, iron, nickel, and cobalt acetates) were separately dissolved in ethanol (5 mL, 95%, W/W) and added to the above mixture under reflux for 120 min. The reaction mixture was cooled to room temperature, precipitated, and filtered, and the resulting solid was washed three times with EtOH and double-distilled water in turn. Subsequently, after re-crystallization in acetone, the solid product was dried under vacuum for 48 h at −20°C to afford the DHM–M(II) complexes.

The DHM–M(II) complexes (0.9050 g) were each added to an acidic solution (20 mL, 1 mL HCl in 100 mL of ethanol, pH = 2) and stirred for 3 h at 50°C under a nitrogen atmosphere. Afterward, the reaction mixtures were moderated to pH = 5–6 by 0.1% aqueous NaOH. After filtering and evaporation, the residues were dissolved in chromatographic methanol and subjected to HPLC analysis.

A 100 mg/mL DHM solution was prepared, to which 3–20 mL of an alkaline solution (pH = 8.2) was added, and the reaction mixture was stirred continuously. After 2 h, the mixture was quenched with 0.1% HCl to pH = 5–6. After evaporating the solvent under vacuum, the residue was dissolved in ethanol and dried over MgSO₄. Subsequently, after filtering and evaporation of the solvent under vacuum, the solution was evaporated to dryness. The residue was separated and purified by column chromatography on silica gel (EtOAc/petroleum ether from 1/25 to 1/3) to afford product A (50.2 mg) as a yellow solid. ¹H-NMR (500 MHz, DMSO-D₆, 25°C, TMS): δ (ppm) 12.51 (s, 1H), 10.82 (s, 1H), 9.36 (s, 1H), 9.25 (s, 2H), 8.83 (s, 1H), 7.25 (s, 2H), 6.38 (s, 1H), 6.19 (s, 1H); ¹³C-NMR (125 MHz, DMSO-D₆, 25°C, TMS): δ (ppm) 176.21, 164.31, 161.17, 156.52, 147.27, 146.16, 136.31, 121.23, 107.60, 103.42, 98.61, 93.66. Anal. Calcd for C₁₅H₁₀O₈: C, 56.61; H, 3.17; O, 40.22; Found: C, 56.45; H, 3.24. ESI-MS m/z (CH₃OH): 318.04; Found: 358 [M+H+K]⁺.

A solution of product A at 1.0 mg/mL was prepared and subjected to HPLC analysis. Chromatographic conditions: Hypersil BDS C18 (4.6^*200 mm, 5 um), Elute: (Acetonitrile / Water (0.1% phosphoric acid) = 24 / 76), Flow rate = 1 ml/min, UV = 254 nm.

Using an air supplier, air, nitrogen, and oxygen were continuously forced into separate alkaline solutions (pH = 8.2) for 10 min. DHM was dissolved in methanol to 0.20 mg/mL, and 100 μL was added to 1 mL of each of the air-, oxygen-, and nitrogen-saturated alkaline solutions with continuous stirring for 1 min. To 100 μL of those solutions, 5,5-dimethyl-1-pyrroline-N-oxide (DMPO) (100 mg/L in methanol, 10 μL) was added, and the resulting mixtures were transferred into silica tubes and subjected to EPR analysis.

Dry DMSO (1 mL) was transferred to a test tube (5 mL), and oxygen was introduced to the system for 10 min at room temperature using an air supplier. Sodium phenolate (100 mmol/L, 100 μL) was immediately added to the system and stirred slowly for 30 min at 37°C under an oxygen atmosphere. DMPO (100 mg/L in methanol, 10 μL) was added to 100 μL of this solution, and the resulting mixture was transferred into a silica tube and subjected to EPR analysis.

To the above described DMSO system, a DHM solution (1.0 mg/mL in dry DMSO) was added with shaking for ~5 min. When the solution changed from colorless to yellow, it was quenched with saturated aqueous NH₄Cl. After filtering (0.45 μm), the solution was subjected to HPLC analysis using the above chromatographic conditions.

Following the above conditions for the DHM generation of superoxide anions, DHM analogs (MYR, quercetin, daidzein, genistein, chrysin, baicalein, rutin hydrate, and kaempferol) were each prepared at the same concentration (2.0 mg/mL). The solutions were transferred to an air-saturated alkaline solution (1 mL, pH = 8.2) with continuous stirring for 1 min. To 100 μL of these solutions DMPO (100 mg/L in methanol, 10 μL) was added. The resulting mixtures were transferred to silica tubes and subjected to EPR analysis.

By following the reported general synthetic procedure (Tu et al., 2016; Maitera et al., 2018; Thangavel et al., 2018; Khater et al., 2019; Gençkal et al., 2020), the DHM–M(II) complexes DHM–Co(II), DHM–Cu(II), DHM–Fe(II), DHM–Zn(II), and DHM–Ni(II) were obtained. The metal complexes were characterized by FT-IR, elemental analysis, UV-vis, and antioxidant capacity. The IR absorption frequency of the functional groups attached to the DHM–M(II) complexes were red-shifted compared with those of free DMY, indicating the complexation of DHM with metal ions (Supplementary Figure 1 and Supplementary Table 1). Elemental analysis demonstrated that the DHM–metal(II) complexes had the expected compositions (Supplementary Table 2). The UV-vis spectra showed red-shifts and changes in intensity of the absorption of the complexes, further evidencing the complexation of DHM with the metal ions under the selected experimental conditions (Supplementary Figure 2). Finally, the antioxidant activities of the DHM–M(II) complexes were confirmed as higher than that of DHM at the same concentration (Supplementary Figure 3 and Supplementary Table 3). Therefore, it is reasonable to assume that the prepared DHM–M(II) complexes are consistent with those in previous studies.

To further understand whether DHM changes during complexation, DHM and its metal complexes were hydrolyzed under the same acidic conditions. After stirring for 3 h at 50°C, the resulting mixtures were subjected to HPLC with a monitoring wavelength of 254 nm. In the LC spectra, we found that free DHM stirred in an acidic solution was stable, without any impurity peaks. However, in the spectra of the DHM–M(II) complexes, many unknown absorption peaks with different intensities appeared (Supplementary Figure 4). This caused us to question the general synthetic procedure for bioflavonoid–metal complexes. Subsequently, to address this, the synthetic conditions of the metal ion, alkalinity, and reflux time were investigated by UV-vis or HPLC to observe their influence on the stability of free DHM.

Reflux time was first investigated by refluxing free DHM for different times in a water–ethanol solution (v/v, 50%) followed by observation by UV-vis spectroscopy. In the resulting spectra, no notable shift in the DHM absorption peak at 325 nm was observed after refluxing for several minutes compared with free DHM without any workup (Figure 3). This indicated that reflux time does not significantly affect the stability of free DHM in an aqueous solution in the absence of alkalinity or metal ions. Therefore, we did not consider reflux time to be a crucial factor affecting the synthesis of the DHM–metal complexes.

To the best of our knowledge, bioflavonoids are generally weakly acidic in ethanol or aqueous solutions, relying on their phenolic hydroxyl moieties to ionize in the liquid phase. In the general synthetic procedure for bioflavonoid–metal complexes, alkalinity promotes the deprotonation of the bioflavonoid phenolic hydroxyl groups to form the corresponding phenolate. Since phenolates can easily donate a pair of electrons, they are better ligands to chelate with metal ions. Thus, alkalinity seems to be a dominant factor in synthesizing DHM–M(II) complexes. A pH of 8.2 for the reaction solution was selected to investigate the influence of alkalinity on free DHM by UV-vis spectroscopy. With longer stirring times at room temperature, the absorbance at 290 nm shifted to 325 nm and increased in intensity (Figure 4), implying that dissolving DHM in an air-saturated alkaline solution decreased the amount of DHM over time. Therefore, the effect of alkalinity on DHM stabilization required further investigation.

HPLC as a quantitative and qualitative analysis method was first utilized to investigate the effect of alkalinity on free DHM stabilization (Supplementary Figures 5–8 and Supplementary Tables 4–6). As shown in Figure 5, the air-saturated alkaline solution was adjusted to a pH of 7.4 by sodium acetate/acetic acid, and free DHM was stirred for different times in the solution at room temperature. The resulting mixtures were quenched with 0.1% HCl to approximately pH = 6 and subjected to HPLC. We found that the concentration of DHM decreased from 0.71 to 0.64 mg/mL after stirring for 120 min. When the solution pH was moderated to 8.2, the DHM concentration decreased more than at pH = 7.4; similarly, the concentration decrease was the most severe for pH = 9.4. This confirmed the alkalinity as a significant factor in the synthesis of the DHM–M(II) complexes.

The metal ion is an important participant in the synthesis of the DHM–M(II) complexes as it supplies empty orbitals for complexation with the ligand. Importantly, most transition metal ions in an electron-loss state are electronegative and are thus generally considered to be oxidants. In addition, bioflavonoids are electron-rich and thus easily lose electrons after deprotonation of their phenolic hydroxyl moieties. The presence of electronegative metal ions with the phenolates derived from deprotonated bioflavonoids can initiate a redox reaction. Therefore, the effect of metal ions with different electronegativities on DHM stabilization deserves further consideration. DHM was stirred for 2 h in the presence of the different metal ions without alkalinity at room temperature, and the resulting mixtures were subjected to UV-vis analysis. As shown in Figure 6, the intensity and position of the DHM absorption peak were affected by the metal ion. With Zn²⁺, the DHM peak red-shifted from 290 to 345 nm, indicating the effective complexation of Zn²⁺ with free DHM. In contrast, Ni²⁺, Cu²⁺, Co²⁺, and Fe²⁺ not only red-shifted the peak at 290 nm to different extents but also changed the intensity, suggesting that the redox of free DHM also occurred. These results showed that the metal ions affect the DHM stability according to their electronegativity in the order of Ni²⁺ > Co²⁺ > Cu²⁺ > Fe²⁺ > Zn²⁺. Therefore, the metal ion is a dominant factor affecting the stability of DHM in DHM–M(II) complexes.

In summary, synthesizing bioflavonoid–metal complexes in air-saturated alkaline solutions is rational in theory. However, in reality, both the alkalinity and metal ions as important participants in the synthesis of DHM–M(II) complexes were shown to be probable reasons for the lack of stability.

Based on the above results regarding the synthetic factors affecting DHM stabilization, it was notable that alkalinity is likely the main factor for the general synthesis of bioflavonoid–metal complexes. In the general synthetic procedure, free DHM was first stirred for different times in an air-saturated alkaline solution (pH = 8.2) for deprotonation into the corresponding phenolate, and the resulting solution was subjected to HPLC. The intensity of the absorption peak (t =15.7 min) at 254 nm increased with stirring time (Supplementary Figure 7). Subsequently, the absorption peak was separated by column chromatography with an eluent of P/E (from 25/1 to 3/1) and identified by ¹H- and ¹³C-NMR, ESI-MS, elemental analysis, and HPLC (Supplementary Figures 9, 10). Interestingly, the component was found to be MYR. As far as we know, MYR is the oxidized product of DHM. Thus, the transformation of DHM into MYR in an air-oxidized alkaline solution required further consideration (Scheme 1).

For the transformation of DHM into MYR, DHM after deprotonation by a base forms its electron-rich phenolate, which undergoes a redox reaction upon exposure to an oxidant. We noted that oxygen molecules from the air can dissolve into the alkaline solution. The collision of diatomic oxygen with the phenolate can promote the univalent reduction (single electron transfer) of diatomic oxygen, resulting in one filled and one half-filled antibonding π^*2p molecular orbitals. This ultimately forms a radical. To validate our proposition, electron paramagnetic resonance (EPR) was applied to observe the radicals generated from DHM dissolved in an air-saturated alkaline (pH = 8.2) solution with DMPO as a spin-trapping electron reagent. DHM dissolved in oxygen- and nitrogen-saturated alkaline solutions at the same pH was also observed for comparison. The EPR results demonstrated that the intensity of the typical signals of the superoxide anion adduct (DMPO-OO–) of DHM in an air-saturated alkaline solution was less than that of the oxygen-saturated alkaline solution, with an overlapping “doublet of a doublet” arising from hyperfine interactions with ¹⁴N (Liu et al., 1996). When DHM was dissolved in a nitrogen-saturated alkaline solution, no typical adduct signals were observed (Figure 7). Thus, DHM dissolved in an alkaline solution in the presence of oxygen molecules generates superoxide-anion radicals (O₂⁻) via univalent reduction.

Next, whether the transformation of DHM into MYR in an air-saturated alkaline solution is associated with O₂⁻ generation was considered. A system of dry DMSO reacted with sodium phenolate in the presence of oxygen molecules to generate O₂⁻ was selected according to a previous report (Branchini et al., 2015). The DMSO system was analyzed by EPR with the aid of DMPO. The typical signals of the DMPO-OO–adduct derived from the DMSO system are shown below (Figure 8).

Subsequently, DHM was added to this system of generated superoxide anions with full shaking. Within a few minutes, the color of the resulting solution gradually changed from colorless to yellow. After quenching with a 0.1% HCl solution, the resulting mixture was subjected to HPLC with MYR as a standard for comparison (Supplementary Figure 11). As a result, the compound in the reaction mixture was identified as MYR from the reaction of DHM with superoxide-anion radicals (O₂⁻). Therefore, we proposed a pathway of DHM transformation into MYR that involves the abstraction of a C3–H atom of phenolate 1 abstracted by O₂⁻ to form the potent radical 2. Phenolate 1 was obtained by the deprotonation of the phenolic hydroxyl moieties of DHM. Subsequently, peroxide 3 was generated from the combination of radical 2 with peroxyl radical HOO·. Then, under acidic conditions, terminal product MYR was formed by successive acidification, dehydration, and enolization (Branchini et al., 2015) (Scheme 2).

Since DHM generates O₂⁻ in an oxygenated alkaline solution, its analogs MYR, quercetin, daidzein, genistein, chrysin, baicalein, rutin hydrate, and kaempferol were also considered as potential ligands for metal complexes based on the multi-hydroxyl moieties attached to their molecular backbones (Figure 9). Thus, their behaviors in an air-saturated alkaline solution were investigated by EPR. All of these bioflavonoids originate from natural products and were tested at the same concentration in an air-saturated alkaline solution, then observed by EPR with DMPO (Figure 10). We found that the intensities of the DMPO-OO–adduct signal could be correlated to the location and/or number of phenolic hydroxyl groups or their configurations.

As shown in Table 1, the more phenolic hydroxyl groups are attached to the molecular skeleton, the higher the signal strength of the DMPO-OO–adduct in general. Comparing DHM with MYR, which have the same number of hydroxyl groups, the adduct signal was stronger for MYR under the same conditions. This indicated that molecular configuration contributes to the capacity for generating superoxide anions, potentially because of the transformation of DHM into MYR in the presence of O₂⁻. For the flavonoids with three hydroxyl moieties, baicalein showed a higher adduct signal than genistein. At the same time, chrysin with two hydroxyl groups at the 5,7-positions of the molecular skeleton exhibited a higher DMPO-OO–adduct signal than daidzein. This indicated that molecular skeletons rich with electrons can promote the capacity for superoxide anion generation because of the +C effect from the hydroxyl functional groups. Rutin hydrate, composed of quercetin with rutinoside, was a particular example, as the adduct signal was markedly weaker than that of quercetin. These experimental results confirmed that the synthesis of bioflavonoid–metal complexes should exclude molecular oxygen from the reaction solution and employ fewer electronegative metal ions. Further investigations on work related to this issue is ongoing in our laboratory.