Exploring the Antioxidant Features of Polyphenols by Spectroscopic and Electrochemical Methods

This paper evaluates the antioxidant ability of polyphenols as a function of their chemical structures. Several common food indexes including Folin-Ciocalteau (FC), ferric reducing antioxidant power (FRAP) and trolox equivalent antioxidant capacity (TEAC) assays were applied to selected polyphenols that differ in the number and position of hydroxyl groups. Voltammetric assays with screen-printed carbon electrodes were also recorded in the range of −0.2 to 0.9 V (vs. Ag/AgCl reference electrode) to investigate the oxidation behavior of these substances. Poor correlations among assays were obtained, meaning that the behavior of each compound varies in response to the different methods. However, we undertook a comprehensive study based on principal component analysis that evidenced clear patterns relating the structures of several compounds and their antioxidant activities.


Instruments
A double beam Perkin Elmer UV/Vis/NIR Lambda 19 spectrophotometer (Waltham, MA, USA) was used to measure antioxidant and anti-radical indexes. Standard and reagent blank solutions were located in the sample and reference holders, respectively. QS quartz high performance cuvettes (10 mm optical path) from Hellma Analytics (Jena, Germany) were used.
Electrochemical studies were performed with a µAutolab system Type (III) (EcoChemie, Utrecht, The Netherlands) attached to a 663VA stand (Metrohm, Herisau, Switzerland). In all cases, a conventional cell was used with Ag|AgCl|KCl (3 mol L −1 ) as the reference electrode, platinum wire as the auxiliary electrode and a screen-printed carbon DRP-110 (Dropsens, Oviedo, Spain) as the working electrode in DPV mode. Data was acquired with a personal computer using the GPES 4.9 software (EcoChemie, Utrecht, The Netherlands).

FC Assay
One mL of water and 250 µL of FC reagent were placed in an amber glass vial. After 8 min, 75 µL of 7.5% (w:v) sodium carbonate aqueous solution and appropriate volumes of polyphenols were added to the vial to get concentrations in the range of 0.2-5 mg L −1 . Water was then added to obtain a final volume of 5 mL. The reaction was developed for 2 h and the absorbance was recorded at 765 nm using the reagent blank as the reference.

FRAP Assay
FRAP reagent was prepared by mixing 20 mmol L −1 FeCl 3 , 10 mmol L −1 TPTZ (containing 50 mmol L −1 HCl) and 50 mmol L −1 formic acid solution in the proportion of 1:2:10 (v:v:v). The reaction was developed with 300 µL of FRAP reagent and appropriate volumes of each polyphenol standards (to provide concentrations in the range of 0.2-5 mg L −1 ), and were diluted with Milli-Q water to obtain a final volume of 2.5 mL. The absorbance resulting after 5 min of reaction was measured at 595 nm using the reagent blank as the reference.

ABTS Assay
A stock solution of the cation radical species referred to as ABTS• + was generated with 20 mL of 7 mM ABTS and 350 µL of 140 mM potassium peroxodisulfate. The mixture was kept in the dark for at least 16 h before use and was stable for 1 week when stored at 4 • C. A working solution was prepared daily by diluting 300 µL of ABTS• + stock solution in 12 mL of ethanol. The assay was carried out by mixing 1.5 mL of ABTS• + , the required volume of polyphenol standard and ethanol up to 2.5 mL. Absorbance values were measured at 734 nm after 25 min of reaction time, using the ABTS• + blank as the reference.

DPPH Assay
A 0.2 mM DPPH stock solution in 50 mL ethanol was prepared and was kept in the dark for 2 h. The assay was carried out in an amber glass vial by mixing 2 mL of the DPPH solution, 1.6 mL 0.1 M phosphate buffer (pH 7.4), the required volume of the standard (providing concentrations in the range 0.2 to 5 mg L −1 ), and ethanol up to 4 mL. The solution was mixed and kept in the dark for 45 min. The absorbance was then measured at 517 nm using the reagent blank as the reference.

Voltammetric Assay
The required volume of standards was added to 25 mL of supporting electrolyte (0.1 M sodium acetate-acetic acid buffer at pH 5) in the electrochemical cell. Differential pulse voltammograms were recorded in the range from −0.2 to +0.9 V. Other conditions included a scan rate 0.1 mV s −1 , modulation time of 0.05 s, and interval time of 0.5 s.

Data Analysis
Excel (Microsoft, Redmond, WA, USA) was used for preliminary correlation and statistical studies. Principal component analysis (PCA) using the PLS-Toolbox (working under MATLAB, Applied Chemometrics, Inc, PO Box 100, Sharon, MA, USA) was applied to a global characterization of selected polyphenols according to spectrophotometric indexes and voltammetric data. The data matrix consisted of 15 rows (selected polyphenols) and 5 columns (slopes from FC, FRAP, ABTS, DPPH, and DPV methods). Data was autoscaled to equalize the influence of each variable in the model.

Results and Discussion
Fifteen polyphenols were chosen as the model compounds to be compared in order to investigate the relationships between the molecule structure and the antioxidant activity. Various food indexes were assessed with regard to their redox, anti-radical and electrochemical properties. Calibration curves from FC, FRAP, ABTS, DPPH and DPV methods were obtained for each polyphenol as described in Section 2.3. In all cases, the working range was 0.2-5 mg L −1 and the resulting sensitivities, in terms of mAU mol −1 , expressed the antioxidant ability of each molecule. In the case of TEAC indexes based on ABTS and DPPH reagents, it should be noted that the slopes were negative as the addition of polyphenols decreased the amount of free radical reagent, thus resulting in a decrease in absorbance with increasing concentration.
The results summarized in Table A1 (Appendix A) show that several compounds such as 3-hydroxybenzoic, 4-hydroxybenzoic, 2,4-dihydroxybenzoic and 2,4,6-trihydroxybenzoic acids displayed, in general, poor activity in all the assays. Other compounds, such as gallic acid, quercetin or luteolin, showed high responses for most of the indexes. Additionally, it was observed that various isomers with differing hydroxylation positions presented quite dissimilar behavior. For instance, 3,5-dihydroxybenzoic acid demonstrated low sensitivities while 2,3-dihydroxybenzoic acid exhibited higher values. Similar results were found when comparing 2,4,6-trihydroxybenzoic acid and gallic acid. It was thus concluded that, apart from the degree of hydroxylation, the antioxidant power depended on structural issues.
In order to gain more information on the role of the hydroxylation on the antioxidant properties of molecules we focused our study on the benzoic acid series because of the simplicity of their structure. The higher activities seemed to be associated with the presence of more than one hydroxyl group conveniently oriented in ortho or para positions while those in meta or monohydroxylated species were less active. The strongest reducing agents according to FC assay were dihydroxy-and trihydroxybenzoic acids with oand p-configurations, while those not following this pattern were less efficient. Similar behaviors were observed for FRAP and DPPH assays. The results of ABTS method were more erratic and independent of the stereochemistry of the molecules. A comparison of the normalized antioxidant activity of all assays ( Figure 1) also showed some subtle differences in the overall behavior of the isomers. The structure of each molecule has been included for a better interpretation of results. structure. The higher activities seemed to be associated with the presence of more than one hydroxyl group conveniently oriented in ortho or para positions while those in meta or monohydroxylated species were less active. The strongest reducing agents according to FC assay were dihydroxy-and trihydroxybenzoic acids with o-and p-configurations, while those not following this pattern were less efficient. Similar behaviors were observed for FRAP and DPPH assays. The results of ABTS method were more erratic and independent of the stereochemistry of the molecules. A comparison of the normalized antioxidant activity of all assays ( Figure 1) also showed some subtle differences in the overall behavior of the isomers. The structure of each molecule has been included for a better interpretation of results. The behavior of the molecules in the process of oxidation, which is closely related to their antioxidant power, can be better interpreted from electroanalytical studies by DPV. The representative examples depicted in Figure 2 indicated that molecules with more than one hydroxyl groups in o-and p-configurations displayed anodic processes at quite low potentials, with peak maxima at ca. 0.2-0.3 V (vs. Ag|AgCl|KCl reference electrode), thus indicating that these substances are highly prone to undergoing oxidation (see the example of 3,4-dihydroxybenzoic acid in Figure  2a). Conversely, meta or monohydroxylated species required higher potentials to be oxidated, with anodic bands at 0.65-0.85 V, which correspond to less labile species in oxidation reactions (see The behavior of the molecules in the process of oxidation, which is closely related to their antioxidant power, can be better interpreted from electroanalytical studies by DPV. The representative examples depicted in Figure 2 indicated that molecules with more than one hydroxyl groups in oand pconfigurations displayed anodic processes at quite low potentials, with peak maxima at ca.  Figure 2a). Conversely, meta or monohydroxylated species required higher potentials to be oxidated, with anodic bands at 0.65-0.85 V, which correspond to less labile species in oxidation reactions (see voltammograms of 2,4-dihydroxybenzoic acid in Figure 2b). A DPV scan of catechin is depicted for illustrative purposes (see Figure 2c). This molecule, with 2 independent dihydroxyphenyl moieties with oand mconfigurations, showed two oxidation peaks at 0.22 and 0.72 V attributed to the respective patterns mentioned above. Apart from the oxidation potential, the current intensity as a function of the concentration was another interesting feature closely related to the antioxidant power. It was considered that species with higher sensitivities could be stronger antioxidants. Thus, quercetin, luteolin and resveratrol acted as powerful substances while some hydroxybenzoic acids (e.g., 3-hydroxybenzoic, 3,4dihydroxybenzoic and 2,5-dihydroxybenzoic acids) displayed poor activities.
To summarize, anodic potentials provide qualitative information about the lability of the substances in (electro)chemical oxidant conditions, thus indicating the threshold from which they will act as antioxidants (i.e., the highest labilities have the lowest potential). In this regard, we guessed that in a mixture of polyphenols, compounds with the lowest potential will be the first to start acting as antioxidants while those with higher potential will remain in reserve until the others are consumed. In contrast, the slope of the calibration curves provides information on the quantitative power of the molecules to combat the presence of oxidant species in the medium.
A preliminary comparison of data from the different indexes was based on correlation studies. For FC vs. FRAP, the data was certainly correlated (r = 0.92), thus suggesting that the reduction of respective Mo(VI) and Fe(III) complexes occurs in analogous circumstances (see Figure 3a). Hence, the two indexes provided similar information regarding the reducing ability of each polyphenol. In the other cases, poorer correlations were obtained, meaning that the data was more dissimilar. However, a more thorough inspection of the scatter plots of various indexes revealed additional details. For instance, FC vs. DPPH (Figure 3b) showed that compounds with high reducing power also display high radical scavenging activity and vice versa. Figure 3b, however, suggested saturation in the anti-radical activity of the most powerful compounds, while molecules with low or moderate activities displayed a more linear relationship. The scatter plot of FC vs. DPV (Figure 3c) showed two characteristic trends depending on the nature of compounds. For instance, flavonoids and stilbenes were more electroactive than phenolic acids. In the case of FC vs. ABTS, the results were more dispersed, although, in general, the most (and less) active compounds were the same in the two systems. Analogous examples considering FRAP instead of FC (results not shown) were studied and the conclusions that were drawn are similar to those presented here.

Comparison of Indexes by PCA
A more comprehensive evaluation of the data from spectroscopic and voltammetric experiments was carried out by PCA. As indicated in the experimental section, an autoscaled model was built to minimize the impact of the differences in magnitude and amplitude of indexes on the description of Apart from the oxidation potential, the current intensity as a function of the concentration was another interesting feature closely related to the antioxidant power. It was considered that species with higher sensitivities could be stronger antioxidants. Thus, quercetin, luteolin and resveratrol acted as powerful substances while some hydroxybenzoic acids (e.g., 3-hydroxybenzoic, 3,4-dihydroxybenzoic and 2,5-dihydroxybenzoic acids) displayed poor activities.
To summarize, anodic potentials provide qualitative information about the lability of the substances in (electro)chemical oxidant conditions, thus indicating the threshold from which they will act as antioxidants (i.e., the highest labilities have the lowest potential). In this regard, we guessed that in a mixture of polyphenols, compounds with the lowest potential will be the first to start acting as antioxidants while those with higher potential will remain in reserve until the others are consumed. In contrast, the slope of the calibration curves provides information on the quantitative power of the molecules to combat the presence of oxidant species in the medium.
A preliminary comparison of data from the different indexes was based on correlation studies. For FC vs. FRAP, the data was certainly correlated (r = 0.92), thus suggesting that the reduction of respective Mo(VI) and Fe(III) complexes occurs in analogous circumstances (see Figure 3a). Hence, the two indexes provided similar information regarding the reducing ability of each polyphenol. In the other cases, poorer correlations were obtained, meaning that the data was more dissimilar. However, a more thorough inspection of the scatter plots of various indexes revealed additional details. For instance, FC vs. DPPH (Figure 3b) showed that compounds with high reducing power also display high radical scavenging activity and vice versa. Figure 3b, however, suggested saturation in the anti-radical activity of the most powerful compounds, while molecules with low or moderate activities displayed a more linear relationship. The scatter plot of FC vs. DPV (Figure 3c) showed two characteristic trends depending on the nature of compounds. For instance, flavonoids and stilbenes were more electroactive than phenolic acids. In the case of FC vs. ABTS, the results were more dispersed, although, in general, the most (and less) active compounds were the same in the two systems. Analogous examples considering FRAP instead of FC (results not shown) were studied and the conclusions that were drawn are similar to those presented here. The plot of loadings provided information about the correlation among variables (Figure 4b). It was found that PC1 mainly described the overall antioxidant activity, with spectroscopic methods with a positive slope (FC and FRAP) to the right and those with a negative slope (DPPH and ABTS) to the left. PC2 was used for modeling differences among spectroscopic (bottom) and voltammetric methods (top), and indicated that correlations among colorimetric indexes and DPV were, in general, poor. The simultaneous interpretation of scores and loadings suggested that, from PC1, compounds located to the right were stronger antioxidants from an overall point of view (basically considering their reducing, anti-radical abilities). PC2 discriminated chemical and electrochemical information. It was observed that the most sensitive electroanalytical species were located at the top while those with smaller slopes were at the bottom. As a result, it was concluded that PC1 mainly explained the polyphenolic behavior according to the spectroscopic indexes and PC2 modeled the voltammetric features of the compounds.

Conclusions
The estimation of the antioxidant ability of food products and identifying the ability of bioactive compounds to reduce oxidative stress are increasingly studied due to their health implications. Unfortunately, information about the antioxidant activity of natural compounds such as polyphenols

Comparison of Indexes by PCA
A more comprehensive evaluation of the data from spectroscopic and voltammetric experiments was carried out by PCA. As indicated in the experimental section, an autoscaled model was built to minimize the impact of the differences in magnitude and amplitude of indexes on the description of samples and variables. The first and second principal components (PC1 and PC2) explained 59.2 and 23.0% of the data variance, respectively. The scatter plot of scores of PC1 and PC2 showed three characteristic patterns regarding the distribution of the samples (see Figure 4a). The group of compounds located to the right side corresponded to di-and trihydroxybenzoic acids having oand p- compounds located to the right side corresponded to di-and trihydroxybenzoic acids having o-and p-oriented hydroxyl groups. Compounds on the left corresponded to benzoic acids with 1 hydroxyl group or several in m-positions. The group in the middle consisted of flavonoids and resveratrol and exhibited intermediate features as they contained both o-and m-moieties in their structures.
The plot of loadings provided information about the correlation among variables (Figure 4b). It was found that PC1 mainly described the overall antioxidant activity, with spectroscopic methods with a positive slope (FC and FRAP) to the right and those with a negative slope (DPPH and ABTS) to the left. PC2 was used for modeling differences among spectroscopic (bottom) and voltammetric methods (top), and indicated that correlations among colorimetric indexes and DPV were, in general, poor. The simultaneous interpretation of scores and loadings suggested that, from PC1, compounds located to the right were stronger antioxidants from an overall point of view (basically considering their reducing, anti-radical abilities). PC2 discriminated chemical and electrochemical information. It was observed that the most sensitive electroanalytical species were located at the top while those with smaller slopes were at the bottom. As a result, it was concluded that PC1 mainly explained the polyphenolic behavior according to the spectroscopic indexes and PC2 modeled the voltammetric features of the compounds.

Conclusions
The estimation of the antioxidant ability of food products and identifying the ability of bioactive compounds to reduce oxidative stress are increasingly studied due to their health implications. Unfortunately, information about the antioxidant activity of natural compounds such as polyphenols The plot of loadings provided information about the correlation among variables (Figure 4b).
It was found that PC1 mainly described the overall antioxidant activity, with spectroscopic methods with a positive slope (FC and FRAP) to the right and those with a negative slope (DPPH and ABTS) to the left. PC2 was used for modeling differences among spectroscopic (bottom) and voltammetric methods (top), and indicated that correlations among colorimetric indexes and DPV were, in general, poor.
The simultaneous interpretation of scores and loadings suggested that, from PC1, compounds located to the right were stronger antioxidants from an overall point of view (basically considering their reducing, anti-radical abilities). PC2 discriminated chemical and electrochemical information. It was observed that the most sensitive electroanalytical species were located at the top while those with smaller slopes were at the bottom. As a result, it was concluded that PC1 mainly explained the polyphenolic behavior according to the spectroscopic indexes and PC2 modeled the voltammetric features of the compounds.

Conclusions
The estimation of the antioxidant ability of food products and identifying the ability of bioactive compounds to reduce oxidative stress are increasingly studied due to their health implications. Unfortunately, information about the antioxidant activity of natural compounds such as polyphenols is sometimes confusing and data from food indexes regarding their reducing or anti-radical power are seldom coherent. Apart from using different reference compounds for expressing such activity (e.g., gallic acid equivalents in FC, trolox equivalents in FRAP or TEAC, or quercetin or rutin in flavonoid indexes), the sensitivities of compounds in the assays are different. This paper has untangled some of the discrepancies that are often encountered in the study of the antioxidant activity of molecules by using a global approach that combines data from various food indexes and electrochemical studies. The PCA model revealed three distinct patterns depending on the number and orientation of hydroxyl groups in the molecules. pand o-dihydroxyphenyl molecules were clearly discriminated from their mcounterparts. Besides, electrochemical data explained both qualitative and quantitative aspects of the antioxidant power of molecules.
In short, we assume that the oxidation of molecules by chemical, radical and electrochemical mechanisms is easier in the case of dihydroxyphenyl moieties with oand porientations. Hence, the corresponding benzoquinones can be easily formed as a first oxidized species. In the case of phenyl moieties with single hydroxylation, the oxidation mechanism also leads to a benzoquinone via a more complex mechanism. The reaction is not as favored as in the previous case and stronger oxidants or higher potential may be required. Finally, processes involving dihydroxyphenols m-substituted have not been so well described but the reactions seem to be more complex and their antioxidant activity may be more limited.