Modelling of Protective Mechanism of Iron ( II )-polyphenol Binding with OH-related Molecular Descriptors

The linear models for the calculation of pIC50, pKa1 and Epa for 12 polyphenolic compounds (catechins, flavonols, catechol and gallol derivatives) were developed. As descriptors we used the number of vicinal (Nv) and non-vicinal (Nnv) OH groups, as well as the number of OH vicinal pairs as possible Fe2+ chelate sites (Nch). The models gave r ˃ 0.9 and standard errors of 0.13, 0.26 and 0.04 for pIC50, pKa1 and Epa, respectively. For modelling of pIC50, Nch is better variable than Nv, and vice versa for modelling of pKa1 and Epa. This result, along with good correlations between pIC50, pKa1 and Epa, suggests two effects for antioxidative activity of polyphenols; their reaction(s) with OH and prevention of Fenton reaction by Fe2+ chelation.


INTRODUCTION
HE protective action of polyphenols is well known, [1][2][3][4][5] but the mode of their action has not yet been sufficiently explained.There are general reaction mechanisms for scavenging of radicals by polyphenols, such as single-step hydrogen atoms transfer (HAT), single electron transfer followed by proton transfer (SET-PT) and sequential proton loss electron transfer (SPLET). [6,7,8]owever, this paper is concerned with the mechanism involving polyphenol interaction with iron(II), [9] preventing in this way Fenton reaction: [10,11]

Fe
H O Fe HO OH The last mechanism is also relevant to DNA damage because hydroxyl radical is the primary cause of cell death under oxidative stress conditions. [12,13]The protective role of polyphenols is thus viewed as being iron(II) chelators, preventing the reduction of Fe 2+ with H2O2.[18][19][20][21] Analogous mechanism was also proposed for copper(I) / copper(II) system, but it proved less efficient than the already mentioned. [22,23]here is yet no general regression model (QSAR or QSPR) for the activity and physicochemical properties of polyphenols, despite many models with various molecular descriptors were tried.The binding of flavonoids to P-glycoprotein was modelled by sophisticated CoMFA and CoMSIA methods [24] and MIFs and VolSurf descriptors, [25] but also using a simple linear model based on zero-order valence molecular connectivity index. [26]The flavonoid toxicity (log cL50) to HL-60 and lamb embrio kidney fibroblast (FLK) cells were correlated to polarographic oxidation half-peak potential (E2/p) and water / octanol partition coefficient (log P) yielding in two-variable linear regression satisfactory correlation for HL-60 (r 2 = 0.915), but not for FLK cells (r 2 = 0.674). [27]Filipović and co-workers correlated VCEAC (antioxidant capacity equivalent to vitamin C concentration) values for 21 polyphenols with a T number of descriptors, but BDE (bond dissociation enthalpy), PA (proton affinity), ETE (electron-transfer enthalpy) and nOHvic (number of vicinal OH groups) proved best. [28]By combining two (BDE, nOHvic) and three descriptors (PA, ETE, nOHvic) they obtained r = 0.957 and r = 0.962 for the first and the second model, respectively.The similar model was used by Amić et al. on the set of 29 flavonoids (r = 0.974). [29]Perron and coworkers used pKa of the most acidic phenolic hydrogen as a sole descriptor for the modeling of inhibition of DNA damage under Fenton reaction conditions, but found the same regression cannot be successfully used for all the investigated polyphenols. [15]owever, pKa proved significantly better descriptor than the reduction or oxidation potential.
The aim of this contribution is to apply on the mentioned DNA system [15] a simpler, but possibly more successful set of molecular descriptors.For that purpose we chose descriptors based on number and position of phenolic and other hydroxyl group (alcoholic, carboxylic) in the molecules of polyphenols.

METHODS
For the given set of 12 polyphenols (catechins, flavonols, catechol and gallol derivatives, Figure 1) we tried to estimate their antioxidative activity by modelling pIC50 (obtained from the percentage of DNA damage inhibition) and Epa (oxidation potential), as well as pKa1 values of the most acidic (phenolic) hydrogen compiled from six different sources. [15]he variables used for modelling pIC50 of polyphenols are related to the number and position of OH groups in molecule viz. the number of vicinal (Nv) and of non-vicinal (Nnv) OH groups, and the number of OH pairs which may form a stable five-membered chelate rings with Fe 2+ (Nch).As carbonyl oxygen of flavonols participates in Fe 2+ chelation, [30] it is treated in the same way as OH group.
Regression calculations, including the leave-one-out procedure (LOO) of cross validation were done using the CROMRsel program. [31]The standard error of the crossvalidation estimate was defined as: where ΔX and N denotes cv residuals and the number of reference points, respectively.

RESULTS AND DISCUSSION
From Table 2 it can be seen that Model 2 (Figure 2) is consistently better than Model 1. Variable Nch, if taken alone, is better descriptor (r = 0.946) than Nv (r = 0.922) when correlated to pIC50.This speaks in favour of the assumption [9,15] that Fe 2+ chelation is the dominant effect for antioxidative activity, for Nch is a measure of chelating capacity of the studied compounds.
As shown previously, [15] pIC50 shows also a good correlation with pKa1 (r = -0.897, Figure 3), but our models proved better and they don't use experimental values as descriptors.Correlation of pKa1 to pIC50 also points to the ability of iron bonding, but pKa1 is not directly related to chelate but rather to inductive effect (electron affinity).This suggests that both effects participate in the antioxidative activity of polyphenols.Comparison of Models 3 and 4 leads to the same conclusion.Model 3 is better than Model 4, and pKa1 correlates better with Nv (r = -0.867)than with Nch (r = -0.818).Namely, Nv is not related directly to the number of chelation sites but to the number of neighbouring OH groups which mutually affect acidity.
We also correlated oxidation potential, Epa, to pIC50 (r = -0.770)and to pKa1 (r = 0.824) (Figures 4 and 5), which is comparable with the previously published results of Epa vs. IC50 for six cateholate compound (r = -0.889). [15]The correlations were substantially improved after removal of EGCG (the highest Nch and the highest pIC50) and other catechins (ECG, EGC, and EC); we obtained r = -0.953and 0.969 for Epa, vs. pIC50 and Epa, vs. pKa1, respectively.Also, for the estimation of Epa, Model 5 proved better than model 6, and Nv gives higher correlation (r = -0.933)than Nch (r = -0.911)with Epa.As the oxidation potential is a measure of ability for releasing of electrons, it is related to the antioxidative mechanisms involving reaction of polyphenols with •OH.We would therefore conclude that the influence of two effects is different for catechins than for other polyphenols in the set.Although EGCG shows the highest activity, its Epa is higher than Epa of Myr, implying that EGCG activity is caused more by chelate effect than the activity of Myr.  2).

Figure 1 .
Figure 1.Structures of polyphenolic compounds and their abbreviations.

Table 2 .
Linear models for the estimation of pIC50, pKa1 and Epa.
(a)Carbonyl oxygen is treated as OH group.