Hooman Shadnia* and James S. Wright

Transcription

1 Chem. Res. Toxicol. XXXX, xxx, 000 A Understanding the Toxicity of Phenols: Using Quantitative Structure-Activity Relationship and Enthalpy Changes To Discriminate between Possible Mechanisms Hooman Shadnia* and James S. Wright Department of Chemistry, Carleton UniVersity, 1125 Colonel By DriVe, Ottawa K1S 5B6, Canada ReceiVed February 13, 2008 Experimental studies of the extended toxicity of substituted phenols are mainly of two types: the toxicity due to phenoxyl radical formation and the toxicity caused by metabolites, for example, the formation of quinones. Quantitative structure-activity relationship (QSAR) studies of phenol toxicity have dealt with the formation of phenoxyl radicals using bond dissociation enthalpy (BDE) of parent phenols, have obtained good correlations with experimental data, and have concluded that phenoxyl radicals are the toxic agent. However, the actual toxic mechanism has remained poorly defined. In this study, we follow the metabolic pathways of monosubstituted phenols to their quinone end products and calculate enthalpy changes for all relevant reactions. These enthalpy changes are first used as descriptors for a QSAR analysis. Many of these new descriptors, including some relevant to quinone formation, are highly correlated with the BDE values of the parent phenols. Therefore, a QSAR analysis by itself is inconclusive as to the mechanism of toxicity. To better define the problem, we have returned to a detailed analysis of net enthalpy changes. We show that the formation of phenoxyl radical is the rate-determining step: This step is slow for electron-withdrawing group substituted phenols (EWG-phenols), whereas it is fast for electron-donating group substituted phenols (EDG-phenols). The study of net enthalpy changes of reactions reveals that once the phenoxyl radical is present, the corresponding quinone is rapidly formed, so that quinone formation may be ultimately responsible for toxicity of EDG-phenols. We then demonstrate how the suggested mechanism (quinone formation) is successful in predicting the toxicity of some complex phenols, which are predicted poorly using the phenoxyl radical argument. We also discuss the toxicities of some estrogens in light of the quinone mechanism. Introduction The toxicity of phenolic compounds in cell cultures or higher organisms has been studied repeatedly (1 11) in an attempt to understand their toxic mechanism(s). Phenolic compounds in biological systems may undergo oxidation reactions leading to the formation of catechol and hydroquinone, followed by further oxidations to the semiquinone and the quinone. The hydroxylation step converting phenol to catechol can be catalyzed by cytochrome P450, using NADPH and oxygen or peroxides as cofactors. Another possible path is via generation of a phenoxyl radical, typically by reaction of the phenol with a peroxyl radical (from lipid peroxidation). Spin density at the ortho position is 0.22 (AM1), and at the para position, it is 0.34, whereas the spin density on the oxygen is only An incoming hydroxyl radical then will attempt to bond where the spin density is highest (ortho or para) to form catechol or hydroquinone. Catechol will react rapidly, for example, with another peroxyl radical or a phenoxyl radical (12, 13), to give the semiquinone. Once the semiquinone has formed, it is transient in aqueous solution at neutral ph, rapidly disproportionating into catechol and the ortho-quinone. Thus, the sequence of events converting phenol to ortho-quinone is complex but well-studied. A simplified reaction diagram illustrating these steps is shown in Figure 1, where the labels for uncatalyzed reactions refer to hydrogen atom transfer (HAT), proton transfer (PT), and * To whom correspondence should be addressed. hooman@ shadnia.com. electron transfer (ET). The catalyzed reaction includes oxygen/ peroxide activation (O/PA) as discussed by Porter and Coon (14). The reaction involves a heme complex and has many steps, but the overall effect is insertion of an OH group into the phenoxyl radical. Hydroxyl radical addition (HRA) is another type of addition reaction of hydroxyl radical to phenoxyl radical, as will be discussed later. It has been shown by many authors that electrophilic quinones derived from phenols are toxic to cells (15 17), mainly through their binding to protein thiols and DNA (18 22) and also to redox cycling (23 25). However, in spite of the studies linking the toxicity of phenols to their quinone end products, a quantitative structure-toxicity relationship (QSTR) study linking quinone formation to cytotoxicity has not been reported previously. Most quantitative structure-activity relationship (QSAR) studies have either looked into general (baseline) toxicity, which is highly predictable by hydrophobicity descriptors such as log P (2, 3), or have discussed what has been termed extended toxicity, that is, the toxicity of phenols that is not predicted by log P and that is predictable using the bond dissociation enthalpy (BDE) of phenols and other similar descriptors (4 11). A study of both baseline toxicity and extended toxicity can be seen in Selassie s work as they measured growth inhibition in mouse lymphocyte leukemia cells (L1210) caused by monosubstituted phenols (4). Usually, the data set is first divided into phenols with electron-withdrawing groups (EWG) and those with electron-donating groups (EDG). Two different models and mechanisms are then assigned to interpret the two groups. In Selassie s data, EWG-substituted phenols have lower toxicities, /tx800058r CCC: $40.75 XXXX American Chemical Society Published on Web 05/23/2008

2 B Chem. Res. Toxicol., Vol. xxx, No. xx, XXXX Shadnia and Wright Figure 1. Possible metabolic reactions of phenols leading to quinones. Figure 2. Overall picture of the metabolism of phenols. which correlate strongly with log P. No other single descriptor seems to have any substantial predictive power on this set. The EDG-substituted phenols as a group possess much higher toxicities. The phenol BDE or related descriptors for this subset are highly correlated with toxicity; for example, the correlation coefficient R 2 ) 0.78 (for our subset, see below) for log(1/c) vs BDE, where C is the concentration of phenol leading to 50% cell death. This excellent correlation apparently led authors to believe that the formation of phenoxyl radicals was responsible for the observed cytotoxicity, although the exact mechanism of action was poorly defined. There are several reasons why a QSTR has not been done on downstream events of the metabolic pathway. First, an analysis based only on the formation of phenoxyl radical has such a high predictive power that further effort seemed unnecessary. Second, the enzymatic step involving O/PA is difficult to model theoretically. Third, calculating enthalpy changes for a large data set of phenols in various oxidation states is a massive undertaking. Fourth, how to estimate relative rates of all relevant reactions given only reaction enthalpy data is not completely obvious, but direct calculation of transition states is tedious and not very accurate. Nevertheless, it seems worthwhile to consider the full range of metabolites and their rates of formation. According to Figure 2, there are a myriad of possibilities for the rate-determining reaction and actual toxic metabolite. While our analysis was not biased for or against any particular possibility, our hypothesis is that quinone metabolites are responsible for the toxicity of monosubstituted phenols. Materials and Methods We chose the data set of Selassie et al. (4) since we found it to be the largest data set on human cells that covered the extended toxicity of phenols. All local minima (i.e., conformers) of 39 monosubstituted phenols from Selassie s data set and all of their metabolites formed according to Figure 1 were determined using MOE2007 (26) with an MMFF94s force field (27, 28). Electronic energies were calculated using a density functional theory (DFT) methodology termed medium-level model 2 (MLM2) (29) using Gaussian 03 (30) running on our local PC cluster and on the HPCVL supercomputing facility (31). Briefly, force field-optimized conformer geometries were reoptimized using DFT with the B3LYP/6-31G(d) functional. Vibrational analysis was done at the same level with frequencies scaled by a factor of and the temperature set to K. A single-point energy was computed at the previously optimized geometry using B3LYP/6-311+G(2d,2p) for closedshell molecules and a restricted open shell treatment (ROB3LYP) for radicals. The electronic energy of the H-atom was set to its exact value, hartree. Gas-phase enthalpies were calculated as the sum of the big basis set energy plus thermal correction for enthalpy. Because there are very few rotatable bonds in monosubstituted phenols, the variation of entropies was small (about 3 cal/mol deg for the entire set, or <1 kcal/ mol at room temperature), so only enthalpy values are reported instead of free energies. Solvation energies were calculated at the gas-phase geometry using HF/6-31G(d) with the CPCM method (solvent ) water) and UAHF radii. Aqueous phase enthalpies were calculated by adding calculated free energy changes of solvation to gas -phase enthalpies. Log P values were calculated using MOE. BDE values for substituted phenols were calculated according to reaction 1: X-Ph f X-Ph + H (1) where X-Ph is the parent phenol and X-Ph is the (monosubstituted) phenoxyl radical. The gas-phase enthalpy of the H-atom is hartree ( kcal/mol), and its solvation energy is kcal/mol, so the aqueous phase enthalpy is kcal/mol. This procedure was used to determine BDE values for phenol, catechol, and semiquinone, which we term BDE 1, BDE 2, and BDE 3, respectively. For a HAT reaction, reactants (phenol, catechol, or semiquinone radical) are assumed to react with allylperoxyl (a simple model of a lipid peroxyl radical) to form allylperoxide, according to reaction 2: X-Ph + ROO f X-Ph + ROO-H (2) The enthalpy change for this reaction for phenol corresponds to the BDE difference between the substituted phenol (X-Ph) and the allyl peroxide (ROOH).

3 Understanding the Toxicity of Phenols Chem. Res. Toxicol., Vol. xxx, No. xx, XXXX C Table 1. QSAR Descriptors Calculated for Selected EDG/EWG Phenols a R Log(1/C) b σ+ b Log P (O/W) BDE1 BDE2 BDE3 H PT1 H PT2 H PT3 H ET1 H ET2 H ET3 H HRA H O/PA EDG 2-NH NH OC 3 H OC 2 H OH OCH OCH CH t-butyl CH C 3 H t-butyl F F CF Br H EWG 4-Br CN F Cl Br CONH CN NO CHO CN NO NO a Values are in kcal/mol. BDE values for phenol, catechol, and semiquinone were , , and kcal/mol, respectively. For other absolute values, refer to Table 4. b Taken from ref 14. Deprotonation enthalpies were calculated for reaction 3, where a proton is transferred to water: X-Ph + H 2 O f X-Ph - + H 3 O + (3) ET enthalpies correspond to transfer of an electron to the allylperoxyl radical, reaction 4: X-Ph - + ROO f X-Ph + ROO - (4) O/PA enthalpies correspond to the addition of an OH group to phenol, with loss of an H-atom. The energetics of this step were modeled using hydrogen peroxide, according to reaction 5: Ph + HOOH f Cat + H 2 O (5) Phenoxyl radical can also lead to catechol in step HRA, corresponding to reaction 6: Ph + HO f [intermediate] f Cat (6) Finally, the semiquinone can form via H-atom abstraction from catechol. At neutral ph, the semiquinone is mostly in its anionic form (pk values for semiquinones are about 4-5). It can react with itself in neutral or anionic form via disproportionation, reaction 7: 2Sq f Cat + Q (7) For the QSAR studies, the calculated enthalpy differences, for example, BDE 1 (aq) etc., were used as descriptors vs Selassie s toxicities, measured as log 1/C 50. Single and double descriptor models were built by linear regression, as shown in Table 2. For all working models, correlation coefficient values of the data (R 2 ) and those of the leave-one-out cross-validation method (q 2 ) are shown in Table 2, along with root mean square of error (RMSE) values and F statistic values. Table 2. Summary of QSAR Model Data for Linear Models: log (1/C) ) a + (b X) a descriptor R 2 a b F RMSE q 2 EDG σ Log P 0.12 BDE BDE BDE H PT H PT H PT H ET H ET H ET H HRA H O/PA 0.32 EWG Log P a RMSE, root mean square error; R 2, correlation coefficient; F, Fischer statistic; and q 2, cross-validated square correlation coefficient. Results and Discussion BDE values for aqueous solution for monosubstituted phenols (BDE 1 ) and the corresponding catechol (BDE 2 ) and semiquinone (BDE 3 ) are reported in Table 1. Values shown are relative to phenol, catechol, and semiquinone for which the absolute BDE values are shown in the footnote. Peroxyl radicals, ROO, are generated in the cell by lipid peroxidation. Their O-H BDE does not vary much, so we chose allyl peroxyl as an example, for which the parent compound allylooh has a BDE(aq) of 91.6 kcal/mol. The BDE of the phenol is calculated to be 93.8 kcal/mol in water (87.4 in gas, Table 1). The overall enthalpy change for reaction 2 is BDE (Ph) - BDE (allylooh) )+2.1 kcal/mol. Table 1 also shows that the OH bond in catechol is weaker than that in phenol by

4 D Chem. Res. Toxicol., Vol. xxx, No. xx, XXXX Shadnia and Wright Table 3. Correlation Coefficients (R 2 ) of Calculated QSAR Descriptors a descriptor BDE1 BDE2 BDE3 H PT1 H PT2 H PT3 H ET1 H ET2 H ET3 H HRA BDE1 1 BDE BDE H PT H PT H PT H ET H ET H ET H HRA a Values greater than 0.6 are shown in bold. Table 4. Values of Net Enthalpy Changes for Phenols and Its Products HAT reactions phenol + ROO f phenoxyl radical + ROOH H )+2.19 (-9.44 to +9.35) catechol + ROO f semiquinone + ROOH H )-6.66 ( to -1.79) semiquinone + ROO f quinone + ROOH H ) ( to ) PT reactions phenol + H 2 O f phenoxide + H 3 O + H ) ( to ) catechol + H 2 O f catechol anion + H 3 O + H ) ( to ) semiquinone + H 2 O f semiquinone anion + H 3 O + H ) (+6.20 to ) ET reactions phenoxide + ROO f phenoxyl radical + ROO - H ) (+4.12 to ) catechol anion + ROO f semiquinone + ROO - H ) (+1.35 to ) semiquinone anion + ROO f quinone + ROO - H )+3.10 (-6.81 to ) other reactions O/PA: phenol + H 2 O 2 f catechol + H 2 O H ) ( to ) HRA: phenoxyl R + 1/2H 2 O 2 f catechol H ) ( to ) phenoxyl R + HO f catechol H ) ( to ) Ranges for substituted phenols are calculated using data from Table 1 and are shown in parentheses. Values are in kcal/mol. ROO is allyl peroxide 8.9 kcal/mol, and in semiquinone, it is weaker by 22.8 kcal/ mol. This renders reactions between peroxyl radical and catechol much more exothermic than with phenol and even more so for the semiquinone. As will be seen, this observation plays an important role in the interpretation of toxicity. Prior to data fitting, two considerations are typically applied to experimental data sets: the separation into EWG and EDG sets and the removal of outliers. The data set was split into two sets of EDG/EWG based on BDE 1. Functional groups that lower BDE 1 in comparison with phenol were considered to be EDG and vice versa. Four molecules that were EDG outliers in almost all models were removed (2-OH, 2-C 3 H 7, 2-Cl, and 4-Cl). The meta-substituted phenols pose a particular problem: They may produce two different types of catechol (1X-2,3 catechol or 1X-3,4 catechol), which could follow two different pathways, resulting in two possible values for BDE 2. One solution is to pick the catechol with the lower enthalpy and follow the corresponding path. However, sometimes, the differences between the two catechols are minute and below resolution of the calculation method. Besides, there is little variation in toxicities in the meta-substituted phenols. To avoid these problems, we removed all meta-substituted compounds from the correlations, which always improved the results. The final subset that we used has 17 data points in EDG and 11 in EWG. 1 For the EDG set, BDE 1 is a good descriptor. An equivalent descriptor is BDE 1, defined as the difference in BDE from phenol, that is, BDE (X-Ph - Ph). This gives the regression equation log(1/c 50 ) ) BDE 1 with a correlation 1 As will be discussed, because the only descriptor that predicted the toxicity on EWG phenols was Log P, meta-substituted compounds are not excluded from this set in Table 2. coefficient of Larger values of log (1/C 50 ) represent increased toxicity; because the BDE 1 values are all negative for the EDG set, the lowest value of BDE 1 is (correctly) the most toxic. On the other hand, a regression equation for the EWG set shows a poor correlation with R 2 ) 0.08 (data not shown). This provides the traditional justification for splitting the data into EDG and EWG sets and using only BDE 1 as a QSTR descriptor for the EDG set. This has led authors to assume that phenoxyl radical is responsible for the toxicity of phenols. However, other downstream descriptors including BDE 2, H ET1, and H ET2 have even higher correlation coefficients of In addition, H HRA, H PT3, and H ET3 are almost as useful descriptors with R 2 values between 0.71 and We checked the validity of omitting the meta compounds by running the same set of regression equations for the data set, which included these compounds. While the nature of models was essentially the same, the quality of regressions was slightly reduced. For example, for BDE 1, the results are as follows: For all data: Log(1/C) ) BDE where n ) 25, R 2 ) 0.73, q 2 ) 0.96, and RMSE ) For meta compounds excluded from Table 2: Log(1/C) )-0.17 BDE where n ) 17, R 2 ) 0.78, q 2 ) 0.73, and RMSE ) As can be seen here, the results are essentially the same; therefore, we omitted the meta compounds. None of the two-descriptor models showed significant improvement over single descriptor models. For example, adding Selassie s log P values to our best descriptors improved R 2 by only 0.03, and no other second descriptor improved R 2 by more

5 Understanding the Toxicity of Phenols Chem. Res. Toxicol., Vol. xxx, No. xx, XXXX E Figure 3. Toxicity of 2,4-substituted phenols. than For this reason, we will restrict our discussion to single-descriptor models. 2 Because there was more than one descriptor that predicted toxicity, the QSTR analysis suggests that any of the metabolic steps corresponding to each descriptor could be the ratedetermining step. That is, the formation of catechol, semiquinone, and quinone, via HAT or ET, is as important to toxicity as the formation of the phenoxyl radical! The only plausible explanation for such a high predictive power of so many descriptors is that they are strongly intercorrelated. In fact, such behavior can be observed in the correlation matrix of the descriptors (Table 3). For example, H ET1 is strongly correlated with BDE 1, with an R 2 value of 0.86, and correlation coefficients of more than half of the intercorrelations are higher than 0.6. This suggests that although there may be only one toxic species in the mechanism, a QSTR analysis using enthalpy descriptors cannot tell us which one it is; that is, QSTR is inconclusive. To gain more insight into the actual toxic mechanism, we should identify known toxic species and return to a study of reaction kinetics to see how fast they are formed. Rate constants can be estimated based on the overall enthalpy change for the reaction and the type of reaction (HAT, PT, ET, or enzymatic). This is done by using the Evans-Polanyi principle, which states that, other things being equal, more exothermic reactions will have lower activation energies (32). Table 4 shows a list of reactions 2 6 and their enthalpy changes for phenol and its metabolites. The ranges of enthalpy changes for substituted phenols are also calculated and shown in parentheses. Table 4 shows that formation of phenoxyl radical from reaction 2 is slightly endothermic for phenol itself (+2.2 kcal/ mol). The EWG groups cause reaction 2 to be more endothermic (up to +9.4 kcal/mol); therefore, the reaction is slow for EWG phenols, since the activation energy must be greater than or equal to the endothermicity. On the other hand, reaction 2 with some EDG is thermoneutral (for example 2- or 4-CH 3 ), but stronger EDGs like OH and NH 2 are exothermic. The Selassie data show clearly that the lower the BDE is, the higher the toxicity is. This seems counterintuitive since high BDE radicals are more reactive and will attack anything; for example, the hydroxyl radical is the most reactive of all biological radicals since it forms water with a very high BDE of 119 kcal/mol. However, if such radicals are formed Very slowly, then their reactivity is irrelevant. In fact, the decreasing barriers for more 2 Adding Log P as a second descriptor improves the quality of correlations for more lipophillic EDG phenols such as estrogens (data not shown). Figure 4. Plot of toxicity (1/C) vsbde 1 values. Compounds seem to show extended toxicity when BDE drops below -2 to-5 kcal. The BDE 1 for phenol is set to zero. exothermic reactions in the EDG group increase the rate of formation exponentially. Thus, for reaction with peroxyl radical, in going from the most endothermic EWG ( BDE 1 )+9.4 kcal/mol) to most exothermic EDG ( BDE 1 )-9.4 kcal/mol), the activation energy for HAT will vary from ca kcal (EWG, endothermicity plus residual 2 kcal barrier) to +2 kcal (EDG, residual 2 kcal barrier), favoring the EDG by a ratio of exp(-11.4/rt)/exp(-2/rt) ) ca at 298 K. Clearly, for current data (substituted phenols with BDE 1 values in range of phenol ( 15 kcal/mol), the rate of formation of a phenoxyl radical is much more important than its reactivity. From Table 4, we can also see that HAT reactions of both catechol and semiquinone are significantly exothermic (-6.7 and kcal/mol, respectively). For catechol, enthalpy values for reaction with peroxyl radical vary from -2 to-19 kcal/ mol, so that even the most EWG is slightly exothermic. This trend is amplified for the semiquinone, where now all of the reactions are highly exothermic. The implication for toxicity is the following: The formation of phenoxyl radical will vary from very slow for EWG phenols (k < 10 0 M -1 s -1 ) and fast for EDG phenols (k > 10 6 M -1 s -1 ). Reactions for the formation of radical from catechol (i.e., the semiquinone) and also from semiquinone (quinone) will also be fast, regardless of substituents. Therefore, once catechol is formed, it will rapidly be converted to the corresponding quinone. Next, consider the formation of catechol directly from phenol, via O/PA. Chemical formation of catechol from phenol using hydroxyl radical is very exothermic, but it has a relatively high barrier [est kcal/mol; ref (31, 33)], and the reaction will not take place unless in the presence of a catalyst (usually Fe, Cu, or Zn) or an enzyme (Cyp450) or at high temperatures (often done in supercritical water T ) C). Inside the cell, this reaction is catalyzed by Cyp450 with or without help of peroxides (O/PA pathway). Actual rates of O/PA depend on concentration of enzyme and cofactors (for some estimates, see ref 2). However, H O/PA correlates poorly with the toxicity of phenols (R 2 ) 0.32, Table 2), while the parallel pathway via phenoxyl radical (HAT 1 ) correlates well (R 2 ) 0.78), suggesting that the rate of catechol formation from O/PA must be slower than HAT rates. To complete the discussion, we need to go from phenoxyl radical to catechol.

6 F Chem. Res. Toxicol., Vol. xxx, No. xx, XXXX Shadnia and Wright Figure 5. Calculated BDE 1 values and predicted toxicities of some estrogens. In Figure 1, HRA to phenoxyl radical results in the formation of catechol. The optimum path requires a concerted migration of the adjacent H-atom with simultaneous insertion of the OH radical ortho to the phenolic OH. Because two radicals recombine to form a closed-shell molecule, the reaction will obviously be very exothermic ( H )-83.2 for phenol and to for entire data set) and should have a low barrier to rearrangement since two covalent bonds are forming. In the absence of a search for the transition state, 3 we will assume, based on the strong exothermic nature of the addition, that the rearrangement barrier is low and OH insertion is fast relative to formation of phenoxyl radical in reaction 2. Thus, the overall picture of reactivity is as shown in Figure 2. Even though the formation of phenoxyl radical is the ratedetermining step, quinone formation is the most likely fate of any formed phenoxyl radical, and quinones should be responsible for toxicity. If our analysis is correct, then it should be possible to significantly reduce the toxicity of phenols by deliberately slowing the formation of quinones. There are two ways to do this: (i) put EWG on the phenol ring and (ii) block positions ortho to the phenolic OH group. So, phenolic compounds with blocked ortho positions should be less toxic than expected. As shown in Figure 3, this is truly the case, where experimental toxicity of such compounds is less than the value predicted using BDE 1 (9). While Selassie et al. used a different QSAR along with different explanation for the toxicity of these compounds, we simply suggest that since these compounds cannot form ortho-quinones, their toxicities are down to baseline toxicities. In fact, the only metabolite for these compounds is their phenoxyl radicals, which apparently does not exhibit any extended toxicity. We suggest that if a compound can only form a phenoxyl radical, it might be reduced back to parent molecule by antioxidant mechanisms, hence demonstrating no extended toxicity. It is only through ortho-quinone formation that phenols exhibit their extended toxicities. Also, Figure 3 shows a comparison between toxicity of 2,6- dimethyl phenol and 2,4,6 trimethyl phenol; 2,6-di-Me phenol 3 Wright, J. S., and Shadnia, H. To be published. can have hydroquinone and p-quinone metabolites, but 2,4,6- tri-me has no hydroxylation site at all, so it does not form any catechol or quinone. If the hydroquinone or p-quinone metabolites are formed and had any toxicity, the di-me compound should have been more toxic than the tri-me compound, but the trend of observed toxicities is the complete opposite meaning hydroquinone and p-quinone do not contribute to toxicity. The nature and the extent of this opposite trend in toxicities can be simply explained by baseline toxicity; the tri-me compound is more lipophilic (log P ) 2.56 vs log P ) 2.19 for dimethyl), which according to the equation for Log P from Table 2 results into 0.28 increase in toxicity of tri-me compound. The observed difference in Log(1/C) values is Another interesting observation is that when plotting the experimental toxicities (1/C) of the entire set vs BDE 1 values (Figure 4), the phenolic compounds begin to show appreciable toxicity at BDE 1 values about 2-5 kcal lower than phenol. In other words, the reaction with the target radical appears to become fast at about a (target radical) BDE of kcal/ mol. Interestingly, using H HAT2 results in similar an estimate for the target radical. This confirms that choosing allyl peroxyl (BDE ) 91.6) as a biological radical to derive the information in Table 4 is a good choice. Our suggested mechanism confirms the link between quinone formation and estrogens, along with the elevated risk of conjugated estrogens. Quinone formation has been blamed for carcinogenicity of estrogens (25, 34 39), particularly conjugated estrogens, which are used in some preparations for the treatment of menopause symptoms and prevention of osteoporosis. As shown in Figure 5, diethylstilbestrol (an estrogen analogue prohibited since 1971 due to teratogenicity) and equilenin (a component of conjugated estrogenic preparations) show slightly higher toxicities. Because these compounds are highly lipophilic, experimental toxicities are even higher than those predicted here without using lipophilic terms (14). Estrogens are phenolic compounds and can become biologically activated to the corresponding catechol and quinone, which has been shown experimentally. Equilenin is a β-naphthol that undergoes conversion to the 1,2-dihydroxynaphthalene readily, followed by rapid conversion to the known carcinogen 3,4- estradiol-quinone (34).

7 Understanding the Toxicity of Phenols Chem. Res. Toxicol., Vol. xxx, No. xx, XXXX G Conclusions We conclude that even though the formation of phenoxyl radicals is the rate-determining step, any phenoxyl radicals formed will continue metabolism through exothermic reactions to their quinone end product, and thus, quinone formation is the most likely reason for extended toxicity of these phenols. We demonstrated how QSAR suggests that reactions related to quinone formation are as relevant as phenoxyl radical formation, but because of high intercorrelations between parameters, QSAR alone was unable to identify the true descriptor and the main step. This was only made possible with the help of calculated net enthalpy changes of the reactions. We demonstrated how quinone formation predicts the lower toxicities of some complex phenols, whose toxicities are overestimated using a phenoxyl radical formation argument. We also discussed the elevated toxicities of estrogens including natural and conjugated estrogens in the light of quinone formation. References (1) Cronin, M. T. D., Zhao, Y. H., and Yu, R. L. (2000) ph-dependence and QSAR analysis of the toxicity of phenols and anilines to Daphnia magna. EnViron. Toxicol. 15 (2), (2) Moridani, M. Y., Sirkani, A., and O Brien, P. J. (2003) Quantitative structure toxicity relationships for phenols in isolated rat hepatocytes. Chem.-Biol. Interact. 145, (3) Escher, B. U., and Schwarzenbach, R. P. (2002) Mechanistic studies on baseline toxicity and uncoupling of organic compounds as a basis for modeling effective membrane concentrations in aquatic organisms. Aquat. Sci. 64, (4) Selassie, C. D., DeSoyza, T. V., Rosario, M., Gao, H., and Hanch, C. (1998) Phenol toxicity in leukemia cells: A radical process? Chem.- Biol. Interact 113, (5) Zhang, L., Gao, H., Hanch, C., and Selassie, C. D. (1998) Molecular orbital parameters and comparative QSAR in the analysis of phenol toxicity to leukemia cells. J. Chem. Soc., Perkin Trans (6) Selassie, C. D., Shusterman, A. J.,., Kapur, S., Verma, R. P., and Zhang, L. (1999) On the toxicity of phenols to fast growing cells. A QSAR model for a radical based toxicity. J. Chem. Soc., Perkin Trans (7) Garg, R., Kapur, S., and Hancsh, C. (2000) Radical toxicity of phenols: A reference point for obtaining perspective in the formulation of QSAR. Med. Res. ReV. 21(1), (8) Selassie, C. D., Garg, R., Kapur, S., Kurup, A., Sajeshwar, R. P., Verma, P., Mekepati, S. B., and Hansch, C. (2002) Comparative QSAR and radical toxicity of various functional groups. Chem. ReV. 102, (9) Selassie, C. D., Verma, R. O., Kapur, S., Shusterman, A. J., and Hansch, C. (2002) QSAR for the cytotoxicity of 2-alkyl or 2,6 dialkyl, 4-X-phenols: The nature of the radical reaction. J. Chem. Soc., Perkin Trans (10) Cheng, Y., and Yuan, H. (2006) Quantitative study of electrostatic and steric effects on physicochemical property and biological activity. J. Mol. Graphics Modell. 24, (11) Hansch, C., McKarns, S. C., Smith, C. J., and Doolittle, D. J. (2000) Comparative QSAR evidence for a free radical mechanism of phenolinduced toxicity. Chem.-Biol. Interact. 127, (12) Kalyanaraman, B., Felix, C. C., and Sealy, R. C. (1985) Semiquinone anion radicals of catechol(amines)s, catechol estrogens, and their metal ion complexes. EnViron. Health Prespect. 63, (13) Basosi, R., Sealy, R. C., Kalyanaraman, B., and Hyde, J. S. (1984) Interaction of semiquinones with paramagnetic metal ions in aqueous solution. Relaxation and complex formation. J. Magn. Reson. 59 (1), (14) Porter, T. D., and Coon, M. J. (1991) Cytochrome P-450 multiplicity of isoforms, substrates, and catalytic and regulatory mechanisms. J. Biol. Chem. 266 (21), (15) Stiborová, M., Suchá, V., Mikśanová, M., Páca, J., Jr., and Páca, J. (2003) Hydroxylation of phenol to catechol by Candida tropicalis: Involvment of cytochrome p450. Gen. Physiol. Biophys. 22, (16) Powley, M. W., and Carlson, G. P. (2001) Cytochrome P450 isozymes involved in the metabolism of phenol, a benzene metabolite. Toxicol. Lett. 125, (17) Schlosser, P. M., Bond, J. A., and Medinsky, M. A. (1993) Benzene and phenol metabolism by mouse and rat liver microsomes. Carcinogenesis 14 (12), (18) Smith, M. T. (1996) The mechanism of benzene-induced leukemia: A hypothesis and speculations on the causes of leukemia. EnViron. Health Perspect. 104 (6), (19) Schlosser, M. J., Shurina, R. D., and Kalf, G. F. (1989) Metabolism of phenol and hydroquinone to reactive products by macrophage peroxidase or purified prostaglandin H synthase. EnViron. Health Perspect. 82, (20) Eastmond, D. A., Smith, M. T., Ruzo, L. O., and Ross, D. (1987) Metabolic activation of phenol by human myeloperoxidase and horseradish peroxidase. Mol. Pharmacol. 30 (6), (21) Smart, R. C., and Zannoni, V. C. (1984) DT diaphorase and peroxidase influence the covalent binding of the metabolism of phenol, the major metabolite of benzene. Mol. Pharm. 26, (22) Irons, R. D. (1985) Quinones as toxic metabolites of benzene. EnViron. Health 16, (23) Ishiharaa, Y., Shibaa, D., and Shimamoto, N. (2006) Enhancement of DMNQ-induced hepatocyte toxicity by cytochrome P450 inhibition. Toxicol. Appl. Pharmacol. 214 (2), (24) Liehr, J. G., and Roy, D. (1990) Free radical generation by redox cycling of estrogen. Free Radical Biol. Med. 8, (25) Nutter, L. M., Wu, Y.-Y., Ngo, E. O., Sierra, E., Gutierrez, P., and Abul-Hajj, Y. J. (1994) An o-quinone form of estrogen produces free radicals in human breast cancer cells: Correlation with DNA damage. Chem. Res. Toxicol. 7, (26) MOE (2007) Molecular Operating Environment, Chemical Computing Group, Montreal. (27) Halgren, T. A. (1996) Merck Molecular Force Field. I. Basis, form, scope, parameterization, and performance of MMFF94. J. Comput. Chem. 17 (5, 6), (28) Halgren, T. A. (1999) MMFF VI. MMFF94s option for energy minimization studies. J. Comput. Chem. 20 (7), (29) Wright, J. S., Johnson, E. R., and DiLabio, G. A. (2001) Predicting the activity of phenolic antioxidants: Theoretical method, analysis of substituent effects, and application to major families of antioxidants. J. Am. Chem. Soc. 123, (30) Frisch, M. J., Trucks, G. W., Schlegel, H. B., Scuseria, G. E., Robb, M. A., Cheeseman, J. R., Montgomery, J. A., Jr., Vreven, T., Kudin, K. N., Burant, J. C., Millam, J. M., Iyengar, S. S., Tomasi, J., Barone, V., Mennucci, B., Cossi, M., Scalmani, G., Rega, N., Petersson, G. A., Nakatsuji, H., Hada, M., Ehara, M., Toyota, K., Fukuda, R., Hasegawa, J., Ishida, M., Nakajima, T., Honda, Y., Kitao, O., Nakai, H., Klene, M., Li, X., Knox, J. E., Hratchian, H. P., Cross, J. B., Bakken, V., Adamo, C., Jaramillo, J., Gomperts, R., Stratmann, R. E., Yazyev, O., Austin, A. J., Cammi, R., Pomelli, C., Ochterski, J. W., Ayala, P. Y., Morokuma, K., Voth, G. A., Salvador, P., Dannenberg, J. J., Zakrzewski, V. G., Dapprich, S., Daniels, A. D., Strain, M. C., Farkas, O., Malick, D. K., Rabuck, A. D., Raghavachari, K., Foresman, J. B., Ortiz, J. V., Cui, Q., Baboul, A. G., Clifford, S., Cioslowski, J., Stefanov, B. B., Liu, G., Liashenko, A., Piskorz, P., Komaromi, I., Martin, R. L., Fox, D. J., Keith, T., Al-Laham, M. A., Peng, C. Y., Nanayakkara, A., Challacombe, M., Gill, P. M. W., Johnson, B., Chen, W., Wong, M. W., Gonzalez, C., and Pople, J. A. (2004) Gaussian 03, revision C.02, Gaussian, Inc., Wallingford, CT. (31) HPCVL, High Performance Computing Virtual Laboratory, Queens University, Kingston, ON; (32) Evans, M. G., and Polanyi, M. (1935) Some applications of the transition state methods to the calculation of reaction velocities, especially in solution. Trans. Faraday Soc. 31, (33) Vincente, J., Rosal, R., and Diaz, M. (2002) Noncatalytic oxidation of phenol in aqueous solutions. Ind. Eng. Chem. Res. 41, (34) Chen, Y., Shen, L., Zhang, F., Lau, S. S., van Breemen, R. B., Nikolic, D., and Bolton, J. L. (1998) The equine estrogen metabolite 4-Hydroxyequilenin causes DNA single-strand breaks and oxidation of DNA bases in vitro. Chem. Res. Toxicol. 11, (35) Bolton, J. L. (2002) Quinoids, quinoid radicals, and phenoxyl radicals formed from estrogens and antiestrogens. Toxicology 177, (36) Pisha, E., Lui, X., Constantinou, A. I., and Bolton, J. L. (2001) Evidence that a metabolite of equine estrogens, 4-hydroxyequilenin, induces cellular transformation in vitro. Chem. Res. Toxicol. 14, (37) Zhang, F., Swanson, S. M., van Breemen, R. B., Liu, X., Yang, Y., Gu, C., and Bolton, J. L. (2001) Equine estrogen metabolite 4-hydroxyequilenin induces DNA damage in the rat mammary tissues: Formation of single-strand breaks, apurinic sites, stable adducts, and oxidized bases. Chem. Res. Toxicol. 14, (38) Liu, X., Yao, J., Pisha, E., Yang, Y., Hua, Y., van Breemen, R. B., and Bolton, J. L. (2002) Oxidative DNA damage induced by equine estrogen metabolites: role of estrogen receptor. Chem. Res. Toxicol. 15, (39) Bolton, J. L., Pisha, E., Zhang, F., and Qiu, S. (1998) Role of quinoids in estrogen carcinogenesis. Chem. Res. Toxicol. 11 (10), TX800058R