- Email: [email protected]
Quantitative structure–activity relationships for estrogen receptor binding affinity of phenolic chemicals
Water Research 37 (2003) 1213–1222
Quantitative structure–activity relationships for estrogen receptor binding affinity of phenolic chemicals Jian-Ying Hua,*, Takako Aizawab b
a Department of Urban and Environment Science, Peking University, Beijing 100871, China Department of Water Supply Engineering, National Institute of Public Health, 4-6-1 Shiroganedai, Minato-ku, Tokyo 108, Japan
Received 18 June 2001; received in revised form 22 July 2002; accepted 23 July 2002
Abstract The estrogen receptor (ER) binding affinities of 25 compounds including 15 industrial phenolic chemicals, two phytoestrogens, three natural steroids and one man-made steroid were detected by a binding competition assay. The 17 industrial phenolic chemicals were selected as objective compounds because they are possibly released from epoxy and polyester–styrene resins used in lacquer coatings of concrete tank and lining of steel pipe in water supply system. A quantitative structure–activity relationship (QSAR) for structurally diverse phenols, nine alkylphenols with only one alkyl group, four hydroxyl biphenyls, bisphenol A and four natural and man-made estrogens was established by applying a quantum chemical modeling method. Logarithm of octanol–water coefficient (log Pow), molecular volume (Vm ), and energies of the highest occupied molecular orbital ðeHOMO Þ and lowest unoccupied molecular orbital ðeLUMO Þ were selected as hydrophobic, steric (Vm ), and electronic chemical descriptors, respectively. Chemicals capable of ER binding had large Vm and high eHOMO ; while the effects of log Pow and eLUMO on the binding affinity could not be identified. The QSAR made successful predictions for the three phytoestrogens. Also, the successful prediction of ERbinding affinity for biochanin A, another phytoestrogen, two indicators of pH (phenolphthalin and phenolphthalein) and one alkylphenolic chemical with three alkyl groups (4-methyl-2,6-di-butyl-phenol), by amending the Vm in the above-mentioned QSAR according to the electron-density distribution (or HOMO density) is an additional step in the elucidation of chemical steric and electronic parameters for predicting the binding affinities of phenolic compounds. r 2003 Elsevier Science Ltd. All rights reserved. Keywords: QSAR; Binding affinity; Phenolic compounds; Endocrine disruption; Quantum chemical modeling
1. Introduction Scientific and public concern heightens over the potential health effects of exposure to environmental pollutants with endocrine disruption potential [1]. While reproductive toxicity studies in animals are typically required for pesticides, other chemicals in use have not been routinely screened for these endocrine activities before being introduced for commercial use. Consequently, the significance of current levels of exposure to
*Corresponding author. Tel./fax: +86-10-62765520. E-mail address: [email protected] (J.-Y. Hu).
environmental estrogens or other hormonally active compounds is unclear. Certain anthropogenic compounds have been shown to elicit estrogenic hormonal activity by binding specifically to the estrogen receptor. One of the most widely recognized compounds of this type is a phenolic chemical group which includes alkylphenols (for example nonylphenol and octylphenol [2–4]), pH indicators [5], biphenyl [6], biphenyl methane (bisphenol A [7]), and stilbene derivatives containing two hydroxyl groups in the para position (4,40 -dihydroxydiethylstibene). While 4,40 -dihydroydiethylstilbene is one of the most potent man-made estrogens, the other phenolic chemicals elicit relatively weak estrogenicity. There are significant
0043-1354/03/$ - see front matter r 2003 Elsevier Science Ltd. All rights reserved. PII: S 0 0 4 3 - 1 3 5 4 ( 0 2 ) 0 0 3 7 8 - 0
1214
J.-Y. Hu, T. Aizawa / Water Research 37 (2003) 1213–1222
differences in the receptor-binding affinity of the various phenolic compounds, so that their biological activity and the significance of exposure to them even to those chemicals with weak estrogenicity are nonetheless important because of their environmental prevalence [8]. To effectively assess the exposure to such chemicals with endocrine disruption activity, the need for a rapid and sensitive screening technique becomes apparent. Although there are several in vitro assays that rely on indicators such as estrogen receptor binding, gene transcription, or cell proliferation, or on short-term in vivo assays such as uterine growth [1], quantitative structure–activity relationships (QSAR) could be invaluable as an initial screening tool for these chemicals prior to in vitro or in vivo assays. There are inherent advantages in the use of such a technique in that QSAR can provide mechanistic information and scaling doses in experiments. Because of the importance of the estrogen receptor (ER) in determining the estrogenicity of a chemical, which mimic or block the activity of natural estrogens by specifically binding to ER, there have been a number of attempts to model the relationship between the structures of chemicals and estrogen receptor binding affinity. Extensive binding studies of 17b-estradiol analogues have proposed a comprehensive binding property [9,10]. That is, the ER can bind with a wide variety of non-steroidal compounds, which are structurally analogous to the alkylsubstituted phenol of 17bestradiol. It was proposed that hydrogen bonding between the phenolic hydroxyl group and the binding site in ER, and also the hydrophobic and steric properties of compounds are important for the binding affinity of a chemical. The structural feature responsible for the estrogenic activity of alkylphenolic chemicals was also isolated based on the results of recombinant yeast screen [11]. The estrogenicity is very dependent on the size and degree of branching of the alkyl group, and its position on the phenol ring [12]. The maximum response is found at eight carbons and a tertiary branched structure. Other authors also reported similar results [13–15]; however, the estrogenicity is dependent on the carbon numbers of the straight alkyl group when the carbon number is less than seven. Besides the above-mentioned structural activity relationships based on simple structural features, more advanced computer-based models have been employed. These include electrostatic models [16], comparative molecular field analysis (CoMFA) [17,18] which considers the overall steric and electrostatic properties of the compound of interest, computer graphic and energy (electrostatic and van der Waals) based models for fit into DNA [19] and common reactivity patterns (COREPA) which reflect the stereoelectronic features [20–22]. While the above several models have been established for deferent chemical classes to estimate their endocrine
disruption potential, no one QSAR model however can simultaneously estimate the disruption potential of a wide variety of structurally diverse phenolic compounds including alkylphenols, hydroxyl biphenyls, hydroxyl biphenyl methanes, phytoestrogens, and natural/synthetic estrogens. In this study, we examined the ER-binding affinities of 25 kinds of structurally diverse phenols such as alkylphenols, hydroxyl biphenyls, bisphenol A together with phytoestrogens, pH indicators and natural and man-made estrogens, and evaluated, by applying a quantum chemical modeling method, their structural requirements for binding to the estrogen receptor which are possibly released from epoxy and polyester–styrene resins used in lacquer coatings of concrete tank and lining of steel pipe in water supply system.
2. Experimental section 2.1. Reagents and chemicals The structures of the 25 objective chemicals examined are shown in Tables 1 and 2. Diethylstilbestrol (DES, 99% pure), 4-n-nonylphenol, and 4-nonylphenol (a mixture kinds of isomers) were purchased from Sigma (St. Louis, MO), and Hayashi Co. (Tokyo, Japan), respectively. Six chemicals, namely 4-tert-octylphenol (95% pure), bisphenol A (99% pure), 4,40 -biphenyldiol (>96% pure), 4-ethylphenol (>97% pure), 3-tertbutylphenol (>96% pure) and, 2-tert-butylphenol (>96% pure), were purchased from Kanto Chemical Co. (Tokyo, Japan). Eight chemicals (4-hydroxybiphenyl (99% pure), 2-sec-butylphenol (98% pure), 2hydroxybiphenyl (99.5% pure), 4-tert-butylphenol (95% pure), phenol, 17b-estradiol, estriol, estrone (biochemical pure)), and four phytoestrogens (genistein, zearalenone, biochanin A, and daidzein) were purchased from Wako Co. (Tokyo, Japan). Two pH indicators, phenolphthalin, phenolphthalein and, two other chemicals, 4-sec-butylphenol (96% pure), 3-hydroxybiphenyl (95% pure) were purchased from Kanto Chemical Co. (Tokyo, Japan). Dimethylsulfoxide (DMSO) stock solutions of all chemicals were prepared to concentrations of 107 nM, except for phenol (108 nM). For binding competition assays, the FP Screen-for-competitor High Sensitivity Kit–Estrogen (Pan Vera Corporation) was purchased from Takarasyuzo Co. (Tokyo, Japan). 2.2. Binding competition assay A human receptor binding competition assay was applied to evaluate the binding affinities of the objective compounds. This assay uses fluorescence polarization to monitor the replacement of a fluorescent ligand with higher affinity from purified recombinant human ER.
J.-Y. Hu, T. Aizawa / Water Research 37 (2003) 1213–1222
1215
Table 1 Structures of training group (18 chemicals)
4-tert-octylphenol
4-n-nonylphenol A
CH3(CH2)8
CH3 (CH3)3CCH2 C CH3
OH
OH
OH
HO
4-hydroxybiphenyl 3-hydroxybiphenyl
OH
2-hydroxybiphenyl 4,4'-biphenyldiol
OH
HO HO
OH
4-sec-butylphenol C
OH
CH3CH2
bisphenol A
phenol
B
4-ethylphenol
H3 C CH3CH2CH2
OH
OH
2-sec-butylphenol
4-tert-butylphenol
HO CH3 CH3CH2CH2
(CH3)3C
OH
3-tert-butylphenol
2-tert-butylphenol
(CH3)3C
(CH3)3C
OH
HO
estriol
17β-estradiol
H3C
OH
H3 C
OH OH
D
HO
HO
estrone
H3C
O
DES HO OH
HO
A: alkylphenols and bisphenol A; B: hydroxybiphenyl; C: butylphenols; D: steroids and DES
The ER-binding affinity was determined as previously described [23]. Briefly, a stock solution of an objective compound was serially diluted using DMSO, and the 2 ml at each concentration was added to a borosilicate test tube followed by adding 48 ml buffer solution and 50 ml human recombinant estrogen receptor a (2 nM)fluorescent ligand (3 nM) complex solution. After mixing, and a 1-h incubation at room temperature, the fluorescence anisotropy of each tube was measured with the aid of a Beacon 2000 Fluorescence Polarization Instrument (Pan Vera Corporation) with 360 nm excitation filters and 530 nm emission filters. The anisotropy values were converted to percent inhibition using the following equation: I% ¼ ðA0 AÞ=ðA0 A100 Þ 100; where I%; A0 ; A100 and A are the percent inhibition, A at 0% inhibition, A at 100% inhibition, and observed A value, respectively. The percent inhibition versus
competitor concentration curves were analyzed by nonlinear least-squares curve fitting and yielded an IC50 value (the concentration of competitor needed to displace half of the bound ligand). The IC50 was defined as the ER-binding affinity of an objective compound. 2.3. Molecular descriptors Four descriptors, consisting of the energies of the highest occupied molecular orbital ðeHOMO Þ; the lowest unoccupied molecular orbital ðeLUMO Þ; the volume, and the logarithm of octanol–water partition coefficient (log Pow) were used for the QSAR evaluation between ER and compounds. Two stereoelectronic parameters, eHOMO and eLUMO ; and one steric descriptor, volume at ( 3, were calculated with the semiempirical meth0.01 e/A od MOPAC [24] using version 3.9 (CAChe Scientific, Oxford) of CAChe’s Macintosh Project Leader interface
J.-Y. Hu, T. Aizawa / Water Research 37 (2003) 1213–1222
1216
Table 2 Structures of test group (seven chemicals) OH
Genistein OH O
OH
Zearalenone
O O
A
HO
HO
O
Biochanin AOH O HO
O
Daidzein O
OCH3
HO
O
O
Phenolphthalein OH
Phenolphthalin OH
OH
4-methyl-2,6-di-tert-butyl-phenol OH (CH3)3C
B
OH
OH O O
C(CH3)3
CH3
O
OH A:phytoestrogen; B: anthropogenic phenolics
(running on a Power Macintosh 7500/100 computer). The PM3 [25] parameterizations of atomic wave functions were used for gas-phase calculations. The log Pows of chemicals were calculated by using ACD/ log Pow Ver.1.0 (Advanced Chemistry Development Inc.).
3. Results and discussion 3.1. Binding affinity To evaluate the affinity binding to ER, the 18 compounds in Table 1 were selected as the training group, and the binding affinities were determined. Fig. 1 shows the competition binding curves of 18 compounds against a human recombinant estrogen receptor a(hrERa)/fluorescent ligand complex. The resulting classic competitive binding curves were analyzed using PRISM software (Ver.2.0; GraphPad Software Inc., USA) in order to assess the potency of the competitor molecule. The concentration that inhibited 50% (i.e. the IC50 value) of each test compound was calculated by nonlinear least-squares regression and is shown in Tables 3 and 4. Except for phenol, which shows very weak hrER-a interaction, the other industrial chemicals including bisphenol A, 4-n-nonylphenol, and 4-tertoctylphenol, which have been shown to elicit estrogenic effects in vitro and in vivo [26–28], show significant hrER-a interaction in this binding assay. The binding affinity of 4,4-diphenyldiol was found to be higher than that of bisphenol A. In order to compare the interaction behavior, the binding affinities of three steroids and one
non-natural steroid, DES in Table 1(D), were also determined. The DES was found to have the strongest affinity. In addition, it was found that the IC50 of 4nonylphenol, a mixture of many kinds of isomers with branched alkyls, was 4.96 103 nM, which is similar to that of 4-n-nonylphenol. It has previously been reported that while 4-nonylphenol elicited the estrogen response, no estrogen response for 4-n-nonylphenol was detected based on the results of a reporter gene expression assay using a yeast two-hybrid method [14]. From the above result, it was supposed that 4-n-nonylphenol might block the effect of naturally occurring estrogens. 3.2. The mechanistic QSAR approach: predictor selection According to the agonist and antagonist binding modes of 17b-estradiol in the ER [10], the 17b-estradiol binds diagonally across the cavity which has a probe ( 3. Hormone recognition is accessible volume of 450 A achieved through a combination of specific hydrogen bonds and the complementarity of the hydrophobic binding cavity to 17b-estradiol’s non-polar character. Based on such a three-dimensional (3-D) structure of the ER, we have attempted to establish a mechanistic QSAR model involving a phenolic chemical and the ER. Eq. (1) shows the general form of this model. log IC50 ¼ bulk þ polarity þ hydrogen bonding:
ð1Þ
The binding affinity of a phenolic compound is described as a linear expression with terms for a bulk effect, a polarity effect, and hydrogen-bonding effects. For the affinity binding to ER, the specific terms and
J.-Y. Hu, T. Aizawa / Water Research 37 (2003) 1213–1222 -20
-10 30 50 2-hydroxybiphenyl 3-hydroxybiphenyl
70 90
4,4'-diphenyldiol 4-hydroxybiphenyl
110
10 4
20 40 60 80 100
130 10 3
0
Inhibition(%)
Inhibition(%)
10
105
10 6
120 2 10
10 7
2-t-butylphenol 2-sec-butylphenol 4-t-butylphenol 3-t-butylphenol 4-t-butylphenol
10 3
Concentration(nM)
Inhibition(%)
40
100
bis-phenolA 4-ethylphenol phenol 4-t-octyl-phenol 4-n-nonyl-phenol
120 1 10
10 2
80
10 5
10 6
107
10 8
10 3
10 4
0
20
60
10 4
Concentration(nM)
0
Inhibition(%)
1217
10 3
10 4
20 40 60 80 100
105
106
10 7
10 8
120 -2 10
Concentration (nM)
DES estradiol estrone estriol
10 -1
10 0
10 1
10 2
Concentration(nM)
Fig. 1. Competition binding curves of various compounds against a human recombinant estrogen receptor a/fluorescent ligand complex (hrER-a-FES-1). Increasing concentrations of competitor were incubated with 2 nM hrER-a-3 nM FES1 for 60 min at room temperature followed by the measurement of fluorescence polarization.
Table 3 Physicochemical properties of training group (18 chemicals) and their binding affinities (IC50) with the estrogen receptor
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18
Chemicals
log Pow
eHOMO (eV)
eLUMO (eV)
( 3) Vm (A
IC50 (nM)
4-n-nonylphenol 4-tert-octyl-phenol 4,40 -biphenyldiol Bisphenol A 4-hydroxybiphenol 4-sec-butylphenol 3-hydroxybiphenol 2-hydroxybiphenol 4-tert- butylphenol 2-sec-butylphenol 4-ethyl-phenol 3-tert- butylphenol 2-tert-butylphenol Phenol Diethylstibestrol 17b-estradiol Estriol Estrone
6.19 5.66 2.42 3.43 3.20 3.35 3.23 2.94 3.17 3.35 2.47 3.17 3.17 1.48 7.03 4.13 2.94 3.69
8.942 8.990 8.441 8.972 8.640 8.980 8.911 8.808 8.997 9.047 8.926 9.093 9.094 9.176 8.653 8.900 8.931 9.005
0.320 0.349 0.31 0.272 0.336 0.333 0.408 0.46 0.347 0.293 0.334 0.314 0.314 0.291 0.027 0.339 0.308 0.237
167.90 155.35 112.88 149.19 108.89 106.93 108.96 108.91 106.66 106.94 82.23 106.83 106.80 58.97 178.01 186.53 190.24 181.52
9.51 103 1.41 104 1.06 105 1.10 105 4.27 105 5.82 105 6.20 105 9.97 105 1.01 106 1.38 106 1.70 106 2.29 106 2.36 106 9.50 107 6 2.30 101 5.00 101 3.31 102
corresponding descriptors are as follows: bulk, volume (Vm ) in cubic angstroms; polarity, logarithm of octanol– water partition coefficient (log Pow); hydrogen bonding, eHOMO or eLUMO : The hydrogen-bonding term is separated into acidic and basic components. Since any hydrogen bond can be considered to have covalent and electrostatic parts, it is necessary to describe the two
bonding contributions using individual descriptor. The covalent contribution to the basic term can associate with the eHOMO which has the electron to form a complex with a proton on a neighboring molecule [29]. In this study, a higher value of eHOMO for a phenolic chemical indicates a greater tendency to form a hydrogen bond with the ER. The electrostatic
J.-Y. Hu, T. Aizawa / Water Research 37 (2003) 1213–1222
1218
Table 4 Physicochemical properties of test group (seven chemicals) and the comparison between observed and calculated binding affinities (IC50) with the estrogen receptor eHOMO
( 3) Vm ( A
( 3) Vm 0 (A
IC50 (nM) Observed
Predicted
Predicted0
Daidzein Genistein Zearalenone Biochanin-A
8.616 8.553 9.375 8.518
139.16 143.13 196.12 156.15
— — — —n
4.27 103 9.12 102 7.08 102 >1.74 107
7.59 103 3.80 103 3.55 102 5.25 102
— — — 8.51 109
Phenolphthalin Phenolphthalein 4-methyl-2,6-di-butyl-phenol
9.138 9.157 8.699
183.79 178.68 164.75
88.8 144.5 106.8
>5.89 107 8.5 105 4.90 105
6.61 103 1.02 104 5.25 103
7.18 106 1.26 105 4.37 105
Vm 0 : the effective bulk which was calculated based on the Eq. (8). Predicted: the binding affinity which was predicted based on the Eq. (4) for anthropogenic phenolics and Eq. (7) for phytoestrogen. Predicted0 : the binding affinity which was predicted based on the Eq. (4) for anthropogenic phenolics and Eq. (7) for phytoestrogen where Vm is replaced by Vm 0 . n It is estimated to be zero.
contribution to the basicity is taken as the magnitude of the most negative formal charge in the molecule. On the other hand, the covalent contribution to the acidic term can be taken as eLUMO ; with a lower value of eLUMO for a phenolic compound indicating a greater tendency to form a hydrogen bond with the ER. In this case, the hydrogen-bonding acidity is applied with this orbital being available to accept electrons from the HOMO orbital of ER. The electrostatic contribution to the acidity is taken as the most positive formal charge, on a hydrogen atom in the molecule. Because the positive formal charge on a hydrogen atom and the negative formal charge on an oxygen atom for the objective phenolic chemicals are similar, the electrostatic contribution to hydrogen binding can be neglected.
where n stands for the number of samples, r2 the variance and s the standard error of the estimate. Fig. 2(a) compares the observed and calculated estrogen receptor binding affinities. It is clear that log Pow was not a proper parameter to predict the ER-binding affinities of 4,40 -biphenydiol and phenol. While the much better correlation between ER-binding affinities and Vm was found, the best regression was acquired by correlating IC50 with Vm and eHOMO simultaneously. As shown in Fig. 2(a) and Eq. (4), the calculated IC50 matched the observed IC50 values well for all of the objective compounds when Vm was combined with eHOMO : log IC50 ¼ 1:46eHOMO 0:032Vm 3:644 r2 ¼ 0:948 n ¼ 14 s ¼ 0:248:
ð4Þ
3.3. QSAR for ER-binding affinity Based on the assumption that the electrostatic contribution to hydrogen-binding can be neglected, correlations were conducted between ER-binding affinity of the phenolic compounds and their volumetric parameter (Vm ), log Pow and eHOMO or eLUMO : It was found that significant correlations could be obtained upon using Vm and log Pow as individual predictors. Eq. (2) expresses the regression between log IC50 and logP ow, and Eq. (3) is the regression between log IC50 and Vm : log IC50 ¼ 0:698 log Pow þ 8:11 r2 ¼ 0:70
n ¼ 14 s ¼ 0:57;
ð2Þ
log IC50 ¼ 0:033lVm þ 9:55 r2 ¼ 0:87
n ¼ 14 s ¼ 0:37;
ð3Þ
It should be noted that it is the HOMO orbital energy, but not LUMO orbital energy, that appears in the significant regression, and the covalent contribution to the basic term of that in Eq. (1) was confirmed. That is, the phenolic compound donates the electron and forms a complex with a proton on the ER. To investigate whether the phenolic chemicals under investigation bind to the ER through a similar molecular mechanism with natural steroids, regressions for the combined set of 14 industrial compounds, three natural and one non-natural steroid (DES) in Table 1 were also conducted. A significant regression could not be obtained upon using only the hydrophobic parameter, log Pow (Eq. (5)). Two significant regressions, Eq. (6) using one parameter of Vm and Eq. (7) using two predictors, Vm ; and eHOMO ; were obtained; but the hydrophobic parameter, log Pow, and eLUMO are also excluded. This result can be interpreted such that the
J.-Y. Hu, T. Aizawa / Water Research 37 (2003) 1213–1222 9.0 Equation(2) Equation(3) Equation(4)
Observed logIC50
8.0
phenol
7.0 6.0 5.0
4,4'-biphenydiol
4.0 3.0 3.0
4.0
5.0
6.0
7.0
8.0
9.0
Calculated logIC50(nM)
(a) 9.0 Equation(5) Equation(6) Equation(7)
Observed logIC50
8.0 7.0 6.0 5.0 4.0 3.0
estrone
2.0
estriol
1.0
17β-estradiol
0.0 0.0
(b)
2.0
4.0
6.0
8.0
Calculated logIC50(nM)
Fig. 2. Plot of observed versus calculated estrogen receptor binding affinity. (a) Fourteen industrial phenolic chemicals; (b) 14 industrial phenolic chemicals, three steroids and DES.
phenolic chemicals bind to ER with a similar mechanism to that for steroids. log IC50 ¼ 0:90 log Pow þ 8:03 r2 ¼ 0:40
n ¼ 18 s ¼ 1:56;
ð5Þ
log IC50 ¼ 0:049Vm 11:10 r2 ¼ 0:89
n ¼ 18 s ¼ 0:69;
ð6Þ
log IC50 ¼ 1:90eHOMO 0:047Vm 6:05 r2 ¼ 0:92
n ¼ 18 s ¼ 0:62:
ð7Þ
3.4. Robust of the QSAR model The test compounds shown in Table 2 were used to check the robustness of the above QSAR models. Of seven compounds, four (genistein, biochanin A, zearalenone and daidzein), are phytoestrogens with structures which are planar and similar to those of steroids; the two pH indicators, phenolphthalin and phenolphthalein, bear some structural resemblance to bisphenol A; and one compound is alkylphenolic with
1219
three alkyl group on the phenolic ring. The determined binding affinities and their molecular predictors are shown in Table 4. The QSAR model (Eq. (7)) predicted the ER-binding affinity with high accuracy for the three phytoestrogens, genistein, zearalenone, and daidzein, suggesting that the binding behavior of the phytoestrogens is similar to that of the objective compounds. From the fact that the QSAR was unsuccessful in predicting the binding affinity of biochanin A, it is reasonable to suppose that besides the eHOMO and Vm ; the QSAR model (Eq. (7)) could be improved by introducing another molecular predictor or by amending the present predictors. Considering the significance of eHOMO in predicting the binding affinity, the electron-density (or HOMO density) distribution of biochanin A was compared with the other three phytoestrogens. Due to the similarity of genistein, zearalenone and daidzein, only genistein was selected, and its electron-density distribution was compared with that of biochanin A (Fig. 3). From the difference in electron-density distribution, it was found that the electron densities on the HOMO orbitals of both biochanin A and genistein are all distributed in the A benzene ring. It should be noted that there is no hydroxyl on the A ring, and no electron density was distributed on the hydroxyl of the B ring in the case of biochanin A, which possibly leads to its lower ERbinding affinity due to the importance of hydroxyl in binding to ER. It is supposed that only part of the molecular bulk of biochanin A contributed to the ERbinding affinity. The distribution range of electron density on the phenolic moiety (B ring), which contributes to the ER-binding affinity, was defined here as the effective bulk (Vm 0 ). In the case of biochanin A, the Vm 0 is near 0 because no electron density was distributed on phenolic moiety (B ring) as shown in Fig. 3. Thus, the predicted binding affinity of biochanin A was calculated to be near 8.51 109 nM (Predicted0 in Table 4) based on Eq. (7). This value is comparatively near its observed detected binding affinity (>1.74 107 nM). In addition, Fig. 4 compares the electron-density distributions on the HOMO of phenolphthalein and phenolphthalin with bisphenol A due to their structural resemblance. Comparing their 3-D structures and electron-density distribution on the HOMO, it was found that while that electron-density distribution ranges between hydroxyl on the A ring to the hydroxyl on the B ring in the case of phenolphthalein, similar to that of bisphenol A, that for phenolphthalin localized on only one ring (A in Fig. 4), a result which significantly differs from that of bisphenol A. The finding also suggested that the electron-density distribution on the HOMO is quite important and is possibly another reasonable predictor in the QSAR model in addition to Vm and eHOMO : To amend the effect of electron-density
1220
J.-Y. Hu, T. Aizawa / Water Research 37 (2003) 1213–1222
Fig. 3. Comparison between electron-density distribution in HOMOs of genistein and biochanin A. Black and dark gray colors show ( the HOMOs. The ranges of electron-density distribution on the HOMOs of genistein and biochanin A are 12.04 and 12.84 A, respectively.
Fig. 4. Comparison among the electron-distribution in HOMOs of phenolphthalein, phenolphthalein, and bisphenol A. Black and dark gray colors show the HOMOs. The ranges of electron-density distribution on the HOMOs of phenolphthalein, phenolphthalin, ( (L in Eq. (8)), 5.66 A ( (L in Eq. (8)) and 9.51 A ( (L0 in Eq. (8)), respectively. and bisphenol A are 9.21 A
distribution on the binding affinity, the following attempt was made. By using bisphenol A as a template, the Vm in Eq. (4) was amended to effective bulk, Vm 0 . Thus, the Vm 0 of phenolphthalin and phenolphthalein ( respectively, are 88.8 and 144.5 A3, Vm0 ¼ L Vm0 =L0 ;
ð8Þ
where Vm0 is the Vm of bisphenol A; L is the range of electron-density distribution on the HOMO, and L0 is that of bisphenol A as shown in Fig. 4. As a result, the predicted binding affinities to ER for phenolphthalin and phenolphthalein are 7.18 106 and 1.26 105 nM, respectively, which match with the observed values (Table 4).
J.-Y. Hu, T. Aizawa / Water Research 37 (2003) 1213–1222
Finally, an attempt to predict the binding affinity of an alkylphenolic compound, 4-methyl-2,6-di-butylphenol was made. When using the absolute value for the Vm of 4-methyl-2,6-di-butyl-phenol in Eq. (4), the predicted IC50 is 5.25 103 nM, which is unfortunately much lower than the observed value of 4.90 105 nM. When the Vm of 2-tert-butylphenol, which was used as the template, was employed instead of that of 4-methyl2,6-di-butyl-phenol, the predicted IC50 was 4.37 105 nM, which is comparable with the observed value as shown in Table 4. This finding suggested that the Vm in Eq. (4) may need to be amended to a volume commensurate with its contribution for estimating the binding affinity of an alkylphenolic compound substituted by more than one alkyl, although more experimental data are required to confirm the above inferential conclusion. In summary, the QSAR can effectively predict the binding affinities of phenolic compounds whose structures were similar to those used in the training group. For the phenolic compounds whose structures were dissimilar to those used in the training group, the values of the steric parameter, Vm ; in this QSAR should be amended according to their status of electron-density distribution on the HOMO or of alkyl-substitution, where a suitable compound in the training group is used as a template.
4. Conclusion We carried out a fundamental study to develop a simplified method for estimating ER-binding affinities of phenolic compounds. The results of correlation analysis indicated that ER-binding affinity can essentially be estimated from the energy of the highest occupied molecular orbital ðeHOMO Þ and steric factor, a molecular volume. This was, especially true in the case where the correlation model was established by amending a molecular volume based on electron-density distributions on the HOMO according to their structural features.
Acknowledgements Financial support from both the Japanese Governmental Research Fund (Ministry of Health Welfare) and the National Natural Science Foundation of China [49925103 and 40024101] is gratefully acknowledged.
References [1] Zacharewski T. In vitro bioassays for assessing estrogenic substances. Environ Sci Technol 1997;31:613–23.
1221
[2] White R, Jobling S, Hoare SA, Sumpter JP, Parker MG. Environmentally persistent alkylphenolic compounds are estrogenic. Endocrinology 1994;135:175–82. [3] Mueller GC, Kim Uh-hee. Displacement of estradiol from estrogen receptors by simple alkyl phenols. Endocrinology 1978;102:1429–35. [4] Soto AM, Lin Tien-Min, Justicia H, Silvia RM, Sonnenschein C. An ‘‘in culture’’ bioassay to assess the estrogenicity of xenobiotics. In: Colborn T, Clement C, editors. Chemically induced alterations in sexual development: the wildlife/human connection. Princeton, NJ: Princeton Scientific Publishing, 1992. pp. 295–309. [5] Berthois Y, Katzenellenbogen JA, Katzenellenbogen BS. Phenol red in tissue culture media is a weak estrogen: implications concerning the study of estrogen-responsive cells in culture. Cell Biol 1986;83:2496–500. [6] Korach KS, Sarver P, Chae Kun, McLachlan JA, McKinney JD. Estrogen receptor-binding activity of polychlorinated hydroxybiphenyls: conformationally restricted structural probes. Mol Pharmacol 1988;33: 120–6. [7] Gould JC, Leonard LS, Maness SC, Wagner BL, Conner K, Zacharewski T, Safe S, McDonnell DP, Gaido KW. Bisphenol A interacts with the estrogen receptor a in a distinct manner from estradiol. Mol Cell Endocrinol 142;1998:203–14. [8] Thiele B, Gunther K, Schwuger MJ. Alkylphenol ethoxylates, trace analysis and environmental behavior. Chem Rev 1997;97:3247–72. [9] Anstead GM, Carlson KE, Katzenellenbogen JA. The estradiol pharmacophore: ligand structure-estrogen receptor binding affinity relationships and a model for the receptor binding site. Steroids 1997;62:268–303. [10] Brzozowski AM, Pike ACW, Dauter Z, Hubbard RE, Bonn T, Engstrom O, Ohman L, Greene GL, Gustafsson J, Carlquist M. Molecular basis of agonism and antagonism in the oestrogen receptor. Nature 1997;389:753–8. [11] Routledge EJ, Sumpter JP. Estrogenic activity of surfactants and some of their degradation products assessed using a recombinant yeast screen. Environ Toxicol Chem 1996;15:241–8. [12] Routledge EJ, Sumpter JP. Structural Features of alkylphenolic chemicals associated with estrogenic activity. J Biol Chem 1997;272(6):3280–8. [13] Blair RM, Fang H, Branham WS, Hass BS, Dial SL, Moland CL, Tong W, Shi L, Perkins R, Sheehan DM. The estrogen receptor relative binding affinities of 188 natural and xenochemicals: structural diversity of ligands. Toxicol Sci 2000;54:138–53. [14] Nishihara T, Nishikawa J, Kanayama T, Dakeyama F, Saito K, Imagawa M, Takatori S, Kitagawa Y, Hori S, Utsumi H. Estrogenic activities of 517 chemicals by yeast two-hybrid assay. J Health Sci 2000;46(4):282–98. [15] Taira Y, Nakai M, Asai D, Yakabe Y, Tahara Y, Shinmyozu T, Noguchi M, Takatsuki M, Shimohigashi Y. Structural requirements of para-alkylphenols to bind to estrogen receptor. Eur J Biochem 1999;262:240–5. [16] Van der Kuur JA, Wiese T, Brooks SC. Influence of estrogen structure on nuclear binding and progesterone receptor induction by the receptor complex. Biochemistry 1993;32:7002–8.
1222
J.-Y. Hu, T. Aizawa / Water Research 37 (2003) 1213–1222
[17] Waller CL, Oprea TI, Chae K, Park HK, Korach KS, Laws SC, Wiese TE, Kelce WR, Gray Jr LE. Ligand-based identification of environmental estrogens. Chem Res Toxicol 1996;9:1240–8. [18] Waller CL, Minor DL, McKinney JD. Using three dimensional quantitative structure-activity relationships to examine estrogen receptor binding affinities of polychlorinated hydroxybiphenyls. Environ Health Perspect 1995;103:702–7. [19] Hendry LB, Chu CK, Copland JA, Mahesh VB. Antiestrogenic piperidinediones designed prospectively using computer graphics and energy calculations of DNA-ligand complexes. J Steroid Biochem Molec Biol 1994;48: 495–505. [20] Bradbury SP, Mekenyan OG, Ankley GT. Quantitative structure-activity relationships for polychlorinated hydroxybipheny estrogen receptor binding affinity: an assessment of conformer flexibility. Environ Toxicol Chem 1996;15(11):1945–54. [21] Schmieder PK, Aptula AO, Routledge EJ, Sumpter JP, Mekenyan OG. Estrogenicity of alkylphenolic compounds: a 3-D structure-activity evaluation of gene activation. Environ Toxicol Chem 2000;19(7):1727–40. [22] Kramer VJ, Giesy JP. Specific binding of hydroxylated polychlorinated biphenyl metabolites and other substances to bovine calf uterine estrogen receptor: structure-binding relationships. Sci Total Environ 1999;233:141–61.
[23] Bolger R, Wiese TE, Ervin K, Nestich S, Checovich W. Rapid screening of environmental chemicals for estrogen receptor binding capacity. Environ Health Perpect 1998;106(9):551–7. [24] Stewart JJP. Optimization of parameters for semiempirical methods, II. Applications. J Comput Chem 1989;10: 221–64. [25] Stewart JJP. MOPAC manual, Frank J. Seiler Research Laboratory, United States Air Force Academy, Colorado Springs CO 80840, 1990. [26] Jobing S, Sheahan D, Osborne JA, Matthiessen P, Sumpter JP. Inhibition of testicular growth in rainbow trout (Oncorhynchus mykiss) exposed to estrogenic alkylphenolic chemicals. Environ Toxicol Chem 1996;15(2):194–202. [27] Sharpe RM, Fisher JS, Millar MM, Jobling S, Sumpter JP. Gestational and lactational exposure of rats to xenoestrogens results in reduced testicular size and sperm production. Environ Health Perspect 1995;103(12):1136–43. [28] Strunck E, Stemmann N, Hopert AC, Wunsche W, Frank K, Vollmer G. Relative binding affinity does not predict biological response to xenoestrogens in rat endometrial adenocarcinoma cells. J Steroid Biochem Molec Biol 2000;74:73–81. [29] Wilson LY, Famini GR. Using theoretical descriptors in quantitative structure-activity relationships: some toxicological indices. J Med Chem 1991;34:1668–74.
Quantitative structure–activity relationships for estrogen receptor binding affinity of phenolic chemicals Jian-Ying Hua,*, Takako Aizawab b
a Department of Urban and Environment Science, Peking University, Beijing 100871, China Department of Water Supply Engineering, National Institute of Public Health, 4-6-1 Shiroganedai, Minato-ku, Tokyo 108, Japan
Received 18 June 2001; received in revised form 22 July 2002; accepted 23 July 2002
Abstract The estrogen receptor (ER) binding affinities of 25 compounds including 15 industrial phenolic chemicals, two phytoestrogens, three natural steroids and one man-made steroid were detected by a binding competition assay. The 17 industrial phenolic chemicals were selected as objective compounds because they are possibly released from epoxy and polyester–styrene resins used in lacquer coatings of concrete tank and lining of steel pipe in water supply system. A quantitative structure–activity relationship (QSAR) for structurally diverse phenols, nine alkylphenols with only one alkyl group, four hydroxyl biphenyls, bisphenol A and four natural and man-made estrogens was established by applying a quantum chemical modeling method. Logarithm of octanol–water coefficient (log Pow), molecular volume (Vm ), and energies of the highest occupied molecular orbital ðeHOMO Þ and lowest unoccupied molecular orbital ðeLUMO Þ were selected as hydrophobic, steric (Vm ), and electronic chemical descriptors, respectively. Chemicals capable of ER binding had large Vm and high eHOMO ; while the effects of log Pow and eLUMO on the binding affinity could not be identified. The QSAR made successful predictions for the three phytoestrogens. Also, the successful prediction of ERbinding affinity for biochanin A, another phytoestrogen, two indicators of pH (phenolphthalin and phenolphthalein) and one alkylphenolic chemical with three alkyl groups (4-methyl-2,6-di-butyl-phenol), by amending the Vm in the above-mentioned QSAR according to the electron-density distribution (or HOMO density) is an additional step in the elucidation of chemical steric and electronic parameters for predicting the binding affinities of phenolic compounds. r 2003 Elsevier Science Ltd. All rights reserved. Keywords: QSAR; Binding affinity; Phenolic compounds; Endocrine disruption; Quantum chemical modeling
1. Introduction Scientific and public concern heightens over the potential health effects of exposure to environmental pollutants with endocrine disruption potential [1]. While reproductive toxicity studies in animals are typically required for pesticides, other chemicals in use have not been routinely screened for these endocrine activities before being introduced for commercial use. Consequently, the significance of current levels of exposure to
*Corresponding author. Tel./fax: +86-10-62765520. E-mail address: [email protected] (J.-Y. Hu).
environmental estrogens or other hormonally active compounds is unclear. Certain anthropogenic compounds have been shown to elicit estrogenic hormonal activity by binding specifically to the estrogen receptor. One of the most widely recognized compounds of this type is a phenolic chemical group which includes alkylphenols (for example nonylphenol and octylphenol [2–4]), pH indicators [5], biphenyl [6], biphenyl methane (bisphenol A [7]), and stilbene derivatives containing two hydroxyl groups in the para position (4,40 -dihydroxydiethylstibene). While 4,40 -dihydroydiethylstilbene is one of the most potent man-made estrogens, the other phenolic chemicals elicit relatively weak estrogenicity. There are significant
0043-1354/03/$ - see front matter r 2003 Elsevier Science Ltd. All rights reserved. PII: S 0 0 4 3 - 1 3 5 4 ( 0 2 ) 0 0 3 7 8 - 0
1214
J.-Y. Hu, T. Aizawa / Water Research 37 (2003) 1213–1222
differences in the receptor-binding affinity of the various phenolic compounds, so that their biological activity and the significance of exposure to them even to those chemicals with weak estrogenicity are nonetheless important because of their environmental prevalence [8]. To effectively assess the exposure to such chemicals with endocrine disruption activity, the need for a rapid and sensitive screening technique becomes apparent. Although there are several in vitro assays that rely on indicators such as estrogen receptor binding, gene transcription, or cell proliferation, or on short-term in vivo assays such as uterine growth [1], quantitative structure–activity relationships (QSAR) could be invaluable as an initial screening tool for these chemicals prior to in vitro or in vivo assays. There are inherent advantages in the use of such a technique in that QSAR can provide mechanistic information and scaling doses in experiments. Because of the importance of the estrogen receptor (ER) in determining the estrogenicity of a chemical, which mimic or block the activity of natural estrogens by specifically binding to ER, there have been a number of attempts to model the relationship between the structures of chemicals and estrogen receptor binding affinity. Extensive binding studies of 17b-estradiol analogues have proposed a comprehensive binding property [9,10]. That is, the ER can bind with a wide variety of non-steroidal compounds, which are structurally analogous to the alkylsubstituted phenol of 17bestradiol. It was proposed that hydrogen bonding between the phenolic hydroxyl group and the binding site in ER, and also the hydrophobic and steric properties of compounds are important for the binding affinity of a chemical. The structural feature responsible for the estrogenic activity of alkylphenolic chemicals was also isolated based on the results of recombinant yeast screen [11]. The estrogenicity is very dependent on the size and degree of branching of the alkyl group, and its position on the phenol ring [12]. The maximum response is found at eight carbons and a tertiary branched structure. Other authors also reported similar results [13–15]; however, the estrogenicity is dependent on the carbon numbers of the straight alkyl group when the carbon number is less than seven. Besides the above-mentioned structural activity relationships based on simple structural features, more advanced computer-based models have been employed. These include electrostatic models [16], comparative molecular field analysis (CoMFA) [17,18] which considers the overall steric and electrostatic properties of the compound of interest, computer graphic and energy (electrostatic and van der Waals) based models for fit into DNA [19] and common reactivity patterns (COREPA) which reflect the stereoelectronic features [20–22]. While the above several models have been established for deferent chemical classes to estimate their endocrine
disruption potential, no one QSAR model however can simultaneously estimate the disruption potential of a wide variety of structurally diverse phenolic compounds including alkylphenols, hydroxyl biphenyls, hydroxyl biphenyl methanes, phytoestrogens, and natural/synthetic estrogens. In this study, we examined the ER-binding affinities of 25 kinds of structurally diverse phenols such as alkylphenols, hydroxyl biphenyls, bisphenol A together with phytoestrogens, pH indicators and natural and man-made estrogens, and evaluated, by applying a quantum chemical modeling method, their structural requirements for binding to the estrogen receptor which are possibly released from epoxy and polyester–styrene resins used in lacquer coatings of concrete tank and lining of steel pipe in water supply system.
2. Experimental section 2.1. Reagents and chemicals The structures of the 25 objective chemicals examined are shown in Tables 1 and 2. Diethylstilbestrol (DES, 99% pure), 4-n-nonylphenol, and 4-nonylphenol (a mixture kinds of isomers) were purchased from Sigma (St. Louis, MO), and Hayashi Co. (Tokyo, Japan), respectively. Six chemicals, namely 4-tert-octylphenol (95% pure), bisphenol A (99% pure), 4,40 -biphenyldiol (>96% pure), 4-ethylphenol (>97% pure), 3-tertbutylphenol (>96% pure) and, 2-tert-butylphenol (>96% pure), were purchased from Kanto Chemical Co. (Tokyo, Japan). Eight chemicals (4-hydroxybiphenyl (99% pure), 2-sec-butylphenol (98% pure), 2hydroxybiphenyl (99.5% pure), 4-tert-butylphenol (95% pure), phenol, 17b-estradiol, estriol, estrone (biochemical pure)), and four phytoestrogens (genistein, zearalenone, biochanin A, and daidzein) were purchased from Wako Co. (Tokyo, Japan). Two pH indicators, phenolphthalin, phenolphthalein and, two other chemicals, 4-sec-butylphenol (96% pure), 3-hydroxybiphenyl (95% pure) were purchased from Kanto Chemical Co. (Tokyo, Japan). Dimethylsulfoxide (DMSO) stock solutions of all chemicals were prepared to concentrations of 107 nM, except for phenol (108 nM). For binding competition assays, the FP Screen-for-competitor High Sensitivity Kit–Estrogen (Pan Vera Corporation) was purchased from Takarasyuzo Co. (Tokyo, Japan). 2.2. Binding competition assay A human receptor binding competition assay was applied to evaluate the binding affinities of the objective compounds. This assay uses fluorescence polarization to monitor the replacement of a fluorescent ligand with higher affinity from purified recombinant human ER.
J.-Y. Hu, T. Aizawa / Water Research 37 (2003) 1213–1222
1215
Table 1 Structures of training group (18 chemicals)
4-tert-octylphenol
4-n-nonylphenol A
CH3(CH2)8
CH3 (CH3)3CCH2 C CH3
OH
OH
OH
HO
4-hydroxybiphenyl 3-hydroxybiphenyl
OH
2-hydroxybiphenyl 4,4'-biphenyldiol
OH
HO HO
OH
4-sec-butylphenol C
OH
CH3CH2
bisphenol A
phenol
B
4-ethylphenol
H3 C CH3CH2CH2
OH
OH
2-sec-butylphenol
4-tert-butylphenol
HO CH3 CH3CH2CH2
(CH3)3C
OH
3-tert-butylphenol
2-tert-butylphenol
(CH3)3C
(CH3)3C
OH
HO
estriol
17β-estradiol
H3C
OH
H3 C
OH OH
D
HO
HO
estrone
H3C
O
DES HO OH
HO
A: alkylphenols and bisphenol A; B: hydroxybiphenyl; C: butylphenols; D: steroids and DES
The ER-binding affinity was determined as previously described [23]. Briefly, a stock solution of an objective compound was serially diluted using DMSO, and the 2 ml at each concentration was added to a borosilicate test tube followed by adding 48 ml buffer solution and 50 ml human recombinant estrogen receptor a (2 nM)fluorescent ligand (3 nM) complex solution. After mixing, and a 1-h incubation at room temperature, the fluorescence anisotropy of each tube was measured with the aid of a Beacon 2000 Fluorescence Polarization Instrument (Pan Vera Corporation) with 360 nm excitation filters and 530 nm emission filters. The anisotropy values were converted to percent inhibition using the following equation: I% ¼ ðA0 AÞ=ðA0 A100 Þ 100; where I%; A0 ; A100 and A are the percent inhibition, A at 0% inhibition, A at 100% inhibition, and observed A value, respectively. The percent inhibition versus
competitor concentration curves were analyzed by nonlinear least-squares curve fitting and yielded an IC50 value (the concentration of competitor needed to displace half of the bound ligand). The IC50 was defined as the ER-binding affinity of an objective compound. 2.3. Molecular descriptors Four descriptors, consisting of the energies of the highest occupied molecular orbital ðeHOMO Þ; the lowest unoccupied molecular orbital ðeLUMO Þ; the volume, and the logarithm of octanol–water partition coefficient (log Pow) were used for the QSAR evaluation between ER and compounds. Two stereoelectronic parameters, eHOMO and eLUMO ; and one steric descriptor, volume at ( 3, were calculated with the semiempirical meth0.01 e/A od MOPAC [24] using version 3.9 (CAChe Scientific, Oxford) of CAChe’s Macintosh Project Leader interface
J.-Y. Hu, T. Aizawa / Water Research 37 (2003) 1213–1222
1216
Table 2 Structures of test group (seven chemicals) OH
Genistein OH O
OH
Zearalenone
O O
A
HO
HO
O
Biochanin AOH O HO
O
Daidzein O
OCH3
HO
O
O
Phenolphthalein OH
Phenolphthalin OH
OH
4-methyl-2,6-di-tert-butyl-phenol OH (CH3)3C
B
OH
OH O O
C(CH3)3
CH3
O
OH A:phytoestrogen; B: anthropogenic phenolics
(running on a Power Macintosh 7500/100 computer). The PM3 [25] parameterizations of atomic wave functions were used for gas-phase calculations. The log Pows of chemicals were calculated by using ACD/ log Pow Ver.1.0 (Advanced Chemistry Development Inc.).
3. Results and discussion 3.1. Binding affinity To evaluate the affinity binding to ER, the 18 compounds in Table 1 were selected as the training group, and the binding affinities were determined. Fig. 1 shows the competition binding curves of 18 compounds against a human recombinant estrogen receptor a(hrERa)/fluorescent ligand complex. The resulting classic competitive binding curves were analyzed using PRISM software (Ver.2.0; GraphPad Software Inc., USA) in order to assess the potency of the competitor molecule. The concentration that inhibited 50% (i.e. the IC50 value) of each test compound was calculated by nonlinear least-squares regression and is shown in Tables 3 and 4. Except for phenol, which shows very weak hrER-a interaction, the other industrial chemicals including bisphenol A, 4-n-nonylphenol, and 4-tertoctylphenol, which have been shown to elicit estrogenic effects in vitro and in vivo [26–28], show significant hrER-a interaction in this binding assay. The binding affinity of 4,4-diphenyldiol was found to be higher than that of bisphenol A. In order to compare the interaction behavior, the binding affinities of three steroids and one
non-natural steroid, DES in Table 1(D), were also determined. The DES was found to have the strongest affinity. In addition, it was found that the IC50 of 4nonylphenol, a mixture of many kinds of isomers with branched alkyls, was 4.96 103 nM, which is similar to that of 4-n-nonylphenol. It has previously been reported that while 4-nonylphenol elicited the estrogen response, no estrogen response for 4-n-nonylphenol was detected based on the results of a reporter gene expression assay using a yeast two-hybrid method [14]. From the above result, it was supposed that 4-n-nonylphenol might block the effect of naturally occurring estrogens. 3.2. The mechanistic QSAR approach: predictor selection According to the agonist and antagonist binding modes of 17b-estradiol in the ER [10], the 17b-estradiol binds diagonally across the cavity which has a probe ( 3. Hormone recognition is accessible volume of 450 A achieved through a combination of specific hydrogen bonds and the complementarity of the hydrophobic binding cavity to 17b-estradiol’s non-polar character. Based on such a three-dimensional (3-D) structure of the ER, we have attempted to establish a mechanistic QSAR model involving a phenolic chemical and the ER. Eq. (1) shows the general form of this model. log IC50 ¼ bulk þ polarity þ hydrogen bonding:
ð1Þ
The binding affinity of a phenolic compound is described as a linear expression with terms for a bulk effect, a polarity effect, and hydrogen-bonding effects. For the affinity binding to ER, the specific terms and
J.-Y. Hu, T. Aizawa / Water Research 37 (2003) 1213–1222 -20
-10 30 50 2-hydroxybiphenyl 3-hydroxybiphenyl
70 90
4,4'-diphenyldiol 4-hydroxybiphenyl
110
10 4
20 40 60 80 100
130 10 3
0
Inhibition(%)
Inhibition(%)
10
105
10 6
120 2 10
10 7
2-t-butylphenol 2-sec-butylphenol 4-t-butylphenol 3-t-butylphenol 4-t-butylphenol
10 3
Concentration(nM)
Inhibition(%)
40
100
bis-phenolA 4-ethylphenol phenol 4-t-octyl-phenol 4-n-nonyl-phenol
120 1 10
10 2
80
10 5
10 6
107
10 8
10 3
10 4
0
20
60
10 4
Concentration(nM)
0
Inhibition(%)
1217
10 3
10 4
20 40 60 80 100
105
106
10 7
10 8
120 -2 10
Concentration (nM)
DES estradiol estrone estriol
10 -1
10 0
10 1
10 2
Concentration(nM)
Fig. 1. Competition binding curves of various compounds against a human recombinant estrogen receptor a/fluorescent ligand complex (hrER-a-FES-1). Increasing concentrations of competitor were incubated with 2 nM hrER-a-3 nM FES1 for 60 min at room temperature followed by the measurement of fluorescence polarization.
Table 3 Physicochemical properties of training group (18 chemicals) and their binding affinities (IC50) with the estrogen receptor
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18
Chemicals
log Pow
eHOMO (eV)
eLUMO (eV)
( 3) Vm (A
IC50 (nM)
4-n-nonylphenol 4-tert-octyl-phenol 4,40 -biphenyldiol Bisphenol A 4-hydroxybiphenol 4-sec-butylphenol 3-hydroxybiphenol 2-hydroxybiphenol 4-tert- butylphenol 2-sec-butylphenol 4-ethyl-phenol 3-tert- butylphenol 2-tert-butylphenol Phenol Diethylstibestrol 17b-estradiol Estriol Estrone
6.19 5.66 2.42 3.43 3.20 3.35 3.23 2.94 3.17 3.35 2.47 3.17 3.17 1.48 7.03 4.13 2.94 3.69
8.942 8.990 8.441 8.972 8.640 8.980 8.911 8.808 8.997 9.047 8.926 9.093 9.094 9.176 8.653 8.900 8.931 9.005
0.320 0.349 0.31 0.272 0.336 0.333 0.408 0.46 0.347 0.293 0.334 0.314 0.314 0.291 0.027 0.339 0.308 0.237
167.90 155.35 112.88 149.19 108.89 106.93 108.96 108.91 106.66 106.94 82.23 106.83 106.80 58.97 178.01 186.53 190.24 181.52
9.51 103 1.41 104 1.06 105 1.10 105 4.27 105 5.82 105 6.20 105 9.97 105 1.01 106 1.38 106 1.70 106 2.29 106 2.36 106 9.50 107 6 2.30 101 5.00 101 3.31 102
corresponding descriptors are as follows: bulk, volume (Vm ) in cubic angstroms; polarity, logarithm of octanol– water partition coefficient (log Pow); hydrogen bonding, eHOMO or eLUMO : The hydrogen-bonding term is separated into acidic and basic components. Since any hydrogen bond can be considered to have covalent and electrostatic parts, it is necessary to describe the two
bonding contributions using individual descriptor. The covalent contribution to the basic term can associate with the eHOMO which has the electron to form a complex with a proton on a neighboring molecule [29]. In this study, a higher value of eHOMO for a phenolic chemical indicates a greater tendency to form a hydrogen bond with the ER. The electrostatic
J.-Y. Hu, T. Aizawa / Water Research 37 (2003) 1213–1222
1218
Table 4 Physicochemical properties of test group (seven chemicals) and the comparison between observed and calculated binding affinities (IC50) with the estrogen receptor eHOMO
( 3) Vm ( A
( 3) Vm 0 (A
IC50 (nM) Observed
Predicted
Predicted0
Daidzein Genistein Zearalenone Biochanin-A
8.616 8.553 9.375 8.518
139.16 143.13 196.12 156.15
— — — —n
4.27 103 9.12 102 7.08 102 >1.74 107
7.59 103 3.80 103 3.55 102 5.25 102
— — — 8.51 109
Phenolphthalin Phenolphthalein 4-methyl-2,6-di-butyl-phenol
9.138 9.157 8.699
183.79 178.68 164.75
88.8 144.5 106.8
>5.89 107 8.5 105 4.90 105
6.61 103 1.02 104 5.25 103
7.18 106 1.26 105 4.37 105
Vm 0 : the effective bulk which was calculated based on the Eq. (8). Predicted: the binding affinity which was predicted based on the Eq. (4) for anthropogenic phenolics and Eq. (7) for phytoestrogen. Predicted0 : the binding affinity which was predicted based on the Eq. (4) for anthropogenic phenolics and Eq. (7) for phytoestrogen where Vm is replaced by Vm 0 . n It is estimated to be zero.
contribution to the basicity is taken as the magnitude of the most negative formal charge in the molecule. On the other hand, the covalent contribution to the acidic term can be taken as eLUMO ; with a lower value of eLUMO for a phenolic compound indicating a greater tendency to form a hydrogen bond with the ER. In this case, the hydrogen-bonding acidity is applied with this orbital being available to accept electrons from the HOMO orbital of ER. The electrostatic contribution to the acidity is taken as the most positive formal charge, on a hydrogen atom in the molecule. Because the positive formal charge on a hydrogen atom and the negative formal charge on an oxygen atom for the objective phenolic chemicals are similar, the electrostatic contribution to hydrogen binding can be neglected.
where n stands for the number of samples, r2 the variance and s the standard error of the estimate. Fig. 2(a) compares the observed and calculated estrogen receptor binding affinities. It is clear that log Pow was not a proper parameter to predict the ER-binding affinities of 4,40 -biphenydiol and phenol. While the much better correlation between ER-binding affinities and Vm was found, the best regression was acquired by correlating IC50 with Vm and eHOMO simultaneously. As shown in Fig. 2(a) and Eq. (4), the calculated IC50 matched the observed IC50 values well for all of the objective compounds when Vm was combined with eHOMO : log IC50 ¼ 1:46eHOMO 0:032Vm 3:644 r2 ¼ 0:948 n ¼ 14 s ¼ 0:248:
ð4Þ
3.3. QSAR for ER-binding affinity Based on the assumption that the electrostatic contribution to hydrogen-binding can be neglected, correlations were conducted between ER-binding affinity of the phenolic compounds and their volumetric parameter (Vm ), log Pow and eHOMO or eLUMO : It was found that significant correlations could be obtained upon using Vm and log Pow as individual predictors. Eq. (2) expresses the regression between log IC50 and logP ow, and Eq. (3) is the regression between log IC50 and Vm : log IC50 ¼ 0:698 log Pow þ 8:11 r2 ¼ 0:70
n ¼ 14 s ¼ 0:57;
ð2Þ
log IC50 ¼ 0:033lVm þ 9:55 r2 ¼ 0:87
n ¼ 14 s ¼ 0:37;
ð3Þ
It should be noted that it is the HOMO orbital energy, but not LUMO orbital energy, that appears in the significant regression, and the covalent contribution to the basic term of that in Eq. (1) was confirmed. That is, the phenolic compound donates the electron and forms a complex with a proton on the ER. To investigate whether the phenolic chemicals under investigation bind to the ER through a similar molecular mechanism with natural steroids, regressions for the combined set of 14 industrial compounds, three natural and one non-natural steroid (DES) in Table 1 were also conducted. A significant regression could not be obtained upon using only the hydrophobic parameter, log Pow (Eq. (5)). Two significant regressions, Eq. (6) using one parameter of Vm and Eq. (7) using two predictors, Vm ; and eHOMO ; were obtained; but the hydrophobic parameter, log Pow, and eLUMO are also excluded. This result can be interpreted such that the
J.-Y. Hu, T. Aizawa / Water Research 37 (2003) 1213–1222 9.0 Equation(2) Equation(3) Equation(4)
Observed logIC50
8.0
phenol
7.0 6.0 5.0
4,4'-biphenydiol
4.0 3.0 3.0
4.0
5.0
6.0
7.0
8.0
9.0
Calculated logIC50(nM)
(a) 9.0 Equation(5) Equation(6) Equation(7)
Observed logIC50
8.0 7.0 6.0 5.0 4.0 3.0
estrone
2.0
estriol
1.0
17β-estradiol
0.0 0.0
(b)
2.0
4.0
6.0
8.0
Calculated logIC50(nM)
Fig. 2. Plot of observed versus calculated estrogen receptor binding affinity. (a) Fourteen industrial phenolic chemicals; (b) 14 industrial phenolic chemicals, three steroids and DES.
phenolic chemicals bind to ER with a similar mechanism to that for steroids. log IC50 ¼ 0:90 log Pow þ 8:03 r2 ¼ 0:40
n ¼ 18 s ¼ 1:56;
ð5Þ
log IC50 ¼ 0:049Vm 11:10 r2 ¼ 0:89
n ¼ 18 s ¼ 0:69;
ð6Þ
log IC50 ¼ 1:90eHOMO 0:047Vm 6:05 r2 ¼ 0:92
n ¼ 18 s ¼ 0:62:
ð7Þ
3.4. Robust of the QSAR model The test compounds shown in Table 2 were used to check the robustness of the above QSAR models. Of seven compounds, four (genistein, biochanin A, zearalenone and daidzein), are phytoestrogens with structures which are planar and similar to those of steroids; the two pH indicators, phenolphthalin and phenolphthalein, bear some structural resemblance to bisphenol A; and one compound is alkylphenolic with
1219
three alkyl group on the phenolic ring. The determined binding affinities and their molecular predictors are shown in Table 4. The QSAR model (Eq. (7)) predicted the ER-binding affinity with high accuracy for the three phytoestrogens, genistein, zearalenone, and daidzein, suggesting that the binding behavior of the phytoestrogens is similar to that of the objective compounds. From the fact that the QSAR was unsuccessful in predicting the binding affinity of biochanin A, it is reasonable to suppose that besides the eHOMO and Vm ; the QSAR model (Eq. (7)) could be improved by introducing another molecular predictor or by amending the present predictors. Considering the significance of eHOMO in predicting the binding affinity, the electron-density (or HOMO density) distribution of biochanin A was compared with the other three phytoestrogens. Due to the similarity of genistein, zearalenone and daidzein, only genistein was selected, and its electron-density distribution was compared with that of biochanin A (Fig. 3). From the difference in electron-density distribution, it was found that the electron densities on the HOMO orbitals of both biochanin A and genistein are all distributed in the A benzene ring. It should be noted that there is no hydroxyl on the A ring, and no electron density was distributed on the hydroxyl of the B ring in the case of biochanin A, which possibly leads to its lower ERbinding affinity due to the importance of hydroxyl in binding to ER. It is supposed that only part of the molecular bulk of biochanin A contributed to the ERbinding affinity. The distribution range of electron density on the phenolic moiety (B ring), which contributes to the ER-binding affinity, was defined here as the effective bulk (Vm 0 ). In the case of biochanin A, the Vm 0 is near 0 because no electron density was distributed on phenolic moiety (B ring) as shown in Fig. 3. Thus, the predicted binding affinity of biochanin A was calculated to be near 8.51 109 nM (Predicted0 in Table 4) based on Eq. (7). This value is comparatively near its observed detected binding affinity (>1.74 107 nM). In addition, Fig. 4 compares the electron-density distributions on the HOMO of phenolphthalein and phenolphthalin with bisphenol A due to their structural resemblance. Comparing their 3-D structures and electron-density distribution on the HOMO, it was found that while that electron-density distribution ranges between hydroxyl on the A ring to the hydroxyl on the B ring in the case of phenolphthalein, similar to that of bisphenol A, that for phenolphthalin localized on only one ring (A in Fig. 4), a result which significantly differs from that of bisphenol A. The finding also suggested that the electron-density distribution on the HOMO is quite important and is possibly another reasonable predictor in the QSAR model in addition to Vm and eHOMO : To amend the effect of electron-density
1220
J.-Y. Hu, T. Aizawa / Water Research 37 (2003) 1213–1222
Fig. 3. Comparison between electron-density distribution in HOMOs of genistein and biochanin A. Black and dark gray colors show ( the HOMOs. The ranges of electron-density distribution on the HOMOs of genistein and biochanin A are 12.04 and 12.84 A, respectively.
Fig. 4. Comparison among the electron-distribution in HOMOs of phenolphthalein, phenolphthalein, and bisphenol A. Black and dark gray colors show the HOMOs. The ranges of electron-density distribution on the HOMOs of phenolphthalein, phenolphthalin, ( (L in Eq. (8)), 5.66 A ( (L in Eq. (8)) and 9.51 A ( (L0 in Eq. (8)), respectively. and bisphenol A are 9.21 A
distribution on the binding affinity, the following attempt was made. By using bisphenol A as a template, the Vm in Eq. (4) was amended to effective bulk, Vm 0 . Thus, the Vm 0 of phenolphthalin and phenolphthalein ( respectively, are 88.8 and 144.5 A3, Vm0 ¼ L Vm0 =L0 ;
ð8Þ
where Vm0 is the Vm of bisphenol A; L is the range of electron-density distribution on the HOMO, and L0 is that of bisphenol A as shown in Fig. 4. As a result, the predicted binding affinities to ER for phenolphthalin and phenolphthalein are 7.18 106 and 1.26 105 nM, respectively, which match with the observed values (Table 4).
J.-Y. Hu, T. Aizawa / Water Research 37 (2003) 1213–1222
Finally, an attempt to predict the binding affinity of an alkylphenolic compound, 4-methyl-2,6-di-butylphenol was made. When using the absolute value for the Vm of 4-methyl-2,6-di-butyl-phenol in Eq. (4), the predicted IC50 is 5.25 103 nM, which is unfortunately much lower than the observed value of 4.90 105 nM. When the Vm of 2-tert-butylphenol, which was used as the template, was employed instead of that of 4-methyl2,6-di-butyl-phenol, the predicted IC50 was 4.37 105 nM, which is comparable with the observed value as shown in Table 4. This finding suggested that the Vm in Eq. (4) may need to be amended to a volume commensurate with its contribution for estimating the binding affinity of an alkylphenolic compound substituted by more than one alkyl, although more experimental data are required to confirm the above inferential conclusion. In summary, the QSAR can effectively predict the binding affinities of phenolic compounds whose structures were similar to those used in the training group. For the phenolic compounds whose structures were dissimilar to those used in the training group, the values of the steric parameter, Vm ; in this QSAR should be amended according to their status of electron-density distribution on the HOMO or of alkyl-substitution, where a suitable compound in the training group is used as a template.
4. Conclusion We carried out a fundamental study to develop a simplified method for estimating ER-binding affinities of phenolic compounds. The results of correlation analysis indicated that ER-binding affinity can essentially be estimated from the energy of the highest occupied molecular orbital ðeHOMO Þ and steric factor, a molecular volume. This was, especially true in the case where the correlation model was established by amending a molecular volume based on electron-density distributions on the HOMO according to their structural features.
Acknowledgements Financial support from both the Japanese Governmental Research Fund (Ministry of Health Welfare) and the National Natural Science Foundation of China [49925103 and 40024101] is gratefully acknowledged.
References [1] Zacharewski T. In vitro bioassays for assessing estrogenic substances. Environ Sci Technol 1997;31:613–23.
1221
[2] White R, Jobling S, Hoare SA, Sumpter JP, Parker MG. Environmentally persistent alkylphenolic compounds are estrogenic. Endocrinology 1994;135:175–82. [3] Mueller GC, Kim Uh-hee. Displacement of estradiol from estrogen receptors by simple alkyl phenols. Endocrinology 1978;102:1429–35. [4] Soto AM, Lin Tien-Min, Justicia H, Silvia RM, Sonnenschein C. An ‘‘in culture’’ bioassay to assess the estrogenicity of xenobiotics. In: Colborn T, Clement C, editors. Chemically induced alterations in sexual development: the wildlife/human connection. Princeton, NJ: Princeton Scientific Publishing, 1992. pp. 295–309. [5] Berthois Y, Katzenellenbogen JA, Katzenellenbogen BS. Phenol red in tissue culture media is a weak estrogen: implications concerning the study of estrogen-responsive cells in culture. Cell Biol 1986;83:2496–500. [6] Korach KS, Sarver P, Chae Kun, McLachlan JA, McKinney JD. Estrogen receptor-binding activity of polychlorinated hydroxybiphenyls: conformationally restricted structural probes. Mol Pharmacol 1988;33: 120–6. [7] Gould JC, Leonard LS, Maness SC, Wagner BL, Conner K, Zacharewski T, Safe S, McDonnell DP, Gaido KW. Bisphenol A interacts with the estrogen receptor a in a distinct manner from estradiol. Mol Cell Endocrinol 142;1998:203–14. [8] Thiele B, Gunther K, Schwuger MJ. Alkylphenol ethoxylates, trace analysis and environmental behavior. Chem Rev 1997;97:3247–72. [9] Anstead GM, Carlson KE, Katzenellenbogen JA. The estradiol pharmacophore: ligand structure-estrogen receptor binding affinity relationships and a model for the receptor binding site. Steroids 1997;62:268–303. [10] Brzozowski AM, Pike ACW, Dauter Z, Hubbard RE, Bonn T, Engstrom O, Ohman L, Greene GL, Gustafsson J, Carlquist M. Molecular basis of agonism and antagonism in the oestrogen receptor. Nature 1997;389:753–8. [11] Routledge EJ, Sumpter JP. Estrogenic activity of surfactants and some of their degradation products assessed using a recombinant yeast screen. Environ Toxicol Chem 1996;15:241–8. [12] Routledge EJ, Sumpter JP. Structural Features of alkylphenolic chemicals associated with estrogenic activity. J Biol Chem 1997;272(6):3280–8. [13] Blair RM, Fang H, Branham WS, Hass BS, Dial SL, Moland CL, Tong W, Shi L, Perkins R, Sheehan DM. The estrogen receptor relative binding affinities of 188 natural and xenochemicals: structural diversity of ligands. Toxicol Sci 2000;54:138–53. [14] Nishihara T, Nishikawa J, Kanayama T, Dakeyama F, Saito K, Imagawa M, Takatori S, Kitagawa Y, Hori S, Utsumi H. Estrogenic activities of 517 chemicals by yeast two-hybrid assay. J Health Sci 2000;46(4):282–98. [15] Taira Y, Nakai M, Asai D, Yakabe Y, Tahara Y, Shinmyozu T, Noguchi M, Takatsuki M, Shimohigashi Y. Structural requirements of para-alkylphenols to bind to estrogen receptor. Eur J Biochem 1999;262:240–5. [16] Van der Kuur JA, Wiese T, Brooks SC. Influence of estrogen structure on nuclear binding and progesterone receptor induction by the receptor complex. Biochemistry 1993;32:7002–8.
1222
J.-Y. Hu, T. Aizawa / Water Research 37 (2003) 1213–1222
[17] Waller CL, Oprea TI, Chae K, Park HK, Korach KS, Laws SC, Wiese TE, Kelce WR, Gray Jr LE. Ligand-based identification of environmental estrogens. Chem Res Toxicol 1996;9:1240–8. [18] Waller CL, Minor DL, McKinney JD. Using three dimensional quantitative structure-activity relationships to examine estrogen receptor binding affinities of polychlorinated hydroxybiphenyls. Environ Health Perspect 1995;103:702–7. [19] Hendry LB, Chu CK, Copland JA, Mahesh VB. Antiestrogenic piperidinediones designed prospectively using computer graphics and energy calculations of DNA-ligand complexes. J Steroid Biochem Molec Biol 1994;48: 495–505. [20] Bradbury SP, Mekenyan OG, Ankley GT. Quantitative structure-activity relationships for polychlorinated hydroxybipheny estrogen receptor binding affinity: an assessment of conformer flexibility. Environ Toxicol Chem 1996;15(11):1945–54. [21] Schmieder PK, Aptula AO, Routledge EJ, Sumpter JP, Mekenyan OG. Estrogenicity of alkylphenolic compounds: a 3-D structure-activity evaluation of gene activation. Environ Toxicol Chem 2000;19(7):1727–40. [22] Kramer VJ, Giesy JP. Specific binding of hydroxylated polychlorinated biphenyl metabolites and other substances to bovine calf uterine estrogen receptor: structure-binding relationships. Sci Total Environ 1999;233:141–61.
[23] Bolger R, Wiese TE, Ervin K, Nestich S, Checovich W. Rapid screening of environmental chemicals for estrogen receptor binding capacity. Environ Health Perpect 1998;106(9):551–7. [24] Stewart JJP. Optimization of parameters for semiempirical methods, II. Applications. J Comput Chem 1989;10: 221–64. [25] Stewart JJP. MOPAC manual, Frank J. Seiler Research Laboratory, United States Air Force Academy, Colorado Springs CO 80840, 1990. [26] Jobing S, Sheahan D, Osborne JA, Matthiessen P, Sumpter JP. Inhibition of testicular growth in rainbow trout (Oncorhynchus mykiss) exposed to estrogenic alkylphenolic chemicals. Environ Toxicol Chem 1996;15(2):194–202. [27] Sharpe RM, Fisher JS, Millar MM, Jobling S, Sumpter JP. Gestational and lactational exposure of rats to xenoestrogens results in reduced testicular size and sperm production. Environ Health Perspect 1995;103(12):1136–43. [28] Strunck E, Stemmann N, Hopert AC, Wunsche W, Frank K, Vollmer G. Relative binding affinity does not predict biological response to xenoestrogens in rat endometrial adenocarcinoma cells. J Steroid Biochem Molec Biol 2000;74:73–81. [29] Wilson LY, Famini GR. Using theoretical descriptors in quantitative structure-activity relationships: some toxicological indices. J Med Chem 1991;34:1668–74.