Endocrinology, doi:10.1210/en.2006-0113
Endocrinology Vol. 147, No. 9 4132-4150
Copyright © 2006 by The Endocrine Society
Quantitative Structure-Activity Relationship of Various Endogenous Estrogen Metabolites for Human Estrogen Receptor 
and ß Subtypes: Insights into the Structural Determinants Favoring a Differential Subtype Binding
Bao Ting Zhu,
Gui-Zhen Han,
Joong-Youn Shim,
Yujing Wen and
Xiang-Rong Jiang
Department of Basic Pharmaceutical Sciences, College of Pharmacy (B.T.Z., G.-Z.H., Y.W., X.-R.J.), University of South Carolina, Columbia, South Carolina 29208; and Julius L. Chambers Biomedical/Biotechnology Research Institute (J.-Y.S.), North Carolina Central University, Durham, North Carolina 27707
Address all correspondence and requests for reprints to: Dr. Bao Ting Zhu, University of South Carolina, Basic Pharmaceutical Sciences, College of Pharmacy, 700 Sumter Street, Columbia, South Carolina 29209. E-mail: btzhu{at}cop.sc.edu.
 |
Abstract
|
|---|
To search for endogenous estrogens that may have preferential binding affinity for human estrogen receptor (ER) 
or ß subtype and also to gain insights into the structural determinants favoring differential subtype binding, we studied the binding affinities of 74 natural or synthetic estrogens, including more than 50 steroidal analogs of estradiol-17ß (E2) and estrone (E1) for human ER and ERß. Many of the endogenous estrogen metabolites retained varying degrees of similar binding affinity for ER and ERß, but some of them retained differential binding affinity for the two subtypes. For instance, several of the D-ring metabolites, such as 16-hydroxyestradiol (estriol), 16ß-hydroxyestradiol-17, and 16-ketoestrone, had distinct preferential binding affinity for human ERß over ER (difference up to 18-fold). Notably, although E2 has nearly the highest and equal binding affinity for ER and ERß, E1 and 2-hydroxyestrone (two quantitatively predominant endogenous estrogens in nonpregnant woman) have preferential binding affinity for ER over ERß, whereas 16-hydroxyestradiol (estriol) and other D-ring metabolites (quantitatively predominant endogenous estrogens formed during pregnancy) have preferential binding affinity for ERß over ER. Hence, facile metabolic conversion of parent hormone E2 to various metabolites under different physiological conditions may serve unique functions by providing differential activation of the ER or ERß signaling system. Lastly, our computational three-dimensional quantitative structure-activity relationship/comparative molecular field analysis of 47 steroidal estrogen analogs for human ER and ERß yielded useful information on the structural features that determine the preferential activation of the ER and ERß subtypes, which may aid in the rational design of selective ligands for each human ER subtype.
|
Introduction
|
|---|
ESTROGENS ARE REALLY fascinating female gonadal hormones with very diverse physiological and pathophysiological functions in different tissues and cell types (1, 2, 3). Ever since the discovery of estradiol-17ß (E2) by Allen and Doisy in the 1920s to 1930s, there has always been a great deal of intense interest in advancing our understanding of the mechanism(s) of action of endogenous estrogens and also in exploring their therapeutic usefulness. Now, it has become common knowledge that many of the well-known hormonal actions of estrogens are mediated by specific estrogen receptors (ERs). The first high-affinity ER, now commonly referred to as ER, was cloned in 1986 from MCF-7 human breast cancer cells, which abundantly expressed this ER subtype (4, 5). For nearly a decade after its cloning, it was believed that the estrogens signal through a single ER. However, a second ER (subtype ß) was later identified in 1996 while studying the roles of estrogens in the prostate, gonads, and the immune system (6, 7). The existence of two distinct ER subtypes indicated that the signaling pathways for endogenous estrogens are significantly more complex than previously thought.
The human ER is a 66-kDa hormone-inducible transcription factor that can act positively or negatively in regulating the expression of genes involved in tissue growth and differentiation (1, 2). Similarly, the human ERß is a 53-kDa hormone-inducible transcription factor that shares high degrees of sequence homology with the human ER, especially in the DNA binding domain (2, 6, 7, 8, 9). Studies have shown that there are a number of functional similarities between human ER and ERß, and both receptor subtypes can bind E2 with similarly high affinities (10, 11). The activated ER and ERß (i.e. receptor bound with an agonist such as E2) can form homodimers (ER-ER or ERß-ERß) or heterodimers (ER-ERß), and these dimerized ERs can bind to various estrogen response elements in highly similar fashions (2, 11, 12, 13). However, there are also significant differences noted for human ER and ERß. It was found that the tissue distribution pattern of these two ER subtypes is quite different (9, 14, 15, 16, 17, 18). In addition, an earlier study has shown that 16ß-hydroxyestradiol-17 (16ß-OH-E2-17; commonly known as 16,17-epiestriol), an endogenous estrogen metabolite, has a preferential binding affinity for human ERß over ER (10). Hence, the possibility exists that some of the endogenously formed estrogen metabolites/derivatives may have differential binding affinity for human ER or ERß, likely contributing to the differential activation of each signaling system in different target sites and/or under different physiological or pathophysiological conditions.
In recent years, we have made considerable effort to systematically characterize the complete profiles of the NADPH-dependent oxidative metabolites of E2 and estrone (E1) that are formed by human liver, nonhepatic tissues, as well as various recombinant human cytochrome P450 isoforms in vitro (for review, see Refs. 19 and 20). A large number of endogenous estrogen metabolites have been identified. As part of our continuing effort to characterize the biological functions that may be associated with some of these estrogen metabolites, we have compared in this study a large number of endogenous E2 metabolites, along with some of their synthetic analogs and phytoestrogens (structures are shown in Fig. 1
and chemical names are listed in Table 1), for their binding affinities for human ER and ERß. The recombinant human ERs used in the present study were produced in a baculovirus expression system that yielded soluble, functionally active recombinant ER proteins with posttranslational modification patterns (mainly phosphorylations and acetylations) similar to those found in mammalian cells (13).
View this table:
[in this window]
[in a new window]
|
TABLE 1. The IC50 and RBA values of various hydroxylated, keto, and dehydrogenated metabolites of E2 and E1 as well as some other natural or synthetic derivatives for the recombinant human ER and ERß
|
|
Here, it should be noted that an earlier study by Fang et al. (21) has probed the structure-activity relationships of the binding affinity of a large number of natural and xenoestrogens for rat uterine cytosolic ERs (mostly ER). The study provided valuable information on the structural requirements that govern ligand binding to the rat uterine ER. Because this earlier study only tested the binding affinity for rat uterine ER and relatively limited members of endogenous estrogen derivatives were included, we thus have placed considerable emphasis in the present study to compare many more endogenously formed estrogen derivatives for their differential binding affinity for human ER and ERß subtypes. We believe these studies will provide valuable information as to whether any of the known endogenous estrogen metabolites is/are preferential agonist(s) for human ER or ERß. In addition, we have also performed detailed computational modeling analyses of the three-dimensional (3D) quantitative structure-activity relationships (QSARs) to probe the precise structural requirements of various estrogen derivatives for the preferential binding of the human ER and ERß. Data from these modeling studies will enable us to gain insights into the structural determinants required for the preferential activation of these two human ER subtypes.
|
Materials and Methods
|
|---|
Chemicals and reagents
E2, E1, and most of their metabolites and derivatives listed in Table 1
were obtained from Steraloids (Newport, RI). 7-(6-Hydroxyhexanyl)-17ß-estradiol [E2-7-(CH2)6OH] and 7-(6-benzyloxyhexanyl)-17ß-estradiol [E2-7-(CH2)6OC6H5] were chemically synthesized in our laboratory (22). For convenience, the structures of various estrogens used in this study are shown in Fig. 1. Dithiothreitol, glycerol, and Tris-HCl were obtained from the Sigma Chemical Co. (St. Louis, MO). Hydroxylapatite and BSA were obtained from Calbiochem (through EMD Biosciences, Inc. San Diego, CA). [2,4,6,7,16,17-3H]E2 (specific activity of 115 Ci/mmol) was obtained from NEN Life Science Products Life Sciences (Boston, MA), and it was purified in our laboratory using a HPLC-based method (20) before its use in the in vitro receptor binding assays.
The recombinant human ER and ERß proteins and BSA were obtained from PanVera Corporation (Madison, WI). According to the supplier, the recombinant human ER and ERß were produced in a baculovirus-mediated expression system, and they were soluble and functionally active, with posttranslational modifications similar to those found in mammalian cells (13).
ER and ERß binding assays
The following buffer solutions were used in the ER binding assays, and they were prepared beforehand and stored at 4 C. The binding buffer consisted of 10% glycerol, 2 mM dithiothreitol, 1 mg/ml BSA, and 10 mM Tris-HCl (pH 7.5). The ER washing buffer contained 40 mM Tris-HCl and 100 mM KCl (pH 7.4), but the ERß washing buffer contained only 40 mM Tris-HCl (adjusted to pH 7.4). The 50% hydroxylapatite slurry was prepared first by vigorously mixing 10 g hydroxylapatite with 60 ml of 50 mM Tris-HCl solution (pH 7.4). Hydroxylapatite was then allowed to settle for 20 min at room temperature, and the supernatant was decanted. The above procedures were repeated 10 times; afterward, hydroxylapatite was kept in the 50 mM Tris-HCl solution overnight at 4 C. Hydroxylapatite slurry was then adjusted to an approximate final concentration of 50% (vol/vol) using the same Tris-HCl solution and stored at 4 C, and the slurry was stable for up to several months.
On the day of performing the ER binding assay, [3H]E2 solution was freshly diluted in the binding buffer, and an aliquot (45 µl) of the [3H]E2 solution was added to a 1.5-ml microcentrifuge tube, giving a final [3H]E2 concentration of 10 nM. Each of the competing ligands (in 50-µl volume) was then added to the mixture for the intended final concentrations at 0, 0.24, 0.98, 3.9, 15.6, 62.5, 250, and 1000 nM. Note that all of the estrogens were initially dissolved in pure ethanol to a stock concentration of 1 mM and then further diluted to 100 µM with 20% aqueous ethanol. In this way, the final ethanol concentration in the incubation mixture was less than or equal to 0.2%. Immediately before the addition of the ER or ERß protein, it was diluted in the binding buffer and mixed gently with repetitive pipettings. An aliquot (5 µl) of the diluted ER or ERß solution was precisely added to the mixture containing 45 µl of the [3H]E2 and 50 µl of the competing ligand, giving a final receptor concentration of 1–2 fmol/ml. The incubation mixture was then mixed gently and thoroughly with repetitive pipettings. Nonspecific binding (NSB) by the [3H]E2 was determined in separate tubes by inclusion of a 400-fold concentration of the nonradioactive E2 (at a final concentration of 4 µM). Based on UV spectrometric monitoring of E2 in water, this estrogen at 4 µM concentration (the highest steroid concentration used) appeared to be readily soluble in the aqueous solution. The binding mixture was incubated at room temperature for 2 h. At the end of the incubation, 100 µl hydroxylapatite slurry was added to each tube, and the tubes were incubated on ice for 15 min with three times of brief vortexing. An aliquot (1 ml) of the appropriate washing buffer was added, mixed, and centrifuged at 10,000 x g for 5 min, and the supernatants were discarded. This wash step was repeated twice. Hydroxylapatite pellets were then resuspended in 200 µl ethanol (followed by another rinse with 200 µl ethanol), and then the content was transferred to scintillation vials (containing 4 ml scintillation fluid) for measurement of 3H-radioactivity with a liquid scintillation counter (Packard Tri-CARB 2900 TR, PerkinElmer Life and Analytical Sciences, Boston, MA).
To calculate the specific binding (picomoles per milliliter) of the human ER or ERß protein at each radioligand concentration, the following equation was used:
The IC50 value for each competing estrogen was calculated according to the sigmoidal inhibition curve, and the relative binding affinity (RBA) was calculated against E2 using the following equation:
It should be noted that the absolute IC50 values are affected by the concentrations of the radioligand ([3H]E2) used. When a lower radioligand concentration is used, the corresponding IC50 value will also be relatively lower, but when the radioligand concentration increases, the corresponding IC50 value will also increase. Because the radioligand [3H]E2 concentration used in this study was 10 nM (which was 10–100 times higher than the previously reported Kd values for human ER), the absolute IC50 values would also be higher if they were compared with values reported in some of the earlier studies where lower concentrations of the radioligand were used. The reason that we chose to use a higher concentration of [3H]E2 was simply because it would yield more reproducible readings of the radioactivity counts. Because the RBA value is a parameter that is independent of the radioligand concentration used, we thus have placed more emphasis on the RBA values instead of the absolute IC50 values in interpreting the physiological meanings of the data from in vitro receptor competition assays.
Computational modeling analyses
All calculations described in the present study were carried out using the SYBYL molecular modeling program (version 6.8 or 7.1; Tripos, Inc., St. Louis, MO) implemented on a Silicon Graphics Octane or Origin 350 workstation (Silicon Graphics Inc., Mountain View, CA).
Molecular models and structural alignment.
A total of 48 compounds, for which the RBA values could be precisely determined for ER and ERß (listed in Table 1, footnote a), were used in our QSAR analyses. They were mainly comprised of A, B, and D-ring metabolites as well as some of the dehydroestrogen metabolites of E1 and E2. For structural building and alignment, we chose to use E2 as template because this endogenous parent hormone had nearly the highest binding affinity for both ER and ERß. The structure of E2 was obtained from its x-ray structure in complex with human ER (23), and all other molecules were similarly constructed based on the structure of E2. The geometry of each molecule was optimized using the MMFF94 molecular mechanics force field (24) with the conjugate gradient method to an energy change convergence criterion of 0.001 kcal/mol·Å. The most stable energy conformation for each compound was attempted by using the conformational analysis, if necessary. The partial atomic charges required to calculate the electrostatic interactions were computed using the MMFF94 method.
After the structures of all molecules were modeled, each of the molecules was aligned using E2 as template, and the following atoms of each molecule were set to be of equal weight to the corresponding atoms of E2 using the rigid-body least squares fitting method: C8, C9, and C11 carbons for ER and C6, C8, C11, and C13 for ERß. After structural alignment, the molecules were then placed in a 3D cubic lattice with a 2-Å spacing. The steric and electrostatic fields were calculated for each molecule at each mesh point using the sp3 carbon probe with +1.0 charge. Any calculated steric and electrostatic energies that were more than 30 kcal/mol were truncated to this value.
3D-QSAR/comparative molecular field analysis (CoMFA).
To develop suitable 3D-QSAR models to probe the relationship between the RBA values (log RBAs) of E2 and its analogs and their molecular structures, the CoMFA methodology was employed. The method of partial least squares regression (25) using the leave-one-out cross-validation procedure (26) was applied to determine the optimum number for the principal components (PCs). In this method, each compound was systematically excluded once from the training set, after which its activity was predicted by a model derived from the remaining compounds. Then, the cross-validated r2 (namely, q2) was calculated using the predicted values. By setting the number of PCs to the optimum number that yielded the smallest standard estimated error or the highest q2 value, the final partial least square analysis was carried out without cross-validation to yield a predictive model and associated conventional r2. This r2 value is an important parameter that reflects the overall degree of correlation between the predicted values and the real data measured for each of the compounds. In the literature (27, 28), a CoMFA model with r2 > 0.8 and q2 > 0.5 is generally considered to be both internally self-consistent and highly predictive.
|
Results
|
|---|
Binding affinities of E1, E2, and estriol (E3; 16-OH-E2) for human ER and ERß
E1, E2, and E3 are three well-known human estrogens. Among all the estrogens analyzed in this study, E2 was found to have nearly the highest binding affinity for both ER and ERß, and its binding affinities for these two ER subtypes were very similar (Fig. 2A and Table 1). Using different concentrations of [3H]E2 as ligand, we have also determined its apparent Kd values for the recombinant human ER and ERß (Fig. 2, E and F). Based on curve regression analysis of the receptor binding data, the Kd of E2 for human ER was 0.7 nM, and its Kd for ERß was 0.75 nM.
View larger version (27K):
[in this window]
[in a new window]
|
FIG. 2. Competition of [3H]E2 binding to recombinant human ER and ERß by E2, E1, E3, or E2-17. The conditions for the in vitro ER binding assay were described in detail in Materials and Methods. For the data shown in A to D, the concentration of the radioactive ligand [3H]E2 was 10 nM, and the concentrations of the competing estrogens were 0, 0.24, 0.98, 3.9, 15.6, 62.5, 250, and 1000 nM. For the data shown in E and F, the concentrations of the radioactive ligand [3H]E2 were 0.025, 0.1, 0.2, 0.4, 0.8, 1.6, 3.2, 6.25, 12.5, and 25 nM. The nonspecific binding (NSB) was determined in the presence of 400-fold excess of cold E2. The Kd values for ER and ERß were calculated according the S-shaped binding curves (curve regression analysis). TB, Total binding; SB, specific binding. Each data point in A to F was the mean of duplicate measurements.
|
|
E1 had 10% of E2’s binding affinity for human ER and had 2% of E2’s affinity for ERß (Fig. 2B and Table 1). E3 also had markedly diminished binding affinity for ER compared with E2 (RBA 10% of E2), but it had rather high binding affinity for ERß (RBA 35% of E2) (Fig. 2C and Table 1). We have also determined, for comparison, the binding affinity of estradiol-17 (E2-17) (a C-17 isomeric analog of E2) for human ER and ERß. Although E2-17 retained considerable binding affinity for human ER (RBA 22% of E2), its binding affinity for ERß was much lower (RBA only 3% of E2) (Fig. 2D and Table 1). Notably, the relative binding affinities and binding preference of E2-17 for human ER and ERß mirror those of E1.
A-ring metabolites
Catechol estrogens.
2-OH-E2 had comparable binding affinity for ER and ERß, and its RBAs for ER and ERß were 22 and 35%, respectively, of E2 (Fig. 3B and Table 1). Notably, the binding affinity of 2-OH-E2 for ER and ERß as determined in the present study is considerably higher than usually thought. The assays were repeated twice, and highly consistent results were obtained. 4-Hydroxyestradiol (4-OH-E2) also had nearly identical apparent binding affinity for ER and ERß, and its RBAs were 70% and 56% of E2, respectively (Fig. 3E and Table 1). In comparison, 2-hydroxyestrone (2-OH-E1) and 4-hydroxyestrone (4-OH-E1) each had markedly weaker binding affinity for ER and ERß. Although 4-OH-E1 had almost identical binding affinity for ER and ERß (Fig. 3D), 2-OH-E1 (the quantitatively predominant endogenous oxidative metabolite of E1) had a substantially higher affinity for ER than for ERß (Fig. 3A). 2-OH-E3 had weak and similar binding affinity for ER and ERß (Fig. 3C).
View larger version (34K):
[in this window]
[in a new window]
|
FIG. 3. Competition of the binding of [3H]E2 to human ER and ERß by various catechol estrogens and methoxyestrogens. The conditions for the in vitro ER binding assay were described in detail in Materials and Methods. The concentration of the radioactive ligand [3H]E2 was 10 nM, and the concentrations of the competing estrogens were 0, 0.24, 0.98, 3.9, 15.6, 62.5, 250, and 1000 nM. Each data point was the mean of duplicate measurements.
|
|
2- or 4-Methoxyestrogens.
All of the monomethylated catechol-E1 metabolites tested in this study (2-methoxyestrone, 2-hydroxyestrone 3-methyl ether, and 4-methoxyestrone) did not have appreciable binding affinity for human ER and ERß at concentrations up to 1000 nM (Fig. 3, F–H). However, the two major monomethylated catechol-E2 metabolites [2-methoxyestradiol (2-MeO-E2) and 4-methoxyestradiol (4-MeO-E2)] each retained weak but similar binding affinities for both ER and ERß (Fig. 3, I and K, and Table 1). In comparison, 2-hydroxyestradiol 3-methyl ether (a close structural analog of 2-MeO-E2) had a substantially weaker binding affinity for ER and ERß than 2-MeO-E2 (Fig. 3J). 4-Methoxyestriol also retained weak but similar binding affinity for ER and ERß (Fig. 3L), and its affinity was comparable with those of 2-MeO-E2 and 4-MeO-E2.
Some other A-ring analogs.
We also compared the binding affinities of several semisynthetic A-ring derivatives of E2 (data shown in Fig. 4). 2-Bromoestradiol (2-Br-E2) markedly decreased the binding affinity for human ER (RBA only 4% of E2), but it decreased the binding affinity for ERß 10 times more (RBA only 0.4% of E2) (Fig. 4A). Interestingly, bromine substitution at the C-4 position of E2 (4-Br-E2) had an effect on the binding affinity for ER and ERß exactly opposite to the C-2 bromine substitution; although it decreased moderately the binding affinity for ERß (RBA 35% of E2), it decreased the binding affinity for ER 5 times more than for ERß, with RBA only 7% of E2 (Fig. 4B and Table 1).
Addition of a methyl group to the C-1 position of E2 (1-methylestradiol) decreased its binding affinity to a similar degree (by 
90%) for ER and ERß (Fig. 4C), but addition of a methyl group to the C-4 position of E2 (4-methylestradiol) decreased its binding affinity for ER more than ERß (similar to that of 4-Br-E2) (Fig. 4D). 2-Ethoxy-E2, an analog of 2-MeO-E2 with strong anticancer activity (29, 30), retained a weak binding affinity for ER and ERß (Fig. 4E), and its affinity is slightly weaker than 2-MeO-E2. The 2,3-dimethylated catechol E1 and E2 derivatives did not have any appreciable binding affinity for ER and ERß (Fig. 4, F and G).
The C-2 or C-4 substitution of an amino (–NH2) or nitro (–NO2) group to E1 significantly diminished its binding affinity for ER and ERß (Fig. 4, H–K). E1-3-sulfate and E2-3-sulfate, two of the major endogenous sulfate conjugates of E1 and E2, had no appreciable binding affinity for ER and ERß (Fig. 4, L and M).
B- and C-ring metabolites
C-6 substituted estrogens.
Among six B-ring hydroxylated or keto metabolites of E2 or E1 tested in this study, all of them retained certain degrees of binding affinity for both ER and ERß (Fig. 5, A–F, and Table 1). 6-Hydroxyestradiol or 6ß-hydroxyestradiol had markedly reduced binding affinities for ER and ERß compared with E2 (Fig. 5, A and B). However, addition of a keto group to the C-6 position of E2 did not markedly affect its original binding affinity for ER and ERß (Fig. 5C). In comparison, addition of a C-6 keto group to E1 differentially altered its binding affinity for ER and ERß (RBAs 23 and 50% of E1, respectively) (Fig. 5D and Table 1). Similarly, addition of a C-6 keto to E2-17 (6-ketoestradiol-17) also markedly reduced its binding affinity for ER (RBA 9% of E2-17), but its binding affinity for ERß was decreased to a relatively lesser extent (RBA 25% of E2-17) (Fig. 5F and Table 1). However, addition of a C-6 keto group to E3 slightly increased its binding affinity for ER (RBA 224% of E3), but it drastically reduced its binding affinity for ERß (RBA 8% of E3) (Fig. 5E and Table 1).
C-11-substituted estrogens.
Addition of a hydrophilic group (such as the hydroxyl or keto group) to the C-11 position (regardless of 11 or 11ß) almost completely eliminated the binding affinities of E2 or E1 for both ER and ERß (Fig. 5, G–J). However, substitution of a lipophilic group with even a bulkier size (such as the acetate or methoxy group) to E2 or 17-ethynylestradiol (17-EE2) did not significantly affect the binding affinity for either ER or ERß (Fig. 5, L and M, and Table 1).
Dehydroestrogens.
In addition to the B- and C-ring substitution metabolites described above, we have also studied several B- or C-ring dehydrogenated E2 or E1 metabolites (data summarized in Fig. 6 and Table 1). Notably, several of the dehydrogenated estrogens tested in this study are the major components (in their conjugated forms) present in Premarin, the commonly used hormone replacement therapy in perimenopausal and postmenopausal women. 6-Dehydroestradiol and 9(11)-dehydroestradiol had similar or somewhat higher binding affinity for human ERß compared with E2 (RBAs 89 and 119%, respectively, of E2), but their binding affinities for ER were slightly reduced, with RBAs 50 and 64% of E2, respectively (Fig. 6, B and F, and Table 1). Interestingly, the binding affinity of 17ß-dihydroequilin (7-dehydro-E2) for ER and ERß was slightly higher than E2 (RBAs 142 and 113%, respectively, of E2) (Fig. 6D). 17-Dihydroequilin (7-dehydro-E2-17) had very similar binding affinities for ER and ERß, but when compared with E2-17, its RBA for ERß was more than 4-fold higher (RBA 447% of E2-17) (Fig. 6I and Table 1).
Compared with E1, 6-dehydroestrone had nearly the same binding affinity for ERß, but its binding affinity for ER was significantly decreased, with its RBA only 10% of E1 (Fig. 6A and Table 1). Equilin and 9(11)-dehydroestrone each had slightly decreased binding affinity for ER compared with E1 (RBAs 45 and 50% of E1, respectively), but they had a drastically increased binding affinity for ERß (RBAs 631 and 316% of E1, respectively). Similarly, D-equilenin (structure shown in Fig. 1) had a weaker binding affinity for human ER than E1 (RBA 20% of E1), but its binding affinity for ERß was much higher than that of E1 (RBA 355% of E1). Also, 17ß-dihydroequilenin (structure shown in Fig. 1) only had a weak binding affinity for ER (RBA 10% of E2), and it retained very high binding affinity for ERß (RBA 100% of E2) (Fig. 6H).
D-ring metabolites
A total of 12 D-ring metabolites/derivatives of E2 and E1 were studied (data summarized in Fig. 7, A–L, and Table 1). The binding affinity of 16-OH-E1 for human ER was twice as high as that of E1, but its affinity for ERß was 18-fold higher than E1 (Fig. 7A). Notably, despite its much higher binding affinities than those of E1 for human ER and ERß, they were still slightly lower than E2 (with RBAs of 56 and 25%, respectively, of E2). Interestingly, whereas 16-ketoestrone (16-keto-E1) only had 18% of E1’s binding affinity for ER, its binding affinity for ERß was 5-fold higher than that of E1 (RBA 501% of E1) (Fig. 7B). Thus, the relative preference of 16-keto-E1 for human ERß over ER is approximately 25 times higher than E1. As already mentioned earlier, E3 (16-OH-E2), a major D-ring metabolite in humans (particularly during pregnancy), had a markedly decreased binding affinity for ER compared with E2 (RBA 11% of E2), but it retained a rather high binding affinity for ERß (RBA 35% of E2) (Fig. 7C and Table 1). Addition of a C-16ß hydroxyl group to E2 (16ß-OH-E2, also referred to as 16-epiestriol) did not noticeably affect its binding affinity for either ER or ERß (Fig. 7D). 16-Keto-E2 and 15-OH-E3 (estetrol) each had a reduced binding affinity for both ER and ERß compared with E2 and E3, respectively (Fig. 7, E and F).
As already mentioned earlier, E2-17 retained considerable binding affinity for ER (RBA 22% of E2), but it had substantially lower binding affinity for ERß (3% of E2) (Fig. 2D or 7G). Interestingly, addition of a hydroxyl group to the C-16 or C-16ß position of E2-17 affected its binding affinity for ER and ERß rather differently (Fig. 7, H and I). 16-Hydroxyestradiol-17 (16-OH-E2-17; 17-epiestriol) had very high, almost identical binding affinities for both ER and ERß (RBAs 71 and 79%, respectively, of E2), which were 3 and 16 times higher, respectively, than its precursor, E2-17. However, 16ß-OH-E2-17 (16,17-epiestriol) had a preferential binding affinity for ERß over ER, and the difference in the binding affinities is 18-fold.
17-EE2, a semisynthetic steroidal estrogen commonly used as an estrogenic component in various oral contraceptives, had very high binding affinity for both ER and ERß compared with E2. The binding affinity of 17-EE2 for ER was twice as high as that of E2, but its affinity for ERß was only about half of that of E2 (Table 1 and Fig. 7J). The same receptor binding assay with this estrogen was repeated twice, and consistent results were obtained. Accordingly, the relative ratio of preference for binding to ER and ERß by 17-EE2 is approximately 4 times of that for E2. Interestingly, the removal of the C-17 hydroxyl group from E2 (17-desoxy-E2, see structure in Fig. 1) did not drastically reduce its binding affinity for human ER and ERß (Fig. 7K). Similarly, 1,3,5(10),16-estratetraen-3-ol, which is also without a C-17 substitution (structure in Fig. 1), had a very similar binding affinity as that of 17-desoxy-E2 for ER and ERß (Fig. 7L).
In summary, most of the D-ring metabolites retained rather high binding affinity for human ER and ERß, but several of them (16ß-OH-E2-17, 16-OH-E2-17, 16-keto-E1, 16-OH-E2, and 16-OH-E1) had markedly increased binding affinity for human ERß over ER compared with their respective precursors (namely, E1, E2, and E2-17).
Antiestrogens, phytoestrogens, and stilbene estrogens
We have also determined, for the purpose of comparison, the binding affinities of a number of steroidal and nonsteroidal antiestrogens, phytoestrogens, stilbene estrogens, and nonaromatic steroids for human ER and ERß. Their data are summarized in Fig. 8 (A–M).
Steroidal and nonsteroidal antiestrogens.
The binding affinities of ICI-182,780 for both ER and ERß were very high and nearly the same (RBAs 45 and 35%, respectively, of E2) (Fig. 8A and Table 1). Similarly, another two synthetic C-7 substituted analogs, E2-7-(CH2)6OH and E2-7-(CH2)6OC6H5 (structure shown in Fig. 1), which have shorter side chains at the C-7 position than ICI-182,780, retained high and similar binding affinities for ER and ERß as the ICI compound (Fig. 8C and Table 1).
Tamoxifen and raloxifene are two well-known nonsteroidal ER antagonists (partial agonists). Tamoxifen had almost identical binding affinities for human ER and ERß (Fig. 8D), although its binding affinities for these two receptors were only 3–4% of those of E2 and 7–10% of ICI-182,780. In comparison, although raloxifene had a similar binding affinity for ERß as tamoxifen, the former had 16-fold higher binding affinity for ER than the latter and was comparable with ICI-182,780 (Fig. 8E and Table 1). Therefore, raloxifene actually had a strong preferential binding affinity for human ER than for ERß.
Phytoestrogens.
Genistein, a well-known isoflavone phytoestrogen abundantly present in soy products, had very high binding affinity for human ERß (almost comparable with that of E2), but its binding affinity for human ER was far lower, only 6% of its binding affinity for ERß (Fig. 8F). Coumestrol had very high binding affinity for human ER and ERß, and its RBA for ERß was slightly higher than its affinity for ER (Fig. 8G). Myricetin basically had no appreciable binding affinity for human ER and ERß (Fig. 8H). Daidzein had very weak binding affinities for both ER and ERß, but its relative affinity for ERß was significantly higher than its affinity for ER (Fig. 8I). Dibenzoylmethane had a weak overall binding affinity for ER and ERß (Fig. 8J).
Stilbene estrogens.
The three well-known nonsteroidal stilbene estrogens [diethylstilbestrol (DES), dienestrol, and hexestrol] had very high binding affinities (similar to that of E2) for human ER and ERß (Fig. 8, K–M, and Table 1). We noted that DES and hexestrol had a slightly higher binding affinity for ERß than for ER, although the difference was only very small.
3D-QSAR/CoMFA of estrogen derivatives for ER and ERß
To probe the structural determinants of various steroids for binding to human ER and ERß, we performed 3D-QSAR/CoMFA analysis of a selected pool of 48 representative steroidal estrogens used in the present study (Table 1, footnote a). A 3D-QSAR/CoMFA model for human ER was developed using 47 compounds excluding 17-desoxy-E2 as an outlier compound, whereas another 3D-QSAR/CoMFA model for human ERß was developed by using 47 compounds excluding 16-keto-E1 as an outlier compound. Of several variations in the structural alignment scheme considered, the best results were obtained by superimposing the backbone carbons in the middle C- and D-rings. It is likely that these two rings of different estrogens experienced less change upon binding to the receptors. The statistical results of coMFA models for both human ER and ERß are summarized in Table 2. The correlations of the predicted log RBA values for ER and ERß are shown in Fig. 9. The r2 values of the CoMFA models for human ER and ERß were 0.963 and 0.971, respectively (Table 2), suggesting high degrees of overall correlation for both ER subtypes between predicted binding affinities and experimental data (as log RBAs) for the selected compounds. The q2 values of the CoMFA models for ER and ERß were 0.531 and 0.634, respectively (Table 2). The q2 values are important parameters that reflect the overall predictive ability of the 3D-QSAR/CoMFA models developed in this study. The individual contributions from the steric/electrostatic fields were 39/61% for the ER CoMFA model and 38/62% for the ERß CoMFA model.
The color contour maps derived from the CoMFA models for ER and ERß are shown in Fig. 10. Note that E2 is shown inside the field only for demonstration purposes. The contours of the steric map are shown in yellow and green, and those of the electrostatic map are shown in red and blue. Green contours indicate regions where a steric bulk substituent would increase the binding affinity of some ligands with the receptor, whereas the yellow contours would indicate areas where a steric bulk substituent would decrease the binding affinity for many of the ligands. The red contours mean regions where a substituent with stronger negative charge would increase the binding affinity, whereas the blue contours show areas where a substituent with strong negative charge would decrease the binding affinity.
|
Discussion
|
|---|
In the present study, we systematically compared the binding affinities of a large number of hydroxylated or keto metabolites of E2 and E1 along with some other natural or synthetic estrogens for the recombinant human ER and ERß proteins. Among all the ligands tested, the parent hormone E2 had nearly the highest and similar binding affinities for both human ER and ERß. The apparent Kd values of E2 for the recombinant human ER and ERß as estimated in this study were 0.7 and 0.75 nM, respectively (Fig. 2), which were slightly higher than some of the earlier measurements (average approximately 0.3 nM) using the crude ER protein preparations from various human tissues or cell lines (31, 32). This difference likely was due to the relatively higher concentration of the recombinant ER and ERß proteins used in our in vitro receptor binding assays and perhaps also due to the absence of other cellular proteins or components that usually partner the steroid receptors in subcellular crude extracts or in vivo.
Notably, E1 and E3 are perhaps the two best known metabolites of E2 in human. Although these two endogenous E2 derivatives had markedly lower binding affinities for human ER and ERß than E2 (Fig. 2), it is of interest to point out that the facile metabolic conversion of E2 to E1 or of E2 to E3 in a woman may confer differential activation of the ER or ERß signaling system under different physiological conditions. For instance, E1 had 4-fold higher RBA for human ER than for ERß, and this estrogen metabolite is present in larger quantities than E2 in circulation as well as in most tissues of a nonpregnant woman, largely due to the actions of high levels of the oxidative 17ß-hydroxysteroid dehydrogenase(s) (33, 34). Hence, the facile metabolic conversion of E2 to E1 would effectively produce a preferential activation of the ER signaling system over the ERß system in most target tissues of a nonpregnant woman. In contrast, E3 has a more than 5-fold preference for the activation of human ERß over ER, and it is a quantitatively predominant estrogen metabolite produced during pregnancy. It is of interest to suggest that the very high levels of E3 present during pregnancy may produce a differential activation of the ERß signaling system in the pregnant woman and fetus for fulfilling various unique physiological functions.
For convenience and clarity, we have separately discussed below the structure-activity relationship for each group of the estrogen metabolites/derivatives tested in the present study. Because the binding affinities of most E1 derivatives are very weak (with a few exceptions, such as 16-OH-E1), our focus will be mostly placed on the structure-activity relationships of various E2 derivatives, many of which still retain high ER-binding affinities.
A-ring derivatives
Addition of a methyl group to the C-1 position of E2 (1-methylestradiol) decreased its binding affinity by about 90% for both ER and ERß. Several earlier studies have also shown that the C-1 substitution of E2 (regardless of polarity of the substituents) all had a negative effect on the binding affinity for crude ER proteins from rabbit or human (35). This influence was thought to be due to a direct interaction of the C-1-substituting group with the ER protein rather than a structural perturbation of the ligand conformations.
2-OH-E2 is the most abundant hydroxy-E2 metabolite formed in human liver (19, 36). Largely because of its rather low estrogenic activity as measured earlier in laboratory animals (ovariectomized or immature rats or mice) and also in cultured human breast cancer cells (32, 36, 37, 38, 39), this catechol-E2 metabolite was generally considered to have very weak estrogenic activity in human. It has been widely accepted the notion that increased metabolic formation of 2-OH-E2 in vivo as opposed to the formation of other oxidative metabolites such as 4-OH-E2, 16-OH-E1, or 16-OH-E2 (E3), would significantly reduce estrogen’s hormonal activity in human and thus would be beneficial for the reduction of breast cancer risk (32, 40, 41). It should be noted that the results of our present study showed that the binding affinities of 2-OH-E2 for human ER and ERß (RBAs 22 and 35%, respectively, of E2) are not too low, and actually they are much higher than usually thought. Despite its relatively high ER binding affinity, 2-OH-E2 may still be a highly beneficial metabolite of E2 in human because of its rapid metabolic O-methylation in vivo that deactivates its hormonal activity and also concomitantly forms the anticarcinogenic 2-MeO-E2 (36, 42). It is of note that although 2-OH-E1 has relatively low binding affinity for human ER and ERß (significantly lower than that of 2-OH-E2), this oxidative E1 metabolite has a significant preference for binding to ER over ERß. Taking together the ER-binding data for E1 and 2-OH-E1, it is of interest to point out that these two quantitatively predominant estrogens normally present in nonpregnant woman would consistently produce a preferential activation of ER over ERß.
Different from 2-OH-E2, 4-OH-E2 is known to retain strong estrogenic activity and high ER binding affinity (36, 37), and our data also showed that this catechol-E2 metabolite retained high and almost identical binding affinity for ER and ERß, with RBAs 70 and 56% of E2, respectively. Similar to 4-OH-E2, 4-OH-E1 (a quantitatively minor metabolite) also retained almost identical binding affinity for ER and ERß, although its binding affinity is significantly lower than that of 4-OH-E2.
Notably, some earlier studies have suggested that substitution of small functional groups at the C-2 and C-4 positions are reasonably well tolerated, whereas larger groups may readily reduce ER binding affinity because they may involve the formation of an intramolecular hydrogen bond with the C-3 hydroxyl group (35). However, we observed that, in some cases, substitution of even a very small group (in the case of 2-Br-E2) drastically reduced its binding affinity for ERß (RBA only <0.5% of E2), whereas this substitution reduced its binding affinity for ER to a relatively lesser degree (RBA 4% of E2). This observation is rather interesting because C-2 bromine substitution had a far stronger negative effect on ER binding (particularly for ERß) than the C-2 hydroxyl substitution. Notably, bromine substitution at the C-4 position of E2 (4-Br-E2) affected its binding affinity for ER and ERß in an opposite manner as what was observed for 2-Br-E2, and it decreased the binding affinity for ER 5 times more than for ERß. Similarly, addition of a methyl group to the C-4 position of E2 (4-methylestradiol) also decreased its binding affinity for ER more than for ERß.
Our data clearly showed that 2-MeO-E2 and 4-MeO-E2 each retained weak but substantial binding affinity for both ER and ERß, with RBAs 1–2% of E2. The estimated binding affinities are considerably higher than earlier measurements using cytosols prepared from human breast cancer. The residual ER-binding activity of 2-MeO-E2 is believed to be mainly responsible for its moderate growth-stimulatory effect in ER-positive human breast cancer cells when exogenous estrogens were not present (38). 2-Ethoxy-E2 also retained detectable binding affinity for ER and ERß, and its affinity was slightly lower than 2-MeO-E2, probably due to the bulkier size of the ethoxy group at the C-2 position compared with a methoxy group. Various dimethylated catechol-E2/E1 metabolites did not have appreciable binding affinity for either ER or ERß, which was in accord with earlier studies using the rat uterine ER protein preparation (37). As expected, several synthetic C-2 or C-4 substitution analogs of E1 (2-aminoestrone, 2-nitroestrone, 4-aminoestrone, and 4-nitroestrone) only retained very weak binding affinities for human ER and ERß. It is of note that the –NO2 and –NH2 substitutions of E1 produced inhibition curves with rather shallow slopes, which likely suggests that these were not pure competitive inhibition.
The C-3 sulfated estrogens (E1-3-sulfate and E2-3-sulfate) were found to be basically devoid of appreciable binding affinity for ER and ERß, which was consistent with earlier findings. Like the C-3 sulfated estrogens, earlier studies have shown that E2 3-methyl ether (43, 44) or 2-desoxy-E2 (45, 46) each had very low binding affinity for ER compared with E2. It was suggested that the C-3 hydroxyl group of E2 functions primarily as an H-bond donor in its interactions with ER and ERß (34). According to more recent x-ray crystallography study of the human ER and ERß bound with E2 (46), it appears that the very low binding affinities of various C-3 modified E2 derivatives are due to a combination of disturbance of H-bond formation and steric hindrance.
B- and C-ring derivatives
Our data showed that addition of a hydroxyl group to the C-6 or C-6ß position of E2 markedly reduced its binding affinity for both ER and ERß, but addition of a keto group to the C-6 positions of E2 or E1 did not significantly affect the original binding affinity of these estrogens for ER or ERß.
Our data with the C-11 position derivatives were rather interesting and revealing. Addition of a hydrophilic group (such as a hydroxyl or keto group) to the C-11 position of E2 or E1 almost completely abolished their binding affinities for both ER and ERß. However, substitution of a lipophilic group with even a bulkier size (such as the acetate or methoxy group) did not significantly affect the binding affinity for either ER or ERß. These data indicated that the drastic decrease in the binding affinities of 11-OH-E2, 11ß-OH-E2, or 11-keto-E2 for human ER and ERß is not due to steric hindrance caused by the C-11 position substitutions, but it is primarily due to alterations of the lipophilicity near the C-11 position. It is of note that earlier studies have also shown that the C-11ß position of E2 was tolerant of even very large substituents, if the polar functional groups were placed at a distance from the steroid core structure (for review, see Ref. 35, 47). These observations agree well with our recent homology modeling data for human ER and ERß showing that there is considerable space near E2’s C-7 binding site that can readily accommodate various estrogen analogs with a rather bulky/lengthy substitution (data not shown).
In addition to the B- and C-ring substitution metabolites described above, we have also studied several common B- or C-ring dehydrogenated E2 or E1 metabolites. We found that most of the dehydroestrogen metabolites retained rather high binding affinities for both ER and ERß compared with their respective nondehydrogenated precursors, and some of them [such as 6-dehydroestradiol and 9(11)-dehydroestradiol] retained high binding affinities for human ERs.
Here, it should be noted that Premarin, the commonly used hormone replacement therapy in perimenopausal and postmenopausal women, contains a mixture of conjugated estrogens obtained from pregnant mare’s urine. The main estrogenic ingredients include sodium E1 sulfate, sodium equilin sulfate, and the sodium sulfate conjugates of E2-17, 17-dihydroequilenin, and 7-dehydro-E2. Equilin had slightly decreased binding affinity for ER (its RBA 40% of E1), but it had drastically increased binding affinity for ERß (its RBA 631% of E1). The binding affinities of 7-dehydro-E2 for human ER and ERß were actually slightly higher than E2 (its RBAs 142 and 113%, respectively, of E2). Similarly, although the binding affinity of 7-dehydro-E2-17 for ER remained the same as that of E2-17, this equine estrogen had a more than 4-fold higher binding affinity for ERß than E2-17. D-Equilenin had a much weaker binding affinity than E1 for human ER (RBA 20% of E1), but its binding affinity for ERß was more than 3 times higher than that of E1. Very similarly, although 17ß-dihydroequilenin had a low binding affinity for ER (35% of E2), it had a high binding affinity for ERß (RBA 100% of E2). Taken together, it is evident that many of the equine estrogens contained in Premarin have a differential binding affinity for human ERß over ER. In comparison, the estrogenic component in most birth control pills is 17-EE2, which has a binding affinity for ER
Top
Abstract
Introduction
