3D Models of MBP, a Biologically Active Metabolite of Bisphenol A, in Human Estrogen Receptor α and Estrogen Receptor β

Bisphenol A [BPA] is a widely dispersed environmental chemical that is of much concern because the BPA monomer is a weak transcriptional activator of human estrogen receptor α [ERα] and ERβ in cell culture. A BPA metabolite, 4-methyl-2,4-bis(4-hydroxyphenyl)pent-1-ene [MBP], has transcriptional activity at nM concentrations, which is 1000-fold lower than the concentration for estrogenic activity of BPA, suggesting that MBP may be an environmental estrogen. To investigate the structural basis for the activity of MBP at nM concentrations and the lower activity of BPA for human ERα and ERβ, we constructed 3D models of human ERα and ERβ with MBP and BPA for comparison with estradiol in these ERs. These 3D models suggest that MBP, but not BPA, has key contacts with amino acids in human ERα and ERβ that are important in binding of estradiol by these receptors. Metabolism of BPA to MBP increases the spacing between two phenolic rings, resulting in contacts between MBP and ERα and ERβ that mimic those of estradiol with these ERs. Mutagenesis of residues on these ERs that contact the phenolic hydroxyls will provide a test for our 3D models. Other environmental chemicals containing two appropriately spaced phenolic rings and an aliphatic spacer instead of an estrogenic B and C ring also may bind to ERα or ERβ and interfere with normal estrogen physiology. This analysis also may be useful in designing novel chemicals for regulating the actions of human ERα and ERβ.


Introduction
One consequence of our industrial society is the presence of novel environmental chemicals that disrupt normal physiological responses in humans, other vertebrates, as well as invertebrates [1,2]. Many of these chemicals are small hydrophobic molecules that resemble steroids, thyroid hormone, retinoids and other lipophilic hormones and, as a result bind to their receptors in vertebrates [3,4,5,6,7]. Some of these chemicals act like hormones, while others act like anti-hormones. In either case, they disrupt normal endocrine physiology.
An endocrine disruptor of much concern is bisphenol A [BPA] because it is widely dispersed in the environment due to the presence of BPA in polycarbonate plastics, which are used in containers for food and water, including baby bottles, as well as the linings of metal cans used for food and beverages [8,9,10]. Leaching of the BPA monomer from these sources into food, milk and the environment exposes humans [11,12,13] and wildlife [2,14] to BPA.
A consequence of the widespread use of BPA is that over 90% of the general population is exposed to BPA [9,13,15]. BPA levels range from 0.3 nM to 40 nM in maternal plasma and fetal human serum [8,10,11]. Moreover, due to the lipophilic nature of BPA, it can accumulate in fat [16].
BPA has some structural similarity to estradiol and diethylstilbestrol [ Figure 1], and, indeed, BPA binds to human estrogen receptor a [ERa] and ERa and is a transcriptional activator of these ERs [17,18,19,20]. However, BPA's binding affinity and transcriptional activity for these ERs is over 1000-fold lower than that of E2 [17,18,19,20], which makes it unlikely that nM concentrations of BPA would disrupt estrogen physiology. Nevertheless, in vivo studies indicate that BPA is active at 1 nM to 10 nM [8,10,15,21], which raises the possibility that BPA is metabolized to a more active endocrine disruptor. One such candidate metabolite is 4-methyl-2,4-bis(4-hydroxyphenyl)pent-1ene [MBP] [ Figure 1], which has about 1000-fold higher estrogenic activity than BPA [22,23]. To begin to understand the structural basis for the high estrogenic activity of MBP and its higher affinity compared to BPA for human ERa and ERb, we constructed 3D models of MBP and BPA in human ERa and ERb. We find that MBP retains key contacts with human ERa and ERb that are important in activation of these receptors by estradiol. We also find that one phenolic ring of BPA can mimic binding of the A ring of E2 to ERa and ERb, which would account for the binding of BPA to these ERs. However, the second phenolic ring on BPA lacks some key contacts that are found between E2 and both ERs, which may explain the lower estrogenic activity of BPA. In addition to elucidating the interaction of MBP and BPA with both human ERs, this analysis may be useful in designing novel chemicals for regulating the actions of human ERa and ERb.

Methods
Human ERa [24] [25,26] and AutoDock Vina [27]. The grid was centered over the estrogen binding site in human ERa. AutoDock 4 was run using the Lamarckian Genetic Algorithm for 250 trials of 5 million energy evaluations. AutoDock Vina was run with a setting of 20 for exhaustiveness and poses for the 100 lowest energies were collected.
The crystal structure of ERb complexed with E2 [PDB:3OLS] [28] was selected for docking MBP and BPA. As was found in other ERb structures in the PDB, 3OLS lacks coordinates for five amino acids corresponding to residues 416-420. To model the missing amino acids, we used the Homology option in Insight II and the 1G50 structure for human ERa as a template. A PDB file of the complete ERb with E2 was refined with Discover 3 with the CVFF force field and a distant dependent dielectric constant of 2 for 50 iterations. We docked MBP and BPA into this PDB file of human ERb with AutoDock 4 and AutoDockVina [27] with the settings used previously for human ERa.
The lowest energy complexes of MBP and BPA in ERa and ERb, as calculated by AutoDock 4 and AutoDock Vina, were refined with the Discover 3 software in Insight II. For this energy minimization step, Discover 3 was used with the CVFF force field and a distant dependent dielectric constant of 2 for 10,000 iterations. During this refinement step, both the amino acids on the ERs and MBP and BPA rearrange their positions so as to lower the Gibbs free energy of the complex.

Docking Energy Analysis
We used X-Score [29,30] and DSX [DrugScore eXtended] [31] to estimate the relative binding energy of MBP and BPA in the various configurations in ERa and ERb. X-Score uses an empirical scoring function to estimate the affinity of a ligand for a protein. DSX uses a knowledge-based scoring function based on the DrugScore formalism [32] to estimate the affinity of a ligand for a protein. In comparing the score of two ligands for a protein, the ligand with the larger negative score has the higher affinity. The spacing between the first and second phenolic hydroxyls on MBP and DES is similar to that between C3 hydroxyl and the 17b-hydroxyl on E2. In contrast, the distance between the two phenolic hydroxyls in BPA is shorter than that in E2. doi:10.1371/journal.pone.0046078.g001 3D Models of MBP in Human ERa and ERb PLOS ONE | www.plosone.org

Docking of MBP and BPA to Human ERa and ERb
Docking of MBP into human ERa and ERb using AutoDock 4 [25,26] and AutoDock Vina [27] gave two symmetric poses, which is not surprising because MBP has a phenolic ring at each end [ Figure 1]. BPA also had two poses for one of the rings in ERa and ERb. We analyzed both poses for MBP and BPA in human ERa and ERb. In our analysis of the 3D models of MBP and BPA in both ERs, we use the term ''first phenolic ring'' to describe the ring that has contacts with ERa and ERb that are similar to the A ring of E2. Analysis of the crystal structures of ERa complexed with E2 [33,34] and other estrogens [35] revealed that Glu-353 and Arg394 have important stabilizing contacts with the C3 hydroxyl on the A ring and His-524 with the 17b-hydroxyl on the D ring. Glu-305, Arg-346 and His-475 on ERb have similar stabilizing contacts with estrogens. As reported below, the presence or absence of these contacts in the 3D models of ERa and ERb with BPA and MBP is important analyzing the interaction between these chemicals and the ERs.

Analysis of MBP in Orientation 1 in Human ERa
In Figure 2A, we show the 3D model of MBP in human ERa in Orientation 1, in which C1 on MBP is closest to first phenolic ring. The distances between MBP and ERa are shown in Figure 2A and Table 1. For comparison, in Figure 3A and Table 2 we show the distances between E2 and human ERa [33,35,36,37,38].
The first phenolic ring on MBP has contacts that are similar to that of the A ring on E2 with human ERa [33,34,39]. The phenolic hydroxyl on MBP is 3.1 Å from Oe2 on Glu-353, 3.1 Å from Ng2 on Arg-394 and 2.8 Å from the backbone oxygen of Leu-387. MBP is 3.9 Å from Ce2 on Phe-404 [ Figure 2A, Table 1]. These contacts are similar to that for E2 with human ERa, except that Cd2 on Leu-387 does not contact the phenolic hydroxyl on MBP, in contrast to the contact between Leu-387 and E2 in human ERa [ Figure 3A, Table 2]. The second phenolic hydroxyl in MBP is 3 Å , 2.8 Å and 4 Å from Nd1 on His-524, the backbone oxygen on Gly-521 and Cb on Leu-525, respectively, on ERa [ Figure 2A]. This phenolic hydroxyl also contacts Met-343 and Met-421 on ERa. These five residues stabilize the D ring on E2 in human ERa [ Figure 3A].
There are, however, differences in some interactions between ERa and MBP compared to that with E2. While Gly-521 and  [24,33,34,36,37,38,39]. The phenolic hydroxyl of E2 contacts Glu-353, Arg-394 and Leu-387. The 17b-hydroxyl contacts His524 and Leu-525. The D ring contacts Met343, Met421, Gly-521 and Ile-424. Favorable van der Waals contacts have a distance of 4.25 Å or less between E2 and amino acids on ERa. B. Interaction of E2 with human ERb [28]. The phenolic hydroxyl of E2 contacts Glu-305, Arg-346 and Leu-339. The 17b-hydroxyl contacts Gly-472, His473 and Leu-476. Met-421 contact the second phenolic hydroxyl on MBP [ Figure 2A], Gly-521 and Met-421 contact C16 on E2 in ERa [ Figure 3A]. While Leu-384 has two van der Waals contacts with MBP, Leu-384 does not contact E2 in ERa. While Thr-347 has a van der Waals contact with MBP, Thr-347 does not contact E2 in ERa. While Cb on Leu-391 is 3.7 Å from the first phenolic hydroxyl on MBP, this contact is absent between ERa and E2. While Ala-350 contacts the linker between the two phenolic rings on MBP, Ala-350 contacts C1 on the A ring in E2 in ERa. Phe-404 and Met-421 have van der Waals contacts with C1 on MBP, which has no equivalent in E2 in ERa.

Analysis of MBP in Orientation 2 in Human ERa
As shown in Figure 2B and Table 1, analysis of ERa with MBP in Orientation 2 reveals that MBP has contacts with Glu-353, Arg-394, Phe-404, Met-343, Leu-384, Met-421, Gly-521, His-524 and Leu-525 that are similar to those found in Orientation 1 of MBP in ERa. Due to the reversed orientation of MBP in ERa, C1 on MBP has a van der Waals contact with Thr-347, and the other part of the linker contacts Ala-350. Analysis of MBP in Orientation 1 in Human ERb Figure 4A shows MBP in Orientation 1 in human ERb. For comparison, in Figure 3B, we show E2 in human ERb [28]. Many of the contacts between MBP and human ERb shown in Figure 4A and Table 3 are similar to that between MBP in Orientation 1 and human ERa [ Figure 2A, Table 2] and between E2 and ERb [ Figure 3B]. Like the A ring in E2, the first phenolic ring on MBP has stabilizing contacts with Glu-305, Arg-346, Phe-356, Leu-339 and Leu-343 in ERb. The second phenolic ring contacts His-475, Gly-472, Leu-476, Ile-373 and Ile-376 [ Figure 4A, Table 3].  Figure 4B shows MBP in Orientation 2 in human ERb. Many of the contacts between MBP in Orientation 2 and human ERb [ Table 3] are similar to that between MBP in Orientation 1 and human ERb [ Figure 4A, Table 3]] and between E2 and ERb [ Table 2]. The backbone oxygen on Leu-339, Cb on Ala-302 and the side chains on Leu-343 contact the first phenolic ring on MBP. These contacts are absent between MBP in Orientation 2 in ERa [ Figure 2B, Table 1].  Figure 5A, Table 4]. However, the backbone oxygen on Leu-387 is 6.4 Å from the phenolic hydroxyl and Leu-391 does not have a van der Waals contact with the phenolic ring.

Analysis of BPA in Orientation 1 in Human ERa
The second phenolic ring does not contact either Gly-521, His-524 or Leu-525 [ Table 4]. Instead, phenolic ring moves so that it contacts Ce2 and Cg on Phe-404, Ce on Met-421 and Cd2 on Ile-424. Also, Ala-350 and Leu-384 and Thr-347 contact the linker on BPA.

Analysis of BPA in Orientation 2 in Human ERa
In Figure 5B, we show BPA in Orientation 2 in human ERa. The first phenolic ring on BPA contacts Oe2 on Glu-353, Ng2 on Arg-394, Cd2 on Phe-404 and Cb on Ala-350 [ Figure 5B,   Table 4]. Leu-391 has a van der Waals contact with the phenolic ring. However, the backbone oxygen on Leu-387 does not contact BPA.

Analysis of BPA in Orientation 1 in Human ERb
In Figure 6A, we show BPA in Orientation 1 in human ERb. The first phenolic ring on BPA contacts Oe2 on Glu-305, Ng2 on Arg-346, Ce2 on Phe-404, Cb on Leu339 and the backbone nitrogen on Ala-302 [ Figure 6A, Table 5]. The backbone oxygen on Leu-339 does not contact the phenolic hydroxyl. Cb on Ala-302 and Cd2 on Leu-343 do not contact the phenolic ring.

Analysis of BPA in Orientation 2 in Human ERb
In Figure 6B, we show the minimized structure of BPA in Orientation 2 in human ERb. The first phenolic ring on BPA contacts Oe2 on Glu-305, Ng2 on Arg-346, Ce2 on Phe-404, Cb on Ala-350, Cd2 on Leu-343 and the backbone oxygen on Leu-339 [ Figure 6B, Table 5].

Docking Energy Analysis
We used X-Score [31] and DSX [31] to estimate the affinity of MBP and BPA in their different orientations in ERa and ERb. Tables 6 and 7 summarize these analyses for the X-Score and DSX. For both algorithms, MBP has an affinity for ERa and ERb that is closer to that of E2 than is BPA for these receptors. This is consistent with previous assays of the activity of MBP and BPA [22,23].

Discussion
The leaching of BPA monomers from polycarbonate containers and from liners of metal containers for food and beverages has contributed to the widespread exposure of humans to BPA [2,11,12,13,15]. The relatively low affinity of BPA, compared to E2, for human ERa and ERb [17,19,20] would, at first glance, make it unlikely that BPA would be a problem as an estrogenic endocrine disruptor at nM concentrations [9,10,11]. However, it is clear that nM concentrations of BPA have estrogenic activity [8,21]. The discovery that MBP, a metabolite of BPA, has a nM affinity for human ERa and ERb, suggests that metabolism of BPA to MBP could explain some of effects of BPA on estrogen physiology [22,23]. There is a structural basis for considering BPA and MBP as potential ligands for ERa and ERb because BPA and MBP have some structural similarities to known synthetic estrogens [ Figure 7]. BPA is a bisphenol linked by one carbon atom [ Figure 1] as are cyclofenil-type estrogens [ Figure 7], some of which have high affinity for ERa and ERb [40]. MBP is a bisphenol linked by three carbon atoms [ Figure 1] as is benzestrol [ Figure 7], which has a high affinity for ERa and ERb [39]. Hexestrol, which is linked by two carbon atoms, also has a high affinity for ERa and ERb [18,39]. Thus, it is reasonable to be concerned about potential endocrine disruption by synthetic bisphenols. However, as discussed below, our 3D model of BPA in ERs indicates that BPA does not have the contacts with an ER as is found between fluorine-substituted cyclofenil derivatives [40]. Studies with a wide variety of synthetic bisphenols [39,40] indicate the length of the carbon linker between bisphenols and side chain substitutents on the cyclohexane ring on cyclofenils are important in establishing contacts that lead to high affinity binding to the ER. This is consistent with the analysis of our 3D models of BPA and MBP in ERa and ERb as discussed below.

MBP Retains Important Contacts found between E2 and ERa and ERb
Our 3D models of MBP and BPA in human ERa and ERb [Figures 2,[4][5][6] identify contacts that can explain MBP's high affinity and BPA's low affinity for both estrogen receptors. A key structural difference between BPA and MBP is the longer spacing between the two phenolic rings in MBP [ Figure 1]. As a result, both phenolic rings on MBP form stabilizing contacts with ERa and ERb that are similar to that between the A and D rings of E2 and human ERa and ERb [28,33,34,35,38,39,41]. These 3D models predict that the second phenolic hydroxyl on MBP has a hydrogen bond with His-524 on ERa and His-475 on ERb. Our 3D models can be tested by investigating transcriptional activation by MBP of ERa and ERb in which His-524 and His-475, respectively, have been mutated.

BPA Lacks some Contacts found between E2 and ERa and ERb
Like the A ring on E2, one phenolic ring on BPA has stabilizing contacts with Glu-353, Arg-394 and Phe-404 in ERa [ Figure 5]. Interestingly, Phe-404 also contacts the second phenolic ring. However, the second phenolic ring on BPA does not contact either Gly-521 or His-524 on ERa, which is significant because contacts between E2 and Gly-521 and His-524 in ERa are important in binding of E2 [28,33,35,39,41]. Also, Leu-387 does not contact the first phenolic ring of BPA in either Orientation 1 or Orientation 2. The loss of these contacts between BPA and ERa may explain the lower affinity of BPA for ERa.
In the 3D model of BPA in Orientation 1 in ERb [ Figure 6A], one phenolic ring on BPA has stabilizing contacts with Glu-305, Arg-346 and Phe-356 in ERb. Moreover, the second phenolic ring on BPA contacts Gly-472, His-475 and Leu-476 on ERb. Thus, some of the key interactions between the ERb and the A and D rings on E2 [ Figure 3B] are conserved for BPA in Orientation 1 in ERb. However, neither Leu-339 nor Leu-343 contacts the first phenolic ring on BPA. The loss of these contacts would be expected to lower affinity of BPA for ERb.

Cellular Context Influences Estrogenic Activity of MBP and BPA
The presence of two phenolic rings in MBP and BPA and the flexibility in the estrogen binding site in ERa and ERb [7,33,35,42,43,44,45,46] are important factors in the binding of MBP and BPA to these ERs. The equilibrium dissociation constant of BPA for ERa and ERb is about 195 nM and 35 nM respectively [18]. Using a different binding assay, IC50s [23] for binding of BPA and MBP to ERa and ERb were reported. BPA and MBP have IC50s of 1.8 mM and 52 nM, respectively  ERa BPA Orientation 1 6.5 Figure 4A ERa BPA Orientation 2 6.5 Figure 4B ERb E2 7.5 Figure 3A ERb MBP Orientation 1 7.1 Figure 5A ERb MBP Orientation 2 7.2 Figure 5B ERb BPA Orientation 1 6.7 Figure 6A ERb BPA Orientation 2 6.7 Figure 6B X-Score Analysis of MBP and BPA in ERa and ERb [29,30]. doi:10.1371/journal.pone.0046078.t006 Table 7. Docking analysis of MBP and BPA in ERa and ERb.
Also important for transcriptional activation of ERs and other nuclear receptors by steroids and endocrine disruptors is the binding of co-regulators to the ligand-receptor complex [47,48,49,50,51,52,53]. Thus, even a low affinity ligand such as 27-hydroxy-cholesterol can have transcriptional activity for the ER in the presence of the appropriate co-activator [44].
The interaction of complexes of ERa and ERb with different co-regulators may explain the report by Yoshihara et al. [23] that transcriptional activation of ERa and ERb by MBP and BPA depended on the cellular context. That is, the estrogenic activity of MBP and BPA is altered in the presence or absence of co-regulators [20,48,50,51,54,55]. In the yeast estrogen screening (YES) assay, the EC50 potencies for transcriptional activation of ERa by MBP and BPA were 0.7 mM and 160 mM, respectively. In experiments, which included the Transcriptional Intermediary Factor 2 [TIF2] co-activator in the assay, the EC50s for transcriptional activation of rat ERa with TIF2 by MBP and BPA were 8.3 nM and 14 mM, respectively, and the EC50s for rat ERb with TIF2, by MBP and BPA were 8.3 nM and 13 mM, respectively. Thus, transcriptional activation of ERs by MBP and BPA in an assay containing TIF2, which mimicked conditions in some mammalian cells, increased by about 10fold compared to the assay in yeast cells.
Further evidence for the importance of cellular context on transcriptional potency of MBP and BPA comes from experiments using an ERE-luciferase reporter assay in 3T3 cells. Yoshihara et al. [23] found that the EC50s for MBP and BPA for ERa were 0.68 nM and 1 mM, respectively. For ERb in 3T3 cells, the EC50s for MBP and BPA were 0.46 nM and 89 nM, respectively. Together these experiments by Yoshihara et al. suggest that MBP is a potential disruptor of physiological responses that are mediated by ERa and ERb.
Although MBP and benzestrol have a three carbon linker between their two phenols, their linkers are different. Despite this difference, nM concentrations of MBP activate transcription of the ER in mammalian cells. This raises the possibility that other environmental chemicals with two phenolic rings connected with novel aliphatic linkers may have a physiologically relevant activity towards ERa and ERb in cells with coactivators that can activate the chemical-ER complex. We also note that the 3D models of MBP in ERa and ERb may be useful in the development of new chemicals for use as selective ER agonists or antagonists. Figure 7. Structures of bisphenols that are potent synthetic estrogens. Bisphenols, linked with one, two or three carbons, can have high affinity for ERs. 3-(3-fluoropropyl)cyclofenil, hexestrol and benzestrol have a higher affinity for ERa that does E2 [18,39,40]. doi:10.1371/journal.pone.0046078.g007