You are on page 1of 9

Journal of Steroid Biochemistry and Molecular Biology 213 (2021) 105966

Contents lists available at ScienceDirect

Journal of Steroid Biochemistry and Molecular Biology


journal homepage: www.elsevier.com/locate/jsbmb

In silico identification of novel inhibitors targeting the DNA-binding domain


of the human estrogen receptor alpha
Huiming Cao 1, Yuzhen Sun 1, Ling Wang, Yu Pan, Zhunjie Li, Yong Liang *
Hubei Key Laboratory of Environmental and Health Effects of Persistent Toxic Substances, Institute of Environment and Health, Jianghan University, Wuhan, 430056,
China

A R T I C L E I N F O A B S T R A C T

Keywords: The human estrogen receptor alpha (ERα) is an important regulator in breast cancer development and pro­
Estrogen receptor alpha gression. The frequent ERα mutations in the ligand-binding domain (LBD) can increase the resistance of anti­
DNA binding domain estrogen drugs, highlighting the need to develop new drugs to target ERα-positive breast cancer. In this study, we
Docking
combined molecular docking, molecular dynamics simulations and binding free energy calculations to develop a
Molecular dynamics simulation
Antiestrogen drug
structure-based virtual screening workflow to identify hit compounds capable of interfering with the recognition
Inhibition of proliferation of ERα by the specific response element of DNA. A druggable pocket on the DNA binding domain (DBD) of ERα
was identified as the potential binding site. The hits binding modes were further analyzed to reveal the structural
characteristics of the DBD–inhibitor complexes. The core structure of the lead molecules was synthesized and was
found to inhibit the E2-induced cell proliferation in MCF-7 cell lines. These findings provide an insight into the
structural basis of ligand-ERα for alternate sites beyond the LBD-based pocket. The core structure proposed in
this study could potentially be used as the lead molecule for further rational optimization of the antiestrogen
drug structure with stronger binding of DBD and higher activity.

1. Introduction antagonists need to be designed to target the new druggable site beyond
the LBD of the receptor.
Breast cancer (BC) is the second leading cause of female death after The ERα DNA-binding domain (DBD) plays an essential role during
ovarian cancer. About 70 % of postmenopausal breast cancer patients transcriptional activation through the specific recognition of the estro­
express estrogen receptor alpha (ERα) [1]. Currently, all antiestrogen gen response element (ERE) [15,16]. This suggests that new anties­
drugs in clinical use target the ligand-binding domain (LBD) of the ERα, trogen drugs that can disrupt the interactions between the DBD and the
thereby competing with endogenous estrogen for the receptor binding ERE could be used to overcome antiestrogens resistance. The architec­
[2]. However, recent studies identified several ERα mutations that ture of the DBD–ERE complex reveals that each DBD monomer contains
frequently occur in the LBD of BC patients resistant to antiestrogen two independent zinc finger domains [17]. The first zinc finger domain
therapy [3–5]. These high rates of polymorphisms are associated with includes the so-called "recognition helix", which is mainly responsible
acquired drug resistance, disease recurrence and increased mortality, for the ERE binding by its interactions with the major groove of each
and therefore represent the main clinical challenge for the current use of half-site in the ERE, while the second zinc finger domain forms a contact
antiestrogen therapy [6–9]. The significant mutations occur within the surface for DBD dimerization [18]. Previous studies [19,20] have indi­
L536Q, Y537S, and D538 G genes, which lead to stabilization of the cated that a series of hairpin polyamides could mainly bind to the minor
active conformation of key helix 12 (H12) and constitutive activity of groove of the ERE with allosteric interference with the receptor–DNA
ERα in the absence of endogenous ligands [10,11]. Moreover, previous contacts. Alternatively to polyamides, a DNA bis-intercalator XR5944
studies have also shown that these mutations decreased the binding [21] was designed to inhibit ERα transcription through competitively
affinities of ERα antagonists by reshaping the LBD architecture [12–14]. binding to the DNA major groove, thereby directly impeding the loca­
These findings suggest that to overcome ERα-LBD mutations, new ERα tion of the DBD in this site. Furthermore, disulfide benzamide and

* Corresponding author at: Hubei Key Laboratory of Environmental and Health Effects of Persistent Toxic Substances, Institute of Environment and Health,
Jianghan University, Wuhan, China.
E-mail address: ly76@263.net (Y. Liang).
1
These authors contributed equally.

https://doi.org/10.1016/j.jsbmb.2021.105966
Received 5 May 2021; Received in revised form 29 July 2021; Accepted 11 August 2021
Available online 17 August 2021
0960-0760/© 2021 Elsevier Ltd. All rights reserved.
H. Cao et al. Journal of Steroid Biochemistry and Molecular Biology 213 (2021) 105966

benzisothiazolone derivatives have been demonstrated to disturb the defects in assessing the binding affinity of a compound because the
ERα binding to the ERE via an electrophilic reaction in the zinc finger protein flexibility and explicit solution environment have not been
domains of DBD, which in turn inhibits the proliferation of the Michigan considered in the docking scores of ligand conformations. Furthermore,
cancer foundation-7 (MCF-7) cell lines [22,23]. However, the afore­ in the current complex system, the screened compounds may have a
mentioned molecules do not selectively disturb the transcriptional trend of free diffusion from the surface site of the ERα-DBD to the so­
activation of ERα, but rather universally interact with the specific lution or exchange between the two states. This means that the top-
response element within the DNA or the zinc finger domains of nuclear scored hits predicted by the docking calculations may be false-positive
receptors. results. Therefore MD simulations were performed to identify poten­
Additionally, by using a high throughput screening based on the tial false positive inhibitors among the top 20 scored compound and to
fluorescence anisotropy microplates, theophylline, 8-[(benzylthio) explore dynamic binding behaviors of these compounds based on the
methyl]-(7Cl,8Cl) (TPBM) was identified to disrupt the E2-bound ERα initial docking poses of DBD-ligand complexes.
complex binding to the ERE [24]. Nevertheless, the underlying struc­ The AMBER ff14SB force field [32] and GAFF force field were
tural basis of the DBD-TPBM complex is still unclear, which limited the adopted to construct the topology structures of the ERα-DBD and li­
understanding of the binding mechanism of DBD inhibitor and the gands, respectively. The partial charges of ligands were generated using
further structural optimization based on the structure-based drug the AM1-BCC method in the antechamber module of the AmberTools17
design. Therefore, utilizing structural information on the DBD–ERE [33]. The DBD–ligand complexes were solvated in a cubic water box of
complex will provide a feasible way to identify the lead DBD inhibitor the TIP3P model [34] and were then neutralized by adding the appro­
compounds and hence optimize the structure of the drug. priate sodium (Na+) counterions. The zinc ion (Zn2+) was treated using
In this study, we combined molecular docking [25,26], molecular the cationic dummy atom (CaDA) approach [35] in which the topology
dynamics (MD) simulations [27,28], and binding free energy calcula­ structures of 16 cysteines were specially set to residue type "CYM" for
tions [29] to develop a structure-based virtual screening workflow to Zn2+ chelation. The particle mesh Ewald method [36] and periodic
identify hit compounds capable of interfering with the DBD–ERE in­ boundaries were used to estimate the long-range electrostatic in­
teractions and their corresponding binding sites. The computational teractions with a cutoff value of 10 Å. All of the hydrogen-containing
model has been demonstrated to be effective for the virtual screening of bonds were constrained using the SHAKE algorithm [37], and the inte­
potential antagonists of Y537S mutation to block the mutant-induced gration time step was set to 2 fs using the Verlet leapfrog algorithm. A
metastatic progression of breast cancer [30]. The binding modes of li­ constant temperature was maintained using the Langevin thermostat
gands were elaborated to explore the structural characteristics of po­ with a 2 ps− 1 collision frequency. The Berendsen barostat with isotropic
tential DBD inhibitors. The structural optimization of lead molecules position scaling was used to maintain constant pressure [38]. For
was further carried out and the promising core structure (CS) was syn­ relaxing the entire system, 5000 step minimizations (2500 steps at the
thesized to evaluate the antiproliferative activity of MCF-7 cell lines. steepest descent and a 2500 step conjugate gradient) were restricted
Generally, these findings provide an insight into the structural basis of with a force constant (k) of 100 kcal mol− 1 Å-2 on the solute, followed by
ligand-ERα for alternate sites beyond the LBD-based pocket. full 5000 step minimizations without any constraints. Subsequently, the
entire systems with a weak restraint on the solute (k = 10 kcal mol− 1
2. Materials and methods Å-2) were gradually heated in the canonical ensemble (NVT) from 0 to
300 K over 500 ps. Likewise, the same weak restraint was applied using
2.1. Structure-based virtual screening 500 ps MD simulations in the NPT ensemble for the systems’ density
equilibrium. Finally, 200 ns MD production simulations were performed
The coordinates of the ERα-DBD structure were extracted from the at 1 atm and 300 K in the isothermal-isobaric (NPT) ensemble NPT
crystal structure of the DBD-DNA complex (PDB ID: 1HCQ) [17]. The ensemble. All conventional MD simulations were carried out using a
missing hydrogen atoms were added, and incomplete residues were mixed single-precision or fixed precision pmemd.CUDA module [39] in
repaired using the protein preparation module of the Discovery Studio the AMBER16 package. Trajectory coordinates were saved every 10 ps
2017 software. Virtual screening was carried out using an online tool for subsequent analysis, ultimately resulting in 10,000 structures per
based on the idock docking engine (http://istar.cse.cuhk.edu.hk/idock/ trajectory. The distance between the two mass centres of the compound
). A total of 23,129,083 compounds were first identified from the ZINC and the DBD structure was monitored during the whole simulation. The
database and filtered according to nine physicochemical molecular distance parameter was adopted to reflect the binding stabilities and
properties including; molecular weight (g/mol): [250, 500], partition residence time of the tested compounds on the surface site of the
coefficient xlogP: [1, 3], rotatable bonds: [0, 6], hydrogen bond donors: ERα-DBD.
[0, 6], hydrogen bond acceptors: [0, 6], net charge: [-1, 0], apolar
desolvation (kcal/mol): [0, 10], polar desolvation (kcal/mol): [-40, 0],
2.3. Binding free energy calculations
and polar surface area tPSA (Å2): [60, 80]. Only the compounds satis­
fying all the nine filtering conditions were submitted to the following
The binding free energies of the studied DBD–ligand complexes were
docking protocol. The whole surface of DBD was defined as the potential
calculated using the molecular mechanics generalized Born surface area
binding site of compounds screened in this work. During the docking
(MM/GBSA) approach [30,40], with the MMPBSA.py.MPI module [41]
calculations, the DBD structure was set to be rigid, while the structures
of the parallel version in AmberTools 17 [33]. The calculated binding
of the tested compounds were flexible for conformations samplings. The
free energy (ΔGcal) was determined as a composition of the corre­
top 100 compounds with the highest binding scores and an idock score
sponding enthalpy (ΔH) and entropy (TΔS), using the Eq. (1):
greater than -10 kcal/mol were selected. The LeDock program was
further adapted to perform the re-docking for the above compounds, ΔGcal = ΔH – TΔS (1)
which was found to give a good performance on the ligand-protein
systems [23,31]. The best-scoring docking poses of the top 20 ligands whereby ΔH can be further described using the Eq. (2):
were kept as the initial structures for the stability evaluations of sub­
ΔH = ΔEMM + ΔGsol = ΔEvdw + ΔEele + ΔEint + ΔGGB + ΔGSA (2)
sequent MD simulations.
The term ΔEMM represents the molecular mechanical (MM) energy,
2.2. MD simulations which includes the contribution of van der Waals interactions (ΔEvdw),
electrostatic forces (ΔEele) and the internal energy (ΔEint). The term
The virtual screening of docking-based calculations has intrinsic ΔGsol symbolizes the solvation free energy, including both polar (ΔGGB)

2
H. Cao et al. Journal of Steroid Biochemistry and Molecular Biology 213 (2021) 105966

and nonpolar contribution (ΔGSA). The ΔGGB can be calculated using the 3. Results and discussion
Onufriev–Bashford–case model (igb = 2), while ΔGSA is evaluated using
the solvent-accessible surface area (SASA) using the equation ΔGSA = The potential ERα-DBD inhibitors identified according to structure-
γSASA + β, where the correction constants γ and β are 0.0072 and 0, based virtual screening, MD simulations and binding free energy cal­
respectively. The remaining parameters were set to the default values. A culations are summarized in Fig. 1. A stepwise filtered strategy was
python script, ante-MMPBSA.py, was used to build the topology struc­ performed to reduce the ZINC database size and facilitate the identifi­
tures for dry DBD–ligand complex, DBD and ligand from their original cation of the lead-like molecules with a high binding affinity for the
solvated topology files for the implicit solvent calculations. All of the DBD. After a physicochemical property-based filtration, the multistep
energy components were calculated using 500 snapshots extracted from docking approaches (idock and Ledock) were used to rank the binding
the 100–200 ns trajectories. The key residues responsible for the affinities of ligands with the active pocket of DBD. Subsequently, the
recognition process were identified using the enthalpy contributions stability of the complex formations was evaluated by the MD simulations
were decomposed into each residue. Due to the high computational cost, in the explicit solvent environment. Further interaction pattern analysis
only 100 snapshots from the last 100 ns trajectories were extracted for was conducted based on the binding free energy calculations of the MM/
the entropy contribution calculations (TΔS) using the NMODE module GBSA method. Finally, the scaffold structure of lead compounds was
of AmberTools17 [42]. synthesized, and its antiproliferative activity was validated in MCF-7
cell lines.
2.4. Cytotoxicity assays
3.1. Screening of the top-scored hits
The MCF-7 breast cancer cell line was grown in Dulbecco’s modified
eagle’s medium with 10 % FBS and cultured in a 5% carbon dioxide
A total of 1,848,620 compounds were screened from the ZINC
(CO2) enriched atmosphere at 37 ◦ C. Exponentially growing cells were
database and subsequently filtered to discover the potential ERα-DBD
seeded into 96-well plates at a density of 8000 cells per well and allowed
inhibitors with high binding affinities. On the basis of high-throughput
to grow for 24 h. Following removal of the cell culture medium, MCF-7
virtual screening, the top scored compounds were retained as possible
cells were incubated for 48 h and treated with various concentrations of
lead molecules. We also predicted the docking score of TPBM as a pos­
the newly synthesized compound (from 1 to 10,000 nM) with or without
itive control using the same idock docking protocol. The result showed
E2 (1 nM). Subsequently, MCF-7 cells were treated with the cell counting
that these top-ranked compounds had more favorable docking scores
kit-8 (CCK-8) reagent at 37 ◦ C for two hours, and the solution was
than TPBM (-6.07 kcal/mol), suggesting that these compounds have
converted to a quantifiable yellow dye using the mitochondrial de­
higher binding affinities with ERα-DBD. Additionally, we refined the
hydrogenases present within viable cells. Absorbency was measured at
binding modes of compounds using the Ledock program. The superim­
450 nm using a microplate reader.
positions of binding poses with best docking scores between idock and
Ledock displayed the low root-mean-square deviation (RMSD) values (<

Fig. 1. Schematic representation of the work-flow in the current study. The virtual screening of ERα-DBD inhibitors from the ZINC database.

3
H. Cao et al. Journal of Steroid Biochemistry and Molecular Biology 213 (2021) 105966

2.5 Å) for heavy atoms of the compounds, which means that the current RMSD value around 2.0 Å. The small amplitude indicated that the
docking results were reliable and could therefore be used in the post- studied complexes rapidly reached the dynamic equilibrium after the
docking analysis. initial simulation stage. Thus, the last 100 ns trajectories were extracted
Following an evaluation of the binding modes used by these com­ to assess the binding affinity of the protein-ligand complex by means of
pounds, we found that most compounds were located on the surface the binding free energy calculations. TPBM was used as a positive con­
pocket of the DBD monomer and showed similar binding patterns trol to screen potential DBD inhibitors with stronger binding free en­
(Fig. 2, left panel). Interestingly, it should be noted that this surface ergies. Among the tested compounds, ZINC49047307, ZINC49089886,
pocket could be recognized by the phosphate backbones and base pairs and ZINC68979694 displayed lower binding free energies when
of the ERE sequence (Fig. 2, right panel). Based on the above results, we compared with TPBM, suggesting that the three hits have relatively
speculated that these lead molecules might competitively bind in the stronger binding affinities on the surface pocket of DBD (Table 1 and
DBD recognition helix, thereby directly obstructing the loading of DBD S1).
on the special DNA fragment and inhibiting the formation of ERα–DNA Subsequent investigations on the nature of dominant binding forces
complex and subsequent transcriptional activation. for the DBD-ligand interactions were performed based on the energy
decomposition calculations for the stable binding simulation systems
3.2. Evaluations of binding stability for the top-scored hits (Table 1). Noticeably, the negative values of ΔEvdw and ΔEele indicate
that the van der Waals and electrostatic interactions of the tested ligands
MD simulations were performed to identify potential false positive favor the DBD recognition. Specifically, the electrostatic interactions
inhibitors among the top 20 scored compounds. Based on the distance (ΔEele) play a significant role in the binding ability of ZINC49047307,
between two mass centers of the compound and DBD structure, the ZINC49089886, and ZINC68979694 at the surface site of the DBD,
binding stabilities and residence time of the tested compounds on the which was counteracted by unfavorable polar contributions of the sol­
surface site of ERα DBD were identified and illustrated in Fig. S1. Two vation free energy (ΔGGB). Conversely, the contributions of ΔEvdw were
binding behaviors were observed for the compounds and divided into observed as the most important contribution for stabilizing the DBD-
cluster 1 and 2. For cluster 1, the distance was lower than 25 Å for most TPBM complex. This meant that relative to the neutral molecule
compounds, suggesting that the compounds have relatively stable TPBM, the top hits have the negatively charged carboxylic acid groups to
binding on the DBD surface. For example, ZINC32180841 maintained a mimic the phosphate backbone of DNA structure, thus leading to the
slight structural fluctuation relative to the initial docking pose (Fig. 3A). appropriate electrostatic attraction between the interfaces of the
Moreover, a relatively large distance of approximately 15 Å was found protein-ligand complex. In addition, the entropy contribution (TΔS)
for some compounds in cluster 1 (such as ZINC49089886), suggesting displayed slight changes from -21.79 to -21.54 kcal/mol among these
that the ligands attempted to accommodate the surface pocket of the compounds, suggesting that the numerical values of TΔS did not influ­
DBD via local conformational adjustments within the solution environ­ ence the relative order rank of binding affinities when screening for
ment (see red dashed box in Fig. 3A). potential DBD inhibitors. This implies that the calculating only the
All the compounds in cluster 2 had a distance larger than 35 Å and enthalpy contribution (ΔH) could accelerate the virtual screening pro­
showed a trend of escaping from the surface of the ERα-DBD ligand, even cess due to the high computational cost of TΔS.
completely entering into the solution and eventually reaching a distance
greater than 45 Å, such as in the case of ZINC57521965 (Fig. 3B). The 3.4. Structural characteristics in the recognition process of potential DBD
dynamic behaviors meant that these compounds were forming unstable inhibitors
complexes and only binding on the DBD surface for a short residence
time. Based on these findings, we concluded that these compounds could The clustering analysis method was used to gain a better under­
not interfere with the ERα-DBD interactions and were therefore classi­ standing of the molecular basis involved in recognition of the DBD-
fied as false positive hits to be excluded from further analysis. ligand complex and its binding characteristics. The respective poses of
ligands from the largest cluster from the last 100 ns trajectories were
3.3. Evaluations of binding free energy for potential DBD inhibitors extracted to elaborate the binding modes of the complexes. As expected,
three of them (ZINC49047307, ZINC49089886, and ZINC68979694)
The RMSD values of the DBD backbone Cα atoms of the compounds occupied the same binding site of the DBD surface and shared a similar
in cluster 1 were further assessed, as shown in Fig. S2. Most complex interaction pattern. Analysis of the binding modes revealed that the
systems showed an increase in the RMSD value in the initial phase of positively charged center consisted of the residues Arg234 and Arg241
simulations and reached the equilibrium after 10 ns simulations with an and provided the important electrostatic attractions with the carboxyl

Fig. 2. The DNA-binding domain (DBD) dimer was shown by monomer structures in blue and cyan (PDB id: 1HCQ). The top 20 score compounds obtained from the
docking-based virtual screening were showed as stick models. The binding poses of compounds were aligned with the ERE on the surface site of DBD structure (red
dashed box).

4
H. Cao et al. Journal of Steroid Biochemistry and Molecular Biology 213 (2021) 105966

Fig. 3. The distance between mass centers of compound and DBD structure were plot during the 200 ns MD simulations for the stable binding molecules (A) and false
positive molecules (B) from the molecular docking results. The representative complex conformations for ZINC49089886 and ZINC57521965 were shown in
red frame.

The interaction contacts showed that Arg234 and Arg241 could form the
Table 1
electrostatic interactions with the phosphate backbone of the G31 ERE
Binding free energy components of different simulation systems based on the
base (Fig. 4E). Next, we built the DBD-ERE complex system to identify
MM/GBSA method (kcal/mol).
the important residues in the DBD surface for the ERE recognition by
Systems ZINC49047307 ZINC49089886 ZINC68979694 TPBM
using the same MD protocol. The energy decomposition results indicated
ΔEvdw − 22.21 ± 4.20 − 21.83 ± 3.66 − 17.26 ± 3.19 − 29.01 ± that Arg234 and Arg241 rendered the top energy contributions among
2.50 all the DBD residues, thus playing a significant role in the ERE binding.
− 272.32 ± − 257.76 ± − 290.98 ± − 13.17 ±
Meanwhile, the energy decompositions of the DBD-ligand systems sug­
ΔEele
20.38 17.46 20.21 6.49
ΔGGB 254.97 ± 16.96 241.13 ± 15.72 272.73 ± 18.83 21.81 ± gested that the two residues have relatively stronger contributions
4.16 compared to other residues, which were comparable with that in the
ΔGSA − 3.65 ± 0.39 − 3.43 ± 0.41 − 3.31 ± 0.33 − 3.42 ± DBD-ERE complex system (Fig. 4F). Consequently, based on the above
0.23
energy analysis, we proposed that the positively charged center con­
ΔH − 43.22 ± 5.81 − 41.89 ± 4.37 − 38.82 ± 4.33 − 23.79 ±
3.53
sisting of Arg234 and Arg241 could attract the carboxyl group of po­
TΔS − 21.74 ± 3.08 − 22.79 ± 3.64 − 22.54 ± 3.41 − 21.65 ± tential DBD inhibitors, which mimics the function of the phosphate
2.50 backbone ERE to form the strong electrostatic recognition, eventually
ΔGcal − 21.48 ± 6.93 − 19.10 ± 6.03 − 16.28 ± 5.82 − 2.14 ± leading to an increase in the binding affinity of DBD-ligand complex,
4.72
thereby blocking the loading of special DNA motifs on the surface
ΔH =ΔEvdw + ΔEele + ΔGGB + ΔGSA, ΔEnon-polar = ΔEvdw+ ΔGSA, ΔEpolar = ΔEele binding pocket of DBD.
+ ΔGGB, ΔGcal = ΔH – TΔS.

3.5. Structural optimization of potential DBD inhibitors


group of the compounds (Fig. 4A-C). On the other hand, the residues
Phe208, Ala207, and Gly204 also stabilized the protein-ligand com­
The compound ZINC49047307 and its isomer ZINC49089886 were
plexes by means of the hydrophobic interactions with the benzene rings
retained as the final candidates of the DBD inhibitors because of their
of ligands. Conversely, TPBM acted as the positive control by adopting
relatively lower binding free energies. Following the superimposition of
an alternative binding pose in the DBD surface, where Tyr219 and
the last 100 ns trajectories, unstable binding behaviour was observed in
Asn217 provided the H-bonding interactions for the ligand binding.
region B of the two compounds, which implies that this molecular region
Compared to the hits, the TPBM molecule moved away from the posi­
could be replaced or modified for further structural optimization, as
tively charged center, leading to a decrease in the binding affinity. This
indicated in Fig. 5. Considering the strong H-bonding interactions of the
might be due to a reduction in the electrostatic attractions during the
carbonyl group in region A, we therefore hypothesized that this region
ligand recognition caused by the neutrality of the TPBM form.
might serve as the CS for the DBD recognition (Fig. 5). The major sta­
Additionally, we scrutinized the crystal structure of the DBD-ERE
bilization energies for the complex formation had been derived from this
complex to investigate the structural features of complex formation.
CS.

5
H. Cao et al. Journal of Steroid Biochemistry and Molecular Biology 213 (2021) 105966

Fig. 4. Binding modes of potential DBD inhibitors (A-D). The key residues for binding were shown in stick mode. The interaction contacts between the residues
Arg234 and Arg241 and the phosphate backbone of G31 base of ERE (E). The energy contributions of Arg234 and Arg241 for the DBD-ligand complex systems (F).

To verify this hypothesis, we implemented the MD simulations to optimization with some modification to the different functional groups.
assess the binding capacity of CS based on the MM/GBSA calculations. A series of substituted CS analogues have been designed to enhance
The result indicated that the binding free energy was -23.41 kcal/mol for the binding affinity of the molecules with DBD. Among these molecules,
the CS, which was relatively superior to that of ZINC49047307 (-21.48 the CS analogue 30 (CS30) showed a relatively more favorable Ledock
kcal/mol) (Fig. 5). On the basis of the energy decomposition analysis score compared to other newly designed molecules (Fig. S3). The MD-
(Table 1 and S2), we found that the enthalpy contribution of based binding free energy calculations displayed a stronger binding af­
ZINC49047307 was more favorably than that of CS (-43.22 vs -40.52 finity for CS30 when compare with CS, suggesting that it is more likely
kcal/mol), whereas the entropy contribution for ZINC49047307 was to disturb the binding between DBD and ERE. Subsequently, we
-21.74 kcal/mol lower when compared with that for CS (-17.11 kcal/ compared the different compositions of binding free energy to identify
mol). The results signified that the remove of unstable fragment of which factors are more likely to improve the binding affinity of CS30
ZINC49047307 decreased the molecular size, thereby leading to a (Table S2). The results indicated that the van der Waals and electro­
reduction in the entropic penalty during the ligand binding and static forces are more favorable for the binding of CS30 relative to CS,
impairing the enthalpy contribution simultaneously. This delicate which means that the substitution of the phenol ring increased the hy­
enthalpy-entropy balance finally improved the total binding free energy drophobic and H-binding interactions by stabilizing the binding of this
of the CS, and eventually, the stably binding on the DBD surface. functional group on the surface pocket of DBD. Although the increased
Therefore, we selected the CS as the initial point for structural molecular size of CS30 aggravated the entropic penalty (-22.54 kcal/

6
H. Cao et al. Journal of Steroid Biochemistry and Molecular Biology 213 (2021) 105966

Fig. 5. Parent structure of potential DBD inhibitors was identified and optimized for the more stable binding and favorable binding free energy.

mol), this adverse effect was compensated by the above-mentioned TPBM and CS are non-hepatotoxic (Table 2).
favorable intermolecular interactions.
3.7. Antiproliferative activities of the newly synthesized compound
3.6. Predicted pharmacokinetic properties of potential DBD inhibitors
The CS structure was ordered and synthesized by the WuXi AppTec
The pharmacokinetic properties of compounds play a significant role (Shanghai) Co., Ltd. and synthetic schemes were summarized as sup­
in drug development processes. By structural optimizations, an ideal plementary data (Fig. S4). The antiproliferative activities of the CS were
drug should be non-toxic as well as outstanding pharmacokinetic tested against the MCF-7 (human breast cancer cell) cell lines using CCK-
properties. To evaluate the adverse effects of potential DBD inhibitors 8 assay, which was used to assess the estrogenic or anti-estrogenic ac­
proposed by our computational study, the in silicon predictions of ab­ tivity of a compound. Inhibition of cell proliferation by the CS with or
sorption, distribution, metabolism, excretion, and toxicity were per­ without E2 (an endogenous ligand of ERα, 10− 9 M) was measured from
formed for the screened and structurally optimized molecules. All the 10-9 to 10-5 M concentrations. The results showed that the CS could
tested molecules showed the high human intestinal absorption above 70 cause significant cytotoxicity of MCF-7 cell lines, in a dose-dependent
% (Table 2). In distribution properties, the blood-brain barrier perme­ manner (Fig. 6, cyan bar). Co-exposure of CS with E2 indicated that
ability was evaluated as logBB, a logarithmic ratio of brain to plasma CS could inhibit the E2-induced cell proliferation, especially at the low
drug concentrations. The numerical values of logBB indicated that concentration (10− 9 M) when compared to that of only E2-exposed MCF-
ZINC49047307 is poorly distributed to the brain, while other molecules 7 cell lines (Fig. 6, grey bar), which means that the CS could cause anti-
were more likely to cross the blood-brain barrier. Moreover, the estrogenic effects through ERα-mediated signaling pathway. To further
assessment of the metabolic behaviour indicated that CYP3A4, an iso­ explore the expression alterations in estrogen-responsive genes, we
form of cytochrome P450, might be the ZINC49047307 and TPBM carried out the real-time PCR experiments to determine whether the CS
target. The drug excretion was measured by the predictions of the he­ treatment resulted in a disruption in ERα-mediated transcription. The
patic and renal clearances. The results indicated a rapid clearance for all details of method were summarized in the Supporting Information (SI).
the molecules due to their drug-likeness behaviour (Table 2). By the comparison of the expression of estrogen-regulated genes (TFF1
For toxicity predictions, the standard lethal dosage 50 value (LD50) and EGR3), we found that the mRNA levels of TFF1 and EGR3 were
was used to estimate the relative toxicity of the different compounds. decreased in a dose-dependent manner after the treatment of CS with or
The LD50 was defined as the quantity of a compound given all at once without E2 (10-9 M). The results suggested that CS could attenuate
that lead to the death of 50 % of a group of test animals. The results endogenous ERα signaling, thus resulting in the disruption of estrogen-
suggested that the designed molecules had low toxicity levels. Addi­ modulated endogenous gene expression in MCF-7 cell lines (Fig. S5).
tionally, the Ames test, a common method used to assess mutagenic Moreover, we speculated that the other potential targets or inhibition
potentials of compounds using bacteria, was negative for all designed mechanisms of CS for the MCF-7 cell lines may provide the multiple
compounds. This means that these compounds are not mutagenic and antagonism effects on the proliferation and progression breast cancer as
will therefore not act as a carcinogen. Furthermore, a compound is well. Based on the above results, we confirmed that CS, as the scaffold
defined as hepatotoxicity if at least one pathological or physiological structure of potential DBD inhibitors, could potentially be used to
liver event is associated with the drug. The results indicated that both develop new antiestrogens therapy for breast cancer.

Table 2
ADMET properties of potential DBD inhibitors.
Absorption Distribution Metabolism Excretion Toxicity
Molecules Intestinal absorption BBB permeability CYP3A4 substrate Total Clearance AMES toxicity LD50 Hepatotoxicity
% log BB Yes/No log ml/min/kg Yes/No mol/kg Yes/No

TPBM 89.356 − 0.501 Yes 0.671 No 2.629 No


ZINC49047307 70.215 − 1.168 Yes 0.248 No 2.427 Yes
CS 95.007 0.008 No 0.725 No 2.169 No
CS30 93.646 − 0.247 No 0.55 No 2.577 Yes

7
H. Cao et al. Journal of Steroid Biochemistry and Molecular Biology 213 (2021) 105966

Zhunjie Li: Visualization, Investigation.


Yong Ling: Reviewing and Editing.

Declaration of Competing Interest

The authors declare no conflict of interest, financial or otherwise.

Acknowledgements

This work was supported by the National Natural Science Foundation


of China (21806058, 21607060, and 21777061).

Appendix A. Supplementary data

Supplementary material related to this article can be found, in the


online version, at doi:https://doi.org/10.1016/j.jsbmb.2021.105966.

References
Fig. 6. Anti-proliferative activity of synthesized compound CS in MCF-7 human
breast cancer cells. Cells were exposed to 0.1 % DMSO (vehicle control), 1 nM
[1] H.S. Rugo, et al., Endocrine therapy for hormone receptor–positive metastatic
E2 (positive control), 10− 9–10-5 M CS, or CS + E2 for 2 days. Results are breast cancer: american society of clinical oncology guideline, J. Clin. Oncol. 34
expressed as the mean ± SD of triplicate measurements in three replicate (2016) 3069–3103.
samples (* means p ≤ 0.05 and ** means p ≤ 0.005). [2] E.A. Ariazi, J.L. Ariazi, F. Cordera, V.C. Jordan, Estrogen receptors as therapeutic
targets in breast cancer, Curr. Top. Med. Chem. 6 (2006) 181–202.
[3] C. Thomas, J. Gustafsson, Estrogen receptor mutations and functional
In a previous report, anacardic acid (C24:1), known as histone ace­ consequences for breast cancer, Trends Endocrin. Met. 26 (9) (2015) 467–476.
tyltransferase inhibitor [43], was shown to inhibit breast cancer cell [4] J.M. Spoerke, S. Gendreau, K. Walter, J. Qiu, T. Wilson, H. Savage, et al.,
Heterogeneity and clinical significance of ESR1 mutations in ER-positive metastatic
proliferation regardless of their endocrine or tamoxifen sensitivity [44]. breast cancer patients receiving fulvestrant, Nat. Commun. 7 (2016) 11579.
This study also demonstrated that C24:1 could inhibit the interactions [5] C. Desmedt, J. Pingitore, F. Rothé, et al., ESR1 mutations in metastatic lobular
between the ERα and pS2 gene in MCF-7 cell lines. Based on the mo­ breast cancer patients, NPJ Breast Cancer 5 (1) (2019) 1–7.
[6] R. Jeselsohn, C. De Angelis, M. Brown, et al., The evolving role of the estrogen
lecular docking prediction, the authors speculated that C24:1 might bind receptor mutations in endocrine therapy-resistant breast cancer, Curr. Oncol. Rep.
on the surface of the zinc DBD fingers rather than in the LBD. Compared 19 (5) (2017) 5.
to C24:1, the potential DBD inhibitors tested in this study displayed an [7] Y. Zhao, M.J. Laws, V.S. Guillen, et al., Structurally novel antiestrogens elicit
differential responses from constitutively active mutant estrogen receptors in
alternative mechanism of action for the CS compounds, whereby the
breast cancer cells and tumors, Cancer Res. 77 (20) (2017) 5602–5613.
exposed surface pocket could bind on the ERα-DBD and act as a [8] N.C. Turner, C. Swift, L. Kilburn, et al., ESR1 mutations and overall survival on
drug-target site with high binding affinity, thereby blocking the tran­ fulvestrant versus exemestane in advanced hormone receptor–positive breast
scriptional activation of E2-induced in MCF-7 cells. Furthermore, the cancer: a combined analysis of the phase III SoFEA and EFECT trials, Clin. Cancer
Res. 26 (19) (2020) 5172–5177.
screened potential small molecule inhibitors for the druggable binding [9] L. Clusan, P.L. Goff, G. Flouriot, F. Pakdel, A closer look at estrogen receptor
site on the AR-DBD surface have been demonstrated to be effective in mutations in breast Cancer and their implications for estrogen and antiestrogen
inhibiting the transcriptional activity of androgen receptor in the pre­ responses, Int. J. Mol. Sci. 22 (2) (2021) 756.
[10] D.R. Robinson, Y.M. Wu, P. Vats, F. Su, R. Lonigro, X. Cao, et al., Activating ESR1
vious studies [45,46]. However, these surface sites of the DBD for the mutations in hormone-resistant metastatic breast cancer, Nat. Gene. 45 (2013)
nuclear receptors are generally shallow with transient pockets. This 1446–1451.
makes it difficult to design a drug that can stably bind in these sites of [11] W. Toy, Y. Shen, H. Won, et al., ESR1 ligand-binding domain mutations in
hormone-resistant breast cancer, Nat. Genet. 45 (12) (2013) 1439.
DBD. Thus, the current computational workflow combining molecular [12] S.W. Fanning, C.G. Mayne, V. Dharmarajan, et al., Estrogen receptor alpha somatic
docking, MD simulations and binding free energy could provide a mutations Y537S and D538G confer breast cancer endocrine resistance by
feasible strategy for the development of potential DBD inhibitors of ERα. stabilizing the activating function-2 binding conformation, elife 5 (2016), e12792.
[13] M. Pavlin, A. Spinello, M. Pennati, N. Zaffaroni, S. Gobbi, A. Bisi, et al.,
A computational assay of estrogen receptor α antagonists reveals the key common
4. Conclusion structural traits of drugs effectively fighting refractory breast cancers, Sci. Rep. 8
(1) (2018) 1–11.
[14] A. Khan, M. Junaid, C.D. Li, et al., Dynamics insights into the gain of flexibility by
In the current study, potential antiestrogens capable of inhibiting the Helix-12 in ESR1 as a mechanism of resistance to drugs in breast cancer cell lines,
binding of DBD to the target DNA were identified through the use of an Front. Mol. Biosci. 6 (2020) 159.
integrated computational approach. The CS molecule of these DBD-ERE [15] C.M. Klinge, Estrogen receptor interaction with estrogen response elements,
Nucleic Acids Res. 29 (2001) 2905–2919.
inhibitors was synthesized, and antiproliferative activities of the newly [16] J.S. Carroll, C.A. Meyer, J. Song, W. Li, T.R. Geistlinger, J. Eeckhoute, A.
designed drug were validated against MCF-7 cell lines. The experimental S. Brodsky, E.K. Keeton, K.C. Fertuck, G.F. Hall, Q. Wang, S. Bekiranov,
results suggested that the CS clearly inhibits E2-induced cell prolifera­ V. Sementchenko, E.A. Fox, P.A. Silver, T.R. Gingeras, X.S. Liu, M. Brown, Genome-
wide analysis of estrogen receptor binding sites, Nat. Genet. 38 (2006) 1289–1297.
tion. Thus, we believe that the CS proposed in this study could be used as
[17] J.W. Schwabe, L. Chapman, J.T. Finch, D. Rhodes, The crystal structure of the
the lead molecule for further structural optimization and further estrogen receptor DNA-binding domain bound to DNA: how receptors discriminate
experimental DBD binding affinities and pharmacokinetic properties of between their response elements, Cell 75 (1993) 567–578.
CS and its analogues will be evaluated in our future work. [18] C. Helsen, S. Kerkhofs, L. Clinckemalie, L. Spans, M. Laurent, S. Boonen,
D. Vanderschueren, F. Claessens, Structural basis for nuclear hormone receptor
DNA binding, Mol. Cell. Endocrinol. 348 (2012) 411–417.
Author statement [19] D. Micah, D. Gearhart, E. Liliane, M. Jennifer, B.D. Christian, E.W. Peter Peter, M.
J. Gottesfeld, Inhibition of DNA binding by human estrogen-related receptor 2 and
estrogen receptor r with minor groove binding polyamides, Biochemistry 44 (2005)
Huiming Cao: Conceptualization, Methodology, Writing-Original 4196–4203.
draft. [20] D.M. Chenoweth, A.H. Daniel, W.P. John, D. Christian, B.D. Peter, Cyclic pyrrole -
Yuzhen Sun: Data preparation and analysis, Software. imidazole polyamides targeted to the androgen response element, J. Am. Chem.
Soc. 131 (2009) 7182–7188.
Ling Wang: Experimental validation. [21] C. Punchihewa, A. De Alba, N. Sidell, D. Yang, XR5944: A potent inhibitor of
Yu Pan: Experimental validation. estrogen receptors, Mol. Cancer Ther. 6 (1) (2007) 213–219.

8
H. Cao et al. Journal of Steroid Biochemistry and Molecular Biology 213 (2021) 105966

[22] L.H. Wang, X.Y. Yang, X. Zhang, K. Mihalic, Y.X. Fan, W. Xiao, O.M. Howard, [34] W.L. Jorgensen, J. Chandrasekhar, J.D. Madura, R.W. Impey, M.L. Klein,
E. Appella, A.T. Maynard, W.L. Farrar, Suppression of breast cancer by chemical Comparison of simple potential functions for simulating liquid water, J. Chem.
modulation of vulnerable zinc fingers in estrogen receptor, Nat. Med. 10 (1) (2004) Physics. 79 (1983) 926–935.
40–47. [35] Y.P. Pang, Successful molecular dynamics simulation of two zinc complexes
[23] L.H. Wang, X.Y. Yang, X. Zhang, P. An, H.J. Kim, J. Huang, R. Clarke, C.K. Osborne, bridged by a hydroxide in phosphotriesterase using the cationic dummy atom
J.K. Inman, E. Appella, W.L. Farrar, Disruption of estrogen receptor DNA-binding method, Proteins 45 (2001) 183–189.
domain and related intramolecular communication restores tamoxifen sensitivity [36] T. Darden, D. York, L. Pedersen, Particle mesh Ewald: an N ⋅ log (N) method for
in resistant breast cancer, Cancer Cell 10 (6) (2006) 487–499. Ewald sums in large systems, J. Chem. Phys. 98 (12) (1993) 10089–10092.
[24] C. Mao, N.M. Patterson, M.T. Cherian, I.O. Aninye, C. Zhang, J.B. Montoya, [37] J.P. Ryckaert, G. Ciccotti, H.J. Berendsen, Numerical integration of the cartesian
J. Cheng, K.S. Putt, P.J. Hergenrother, E.M. Wilson, A.M. Nardulli, S.K. Nordeen, D. equations of motion of a system with constraints: molecular dynamics of n-alkanes,
J. Shapiro, A new small molecule inhibitor of estrogen receptor alpha binding to J. Comput. Phys. 23 (3) (1997) 327–341.
estrogen response elements blocks estrogen-dependent growth of cancer cells, [38] H.J. Berendsen, J.P.M. Postma, W.F. van Gunsteren, A. DiNola, J. Haak, Molecular
J. Biol. Chem. 283 (19) (2008) 12819–12830. dynamics with coupling to an external bath, J. Chem. Phys. 81 (8) (1984)
[25] J. Tu, L.T. Song, H.L. Zhai, et al., Selective mechanisms and molecular design of 2, 3684–3690.
4 Diarylaminopyrimidines as ALK inhibitors, J. Int. J. Biol. Macromol. 118 (2018) [39] R. Salomon-Ferrer, A.W. Gotz, D. Poole, G.S. Le, R.C. Walker, Routine microsecond
1149–1156. molecular dynamics simulations with AMBER on GPUs. 2. explicit solvent particle
[26] Z. Qiu, K. Qu, F. Luan, et al., Binding specificities of estrogen receptor with mesh ewald, J. Chem. Theory Comput. 9 (2013) 3878–3888.
perfluorinated compounds: a cross species comparison, Environ. Int. 134 (2020) [40] T. Hou, J. Wang, Y. Li, W. Wang, Assessing the performance of the molecular
105–284. mechanics/Poisson Boltzmann surface area and molecular mechanics/generalized
[27] M. Bello, L. Saldaña-Rivero, J. Correa-Basurto, et al., Structural and energetic basis born surface area methods. II. The accuracy of ranking poses generated from
for the molecular recognition of dual synthetic vs. Natural inhibitors of EGFR/ docking, J. Comput. Chem. 32 (2011) 866–877.
HER2, J. Int. J. Biol. Macromol. 111 (2018) 569–586. [41] B.R. Miller III, T.D. McGee Jr., J.M. Swails, N. Homeyer, H. Gohlke, A.E. Roitberg,
[28] H. Cao, Z. Zhou, L. Wang, G. Liu, Y. Sun, Y. Wang, T. Wang, Y. Liang, Screening of MMPBSA. py: an efficient program for end-state free energy calculations, J. Chem.
potential PFOS alternatives to decrease liver bioaccumulation: experimental and Theory Comput. 8 (2012) 3314–3321.
computational approaches, Environ. Sci. Technol. 53 (2019) 2811–2819. [42] H. Cao, X. Li, W. Zhang, L. Wang, Y. Pan, Z. Zhou, M. Chen, A. Zhang, Y. Liang,
[29] T. Hou, J. Wang, Y. Li, W. Wang, Assessing the performance of the MM/PBSA and M. Song, Anti-estrogenic activity of tris(2,3-dibromopropyl) isocyanurate through
MM/GBSA methods. 1. The accuracy of binding free energy calculations based on disruption of co-activator recruitment: experimental and computational studies,
molecular dynamics simulations, J. Chem. Inf. Model. 51 (2010) 69–82. Arch. Toxicol. 92 (2018) 1471–1482.
[30] M. Pavlin, L. Gelsomino, I. Barone, A. Spinello, S. Catalano, S. Andò, A. Magistrato, [43] K. Balasubramanyam, V. Swaminathan, A. Ranganathan, T.K. Kundu, Small
Structural, Thermodynamic, and kinetic traits of antiestrogen-compounds molecule modulators of histone acetyltransferase p300, J. Biol. Chem. 278 (21)
selectively targeting the Y537S mutant estrogen receptor α transcriptional activity (2003) 19134–19140.
in breast cancer cell lines, Front. Chem. 7 (2019) 602. [44] D.J. Schultz, N.S. Wickramasinghe, M.M. Ivanova, S.M. Isaacs, S.M. Dougherty,
[31] H. Cao, F. Wang, Y. Liang, H. Wang, A. Zhang, M. Song, Experimental and Y. Imbert-Fernandez, A.R. Cunningham, C. Chen, C.M. Klinge, Anacardic acid
computational insights on the recognition mechanism between the estrogen inhibits estrogen receptor alpha-DNA binding and reduces target gene transcription
receptor α with bisphenol compounds, Arch. Toxicol. 91 (2017) 3897–3912. and breast cancer cell proliferation, Mol. Cancer Ther. 9 (3) (2010) 594–605.
[32] J.A. Maier, C. Martinez, K. Kasavajhala, L. Wickstrom, K. Hauser, C. Simmerling, [45] K. Dalal, M. Roshan-Moniri, A. Sharma, H. Li, F. Ban, M. Hessein, P.S. Rennie,
ff14SB: improving the accuracy of protein side chain and backbone parameters Selectively targeting the DNA-binding domain of the androgen receptor as a
from ff99SB, J. Chem. Theor. Comput. 11 (2015) 3696–3713. prospective therapy for prostate cancer, J. Boil. Chem. 289 (38) (2014)
[33] D.R. Roe, T.E. Cheatham, PTRAJ and CPPTRAJ: software for processing and 26417–26429.
analysis of molecular dynamics trajectory data, J. Chem. Theory Comput. 9 (2013) [46] H. Li, F. Ban, K. Dalal, E. Leblanc, K. Frewin, D. Ma, A. Cherkasov, Discovery of
3084–3095. small-molecule inhibitors selectively targeting the DNA-binding domain of the
human androgen receptor, J. Med. Chem. 57 (15) (2014) 6458–6467.

You might also like